Protein oligomerization in the bacterial outer membrane (Review)

Abstract

The formation of homo-oligomeric assemblies is a well-established characteristic of many soluble proteins and enzymes. Oligomerization has been shown to increase protein stability, allow allosteric cooperativity, shape reaction compartments and provide multivalent interaction sites in soluble proteins. In comparison, our understanding of the prevalence and reasons behind protein oligomerization in membrane proteins is relatively sparse. Recent progress in structural biology of bacterial outer membrane proteins has suggested that oligomerization may be as common and versatile as in soluble proteins. Here we review the current understanding of oligomerization in the bacterial outer membrane from a structural and functional point of view.

• Bacterial outer membrane
• oligomerization
• membrane protein
• structural biology
• β-barrel

Introduction

Protein folding and assembly generally follow a series of hierarchical events from translation of the genetic information into a linear polypeptide, to formation of secondary structural elements and arrangement of these into well-defined three-dimensional structures. However, a large fraction of proteins go one step further, oligomerizing to adopt a defined quaternary structure, whereby multiple copies of the same or different protein subunits can organize into homo- and hetero-oligomeric complexes. In particular, homo-oligomerization is widespread, found in over 50–70% of known complexes [1], [2]. Such multimerization can bring several functionally important advantages [1], [3–5]: (1) Oligomerization can give shape to active sites or even lead to their compartmentalization. It has been estimated that about one sixth of oligomeric enzymes have their active sites located at oligomeric interfaces; (2) Oligomerization can allow cooperativity between subunits, enabling allosteric regulation as an additional level of control; (3) Oligomerization can serve as a tool to create multivalency in active or interactive sites, increasing affinity of the complexes for substrates or binding partners; (4) Oligomerization can enhance protein stability; (5) For those proteins that have different activities in their oligomeric and monomeric states, oligomerization can provide an additional level of regulation; and (6) For hetero-oligomeric complexes, oligomerization may allow the formation of enzymatic and signaling cascades.

Our present understanding of protein oligomerization as detailed above stems almost entirely from structural and biophysical studies of soluble proteins. Whether or not these same principles apply to membrane proteins and to which extent remains to be evaluated. Membrane proteins, by definition, reside in an apolar milieu. This specific environment imposes restriction on the tertiary, and possibly quaternary structures that can be adopted. Available structures show transmembrane stretches in integral membrane proteins are confined almost exclusively to α-helix bundles in the case of cytoplasmic membrane, golgi and endoplasmic reticulum proteins, or closed β-barrels in the case of plant plastid, mitochondrial and bacterial outer membrane proteins (OMPs). For both families, oligomerization is observed quite frequently: 105 and 22 out of 159 and 56 non-redundant high-resolution structures of α-helical and β-barrel proteins, respectively, show oligomers. For clarity, only intramembranous multimerizations (e.g. with each subunit spanning the membrane) have been considered, not taking into account complexes with soluble protein partners. Out of these, 95 and 19 α-helical and β-barrel membrane proteins are homo-oligomeric complexes, respectively (an exhaustive list of available structures can be found at the membrane protein orientation database (http://opm.phar.umich.edu/)).

In this review, we focus on the recent advances in structural biology of bacterial integral OMPs. With the sole exception of Wza [6], OMPs of known structures cross the lipid bilayer as β-barrels [7]. These β-barrels form pores of various diameters, often obstructed by soluble domains or loops. In Wza, the outer membrane (OM) is traversed by amphiphylic α-helices, organized side-by-side into a pore-like structure [6]. According to their pore-forming topology in the membrane, OMP oligomers can be categorized into two main classes: ‘single-pore’ and ‘multiple-pore’ oligomers. Single-pore oligomers can be divided into three subgroups: trimeric β-barrels, multimeric ‘super’ channels and the α-helical barrel as observed in Wza (Figure 1A, 1B, 1C). All have two features in common: subunits on their own are unstable and the multimeric trans-membrane channel can only be formed via oligomerization. Therefore, these proteins are obligate oligomers when crossing the bilayer. In contrast, in multiple-pore oligomers, each protein subunit folds into a separate transmembrane channel, which then assembles together into a multiple-pore complex (Figure 1D, 1E). In these oligomers, the oligomerization is not a requirement for structural integrity and membrane insertion per se, but is important for the functionality of these OMPs.

Figure 1.  Protein oligomerization in the bacterial outer membrane. Top and side views of the oligomerization topologies observed in known outer membrane protein structures. OM single pores: (A) Trimeric β-barrels: Hia (left, PDB code: 2GR7) and TolC (right, PDB code: 1EK9) (B) Multimeric superchannels: PulD secretin, obtained from [83] and (C) α-helical barrels: Wza (2J58). OM multipores: (D) Twinned pore oligomers: PapC (PDB code: 2VQI). (E) Triplet pores: LamB (PDB code: 1MAL). All the structures presented in this figure are scaled against the estimated width of outer membrane bilayer (36 Å). Figure is prepared using pymol [84].

Single-pore oligomers

Trimeric β-barrels

The first example of how bacterial OMPs oligomerize into a single β-barrel topology was revealed in the structure of Escherichia coli TolC [8]. The TolC family proteins form the OM partner of the three-component multidrug efflux and protein export machineries in Gram-negative bacteria, also known as Type 1 secretion systems (T1SS) [9]. TolC family proteins form homotrimeric complexes in the outer membrane, where three ∼470 residue monomers assemble into a 140 Å long ‘channel-tunnel’ that traverses both the outer membrane and periplasmic space (Figure 1A) [8]. The periplasmic ‘tunnel’ consists of twelve α-helices, four from each monomer, that organize into a sealed conical structure by pair-wise helical coiled-coil interactions. In the outer membrane twelve β-strands, four from each monomer, give rise to a single, hollow β-barrel ‘channel’ with an inner diameter of 25 Å. The oligomerization of the TolC β-barrel is mediated by the juxtaposition of a 4-stranded antiparallel β-sheet in each of the three monomers.

A similar trimerization was recently reported in the structure of the translocator domain of the Haemophilus influenzae adhesin (Hia) [10], [11]. Hia adhesin belongs to the autotransporter (AT) family, a large and diverse group of mostly virulence proteins in Gram-negative bacteria that expose or release adhesive or catalytic domains to the bacterial exterior [12], [13]. All autotransporter proteins are expressed as precursor proteins with three basic functional domains: an N-terminal signal peptide, a soluble domain termed ‘passenger domain’ and a C-terminal translocator domain [12], [13]. The C-terminal translocator domain is embedded in the outer membrane and facilitates delivery of the passenger domain to the bacterial surface. In conventional autotransporters, the C-terminal translocator domain contains approximately 300 amino acids, which fold into a single 12-stranded β-barrel [14], [15]. In contrast, in trimeric autotransporters, the translocator domain contains ∼60–70 amino acids and forms SDS-resistant trimers in the outer membrane [16], [17]. The structure of the Hia translocator domain showed it forms a trimeric β-barrel, with 12 transmembrane β-strands, each monomer contributing four strands (Figure 1A). The Hia trimeric barrel is remarkably similar to its counterparts in monomeric autotransporters such as NalP [15] and EspP [18]: all share the same number of strands, the same angle of strands relative to the axis of the barrel, and very similar pore dimensions, suggesting that both conventional and trimeric AT adopt a similar secretion mechanism to cross the outer membrane [11].

Despite their oligomeric nature, the TolC and Hia β-barrels share the typical common features observed in ‘classical’ OMPs: (1) β-strands, which contain an average of 10–13 residues, undergo a right-handed twist and pack with a positive inclination perpendicular to the plane of the bilayer; (2) The inner face of the β-barrel is lined with mostly hydrophilic side chains, enriched in small or unbranched residues. The outer, lipid-facing side of the barrel is lined with mostly branched, hydrophobic residues, mediating the interaction between the barrel and the hydrophobic environment; (3) The base of the β-barrel, facing the inner, periplasmic leaflet of the lipid bilayer, is lined with a ring-like cluster of aromatic residues, a feature that is thought to help orientate OMPs in the OM. Available structures show OMP β-barrels comprise between 8 and 24 β-strands. The examples of TolC and trimeric autotransporters illustrate the fact that sequences with fewer β-strands can still assemble into β-barrels through homo-oligomerization. Another example is the Mycobacterial OM protein MspA, where each monomer shares 2 β-strands to form a composite, octameric 16-stranded β-barrel [19].

Given the high structural similarity between oligomeric and monomeric β-barrel proteins, what could be the advantages of maintaining an oligomeric structure? For trimeric autotransporters, oligomerisation has been suggested to increase binding affinity of the extracellular passenger domains for their cellular targets through receptor multivalency or cooperativity [10], [20], [21]. In TolC family proteins, the sequences and 3D structures suggest the subunits in the trimer may have evolved from gene-duplication of a Helix-Helix-Strand-Strand-Helix structural repeat unit, which probably functioned as a hexameric channel-pore in a common ancestor [8]. The adoption of a trimeric pore in present-day TolC family proteins, rather than a monomeric pore, may facilitate interactions with their reaction partners. Indeed, during type 1 protein secretion and multidrug efflux, TolC functions together with a periplasmic adaptor or membrane fusion protein and with an inner membrane traffic ATPase or a proton antiporter (a representative model system is shown in Figure 2A). In particular, proton antiporters like AcrB in the acridine efflux pump form trimeric complexes in the inner membrane [22]. Drug export in AcrB occurs according to a ‘peristaltic’ mechanism, whereby AcrB monomers cycle sequentially through different conformational states [23], [24]. This gives rise to a functionally essential asymmetry in the AcrB trimer. During drug efflux, AcrB connects to TolC via the AcrA adaptor, which binds in a shallow groove extending from the periplasmic lumen to the equatorial domain of TolC. Interestingly, an asymmetry has been observed in these grooves, arising from the various monomers in the partially open trimeric TolC [25]. AcrA-TolC binding has been suggested to open the AcrA-AcrB-TolC efflux pump for export [25].

Figure 2.  Functional significance of protein oligomerization in the bacterial outer membrane. (A) Model for the TolC-AcrA-AcrB acridine efflux pump [85]. The three monomers in TolC, AcrA and AcrB trimers are colored yellow, magenta and cyan, red, orange, maroon, slate, greencyan. (B) Model of three-subunit assembly intermediate of the type 1 pilus bound to the FimD usher twinned pore [50]. The twinned pores are shown in yellow and magenta, with the FimH, FimG and FimF subunits and FimC chaperone shown in skyblue, cyan, blue, and green, respectively. The model includes a newly recruited chaperone subunit complex (FimC:FimA in green:dark blue) bound to the N-terminal domain of the second usher pore/monomer. Figure is prepared using pymol.

Multimeric superchannels

Secretins and secretin-like proteins are the major OM component of type II and type III secretion apparatus (Klebsiella oxytoca PulD and, Pseudomomas aeruginosa XcpQ (type II); Yersinia enterocolitica YscC (type III)) and are also essential to type IV pilus biogenesis (Neisseria meningitides or P. aeruginosa PilQ) or filamentous phage release (bacteriophage f1 pIV protein). They can oligomerize to form SDS-resistant homo-multimeric OM assemblies composed of 12 or 14 subunits [26]. Genetic and biochemical analysis, sequence alignments and secondary structure predictions reveal two major domains: the N-domain and the ß-domain [27], [28]. The ß-domain, corresponding to the C-terminal half of the protein, is highly conserved and is predicted to comprise 10-14 amphipathic ß-strands spanning through the OM [27]. The N domain is predicted to be located in the periplasm and to interact specifically with different inner-membrane counterparts or substrates in different secretion systems [29], [30]. Though the N-domain contributes to multimer formation, it was shown that the ß-domain is a major determinant for secretin multimerization and its stability [28]. Secretins can interact with a lipidated pilot protein, known as pilotin, which is important for targeting secretins into the OM [30–33]. In some secretins, the pilotin-secretin interaction is also required for oligomerization [33], [34]. However, it was recently reported that PulD multimers could be assembled and inserted into membranes in vitro, in absence of the PulS pilotin [35]. Interestingly, it has been shown that, in vivo, in the absence of the PulS pilotin, PulD co-purifies with the inner membrane. This has led to the hypothesis that PulS might prevent premature multimerization of PulD and accompany it through the periplasm to the outer membrane [36].

Low-resolution structures of secretins (PulD in a complex with its pilotin partner PulS, the bacteriophage pIV and Neisseria meningitides PilQ determined by cryo-electron microscopy, revealed a cylindrical mega-dalton oligomer with a rotational 4-, 12-, or 14-fold symmetry for PilQ, pulD, pIV secretins, respectively (Figure 1B). The 12-mer or 14-mer oligomerization structurally allows the formation of a massive translocation channel that ranges in size from 5 nm (PilQ, YscC) to 10 nm (XcpQ) [37–39], large enough to accommodate their sizable substrates [39–41]. Due to the lack of atomic resolution structures, it is unclear how the secretin monomers are associated with each other and subsequently assemble into the multimeric superchannel with two chambers on either side of the complex (Figure 1B). One hypothesis for secretin oligomerization can be derived from the structures of trimeric TolC and Hia (Figure 1A), in which the monomers are assembled circumferentially, via extensive strand-strand interactions, to form a single SDS-resistant β-barrel across the outer membrane [8], [11]. Assuming the secretin monomers indeed comprise 10–14 amphipathic transmembrane β-strands as predicted by Genin et al. [27], their side-by-side juxtaposition would give rise to a superchannel that consists of 120–196 β-strands. Based on available OMP structures, we calculated an average channel circumference of 5.9 Å per β-strand for classical and trimeric β-barrels (ranging from 5.4–7.5 Å depending on the inclination angle, or shear number, of the β-strands to the barrel axis). Cryo-electron images of secretin channels show 5–10 nm pores, corresponding to a calculated circumference of 15.7–31.4 nm. Assuming the predicted 120–196 β-strands form the channel, this would give outer values for a circumference per β-strand ratio of ∼0.8–2.6 Å. These are significantly smaller than what is routinely observed in available OMP β-barrel structures, suggesting either that not all predicted amphipathic β-strands in the secretin monomers form part of the β-barrel pore, or that secretins might adopt a pore structure that significantly differs from the classical OMP β-barrels, possibly adopting a double-walled β-barrel structure [42] where a β-pore could be formed by juxtaposition of β-sandwiches rather than juxtaposition of single β-sheets. However, in the absence of a crystal structure and given the precedent of Wza, we cannot exclude the possibility that secretins might be at least in part α-helical.

α-helical barrel

It was suggested that 2–3% of the genes in Gram-negative bacterial genomes encode OMPs [43–45]. Secondary structure predictions on bacterial OMPs and available three-dimensional structures consistently show these to be β-barrels [7], [46]. β-barrel topology, hence, has long been assumed to be the only structural design for bacterial OMPs [7]. In that context, the structure of E. coli Wza revealing a novel α-helical barrel [6] stands out.

Wza forms the outer membrane lipoprotein component of the capsular polysaccharide (CPS) biosynthesis pathway, which is utilized by Gram-negative bacteria to synthesize and export over 80 structurally different lipid-linked polysaccharides across the periplasm and OM to form an extracellular capsule. The crystal structure of E. coli Wza, determined by Dong et al., reveals a 340 kDa octameric complex. Three consecutive folded domains in the Wza monomers arrange into stacked octameric rings around a central 8-fold rotational axis (Figure 1C), forming a large cage-like complex that resides almost entirely in the periplasm. The mature polypeptide carries an acylation and diacylglycerol modification on its N-terminal Cys residue, which is thought to be involved in anchoring the protein to the OM. In the crystal structure of Wza, the lipidated N-terminus resides near the top of the amphora-shaped complex, where it can insert into the inner leaflet of the OM. Surprisingly, contrary to classical OMPs, the complex traverses further the OM via a novel α-helical barrel, formed by eight C-terminal amphipathic helices, one from each monomer, that assemble into a cone-shape channel via oligomerization (Figure 1C). Inside the channel, the interactions between the helices are mediated by direct and indirect side chain hydrogen bonds via water molecules. On the side of the lipid bilayer, the helices are decorated with hydrophobic residues including a Trp residue. Oligomerization in the case of Wza gives rise to a Trp-ring in the inner leaflet of the outer membrane, reminiscent of the ring-like cluster of aromatic residues at the base of β-barrel OMPs [7].

Based on its structure, oligomerization in Wza seems to be largely driven by the soluble, periplasmic domains. Most of the conserved residues are located at the extensive subunit-subunit interfaces [6]. Oligomerization allows the formation of a compartmentalized complex that encloses a large internal cavity of 15,000 ų (Figure 1C). This cavity opens up to the extracellular milieu via the α-helical barrel in the OM and is proposed to work as a conduit for the passage of chemically distinct CPSs across the periplasm and OM [6]. The enclosed cavity in the Wza octamer is closed off on the periplasmic side, but during transport the channel conduit across the periplasm and OM is thought to be actively gated by the inner membrane ATPase Wzc [47].

Multipore oligomers

Twinned pores

Bacterial ushers are OM transporters responsible for the assembly of adhesive multisubunit fibres, called pili or fimbriae, on the outer surface of a variety of Gram-negative bacteria [49], [48]. During pilus assembly, the usher recruits periplasmic chaperone:subunit complexes to the OM, catalyzes the ordered subunit polymerization and translocates the nascent pilus to the cell surface. Ushers are ∼800 residue proteins comprising a central β-barrel domain flanked by periplasmic soluble domains of ∼125 and ∼170 residues at the N- and C-terminus, respectively [50]. The N-terminal domain forms the recruitment site for chaperone:subunit complexes [51]. The function of the C-terminal domain is unknown, though a possible role in an early activation event has been suggested. EM imaging of 2D crystals of lipid-reconstituted ushers as well as genetic complementation studies have established that ushers function as twinned (dimeric) pore complexes [52], [53]. The crystal structure of a fragment of the PapC usher shows the usher translocation pore consists of a 24-stranded β-barrel that gives rise to a channel with inner diameter of ∼45 by 25 Å, which is compatible with the translocation of folded pilus subunits [49]. The crystals of the PapC translocation domain also reveal the dimerization interface along the flat surface of the kidney-shaped β-barrel (Figure 1D), very similar to the twinned pore observed by electron microscopy of 2D crystals of lipid-reconstituted full-length PapC [29], [50]. In combination with the single particle cryo-electron microscopy imaging of the type 1 pilus usher (FimD) bound to a translocating three-subunit assembly intermediate, these structures captured a twinned pore translocation machinery in action (Figure 2B). Unexpectedly, only a single pore is used for pilus secretion [50]. However, both usher molecules are required for pilus subunit polymerization through a mechanism of alternating chaperone-subunit complex recruitment. At each step of the incorporation cycle, the N-terminal domain of one of the two pores is engaged in binding the chaperone:subunit complex at the base of the growing fiber, whilst the N-terminal domain of the opposing pore forms the docking site for a newly incoming chaperone:subunit complex. During polymerization, the chaperone at the now penultimate site in the fibre is released, thereby liberating that usher's N-terminal domain for a new recruitment step. In this way, alternating recruitment of chaperone:subunit complexes at either usher N-terminal domain allows the step-wise incorporation of new pilus subunits, with the growing fibre translocating through the activated channel in the twinned pore complex (Figure 2B).

A second interesting example of the dimerization of β-barrels in the OM is seen in the outer membrane phospholipase A (OMPLA). OMPLA is responsible for the hydrolytic degradation of phospholipids in the outer leaflet of the OM, an activity that alters cell wall integrity and can contribute to the release of bacteriocins (to inhibit the growth of similar or closely related bacterial strains) and hence enhance bacterial virulence [54], [55]. OMPLA is present in the OM in a monomeric form, but assembles into a reversible, substrate-induced active dimer [56]. Structures of monomeric and dimeric E. coli OMPLA reveal how the hydrolase activity is regulated by dimerization in the outer membrane [57]. The OMPLA monomer consists of a 12-stranded β-barrel. Chemical modification and site-directed mutagenesis identified that the active site is formed by the catalytic triad Asn156-His142-Ser144 [58], located on the outside of the barrel [57]. Snijder et al. trapped OMPLA in its active, dimeric state by means of the covalent inhibitor hexadecansulphonylfluroride. The structure of the dimeric OMPLA shows that the dimeric interface resides almost exclusively in the membrane-embedded parts along the flat side of the barrel. This intimate dimeric oligomerization is essential for OMPLA's catalytic activity [57]: indeed, the interface between monomers contains the functional oxyanion holes, the substrate-binding pockets and is also the site for binding of a catalytically important Ca²⁺ ion. Hence, this structure provided a unique example to show how oligomerization in OMPs can be utilized to engineer a catalytic compartment.

Triplet pores

Porins comprise the oldest example of OMP multimerization. The X-ray structures of the Rhodobacter capsulatus porin and E. coli OmpF and PhoE [59], [60] revealed porin trimers are formed by three independently folded β-barrels that are clustered together into a triplet pore complex. Monomers interact inside the membrane via inter-digitating hydrophobic side chains on the outer surface of the β-barrels (of which ∼35% of the surface area is buried) as well as via a hydrophilic network in a structurally conserved loop outside the bilayer. Very similar β-barrel/β-barrel associations (Figure 1E) have since been observed in other members of the porin family [60–69]. These bacterial trimeric porins provide the passive diffusion passages for small and polar molecules up to the size of 600 Da through the membrane, either in a substrate-specific (e.g. LamB or ScrY) or non-specific manner (e.g. OmpF, PhoE).

Oligomerization of the trimeric porins is very different in principle to the single-pore OMP oligomerizations discussed elsewhere in the review. Each monomer in trimeric porins contains either 16 or 18 transmembrane β-strands [67], [70] and appears to be able to form a complete barrel without the need of an adjacent monomer (Figure 1E). The role of oligomerization in porins is still a matter of debate. Contrary to the single-pore oligomers, this oligomerization does not appear to be a structural requirement for the stability and insertion of the monomers inside the bilayer [71]. It is also unclear how multimerization would influence porin function in terms of solute passage. No evidence exists for a conformational communication between pores in trimeric porins during solute passage. The size and, in some cases, specificity of the channel is restricted by the insertion of an extracellular loop (L3) into the channel. Conserved residues such as the aromatic residues in LamB in the channel form a so-called ‘grease slide’, to guide substrate diffusion across the outer membrane [70]. All structural elements that give rise to the diffusion channel originate from within the individual monomers in the triplet pore. Hence, based on crystal structures it would appear that the subunits could function independently from each other. The existence of functional monomeric porins is demonstrated by OmpG, a 14-stranded β-barrel that functions as a non-specific diffusion channel in the OM [72].

Higher order multipores

A number of biochemical and low resolution observations point to the existence of higher order multipores in the OM. Omp85, an evolutionary conserved outer membrane protein essential for folding and inserting β-barrel OMPs in the bacterial or mitochondrial outer membranes [72], is suggested to form tetrameric complexes [73]. E. coli Omp85 has been observed to be present predominantly as a tetrameric species by Blue Native PAGE and size exclusion chromatography (SEC) [74]. For Haemophilus influenzae HMW1B, another member of the Omp85/TpsB superfamily, SEC and negative stain electron microscopy have identified the existence of dimeric as well as tetrameric complexes [75], [76]. Under the EM, these appear like twinned and quadruplet pore complexes of 100 Å by 50 Å and 100 Å by 100 Å, respectively, formed of two or four ring-like structures, each with a central cavity of 2.5 nm [75], [76]. The diameter of the individual ring is consistent with that of the 16-stranded β-barrel observed in the X-ray structure of Bordetella pertussis TpsB protein FhaC [77]. Sequence alignment between Omp85 and FhaC indicates the conservation of the 16-stranded β-barrel in Omp85/TpsB superfamily. This, together with the tetrameric complexes observed in HMW1B and Omp85, further suggests that these oligomers would form quadruplet pores in the OM. The functional significance of these quadruplet pores for OMP folding and membrane insertion in the case of Omp85 proteins or protein secretion in the case of TpsB proteins is unknown.

Finally, higher order multipores have also been reported for certain members of the autotransporter family. Veiga et al. reported the observation of a hexameric transmembrane multipore in the case of the Neisseria gonorrhoeae IgA protease [78]. Here again, the functional significance of the observed multipore complex is unclear.

Conclusion

The examples above demonstrate the versatility and abundance of homo-oligomerization in outer membrane proteins. Of the 57 non-redundant OMP structures, more than 20 proteins form multimeric complexes. As is the case for soluble proteins, the reasons behind OMP multimerization are diverse. These include catalytic cooperativity (OmpLA, usher), active site multivalency (trimeric autotransporters), structural scaffolding (TolC) and compartmentalization (secretins, Wza). In other instances, reasons behind oligomerization are less clear (Omp85/TpsB). Ironically, trimeric porins form the oldest example of OMP multimerization. Yet, despite the strict conservation of trimerization, the reasons for porins to do so are still unclear.

This review concentrated on homomeric complexes in the OM. To what extent OMPs can also form hetero-oligomeric complexes is yet unclear. Thus far, the only examples are restricted to complexes between OMPs and lipoproteins. For example, in E. coli, it has been shown that a heteromeric functional complex can be formed between YaeT, an Omp85 homolog, and several lipoproteins such as YfiO, YfgL, NlpB and SmpA (see below) [79]. In secretins, the heteromeric association with lipoprotein is required for OM targeting [31]. In Type 4 secretion systems, a macromolecular machinery secreting proteins and/or nucleoproteins across membranes in Gram-negative bacteria, the hetero-oligomeric complex between VirB7, a lipoprotein, and VirB9 is thought to form a channel in the OM [80]. Further work is required to uncover the extent and the structural basis of hetero-oligomerization in the OM.

Finally, the molecular pathways that target OMPs to the outer membrane and that ensure their correct folding and insertion into the lipid bilayer have only recently started to be unravelled [73], [79], [81], [82]. Where and when the OMPs obtain their oligomeric form along this pathway is unknown. In case of multipore oligomers, the distribution of apolar side chains on the β-barrel exterior does not exclude that these could, at least transiently, reside in the OM in their monomeric form. In the case of single-pore oligomers, however, the transmembrane domains of the monomers expose large regions of polar surface. For such composite β-barrels, it seems rather more likely that oligomerization occurs before insertion/release into the OM bilayer. Most, if not all, β-barrel OMPs sort into the OM in a pathway dependent on the BAM complex, an assembly platform in the OM composed of Omp85/YaeT (renamed as BamA) and the lipoproteins YfgL, YfiO, NlpB and SmpA (renamed as BamB, C, D & E, respectively) [79]. In conjunction with the BAM complex, at least three periplasmic chaperones (SurA, Skp and DegP) have been directly implicated with OMP transit in the periplasm [81]. Determining whether nascent composite β-barrel OMPs are adopting their oligomeric form during contact with the periplasmic factors or at the BAM complex, will provide valuable information on the OMP assembly pathway.

Acknowledgements

This work was funded by grant 08087 from the Wellcome Trust to GM and GW. We thank Dr Ben Luisi for sharing the model of the TolC-AcrA-AcrB acridine efflux pump that was used to prepare Figure 2A. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.