Structure and Function of an Essential Component of the Outer Membrane Protein Assembly Machine

See allHide authors and affiliations

Science  17 Aug 2007:
Vol. 317, Issue 5840, pp. 961-964
DOI: 10.1126/science.1143993


Integral β-barrel proteins are found in the outer membranes of mitochondria, chloroplasts, and Gram-negative bacteria. The machine that assembles these proteins contains an integral membrane protein, called YaeT in Escherichia coli, which has one or more polypeptide transport–associated (POTRA) domains. The crystal structure of a periplasmic fragment of YaeT reveals the POTRA domain fold and suggests a model for how POTRA domains can bind different peptide sequences, as required for a machine that handles numerous β-barrel protein precursors. Analysis of POTRA domain deletions shows which are essential and provides a view of the spatial organization of this assembly machine.

Although most biological membranes contain exclusively α-helical proteins, the outer membrane of Gram-negative bacteria and the organellar membranes of mitochondria and chloroplasts contain β-barrel proteins (1). These integral β-barrel proteins, called outer membrane proteins (OMPs), are folded and inserted into membranes by a process, conserved between prokaryotes and eukaryotes (24), that involves the action of a multiprotein machine (5, 6). Genetic and biochemical experiments have identified many parts of this machine in several organisms, including Saccharomyces cerevisiae and E. coli (313). The only conserved component in prokaryotes and eukaryotes is an integral β-barrel membrane protein, represented by YaeT in E. coli, Sam50 in mitochondria, and a Toc75 isoform in chloroplasts. A substantial region of all three proteins projects into the intermembrane space and contains one or more predicted polypeptide transport–associated (POTRA) domains (3, 4, 14).

Proteins destined for the outer membrane of E. coli are synthesized in the cytoplasm and transported across the inner membrane through the SecYEG protein secretion machinery (Fig. 1) (15). The signal sequence targeting them for secretion is removed at the outer face of the inner membrane. The processed OMP then traverses the periplasmic compartment to β-barrel assembly sites in the outer membrane. Chaperones may assist in periplasmic passage (16). It is presumed that the processed OMPs contain structural features that allow them to be recognized by the β-barrel assembly machinery, which in E. coli consists of at least five interacting components: four lipoproteins (YfgL, YfiO, NlpB, and SmpA) and the conserved integral membrane protein, YaeT (5, 13).

Fig. 1.

Diagram of bacterial outer membrane protein (OMP) biogenesis.

There are homologs of YaeT in organisms from bacteria to humans (17). Recent experiments with E. coli YaeT and S. cerevisiae Sam50 have shown that these proteins are essential for viability. Furthermore, levels of folded β-barrel proteins decrease and levels of misfolded β-barrel proteins increase when they are depleted (4, 5, 7, 8, 18, 19). YaeT was reported to bind C-terminal peptides of OMPs (20). The POTRA domain in Sam50 was shown to bind unfolded β-barrel precursors, suggesting that this POTRA domain plays an important role in assembling other β-barrel proteins in the mitochondrial membrane (21). Biochemical studies of truncated variants of Toc75 have also implicated its POTRA domains as docking sites for proteins destined to be targeted to, or across, biological membranes (22). No structure of a POTRA domain has yet been reported.

We expressed and purified the periplasmic domain of E. coli YaeT containing all five POTRA domains (YaeT21-420) (23, 24). Crystallization of this construct was unsuccessful, but a shorter fragment containing four POTRA domains (residues 21 to 351) yielded well-ordered crystals with diffraction to spacings of 2.2 Å (23, 24).

The overall structure of YaeT21-351 has a fishhook-like shape, with successive POTRA domains rotated in a right-handed direction (Fig. 2, A and B). Despite having low sequence similarity, the POTRA domains have similar folds, comprising a three-stranded β sheet overlaid with a pair of antiparallel helices (Fig. 2C). The order of secondary-structure elements is β-α-α-β-β (disproving a previous prediction) (14); the first and second β strands form the two edges of the sheet, with the β3 strand sandwiched between them. The conserved residues that define the POTRA domains are primarily in the hydrophobic core or loop regions, suggesting that they are important for the structural integrity of POTRA domain (Fig. 2, C and D).

Fig. 2.

Structure of YaeT. (A) Domain organization. (B) X-ray structure of YaeT21-351. POTRA domains P1 to P4 are colored yellow, green, blue, and red, respectively. The eight residues from P5 are colored gray. The missing electron density in the P3 domain is represented by a dashed line. (C) Ribbon diagram of a POTRA domain (P2) with side chains of the conserved residues shown. (D) Sequence alignments of POTRA domains from selected members of the YaeT/Omp85, Sam50, and Toc75 families, found in Gram-negative bacteria, mitochondria, and chloroplasts or cyanobacteria, respectively [adapted from Sánchez-Pulido et al. (14)]. Conserved residues are highlighted (28). The intensity of the orange color reflects the level of conservation in physicochemical properties. (E) X-ray structure of the dimer. The POTRA domains in one monomer are colored as in (B); the other monomer is purple. (F) Dimer interface showing the C-terminal residue contacts of one monomer (gray) to the P2 (light green) and P3 (light blue) domains of the other monomer. Labels represent hydrophobic residues. L, Leu; Y, Tyr; F, Phe; V, Val; I, Ile; T, Thr.

YaeT21-351 is a dimer in the crystal (Fig. 2E). The two monomers are intertwined, burying 1900 Å2 of solvent-accessible surface of each monomer. The longest contiguous set of contacts between monomer units involves a series of main-chain hydrogen bonds between the β2 edge of the P3 domain of one monomer (Asp241 to Leu247) and the first residues (Asn345 to Lys351) of the truncated P5 domain of the other monomer (Fig. 2F). These residues form a parallel β strand with respect to the β2 edge of the P3 domain and bury ∼1000 Å2, more than half the total buried surface. There are no other extensive contacts between monomers, suggesting that dimerization is mediated by this parallel β-stranded interface. Formation of this interface may have been necessary for growth of well-ordered crystals given that slightly shorter (YaeT21-348) or longer (YaeT21-355) constructs failed to crystallize. Nonetheless, highly ordered contacts are conserved at the interfaces between successive POTRA domains (fig. S1), suggesting that the fishhook conformation is present in the monomer.

We do not think that the dimer is physiologically relevant for several reasons. First, YaeT21-351 elutes as a monomer from a size exclusion column (fig. S2), implying that the stability of the dimer observed in the crystal is weak. Second, the N terminus of P5, which forms one of the β strands of the dimer interface, would not be available to interact with P3 in the full-length protein because the interacting residues would be buried in the P5 hydrophobic core. Nevertheless, the dimer interface shows that one way in which other polypeptides can interact with POTRA domains is by β augmentation (25).

The lipoproteins in the OMP assembly complex reside in the periplasmic space along with the five POTRA domains of YaeT. One function of the POTRA domains in YaeT could be to provide a scaffold to organize these lipoproteins. Using the crystal structure as a guide, we prepared five N-terminally His-tagged YaeT deletion constructs, each lacking a POTRA domain. All five deletion constructs (YaeTΔP1 to YaeTΔP5) could be expressed in an E. coli strain containing a wild-type chromosomal yaeT gene; all were targeted to the outer membrane and folded as judged by heat modifiability (Fig. 3A). Each deletion construct was purified by Ni-affinity chromatography, and eluents were assayed to determine which lipoproteins were present. Any of the first four POTRA domains can be deleted without disrupting the interactions with YfiO, NlpB or SmpA; however, the P5 deletion loses all three of these lipoproteins (Fig. 3B). YfgL disappears when any POTRA domains except P1 are deleted (Fig. 3B). These studies show that the periplasmic portion of YaeT scaffolds the other four proteins; and the studies also outline the spatial organization of the OMP assembly complex. Although YaeT purified from inclusion bodies is reported to form higher-order oligomers (20), the multiprotein OMP assembly complex behaves as a monomer. It has a mobility on Blue-Native polyacrylamide gel electrophoresis (PAGE) corresponding to a mass less than 230 kD (Fig. 3C). Furthermore, wild-type YaeT does not associate with the His-tagged YaeT POTRA domain deletion mutants (Fig. 3D).

Fig. 3.

(A) SDS-PAGE analysis of YaeT wild-type (wt) and deletion mutants from whole-cell lysates, without (–) and with (+) prior heat treatment. Proteins were detected by Western blot analysis with the use of an antibody recognizing the His tag. (B) His-tagged YaeT wild-type or deletion mutants (ΔP1 to ΔP5) and associated proteins following Ni-affinity chromatography. Eluted samples were blotted against His-tag, YfgL, NlpB, SmpA, and YfiO antibodies. (C) The purified YaeT complex run on a Blue-Native PAGE with molecular weights from a standard lane indicated. (D) Same as in (B), but YaeT was blotted with an antibody to YaeT. YaeTΔP1 cannot be detected with our YaeT peptide antibody. (E) His-tagged wild-type YaeT and P3 mutants after purification by Ni-affinity chromatography and analysis, as in (B).

To assess the functional importance of each POTRA domain, we constructed five POTRA domain deletion mutants without His tags for complementation studies in an E. coli YaeT-depletion strain. The ΔP1 and ΔP2 mutant proteins retained partial function: Strains expressing these proteins can survive YaeT depletion but grow poorly (Fig. 4). Strains producing the ΔP3 and ΔP4 mutant proteins did not survive YaeT depletion (Fig. 4), showing that P3 and P4 are essential for viability even though neither scaf-folds an essential lipoprotein. The ΔP5 construct could not be introduced into the YaeT-depletion strain even under conditions where wild-type YaeT was expressed. Apparently, the ΔP5 mutant protein is toxic to cells in this context. Because we cannot detect an interaction between the mutant protein and wild-type YaeT or any of the lipoproteins, we suggest that ΔP5 mishandles nascent β-barrel substrates, producing harmful misfolded or aggregated OMPs.

Fig. 4.

Essentiality of POTRA domains. Cultures were grown with l-arabinose (A) or d-fucose (B) to induce or inhibit wild-type yaeT expression, which is driven by the ara PBAD promoter (5). Plasmid-borne yaeT variants were constitutively expressed. Samples taken after 6 hours were subjected to Western analysis. (A) Strains expressing plasmid-borne yaeT variants grew normally when wild-type yaeT was expressed. YaeTΔP1 cannot be recognized with our YaeT peptide antibody (Fig. 3). Strains have low levels of DegP and normal OMP levels (LamB and OmpA). (B) When wild-type YaeT is absent, strains producing mutant YaeT variants exhibit growth defects. Strains expressing ΔP1 and ΔP2 grow better and have higher levels of OMPs than ΔP3, ΔP4, and the vector-only control. Although levels of ΔP1 cannot be quantified, ΔP2 is stable, indicating insertion into the membrane even in the absence of wild-type YaeT. Nevertheless, all strains lacking wild-type YaeT exhibit a strong extracytoplasmic stress response (increased DegP) indicative of OMP-assembly defects. Asterisk in (B) corresponds to proteolyzed DegP. OD, optical density.

P3 has a feature not present in the others—a β bulge (Ile240 and Asp241) in strand β2. This strand is at the edge that binds the vestigial residues of P5, and the bulge appears to expose the strand for β augmentation. To determine whether this feature of P3 is involved in an essential function of YaeT or in its association with YfgL, we moved Asp241 two and four residues along the β strand to alter the likely location of the bulge and to reduce or disrupt the potential for β augmentation. These bulge translation mutants were expressed at wild-type levels. The two- and four-residue shifts decreased and abolished, respectively, binding to YfgL (Fig. 3E), but both mutants complemented the YaeT deletion strain. These results show that the edge of P3 participates in binding YfgL but that the essential functions of P3 do not involve the modified edge of the domain, nor do they require its interactions with YfgL, as expected from the nonessential nature of this lipoprotein.

The crystal structure may also hold clues to other functionally important regions of P3. The only residues in the polypeptide chain that are not resolved in the crystal structure are located within the loop between the α1 and α2 helices of P3. We have previously isolated a mutant that encodes a YaeT variant, YaeT6, which contains atwo–amino acid insertion in the same region of the α1-α2 loop (12) of P3. YaeT6, which retains the ability to bind YfgL (Fig. 3E) as well as the other three proteins of the OMP assembly complex, compromises OMP assembly in a wild-type background, but suppresses the outer membrane permeability defects conferred by imp4213, a mutant allele of an essential gene that encodes an OMP that is required for lipopolysaccharide assembly (26). The α1-α2 loop of P3 may interact with Imp, providing an explanation for why mutations that alter the loop suppress the permeability defects caused by imp4213.

Notably, β-strand augmentation (25), observed in the dimer interface of the YaeT crystal structure, occurs in other complexes that bind unfolded OMPs—for example, the PDZ domain of DegS, which helps clear misfolded OMPs from the periplasm (27). We have shown that P3 may bind YfgL in this way, and it is possible that other POTRA domains, which also contain exposed edges, interact with polypeptides by β-strand augmentation. This mode of capture would allow POTRA domains to participate in assembling the β barrels of OMPs in a manner that is insensitive to the diversity of their primary sequences but dependent on their common hydrophobic periodicity.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article