The best-characterized class of outer membrane proteins are the β-barrels. In this well-defined structure, the polypeptide chain folds into a series of antiparallel β-strands that join through hydrogen bonds between the first and last β-strand, to form a barrel (Fig. 1a). In cases such as LamB, the N-terminus is exposed to the periplasm (Fig. 1b), while in some β-barrel structures the N-terminus is folded back into the barrel lumen. The outer surface of the β-barrel is formed from amino acid residues with hydrophobic side chains, and the central space of the barrel filled by more hydrophilic (including charged) residues. If the diameter of the barrel is large enough, it will form a water-filled cavity. The cavity can form a pore through the outer membrane, but can be partially (or fully) occluded by the infolding of interstrand loops and the N-terminus (Cowan et al., 1992; Buchanan, 1999; Basle et al., 2006). This provides several possibilities to regulate the semi-permeable nature of the outer membrane, such that β-barrel membrane proteins can define the barrier characteristics of the cell wall in Gram-negative bacteria.

The β-barrel outer membrane proteins are made as precursors with an N-terminal leader sequence, and these precursor proteins interact with the cytoplasmic factors SecB and SecA, in a process that can also be facilitated by the cytoplasmic chaperone DnaK (Qi et al., 2002; Scott & Barnett, 2006; Chen et al., 2007; Driessen & Nouwen, 2007; Papanikou et al., 2007). After translocation through the SecYEG complex in the inner membrane, the leader sequence is processed by the SP-I signal peptidase and the unfolded protein then associates with the chaperone SurA (Tamm et al., 2004; Mogensen & Otzen, 2005). A second chaperone, Skp, can bind these substrate proteins but perhaps only if they fail to productively interact with SurA (Bitto & McKay, 2003; Kleinschmidt, 2003; Walton & Sousa, 2004; Sklar et al., 2007a). SurA transfers partially folded β-barrels to an ‘insertion and assembly complex’ in the outer membrane. In E. coli, the core subunit of this complex is YaeT, a member of the Omp85 protein family, and it is associated tightly with four lipoproteins called YfgL, YfiO, NlpB and SmpA (Wu et al., 2005; Ruiz et al., 2006). This complex, recently dubbed the β-barrel assembly machinery complex (forBAM; Misra, 2007), completes the insertion and assembly of β-barrel protein substrates.

To determine whether equivalent components are found ubiquitously and particularly in the Alphaproteobacteria, we made use of hidden Markov models (HMMs) developed to describe each of the known components of the β-barrel assembly pathway. Table 1 summarizes the results for the intracellular pathogens R. prowazekii and Brucella melitensis and the free-living species Caulobacter crescentus. For more broad comparison, we included in the analysis a betaproteobacterium Neisseria meningitidis; two seminal studies on outer membrane assembly in N. meningitidis first identified Omp85 (Genevrois et al., 2003; Voulhoux et al., 2003). The process of β-barrel protein assembly is highly conserved with Omp85 family members found in all bacteria that have outer membranes (Gentle et al., 2004; Cavalier-Smith, 2006), and many of the other components of the pathway are also highly conserved. However, several distinctions are apparent, as discussed in the following sections.

Members of the Skp protein family are present in diverse bacteria, including all proteobacterial groups and other bacteria that have outer membranes including Treponema pallidum, Bacteroides thetaiotaomicron, Chlamydophila caviae and Deinococcus radiodurans (Alcock et al., 2008). The cyanobacteria have an outer membrane and Omp85 (Bölter et al., 1998; Gentle et al., 2004; Schleiff & Soll, 2005), but their genomes do not encode a Skp homologue (Alcock et al., 2008).

SurA sequences have been identified in many groups of bacteria, such as the Chlorobiaceae and groups typified by D. radiodurans and Cytophaga hutchinsonii and are highly conserved suggesting that this chaperone, like Skp, was derived early in evolution. However, at least two genera of Alphaproteobacteria (Anaplasma and Ehrlichia) have no sequence that matches the characteristic features of SurA. It is likely then that a secondary loss of SurA has occurred in these groups, reflecting the common trend of gene loss that occurs in intracellular Alphaproteobacteria (Batut et al., 2004; Sallstrom & Andersson, 2005). In the absence of SurA, either the Skp and DegP chaperones that are present in Anaplasma and Ehrlichia perform the essential function provided by SurA in other organisms or, perhaps, a novel chaperone might work in the place of SurA.

Omp85 proteins are defined by the occurrence of two domains: an N-terminal region containing multiple polypeptide-transport-associated (POTRA) domains and a C-terminal β-barrel domain. The recent structure of a related protein, FhaC, provides a reasonable model for the barrel domain of Omp85 (Clantin et al., 2007). In E. coli and in Neisseria, deletion of any one of the POTRA domains has at least partial effects on outer membrane protein assembly (Robert et al., 2006; Kim et al., 2007). Recently, the three-dimensional structure of the POTRA domains from the Omp85 protein YaeT were solved and provide the basis for a model for how Omp85 can bind diverse peptide sequences, enabling it to handle its numerous β-barrel protein substrates (Kim et al., 2007; Misra, 2007). The alphaproteobacterial homologs of Omp85 all appear to have five POTRA domains.

The gene encoding BamD is essential for viability of E. coli (Onufryk et al., 2005; Wu et al., 2005; Malinverni et al., 2006) and the BamD lipoprotein is found ubiquitously in Gram-negative bacteria (Malinverni et al., 2006; Table 1). This includes non-Proteobacteria such as Treponema, Chlorobium and Chlamydophila. In N. gonorrhoeae the homologous protein is a peptidoglycan-associated lipoprotein called competence lipoprotein (ComL) and, as a consequence, BamD homologs are often annotated as encoding a ‘DNA uptake lipoprotein’; a transposon insertion into the middle of ComL resulted in reduced cell size, aberrant cellular morphology and transformation deficiency (Fussenegger et al., 1996), presumably a result of the altered outer membrane properties of these mutants. The BamD protein in Rickettsia and other Alphaproteobacteria each strongly predict to have at least three tetratricopeptide repeat (TPR) motifs, but comparisons of all these sequences in multiple sequence alignments suggest each protein probably contains six TPR helix-turn-helix structures. The consensus sequence for the TPRs is not stringent, and is not well conserved in the BamD from E. coli, however, overall sequence similarities would suggest a homologous TPR-rich structure is present in all BamD proteins (Fig. 4a). TPRs are structural elements that enable protein–protein interactions and have been found operating in a number of protein transport pathways (Blatch & Lassle, 1999; D'Andrea & Regan, 2003). For example, the mitochondrial protein import receptor Tom70 is built from multiple TPR elements (Chan et al., 2006; Wu & Sha, 2006) and binds β-barrel substrate proteins en route to the mitochondrial equivalent of the BAM complex (Chan et al., 2006). A TPR-rich structure might enable BamD to bind partner proteins (like BamC) and/or substrate proteins.

Escherichia coli mutants lacking BamC have mild defects in outer membrane biogenesis (Onufryk et al., 2005; Wu et al., 2005; Malinverni et al., 2006; Sklar et al., 2007b. A BamC homolog has not been found in any of the Alphaproteobacteria (Table 1), a distinct (or highly diverged) protein might fulfil the function of BamC. Alternatively, if the function of BamC is redundant another component of the BAM complex might compensate in these organisms.

Mutants in which the yfgL gene, encoding BamB, has been deleted have reduced levels of outer membrane proteins (Onufryk et al., 2005; Ruiz et al., 2005). The BamB protein from Alphaproteobacteria has seven or eight predicted Pyrrolo-quinoline quinone (PQQ) enzyme repeats, a motif representative of β-propeller structures (Jawad & Paoli, 2002) found in some enzymes and in protein domains involved in protein-protein interactions. Homology searches suggest that both the alphaproteobacterial and the E. coli BamB fit the profile of domains with an eight-bladed β-propeller fold (Fig. 4b). The PQQ enzyme active site residues are not present in BamB, indicating this protein is unlikely to be involved in enzymatic catalysis. If BamB does have a β-propeller structure, it could interact with β-strands in the POTRA of BamA and/or assist in stabilizing nascent β-strands in substrate proteins. While BamB is found in several species of Alphaproteobacteria including Caulobacter and Rickettsia, no related sequence was found in Brucella. BamB is also absent from the Omp85 complex in Neisseria (Table 1; Bos et al., 2007). It was recently shown that cargo-bound SurA can bypass interaction with BamB and interact directly with BamA (Vuong et al., 2008). It may be that only some protein substrates interact with BamB prior to assembly.

Biophysical evidence suggests the substrate proteins would be in a partially structured form in the periplasm (reviewed by Tamm et al., 2004) and that periplasmic chaperones such as SurA are needed to maintain the ‘periplasmic intermediate’ (Lazar & Kolter, 1996; Bulieris et al., 2003; Hennecke et al., 2005). In vitro models of outer membrane protein insertion predict a so called ‘molten disc’ intermediate at the bilayer interface, with β-strands sitting flat on the membrane (Kleinschmidt & Tamm, 2002; Tamm et al., 2004). SurA is, therefore, the prime candidate to assist in process (1).

The β-strands in outer membrane proteins have a sequence Φ–X–Φ (where Φ represents an aromatic residue) at least in the C-terminal strand, and mutations in this motif impair assembly of outer membrane proteins (Struyve et al., 1991; de Cock et al., 1997; Robert et al., 2006). SurA has been shown to bind directly to the C-terminal Φ–X–Φ motif (Bitto & McKay, 2003), as has Omp85 (de Cock et al., 1997; Robert et al., 2006; Bos et al., 2007). The POTRA domains of Omp85/BamA provide a means to receive substrates from chaperones like SurA. The POTRA domains have been suggested to facilitate templating of β-strands in the nascent outer membrane protein substrate and, if the lipoprotein BamB has a β-propeller structure, it too might facilitate strand formation in this way to assist in the completion of process (2). If its role were simply to assist the function of SurA and Omp85, this redundancy of function would explain why BamB is absent in some bacteria and why corresponding yfgL mutants of E. coli show only relatively minor defects in outer membrane assembly.

The third process is perhaps the one we understand least about. What mechanism is used to insert proteins into the hydrophobic core of the outer membrane? The β-barrel domain of Omp85/BamA might increase the kinetics of strand insertion by providing some local distortions in the lipid population, assisting intermediate forms of a barrel to gradually assemble and enter the membrane (3). Alternatively, the proteinaceous environment created by the components of the BAM complex in the periplasm might favour a barrel forming, to enable a more dramatic en bloc insertion of the barrel into the plane of the outer membrane.

Despite being imported from the cytosol, mitochondrial β-barrel proteins are still inserted into the outer membrane from the inner surface (Fig. 5). This requires that proteins be recognized and translocated through the outer membrane by the protein translocase in the outer mitochondrial membrane (TOM complex). The TOM complex has no counterpart in bacteria. In the intermembrane space, topologically equivalent to the bacterial periplasm, β-barrel proteins are chaperoned by tiny TIM chaperones (Fig. 5). While the tiny TIM proteins show some structural similarities to Skp chaperones (Webb et al., 2006), and functional similarities to SurA (Alcock et al., 2008), there is no obvious ancestry to relate the mitochondrial and bacterial chaperones. The assembly of β-barrel proteins into the mitochondrial outer membrane is driven by the sorting and assembly machinery (SAM complex) and at the core of the SAM complex is a protein called Sam50, a member of the Omp85 protein family (Kozjak et al., 2003; Paschen et al., 2003; Gentle et al., 2004). Reflecting their ancestry, mitochondrial Sam50 proteins are most closely related to the Omp85 proteins from Alphaproteobacteria (Gentle et al., 2004).

The mitochondrial SAM complex receives substrate proteins from the tiny TIM chaperones, and in no eukaryotes have homologs of SurA or Skp been found (Alcock et al., 2008). This demonstrates that Omp85 proteins do not need to contact a SurA chaperone to productively receive substrate proteins. The hidden Markov model searches reported here find no evidence for BamB, BamC, BamD or BamE homologs in eukaryotes. In a sense then, mitochondria can be considered as representing an end-point to the gene loss seen in Rickettsia and Brucella in regards to components of the outer membrane assembly machinery.

With the availability of vast genome sequence data from so many species of bacteria it has become possible to make comparative sequence analyses, to add further value to the work carried out with model bacteria such as E. coli and N. meningitidis. Some protein features, the presence of multiple TPR motifs in BamD and of multiple β-propeller signatures in BamB, are not readily apparent in an analysis of a single polypeptide. However, these structural features become strikingly apparent when a set of homologous proteins from across bacterial species is analysed as a group.

This is an exciting time for the study of membrane proteins. An increase, slow but sure, in the number of structures solved for membrane proteins has increased our appreciation of how polypeptides can be organized in biological membranes. Improved assays for tracking the assembly of membrane proteins in distinct systems provides a means to cross-fertilize our knowledge of any given system. A prime example of this is how the discovery of Omp85 in bacterial outer membranes led to the characterization of the SAM complex in the outer membrane of mitochondria, and the further understanding of the ‘BAM complex’ in bacteria. Three clear questions now present themselves, with answers that would largely complete our understanding of β-barrel membrane protein assembly.

Firstly, where and how does folding of a polypeptide into a β-barrel occur? The barrels may fold into the lipid environment of the outer membrane, but it seems increasingly likely that folding occurs before the encounter with the bilayer. But would this be in the periplasm per se, assisted by soluble chaperones such as SurA, or in the grasp of the various periplasmically exposed components of the BAM complex?

Secondly, how do the external loops of β-barrels fold? These loops are crucial for the function of β-barrel membrane proteins often forming the occlusions that provide selectivity to the β-barrel pores. In mitochondria, there are external factors, the metaxins, bound to the SAM complex that might assist these folding reactions (Pfanner et al., 2004; Paschen et al., 2005; Dolezal et al., 2006; Chan & Lithgow, 2008), but we know of no specific factors bound to the outer surface of the bacterial BAM complex.

Thirdly, can Omp85 do more than just assemble β-barrels? While it is clear Omp85 plays a direct role in outer membrane protein insertion (Voulhoux et al., 2003; Doerrler & Raetz, 2005), other studies have shown that cells in which the expression of Omp85 has been shut down accumulate lipopolysaccharide and phospholipids in their inner membranes, likely en route to the outer membrane. This defect could be a secondary effect of improper outer membrane protein assembly, but it remains possible that Omp85 also aids lipid insertion via a similar mechanism to protein insertion: through creating a ‘disturbance’ in the outer membrane that could accelerate the transfer of substrates into the hydrophobic core.