Insertion of proteins and lipopolysaccharide into the bacterial outer membrane

Abstract

The bacterial outer membrane contains phospholipids in the inner leaflet and lipopolysaccharide (LPS) in the outer leaflet. Both proteins and LPS must be frequently inserted into the outer membrane to preserve its integrity. The protein complex that inserts LPS into the outer membrane is called LptDE, and consists of an integral membrane protein, LptD, with a separate globular lipoprotein, LptE, inserted in the barrel lumen. The protein complex that inserts newly synthesized outer-membrane proteins (OMPs) into the outer membrane is called the BAM complex, and consists of an integral membrane protein, BamA, plus four lipoproteins, BamB, C, D and E. Recent structural and functional analyses illustrate how these two complexes insert their substrates into the outer membrane by distorting the membrane component (BamA or LptD) to directly access the lipid bilayer.

This article is part of the themed issue ‘Membrane pores: from structure and assembly, to medicine and technology’.

1. Introduction

A cell envelope consisting of an inner membrane, a periplasmic space, and an outer membrane protects Gram-negative bacteria from environmental stresses and toxins. While phospholipids make up both leaflets of the inner membrane, the outer membrane is an asymmetric structure where the inner leaflet contains phospholipids, while the outer leaflet consists of lipopolysaccharides (LPS) coordinated by divalent metal ions such as calcium and magnesium. The periplasmic space separating the inner and outer membranes accommodates a layer of peptidoglycan, and has been estimated to have an average thickness of 150 Å [1].

To maintain the integrity of the outer membrane, both LPS and certain outer-membrane proteins (OMPs) are required for viability, and they must be transported across the cell envelope and positioned correctly in the outer membrane. As described in the following sections, both proteins and lipids are thought to be inserted laterally into the outer membrane through openings in the protein and LPS biogenesis machineries. Lateral migration of lipids and proteins is a common theme that has been observed before [2]. The lateral movement of proteins into a lipid bilayer has been demonstrated for the Sec translocon, where individual hydrophobic, membrane-spanning α-helices are co-translationally translocated into the inner membrane, where the final assembly of α-helical inner membrane proteins takes place [3,4]. Similarly, two β-barrel OMPs can transfer hydrophobic substrates from the lumen of the β-barrel to the outer membrane. FadL binds long-chain fatty acids first at a low-affinity (extracellular) site, and then the substrate is moved by diffusion to a high-affinity site within the barrel lumen. A conformational change occurs to release the substrate from the high-affinity site and provides free access to a ‘hole’ in the barrel wall, allowing the substrate to diffuse laterally into the bilayer through the barrel opening [5–7]. A second example is found in PagP, an outer-membrane acyltransferase that protects bacteria against the host immune response by transferring a palmitate chain from a phospholipid to Lipid A. When substrate binds, His, Asp and Ser residues located on the extracellular surface of the barrel facilitate transfer of the palmitoyl group from the donor phospholipid to the Lipid A acceptor molecule [8,9].

In the outer membrane, the only essential OMPs in almost all Gram-negative bacteria are the biogenesis machineries that function to insert LPS and OMPs. This review describes recent structural and functional studies that illustrate the complex processes of protein and LPS insertion into the outer membrane.

2. The lipopolysaccharide-transporting complex LptDE

The biogenesis pathway for LPS has been elegantly defined in recent years [10]. LPS consists of three components: Lipid A contains the hydrophobic hydrocarbon chains that anchor the molecule in the outer membrane, while a core oligosaccharide and a variable O-antigen carbohydrate chain, containing up to 200 sugars, are hydrophilic and eventually reside on the cell surface [11]. Both the Lipid A and oligosaccharide molecules are independently synthesized in the cytoplasm and translocated to the outer leaflet of the inner membrane, where the bulky O-polysaccharide is assembled on the Lipid A core, yielding mature LPS [11].

Mature LPS is then extracted from the inner membrane, transported across the periplasmic space, and inserted into the outer membrane by the LPS transport system LptABCDEFG [12]. All seven Lpt components are essential because they form a continuous protein network spanning the entire cell envelope (figure 1). LptB₂FG is an ABC transporter that hydrolyses ATP to provide energy for LPS extraction from the inner membrane [13], where it is transferred to LptC [14]. LptC is thought to interact with a polymer chain of LptA molecules that span the periplasmic space, with the terminal LptA linking to the periplasmic domain of LptD in the outer membrane [15]. LptA, C and D share a common fold with a hydrophobic groove that is predicted to shield Lipid A from the aqueous environment of the periplasm during transit. The Lpt components thus form a trans-envelope scaffold that drives transport of LPS by hydrolysing ATP at the inner membrane [16,17]. Once it reaches LptD, the LPS molecule is inserted into the outer membrane by the LptDE complex [18].

Figure 1.

Transport of LPS and nascent OMPs to the outer membrane. Schematic pathways are shown for the biogenesis of β-barrel outer-membrane proteins and lipopolysaccharide in Gram-negative bacteria.

3. Description of the LptDE structures

Four LptDE structures were recently determined: the structure of full-length Klebsiella pneumoniae LptDE, and three N-terminally truncated LptDE core complexes from K. pneumoniae, Pseudomonas aeruginosa and Yersinia pestis [19]. In addition, a full-length LptDE structure was determined from Shigella flexneri [20] and a core complex from Salmonella typhimurium [21], as well as a core complex from Escherichia coli (PDB code 4RHB). With 26 β-strands, LptD is the largest known single-protein β-barrel. Also, it is the only protein to use a separate lipoprotein, LptE, as its ‘plug’ domain. Most of the globular lipoprotein LptE is nested in the barrel lumen and around a quarter of its surface is exposed to the periplasm. This architecture is preserved in all currently determined LptDE crystal structures (figure 2 and table 1).

Figure 2.

Structure of the Klebsiella pneumoniae full LptDE complex. The LptD β-barrel is shown in orange, with the N-terminal domain in green. The LptE plug is shown in blue. The first and last β-strands of the barrel are highlighted in red.

Table 1.

Table of Lpt and BAM structures discussed in the manuscript.

molecule	organism	resolution (Å)	PDB code	reference
LptE and LptDE complexes
LptDE core	Klebsiella pneumoniae	2.94	5IV8	[19]
LptDE full complex	Klebsiella pneumoniae	4.37	5IV9	[19]
LptDE core	Pseudomonas aeruginosa	2.99	5IVA	[19]
LptDE core	Yersinia pestis	2.75	5IXM	[19]
LptDE core	Salmonella enterica	2.8	4N4R	[21]
LptDE full complex	Shigella flexneri	2.39	4Q35	[20]
LptDE core	Escherichia coli	3.36	4RHB	[22]
LptE	Caulobacter crescentus	2.4	4KWY	[23]
LptE	Neisseria meningitidis	2.6	3BF2	[24]
LptE	Esherichia coli	2.34	4NHR	[25]
LptE	Shewanella oneidensis	2.6	2R76	[26]
LptE	Nitrosomonas europea	NMR	2JXP	[27]
LptE	Pseudomonas aeruginosa	NMR	2N8X	[28]
BamA and BAM complexes
Bam ACDE	Escherichia coli	3.39	5EKQ	[29]
Bam ABCDE	Escherichia coli	2.9	5D0O	[30]
Bam ACDE	Escherichia coli	3.5	5D0Q	[30]
Bam ABCDE	Escherichia coli	3.56	5AYW	[31]
Bam A (+2 POTRA)	Haemophilus ducreyi	2.91	4K3C	[32]
Bam A (+5 POTRA)	Neisseria gonorrhoeae	3.2	4K3B	[32]
Bam A (+1 POTRA)	Escherichia coli	3.0	4C4 V	[33]
Bam A (barrel only)	Escherichia coli	2.6	4N75	[34]

LptD and LptE interact very strongly with one another. These two proteins co-purify as a complex from detergent-solubilized outer membranes, they migrate as heterodimers on size exclusion columns, and are resistant to dissociation in solution, except when heated. The co-purified complex also migrates as a single band on a denaturing gel [18,35].

The dimensions of the elliptical transmembrane LptD β-barrel are around 50 × 50 × 70 Å, with three to five hydrogen bonds between the first (β1) and last (β26) strands closing the β-barrel. Strand β1 of the barrel is preceded by a periplasmic β-jellyroll domain consisting of two 11-strand β-sheets arranged in a V shape that form a hydrophobic groove, homologous to LptA. Terminal strand β26 contains a short C-terminal extension that resides in the lumen of the β-barrel. The β-barrel is also slightly tapered toward the extracellular side, with its periplasmic diameter around 10 Å larger than its extracellular diameter. LptE exhibits an α/β fold with two α-helices and five β-strands. Despite low sequence identity, individual LptE crystal structures from Shewanella oneidensis (PDB code 2R76), Caulobacter crescentus (PDB code 4KWY), Neisseria meningitidis (PDB code 3BF2), Escherichia coli (PDB code 4NHR) [25] and the solution structures from Nitrosomonas europeae (PDB code 2JXP) and Pseudomonas aeruginosa (PDB code 2N8X) [28] are almost identical to their LptDE-complexed counterparts (table 1). LptE has several distinct functions: assembly of the LptD β-barrel, stabilization of the LptDE complex by forming a plug and LPS translocation [18,25,36–38]. The N-terminus of LptE is anchored to the inner leaflet of the outer membrane by an N-acyl-S-diacylglycerylcysteine moiety that is not visible in any of the crystal structures. While the β-barrel lumen is mainly hydrophilic, the LptE surface has both hydrophobic and hydrophilic patches, making it highly unlikely to bind the hydrophobic portion of LPS. The size of the LptDE lumen is large enough to accommodate an Ra LPS molecule (Ra LPS contains Lipid A, an inner core of oligosaccharide consisting of at least one molecule of 3-deoxy-d-manno-oct-2 ulosonic acid (KDO) linked to Lipid A, and an outer core of oligosaccharide) [20].

The available LptDE structures exhibit an identical fold, with the Pseudomonas LptDE structure being slightly different due to two insertion loops. These insertion loops are located in the extracellular loop region of the barrel, conferring a larger molecular surface to the Pseudomonas LptDE core complex [19]. The significance of the larger molecular surface of LptDE for this species is not clear, but may be related to the different O-antigen structure of Pseudomonas LPS.

4. Insertion mechanism

LPS translocation differs from protein translocation by BamA [32] or long-chain fatty acid transport by FadL [39]. The β-barrel opens only partially between strands β1 and β26, forming a lateral gate that permits diffusion of the Lipid A domain of LPS into the outer leaflet of the outer membrane. This can be attributed to the weakened hydrogen bonding interactions between strands β1 and β26 caused by two proline residues, P231 and P246 [20]. Mutation of these proline residues yields reduced viability in in vivo assays and closer association of the β1 and β26 strands in MD simulations [19]. Double cysteine mutations that link β1 to β26, and block the opening of the lateral gate, are lethal to the bacteria [21]. The LptDE lateral gate opening is also limited by the presence of the LptD N-terminal domain and the conserved disulfide bond between the β24-β25 turn and the N-terminal domain. This N-terminal domain can rotate up to 21° around the hinge region that connects it to the β-barrel, as observed by comparing the two full-length LptDE structures [19,20] and it can physically block LPS from entering the inner leaflet of the outer membrane.

The tapering lumen of the LptD barrel is lined by conserved, increasingly electronegative residues that create unfavourable interactions with the negatively charged sugar moieties of the LPS molecule. Once the β-barrel outer loops open and the O-antigen portion of LPS is translocated to the outer leaflet of the outer membrane, the charge repulsion is resolved. Mutagenesis of the lumen residues showed that a strongly electronegative lumen is not required for transport, but an electropositive lumen does inhibit LPS translocation [19]. Recent studies using UV photoaffinity cross-linking delineate steps in the pathway of LPS transport through LptDE, showing that LPS accesses the periplasmic N-terminal domain, the hole between the first and last β-strands, the luminal gate, the lumen of the LptD β-barrel, and extracellular loops 1 and 4 [40].

One important unanswered question for the Lpt system is the following: how is the transport of the O-antigen portion of LPS coordinated through the periplasm? LPS can have an O-antigen region composed of up to 200 sugar rings [11]. This polysaccharide chain can be of a considerable volume that must be translocated through the periplasm and an LptD lumen having dimensions of around 35 × 45 Å. The peptidoglycan in the periplasm also forms a mesh-like barrier that would not allow easy passage of such a large sugar chain. In a current model [10], the LPS molecules are lined up like PEZ candy in the LptC-LptA-LptD trans-periplasmic bridge that resembles a PEZ dispenser (electronic supplementary material, figure S1). After the topmost LPS molecule is loaded into the outer leaflet of the outer membrane and its respective O-antigen completely translocated to the extracellular space, a new LPS molecule can be loaded on the inner membrane side of the bridge. This model works best for LPS molecules that have very short or no O-antigens. However, bulkier O-antigen head groups might limit the close alignment of the Lipid A moieties from LPS in the trans-periplasmic bridge, suggesting that the translocation of the O-antigen would constitute the rate-limiting step in LPS translocation to the outer membrane. LptE may play a role in opening up the β-barrel lumen to facilitate O-antigen translocation. However, many (most?) of the molecular details of this system remain to be worked out.

A second important question is how a new class of peptidomimetic antibiotics interacts with LptD and other OMPs. One such compound was initially thought to specifically interact with the N-terminal domain of Pseudomonas aeruginosa LptD [41]; however, a recently described peptidomimetic compound targeting E. coli appears to disrupt the outer membrane by interacting with various OMPs, including LptD and BamA [42]. The molecular mechanism by which these antibiotics interact with OMPs to disrupt the outer membrane remains to be established.

5. The outer-membrane protein insertion complex BamABDCE

Just as LPS is synthesized in the cytoplasm and transported across the entire cell envelope to reach the outer membrane, proteins targeted to the outer membrane make a similar journey. Proteins residing in the outer membrane adopt a β-barrel fold, with antiparallel β-strands connected to one another through extensive hydrogen bonding, where the first and last β-strands connect to form a β-barrel with a hydrophobic exterior and hydrophilic lumen. The individual strands of the β-barrel are amphipathic, with one side hydrophilic and the other side hydrophobic. Only when the β-barrel is fully formed, is a continuous hydrophobic surface presented to the lipid bilayer. This architecture presents complications for folding of nascent OMPs prior to insertion into the membrane.

Proteins destined for the outer membrane are synthesized in the cytoplasm and secreted across the inner membrane in an unfolded state by the Sec translocon [43]. In the periplasm, the nascent OMP interacts with chaperones such as SurA, Skp, FkpA and the bifunctional protease/chaperone DegP [44–46]. These protein–chaperone complexes are then targeted to the β-barrel assembly machinery (BAM) complex (figure 1). In E. coli, the BAM complex consists of BamA, an integral outer membrane protein with an extended periplasmic domain, and BamB, BamC, BamD and BamE, all lipoproteins anchored to the inner leaflet of the outer membrane [35,47]. The BAM complex is thought to help fold and insert the nascent protein into the outer membrane, but details of the mechanism are still being worked out. In the following sections, we discuss what is currently understood from a structural perspective and which issues remain to be resolved.

6. BamA catalyses insertion of nascent outer-membrane proteins into the outer membrane

Structures were determined for full-length BamA from Neisseria gonorrhoeae and an N-terminally truncated BamA from Haemophilus ducreyi (figure 3) [32]. In these organisms as well as in E. coli [33,34], the structure consists of five periplasmic polypeptide translocation-associated (POTRA) domains followed by a membrane-inserted β-barrel with 16 antiparallel β-strands. Studies of POTRA-barrel chimeras have demonstrated that the β-barrel domain is often interchangeable, while the POTRA domains show more species specificity [48]. The β-barrel is a very unusual OMP, displaying a reduced hydrophobic thickness where strands β1 and β16 are joined compared with the opposite side of the β-barrel. This feature was observed by using the crystal structures to measure the distances between aromatic residues on the outside of the barrel [32], by molecular dynamics simulations that showed a reduced distance between lipid head groups and a corresponding increased disorder of surrounding lipid chains [32], and by direct observation using electron microscopy of lipid vesicles with BamA inserted [49]. The hydrophobic thickness mismatch between BamA and the outer membrane appears to locally thin and disorder the lipid bilayer, creating a region of the bilayer where newly folded OMPs may be inserted more efficiently. In vitro folding experiments show that BamA catalyses the folding/insertion of OMPs by lowering the energetic barrier imposed by phosphatydylethanolamine lipids found in the inner leaflet of the outer membrane [50]. BamA presumably accomplishes this feat by creating bilayer defects.

Figure 3.

Structure of Neisseria gonorrhoeae and Haemophilus ducreyi BamA. The Neisseria structure has five POTRA domains (P1 to P5), while in the Haemophilus only two are visible. The different number of hydrogen bonds (yellow dashed lines) between strands β1 and β16 (red) are shown in the insets.

Another unusual property of the BamA β-barrel may also influence the folding and/or insertion of nascent OMPs. When the structure of N. gonorrhoeae BamA was determined, it showed that strands β1 and β16 were connected by only two hydrogen bonds near the extracellular surface [32], in contrast with other pairs of strands in the β-barrel that are generally linked by eight or more hydrogen bonds (inset, figure 3). In fact, most of strand β16 adopts a random coil secondary structure and is tucked inside the β-barrel lumen. Molecular dynamics simulations analysing the stability of the β-barrel demonstrated that strands β1 and β16 separate completely during a 1 µs simulation; they also re-associate during the same time period. Simulations were performed at 340 K or 310 K using a symmetric dimyristoyl-phosphatydylethanolamine (DMPE) lipid bilayer; under these conditions the LptDE β-barrel does not open, suggesting greater β-barrel instability, or propensity for β-strand separation, for BamA. Biochemical experiments subsequently confirmed that this lateral opening event is required for BamA function by engineering pairs of cysteine residues to form a disulfide bond at the extracellular surface, in the middle of the bilayer, or at the periplasmic surface. In all cases, tethering BamA closed abolished function, and reducing the disulfide to release the tether restored activity [51]. The lateral opening between strands β1 and β16 exposes the edges of those β-strands, which is energetically unfavourable and may allow for β-augmentation, or strand pairing, with other β-strands, such as those of the nascent OMP. This could potentially allow for the coordinated folding and insertion of a nascent OMP by directly budding it off from the BamA opening [32]. The opening also creates a pathway from the lumen of the β-barrel directly into the lipid bilayer, although exactly what this interface looks like is currently unknown. Located just above the barrel opening is an ‘exit pore’ defined primarily by alternating movements of extracellular loop 1 [51]. The exit pore may facilitate transport of extracellular loops (which can be as long as 100 residues [52]) of the nascent OMP to the cell surface. These features have been found in all BamA structures solved to date (table 1), and may play a role in folding nascent OMPs or inserting them into the outer membrane, or perhaps both.

7. Structures of the BAM complex reveal further BamA movements

By 2011, the structures of all of the lipoproteins, BamB, BamC, BamD and BamE, had been determined and a model for assembly of the BAM complex was formulated (for a review, see [53]). Recent structures of BamB and BamD bound to a periplasmic fragment of BamA, as well as an earlier structure of a BamCD complex, confirmed the model and helped to visualize component assembly [54–56]. It has become clear that the BamA POTRA domains can exhibit flexibility in the absence of other BAM components [57–59] but they probably interact with the periplasmic surface of the outer membrane in vivo, modulating their flexibility [60]. BamD plays an important role in outer-membrane protein biogenesis [61,62], while the trapping of a folding intermediate suggested that BamA and BamD work together to catalyse folding [63].

Although the model correctly predicted many of the interactions made by individual lipoproteins with the BamA POTRA domains, and the recent structural and functional studies illustrated a number of important features, we could not have anticipated the unusual conformational changes in BamA occurring upon BAM complex formation.

In 2010, an intact BAM complex was prepared by making sub-complexes containing BamAB and BamCDE, and then reconstituting the BAM complex into lipid vesicles. Upon addition of the chaperone SurA and an unfolded OMP substrate (OmpT), the BAM complex was able to fold and insert the OMP substrate into vesicles without any additional protein components or energy requirement [64]. This in vitro system was recently improved upon by expressing all five Bam components from a single expression vector, resulting in a BAM complex with markedly increased folding activity [65]. The improved method for producing an intact BAM complex assisted three groups in determining crystal structures of E. coli BamACDE and BamABCDE (table 1 and figure 4) [29–31].

Figure 4.

Structure of the BAM complex. This view was generated by merging the crystal structures of BamACDE (PDB code 5EKQ) and BamAB (4PKI). BamA is shown in cyan, the POTRA domains in different shades of teal, BamB in grey, BamC in green, BamD in gold and BamE in violet. The first and last β-strands of the BamA barrel are highlighted in red.

Purification of the BAM complex resulted in stoichiometric amounts of BamA, BamB, BamC, BamD and BamE after size exclusion chromatography [29], but BamB dissociates or is degraded during crystallization. Of the four recently published BAM complex structures, two contain BamB [30,31] and two do not [29,30]. The Bam lipoproteins assemble on the BamA POTRA domains essentially as predicted from earlier structural and functional/genetics studies (reviewed in [53]), with BamB on one side primarily interacting with POTRA 3 and BamCDE on the other side, where BamD interacts primarily with POTRA 5 (with a smaller number of interactions with POTRA 2) and BamE also interacts with POTRA 5. BamC binds to BamD and does not make direct contact with the BamA POTRA domains in three of the four structures. However, in one of the structures, the C-terminal domain of BamC interacts with POTRA 2 [30].

In the two BamABCDE structures, the BamA β-barrel is fully closed, and POTRA 5 is rotated about 30° from the β-barrel domain, allowing access to the β-barrel lumen. In the two BamACDE structures, POTRA 5 is located directly beneath the β-barrel lumen, excluding access, and the barrel itself is more open than was observed in the structure of BamA alone (figure 5). Binding of BamCDE to BamA induces a large conformational twist in the BamA β-barrel, where the first eight β-strands rotate away from strand β16. The most dramatic change is observed in strand β1, which shows a rotation of about 45°. The resulting BamA β-barrel makes no hydrogen bonds at all between strands β1 and β16, with only a single hydrogen bond linking the beginning of strand β1 to the periplasmic loop connecting strands β14 and β15. The movement changes the angle of the affected β-strands in the membrane, opens the exit pore through conformational changes in extracellular loops 1–3, and causes a large rearrangement of the β-barrel opening to the lipid bilayer (electronic supplementary material, movie S1). Whether BamB is actively modulating these conformational changes, and exactly how these movements assist folding and insertion of nascent OMPs, is unknown.

Figure 5.

Conformational changes in the BamA β-barrel and POTRA 5. A rotational movement of the first eight strands of the β-barrel and a translational movement of POTRA 5 are observed by comparing the BamACDE structure (open barrel, PDB code 5EKQ) with the BamABCDE structure (closed barrel, PDB code 5D0O).

There are many questions remaining in the BAM field, and the structures of Bam components and BAM complexes provide a framework for future experiments. Among the unresolved issues are the following: how folded or unfolded is the substrate OMP when bound to chaperones? How many copies of a chaperone bind to a single OMP? As different classes of OMPs display dependencies on different chaperones [45], does that indicate that there are multiple folding/insertion mechanisms facilitated by the BAM complex? What does the interaction between a chaperone–OMP complex and the BAM complex look like? What exactly is the BAM complex doing: does it actively fold OMPs, or simply facilitate insertion by local destabilization of the outer membrane, or does it participate in both folding and insertion? How similar is biogenesis of OMPs in mitochondria and chloroplasts [66,67]?

8. Conclusion

We have reviewed two major players of outer-membrane biogenesis: the LPS and the bacterial assembly machineries. Both complexes are essential for the integrity and survival of Gram-negative bacteria, and while both translocate basic building blocks of the bacterial outer membrane, their mechanism of action is markedly different. A series of crystal structures complemented by mutagenesis and molecular dynamics simulations offer a glimpse into the intricate mechanisms of these highly complex molecular machines. Further experiments are needed to fully understand the details of these processes and permit the design of molecules that target specific parts of the BAM and Lpt pathways, with potential therapeutic benefits.

Note added in proof

A recent cryo-EM structure of the BAM complex contains all five Bam proteins and shows a closed BamA barrel (Iadanza et al., Nat. Comm. 7:12865 (2016).

Authors' contributions

I.B., N.N. and S.K.B. wrote the manuscript and approved the final version.

Competing interests

We declare we have no competing interests.

Funding

I.B. and S.K.B. are supported by the Intramural Research Program of the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases (ZIA DK036139-10 LMB). N.N. is supported by the Department of Biological Sciences at Purdue University and by the National Institute of Allergy and Infectious Diseases (1K22AI113078-01).