The multifarious roles of Tol-Pal in Gram-negative bacteria

1The cell envelope of Gram-negative bacteria is comprised of a symmetric inner membrane and an asymmetric outer membrane with an intervening layer of peptidoglycan (PG) in the periplasm. The outer membrane is a robust protective barrier that shields the bacterium from the immune system and excludes major classes of antibiotics such as vancomycin thereby contributing to multidrug resistance. The outer membrane is not energised and there is no ATP in the periplasm so active processes must be coupled either to ATP hydrolysis in the cytoplasm or the proton motive force (PMF) across the inner membrane. The Tol-Pal system straddles the three layers of the cell envelope, is coupled to the PMF and plays a major role in constricting the outer membrane (Egan ).

2tol (tol-pal) genes were originally identified through mutations that engendered Escherichia coli tolerance towards colicins and bacteriophages (Gratia; Reeves; Hill and Holland; Nagel de Zwaig and Luria ). The tol-pal operon in E. coli is composed of seven genes, five of which are generally regarded as comprising the core Tol-Pal system in bacteria and in the following order: tolQ, tolR, tolA, tolB and pal. Deletion of these core genes generates the classical tol phenotype of outer membrane instability (see below) and all, with the exception of pal, also result in tolerance towards group A colicins and filamentous bacteriophages. Colicins are E. coli-specific, multidomain bacteriocins that harness the PMF through the Tol-Pal or ExbB-ExbD-TonB (Ton) systems to promote translocation of their cytotoxic domains across the OM (Cascales et al.; Kleanthous ). Phages also exploit these systems and appear to use similar strategies to colicins to deliver epitope signals into the cell (Jakes, Davis and Zinder; Riechmann and Holliger ). The likely reason E. coli Pal is not targeted by colicins is because it is not coupled to the PMF, which is needed for outer membrane translocation (Jetten and Jetten; Hancock and Braun; Lieberman and Hong; Braun and Herrmann ). The other genes that are part of the tol-pal operon, but which do not yield the same phenotypes as tol-pal mutations, are ybgC, a cytoplasmic lipid thioesterase, and cpoB/ybgF, a periplasmic regulator of peptidoglycan (PG) peptide crosslinking. The majority of this review is focused on core Tol-Pal proteins but additional components are included where their functions intersect with those of the Tol-Pal assembly.

3The tol-pal operon is found in all subclasses of proteobacteria and prominent in other phyla, principally the Chlorobi, Chlamydiae and Acidobacteria (Krachler et al. ). With the exception of ybgC and cpoB/ybgF, which are sometimes absent or replaced by other genes, the order of tol-pal genes is also highly conserved (Sturgis ). The essentiality of tol-pal genes varies in different species; the operon is not essential in E. coli K-12 but is essential in Caulobacter crescentus (Yeh et al. ) and Pseudomonas aeruginosa (Dennis, Lafontaine and Sokol; Lo Sciuto et al. ). In P. aeruginosa, tol-pal expression is modulated by iron in the medium and the growth phase of the organism (Lafontaine and Sokol; Duan et al. ). Beyond early studies suggesting that tol-pal expression is induced by RcsC in E.coli (Clavel et al. ) and quantitative proteomics studies showing all the components are expressed in both rich and defined media (Li et al. ) surprisingly little is known about how the system is regulated.

4The pleiotropic outer membrane instability phenotype typically associated with tol-pal mutations has been well-characterised but remains poorly understood, primarily because of the difficulties in differentiating traits attributable directly to tol-pal genes from those that are downstream effects, such as the activation of cell envelope stress responses. Lopes, Gottfried and Rothfield first characterised ‘leaky’ (lky) mutants in E. coli and Salmonella typhimurium (Lopes, Gottfried and Rothfield ) that were later mapped to the tol-pal operon (Lazzaroni and Portalier ). lky cells have a permeabilised outer membrane that releases ribonuclease I from the periplasm. In 1976, Weigand and Rothfield demonstrated that Salmonella cells with a standard lky mutant phenotype display a defect in outer membrane invagination during formation of the septum. Using electron microscopy, they showed ‘ballooning’ of the outer membrane from the septal region, with the formation of the large bleb on the surface of the cell (Weigand and Rothfield ). tol-pal cells bleb during division and produce copious outer membrane vesicles (OMVs). Indeed, tolR mutants are used to increase yields of OMVs for the production of vaccines against nontyphoidal Salmonella (Micoli et al. ) and Shigella flexneri (Pastor et al. ).

5E. coli K-12 tol-pal strains grow normally at 37°C in high salt growth media, but produce mucoid colonies at 30°C and are not viable at 42°C (Nomura and Witten; Bernstein, Rolfe and Onodera; Yakhnina and Bernhardt ). In LB media lacking salts at 30°C, E. coli tol-pal cells filament (Gerding et al. ). Conversely, E. coli tolA mutants grow in chains in both high and low osmolarity media (Meury and Devilliers ) and are unable to grow at high hydrostatic pressure (Black et al. ). tol-pal deficient cells tend to have increased sensitivity to surface active compounds such as bile salts (for example, deoxycholic acid), detergents such as SDS (Nagel de Zwaig and Luria ) and drugs such as polymyxin B (Lazdunski and Shapiro ). tol-pal cells are also more sensitive to β-lactam antibiotics (Davies and Reeves ), and susceptible to vancomycin (Onodera, Rolfe and Bernstein ) and novobiocin (Foulds and Barrett ), phenotypes that are consistent with the barrier function of the outer membrane being compromised.

6Another well-documented effect of tol-pal mutations is their impact on the ratio of phospholipids-to-lipopolysaccharide (LPS) in the outer membrane (Shrivastava, Jiang and Chng; Masilamani, Cian and Dalebroux ). tolA cells are unable to express full O-antigen on their surface (Gaspar et al.; Vines et al. ), probably due to a shortened core LPS (Anderson, Wilson and Oxender ). When the O-antigen is shortened (Rottem and Leive ) or the ratio of phospholipids-to-LPS is increased, the outer membrane becomes more fluid and thus more susceptible to mechanical stress as suggested by course-grain simulations (Jefferies, Shearer and Khalid ).

7tol-pal mutations activate two main cell envelope stress response pathways, Rcs and σE. The Rcs pathway senses lateral interactions between LPS molecules (Konovalova, Mitchell and Silhavy ) and modulates the expression of genes responsible for production of biofilm, capsule or modification of lipids (Wall, Majdalani and Gottesman ). Clavel et al. demonstrated that a mutation in E. coli RcsC increases its kinase activity and downregulates tolQRA expression (Clavel et al. ). When tolA (Clavel et al. ) or tolB (Mouslim, Latifi and Groisman ) genes are deleted, cells upregulate capsule production in a RcsC-RcsB-dependent manner, resulting in mucoid colonies at low temperatures (Bernstein, Rolfe and Onodera ). Similar activation of the Rcs pathway is seen in S. typhimurium tol-pal mutants (Masilamani, Cian and Dalebroux ). The σE stress response is activated by both misfolded outer membrane proteins and LPS that is retained in the periplasm (Lima et al. ), resulting in transcription of genes involved in outer membrane protein folding (the Bam complex) and degradation of misfolded proteins (DegP) (for a review of stress systems see (Mitchell and Silhavy ). Vines et al. demonstrated that both tolA and pal mutants increase their expression of degP, consistent with increased outer membrane fluidity and problems with outer membrane protein insertion (Storek et al. ).

8Tol-Pal is required for pathogenesis and virulence in many species of Gram-negative bacteria, including uropathogenic E. coli (Hirakawa et al. ), Edwardsiella ictaluri (Abdelhamed et al. ), Salmonella typhimurium (Bowe et al. ), Erwinia chrysanthemi (Dubuisson et al. ) and Haemophilus ducreyi (Fortney et al. ). Pseudomonas putida tolB cells are less efficient at forming biofilms (Lopez-Sanchez et al. ) with similar phenotypes reported for Burkholderia pseudomallei tolB (Khan et al. ) and E. coli tolA cells (Ranjith et al. ). Indeed, in P. aeruginosa and Xylella fastidiosa, tol-pal genes are overexpressed during biofilm formation (Whiteley et al.; Santos et al. ). Pal has been reported to be essential for persister cell survival during antibiotic treatment of E. coli (Sulaiman, Hao and Lam ). Uropathogenic E. coli pal-deficient cells are unable to produce capsule and are sensitive to serum (Diao et al. ). Finally, an aspect of Pal biology that is not well-understood is its apparent dual orientation in the outer membrane of some species, which has been exploited to produce vaccines. Pal is an abundant lipoprotein, normally inserted in the inner leaflet of the outer membrane by the Lol system (Ichihara, Hussain and Mizushima ). However, some Pal molecules are seemingly exposed on the surface of bacteria. This feature has enabled Pal-directed vaccines to be developed against Haemophilus influenzae (McMahon et al. ), Legionella pneumophila (Mobarez et al. ) and Acinetobacter baumannii (Lei et al.).

9In order to understand the principal functions of the Tol-Pal system a complete picture of its structural biochemistry is needed to which specific cellular phenotypes can be linked. The following sections summarise what is known of the structure and the function of core components of the Tol-Pal system; TolQ, TolR, TolA in the inner membrane, TolB in the periplasm and Pal in the outer membrane. While some structures are known (Fig. 1) others, such as the TolQ-TolR-TolA complex, are not. In these cases, inferences are made from past mutational, biochemical and in vivo studies along with functional similarities to homologous systems, primarily the ExbB-ExbD-TonB and MotA-MotB assemblies in Gram-negative bacteria.

10All three systems are PMF-linked nanomachines that drive mechanical processes at or beyond the outer membrane of Gram-negative bacteria and have related inner membrane stator complexes. The MotA-MotB complex drives rotation of the bacterial flagellum (Berg ), ExbB-ExbD powers TonB-mediated nutrient uptake through outer membrane transporters (Noinaj et al. ) and TolQ-TolR energises TolA to dissociate TolB-Pal complexes at the outer membrane (Szczepaniak et al. ). These simple comparisons emphasise three important points. First, that related stators can have very different biological functions. Second, that the three stator complexes have conserved residues implicated in proton transfer and almost certainly share common folds. Third, that the mechanical mechanism used by these stators to generate PMF-induced force is likely to be common to all of them, and indeed shared with other stators such as that used to drive gliding motility in Myxococcus xanthus (Youderian et al. ).

11Here, we focus on Tol-Pal but reference other systems where they illuminate aspects of Tol-Pal structure-function. Before detailing the structural biochemistry of each Tol-Pal component we briefly summarise current understanding of the biological function of Tol-Pal. Tol-Pal's PMF-mediated disruption of TolB-Pal complexes in the outer membrane is the means by which the system accumulates Pal at the division site. TolB and Pal are unique to the Tol-Pal system. Pal binds the cell wall so by increasing its concentration at the division site the cell has a way of preventing the ballooning associated with tol-pal mutations. TolB is the means by which the PMF-linked inner membrane stator ensures Pal is displaced wherever the stator is located, which, as we shall see, is linked to formation of the division septum.

12TolQ is 25-kDa protein composed of three transmembrane helices (Kampfenkel and Braun; Vianney et al. ). Although little structural or biochemical data are available for TolQ a number of mutational and crosslinking studies have established connections to partner proteins TolR and TolA (Derouiche et al.; Lazzaroni et al.; Germon et al.; Journet et al.; Zhang et al. ). Given the paucity of structural information we base the following on recent studies of the ExbB-ExbD complex. TolQ and TolR are homologues of ExbB and ExbD, with 35% and 29% sequence identity, respectively, between each homologue. Importantly, overexpression of tolQ-tolR complements a strain in which exbB and exbD are deleted, and vice versa, hinting at a common mechanism (Braun and Herrmann ). Six structures of ExbB, some in complex with ExbD, have been published often with differing subunit stoichiometry (Celia et al.; Maki-Yonekura et al.; Celia et al. ). ExbB is comprised of seven α helices, three of which are transmembrane (α2, α6 and α7). The transmembrane helices of individual ExbB molecules extend into the cytoplasm where they contribute to five-helix bundles that combine to form a chamber (Celia et al. ). In the pentameric structures of ExbB the chamber is a closed cavity with five-fold symmetry (Celia et al. ). A hexameric structure of the ExbB-ExbD complex has also been reported, when crystallised at high pH (pH 9.0) (Maki-Yonekura et al. ), but how relevant this structure is to the in vivo functioning of the complex is unclear.

13In the most recent cryo-EM structures, a 5:2 complex of ExbB-ExbD is observed in which the two transmembrane helices of the ExbD dimer reside within the pore formed by the ExbB pentamers rather than the membrane (Celia et al. ). We refer to this region of ExbD as the transpore helix (TPH). Although the TPH shows varying degrees of conservation amongst the stators (TolR and MotB are 66% and 19% identical, respectively, to the TPH of ExbD; Fig. 2A) they all possess conserved aspartic acid and phenylalanine residues (Asp23 and Phe32 in TolR). Native mass spectrometry data for the complex ejected directly from the native E. coli inner membrane are also consistent with a 5:2 stoichiometry for ExbB-ExbD (Chorev et al. ). Given these complementary data, the following discussion of Tol-Pal literature assumes TolQ-TolR also forms a 5:2 complex.

14A homology model of TolQ-TolR based on 5:2 subunit stoichiometry is shown in Fig. 2B–D. The TolQ pentameric assembly similarly forms a large cytoplasmic chamber but with distinct charge patterning on its inner surface relative to that seen in ExbB (Fig. 2E). ExbB has a band of positive charge in the middle of the chamber followed by a negative band beneath whereas the TolQ chamber is exclusively negatively charged. Whether these differences in electrostatics are physiologically relevant is not known. Notwithstanding these differences, however, the electrostatic charge state of the transmembrane regions of ExbB and TolQ are similarly neutral. A series of mutagenesis studies have identified several residues within TolR and TolQ as functionally important, some presumed to be part of the proton conducting pathway through the complex. These include Asp23 in the TolR TPH and Thr145, Thr178 and Pro187 in the second and third transmembrane helices of TolQ (Cascales, Lloubes and Sturgis; Goemaere, Cascales and Lloubes; Goemaere et al.; Zhang et al. ). The model presented in Fig. 2D shows how TolQ Thr145 and Thr178 are in close proximity to TolR Asp23. A similar constellation of residues are found in the ExbB-ExbD and MotA-MotB stator complexes (Braun and Herrmann ).

15Several studies have demonstrated that TolR is dimeric. In vivo disulphide crosslinking centred on TolR TPH residues are consistent with the TPH forming a homodimer (Zhang et al. ). Other in vivo studies have demonstrated that the periplasmic domain of TolR is also dimeric but likely to undergo substantial structural changes in response to the PMF and interactions with TolQ (Journet et al.; Goemaere et al. ). NMR and crystallographic studies of the periplasmic domains from H. influenzae and E. coli TolR, respectively, both show dimer structures (Parsons, Grishaev and Bax; Wojdyla et al. ). The homologous proteins ExbD and MotB have also been shown to be dimeric, by DEER spectroscopy and X-ray crystallography, respectively (O'Neill et al.; Celia et al. ).

16The structures of H. influenzae and E. coli TolR periplasmic domains reveal substantially different dimer interfaces suggesting they represent alternative structural states for the protein. The NMR structure of H. influenzae TolR was determined using a construct (residues 59–130) in which both the N- and C-termini of the periplasmic domain were truncated. (Fig. 1A). The β-sheets of each monomer contribute to form a deep cleft similar to a baseball mitt (Parsons, Grishaev and Bax ). Wojdyla et al. demonstrated that these sequences in the intact E. coli periplasmic domain (residues 36–142) form a strand-swapped dimer in which two additional β-strands and α-helix stabilise the dimer interface and obliterate the deep cleft observed in the truncated H. influenzae structure (Fig. 1B) (Wojdyla et al. ). Notwithstanding these additional sequences, the overall fold of the TolR domain is very similar in the two structures except that the subunits are rotated ∼180° relative to each other. The conformation of the strand-swapped E. coli dimer is consistent with earlier in vivo cysteine crosslinking studies showing that a spontaneous disulphide formed in vivo when Tyr117 was substituted for cysteine. The two residues are only close enough to form a disulphide in the full-length E. coli structure, but too far apart in the rearranged (truncated) H. influenzae structure. Moreover, formation of the disulphide inactivates Tol-Pal in vivo and blocks proton transport (Goemaere et al. ), consistent with the structural changes associated with the TolR dimer being linked to PMF activation of the stator complex.

17TolR binds PG but the molecular details are not yet known. Wojdyla et al. found that only the truncated form of E. coli TolR could bind to PG (in the form of isolated sacculi) whereas the full-length, strand-swapped dimer had no PG binding activity (Wojdyla et al.; Boags, Samsudin and Khalid ). Similar findings have also been reported for MotB (Roujeinikova; O'Neill et al.; Kojima et al. ). The picture emerging is one in which the shortened forms of MotB and TolR (and possibly ExbD, although this has not been demonstrated directly) bind to PG whereas PG binding is inhibited in the full-length proteins, most likely due to the combined effects of conformational rearrangement and occlusion of the PG binding site. The N-terminal sequences of the periplasmic domain in the full-length versions of MotB and TolR are of sufficient length when extended as disordered sequences to allow the PG-binding dimer to reach the cell wall, ∼90 Å from the inner membrane. While the molecular details of PG recognition by these dimers remains to be established, the structure of H. pylori MotB bound to the N-acetyl muramic acid (NAM) moiety of PG offers some clues (Roujeinikova ). This crystallographic/modelling study suggested glycan chains sit in grooves either side of the dimer interface, with the peptide cross-bridge connecting the two chains bound within the groove although there is no direct evidence for this binding mode.

18The available data point to TolR and MotB (and possibly ExbD) dimers existing in either closed or open states. In the closed state (equivalent to the full-length E. coli TolR structure; (Wojdyla et al. )), the strand-swapped dimer sits close to the surface of the stator partner in the inner membrane which is also thought to close the proton pore of the channel (Goemaere et al. ). In the open state, (equivalent to the truncated forms of MotB and TolR; (Parsons, Grishaev and Bax; Roujeinikova ), the N-terminal linker residues connecting the periplasmic domains to the TPH helix of each stator unravel enabling the restructuring of the dimer and binding of the cell wall. The PMF activates these large-scale structural transitions by moving protons between the TPH of TolR/ExbD/MotB and the specific inner membrane stator protein partner (TolQ/ExbB/MotA). We note that interatomic distances of residues present in both the TPH dimer and strand-swapped dimer of TolR differ significantly; the distance between Pro37 within these dimers is ∼7 Å and ∼40 Å, respectively. We speculate these differences may reflect changes within the stator that are linked to proton flow and PG binding by the periplasmic domain.

19TolA is a monomeric 40-kDa inner membrane protein comprised of three domains; a transmembrane helix (TolAI), a helical domain rich in alanine and charged residues that is thought to span the periplasm (TolAII) and a C-terminal, 12-kDa globular domain (TolAIII) (Levengood, Beyer and Webster; Witty et al. ). There are no structures available for intact TolA.

20Studies from a number of laboratories have shown that TolA is a protein-protein interaction hub, able to form complexes with Tol proteins (Derouiche et al. ) and CpoB/YbgF in the periplasm (Walburger, Lazdunski and Corda, Krachler et al. ) as well as being targeted by bacteriophages and bacteriocins (Cascales et al.; Kleanthous ), to promote their entry into cells (Fig. 3 and Box 1). Most of these interactions have been validated through structural and biophysical analysis. Other TolA interactions however have proven controversial either for lack of corroborating biochemical data or because they are contradicted by other work. For example, crosslinking studies have implicated TolAI as interacting with both TolQ and TolR, forming a TolQ-TolR-TolA complex in the inner membrane (Germon et al. ). Yet, as described above, the TolR TPH probably does not reside in the membrane (Celia et al.; Celia et al. ). Similarly, TolAIII was shown by crosslinking and immunoprecipitation assays to interact with Pal in the outer membrane (Cascales et al. ). However, no such interaction is observed by a range of biophysical methods using purified proteins (Bonsor et al. ). Moreover, Pal residues purportedly involved in binding TolAIII (Cascales and Lloubes ) in fact form the high affinity binding site for TolB (Bonsor et al.; Kleanthous ).

21Group A colicins and filamentous bacteriophages use protein-protein interactions to hijack the energised Tol-Pal system for entry into Escherichia coli cells (Cascales et al.; Kleanthous; Atanaskovic and Kleanthous ). Group A colicins bind to a specific surface receptor on the target bacterium from where they recruit an outer membrane porin (OmpF or OmpC), the pores of which are used to reach either TolA or TolB in the periplasm (Loftus et al.; Housden et al.; Housden et al. ). Filamentous bacteriophages (f1, fd, M13)) use the conjugating F-pilus as their receptor and thereafter target TolA. Some bacteriophages and group B colicins parasitise the Ton system for entry but are not dealt with further here (see (Cascales et al. )). Phages appear not to require Tol-Pal be coupled to the proton motive force for cell entry (Samire et al. ). Involvement of the PMF in colicin translocation remains controversial (Cramer, Sharma and Zakharov ). Several studies however clearly point to its requirement in the early stages of import across the outer membrane (Bonsor et al.; Vankemmelbeke et al. ).

22Fig. 3 shows the modes of binding of Ff phage coat protein g3p (a) and colicin A (b) with TolAIII, and colicin E9 with the β-propeller domain of TolB (c). The small N1 domain of g3p interacts with β3 of TolAIII through β-strand augmentation (Lubkowski et al. ), but adopts an anti-parallel orientation compared to the parallel orientation seen in the endogenous TolAIII-TolB complex (Szczepaniak et al. ) (compare with Fig. 1D). Colicin A on the other hand binds through an alternative β-strand augmentation site, interacting with β2 on the opposite side of the β-sheet (Fig. 3B) (Li et al. ). Bacteriocins such as colicins E2-E9 target TolB using an intrinsically disordered protein epitope that is part of the bacteriocin's translocation domain (residues 32–47) (Bonsor et al. ). Colicin E9 mimics interactions of Pal with TolB but without inducing the structural changes in TolB that normally diminish its binding to TolA (Loftus et al.; Bonsor et al. ). This stealth mechanism enables the surface-bound colicin E9 to connect itself to the PMF, which is required for the early stages of import (Rassam et al. ).

23Both TolAI and the equivalent transmembrane region of TonB have a conserved Ser-His-Leu-Ser motif (Koebnik ). Germon et al. found that mutating Ser18 and His22 in TolAI diminished TolQ binding (as determined by formaldehyde cross-linking and immunoblotting). Suppressor analysis has identified the first transmembrane helix of TolQ as the likely interaction site for the TolAI motif (Germon et al. ).

24One of the most important interactions of TolA is with the N-terminus of TolB in the periplasm. Suppressor mutation analysis and yeast two hybrid screens originally showed TolAIII binds to the N-terminal domain of TolB (Lazzaroni, Dubuisson and Vianney; Walburger, Lazdunski and Corda ). The interaction site on TolB was established definitively by deletion and biochemical analysis; removal of the N-terminal 12-amino acids of TolB generates a tol phenotype, inhibits binding of TolB to TolAIII (Bonsor et al. ) and abolishes accumulation of Pal at division sites (Szczepaniak et al. ). Biophysical studies have shown that the TolAIII-TolB complex has a low affinity (Kd∼40–200 μM, depending on the species), a consequence of the structural rearrangements in TolAIII. The recent solution state structure for Pseudomonas aeruginosa TolAIII bound to a TolB peptide shows a β-strand augmentation binding mechanism; the C-terminal helix of TolAIII is displaced by the N-terminal residues of TolB, which form a parallel β-strand (Szczepaniak et al. ) (Fig. 1D). The resulting complex is structurally similar to that of TonB bound to the TonB box of TonB dependent transporters (TBDTs) in the outer membrane (2.2 Å rmsd). The architecture of TonB-TBDT complexes, which can also have similarly low affinities, makes them mechanically stable (Chen, Radford and Brockwell; Hickman et al. ). The role of the PMF-coupled ExbB-ExbD stator complex is to exploit this mechanical stability to dislodge the plug domains of ligand-bound TBDTs via their complexes with TonB. We suggest that TolA through its coupling with the PMF-linked TolQ-TolR stator complex adopts a similar role, but one in which TolB is dislodged from its complex with Pal at the outer membrane via a TolA-TolB-Pal ternary complex (Szczepaniak et al. ). The biological rationale for energised dissociation of the TolB-Pal complex is explored below.

25There is no structural information available for the central 25-kDa domain of TolA (TolAII), > 50% of which is alanine, lysine and glutamate; the motif Lys-Glu–Ala3–Glu/Asp is repeated thirteen times (Levengood, Beyer and Webster; Derouiche et al.; Schendel et al. ). Solution X-ray scattering and far-UV circular dichroism predict that TolAII has an elongated helical structure, possibly involving a three-helix bundle (Derouiche et al.; Witty et al. ). Several pieces of evidence shed light on how TolAII might function. First, deletion analysis suggests the length of the domain is important (Schendel et al. ). Second, TolA undergoes structural changes in response to the PMF although the details of these rearrangements are obscure (Germon et al. ). Third, fluorescence recovery after photobleaching (FRAP) and single particle tracking (SPT) data show that GFP-TolA displays unrestricted Brownian motion in the inner membrane of non-dividing cells (Rassam et al. ) prior to its recruitment to the divisome (Gerding et al.; Rassam et al. ). Fourth, microscopy data suggest TolA can fully extend through the periplasm, as demonstrated by the capture and restriction of GFP-TolA in the inner membrane by colicin-bound TolB at the outer membrane (Rassam et al. ). Cumulatively, these data suggest that while TolAII can extend through the PG layer to reach the outer membrane it cannot be permanently extended as this would entrap TolA in the holes that exist in the PG (Turner et al. ) – estimated to be ∼50–100 Å diameter (Turner et al. ) – restricting its diffusion.

26We speculate that cycles of TolAII extension and retraction are linked to proton flux through the TolQ-TolR stator complex. Extension-retraction may also be a feature of TonB activity. A number of studies have shown that TBDT ligands activate transcription of their respective TBDT gene through a specific sigma factor-anti-sigma factor regulatory complex at the inner membrane (Noinaj et al. ). For example, the N-terminal periplasmic domain of the ferric citrate TBDT FecA interacts with the C-terminal periplasmic domain of the FecR regulator in the inner membrane, activating transcription of fecABCDE transport genes (Enz et al. ). Given the similarities between the Ton and Tol-Pal systems we speculate that such signalling might be based on TonB pulling the TonB box of a TBDT through the PG layer so its associated N-terminal signalling domain can physically interact with transcriptional regulators in the inner membrane. Below we explore how extension-retraction of TolA could be linked to the outer membrane stabilising role of Tol-Pal.

27TolB is a 45-kDa soluble periplasmic protein that is also an interaction hub. The structure of E. coli TolB (Abergel et al.; Carr et al. ) (Fig. 1E) and TolB in complex with Pal (Bonsor et al.; Bonsor et al. ) (Fig. 1F), TolA (Szczepaniak et al. ), colicin A (Zhang et al. ) (note, colicin A binds to both TolA and TolB) and colicin E9 (Loftus et al. ) (Fig. 3C) have all been reported. Other interaction partners have also been identified in vivo (including Lpp and OmpA) (Clavel et al. ), but these have not been validated in vitro nor structurally characterised. TolB is composed of two distinct domains; an N terminal α/β domain which binds TolA and a C-terminal, six-bladed β-propeller domain that binds Pal (Fig. 1F). The two domains are connected by a 9-residue linker sequence. The different structures of TolB combined with biophysical studies demonstrate that the protein is in conformational equilibrium, its different states favoured by specific binding partners. Pal stabilises largescale structural changes in TolB relative to the unbound state in which TolB's N-terminus becomes sequestered between its two domains (Bonsor et al.; Bonsor et al. ) (Fig. 1F). Consequently, Pal diminishes TolB's interaction with TolA since the TolB N-terminus constitutes the TolA binding site. Conformational changes in TolB's two domains ensue when the protein dissociates from Pal, releasing its N-terminus from its interdomain binding site and promoting binding to TolA.

28Using in vitro chemical crosslinking, Bonsor et al. also found a third, lowly-populated (presumably high energy) conformational state involving a ternary TolA-TolB-Pal complex (Bonsor et al. ). In this Pal-bound state, the N-terminus of TolB becomes dislodged from its interdomain binding site, enabling binding to TolA. Although not understood at the time the ternary complex likely plays a central role in the postulated force-dependent dissociation of the TolB-Pal complex (see below). Steered molecular dynamics simulations suggest the force required to dissociate the TolB-Pal complex is greater when the N-terminus of TolB is bound between its two domains (Szczepaniak et al. ). These simulations are consistent with the need for TolB's N-terminus to become dislodged from the body of TolB in the Tol-Pal complex to enable force-dependent dissociation. Moreover, they reveal that several conserved TolB linker residues mediate communication between TolB's N-terminal domain, where force in vivo is presumably applied, and the C-terminal β-propeller domain where Pal is bound. Mutation of these residues generates tol-like phenotypes consistent with such a role in vivo (Szczepaniak et al.).

29Pal (peptidoglycan associated lipoprotein) is attached to the inner leaflet of the outer membrane by an N-terminal lipid anchor from where it binds either PG (Lazzaroni and Portalier ) or TolB (Bouveret et al.; Clavel et al. ). Both crystal and NMR structures of Pal have been reported (Abergel et al. ), the latter bound to a fragment of PG (Parsons, Lin and Orban ). Pal has an α/β sandwich fold (Fig. 1G), the loops connecting its elements of secondary structure comprising the PG-binding site (Fig. 1H). Pal is a member of the same large family of PG-binding proteins that includes TolR and MotB but in contrast to these proteins is monomeric. Pal binds the diaminopimelic acid residue (mDAP) of non-crosslinked stem peptides within PG, utilising conserved aspartic acid and arginine residues (Asp71 and Arg73 in H. influenzae Pal) (Parsons, Lin and Orban ).

30One of the consequences of Pal being simultaneously tethered to the outer membrane and bound to the PG layer is that its lateral diffusion is severely restricted (Szczepaniak et al. ). Yet a key aspect of Tol-Pal function is the accumulation of Pal at division sites during cell division, showing that the protein is nevertheless mobile on the timescale of cell growth and division (Gerding et al.; Petiti et al.; Szczepaniak et al. ). Pal mutations or deletions that inhibit PG binding lead to faster and unrestricted diffusion in the outer membrane but also block outer membrane stabilisation and prevent the protein's accumulation at division sites (Petiti et al.; Szczepaniak et al. ). Pal employs the same residues to bind TolB as are used to bind PG (Bonsor et al. ). TolB is therefore key to Pal's accumulation at division sites where its role is two-fold; to block Pal binding to PG, thereby increasing its mobility in the outer membrane, and to render the complex a target for force-mediated dissociation by PMF-linked TolQ-TolR-TolA in the inner membrane (Szczepaniak et al. ).

31The pleiotropic nature of the tol-pal phenotype has confounded efforts to determine the physiological role of Tol-Pal in bacteria since discovery of the tol-pal genes. Some involvement in outer membrane stabilisation has always been envisaged but its nature was obscure. In addition, tol-pal genes are not essential in some Gram-negative bacteria, which is counter intuitive if the system is required for outer membrane stabilisation. With hindsight, the outer membrane blebbing frequently observed at mid-cell positions of dividing tol-pal mutants was an important clue (Weigand and Rothfield; Weigand, Vinci and Rothfield; Fung, MacAlister and Rothfield; Fung et al. ), which implied that the outer membrane at the constriction zone was dissociating from the cell wall. It was not until 2007 however, when it was demonstrated that all Tol-Pal proteins are recruited to the divisome (Gerding et al. ), that a role in outer membrane invagination at septation sites seemed likely. This role was originally thought to be that of an energised tether between TolA in the inner membrane and Pal in the outer membrane based on earlier in vivo crosslinking data (Cascales et al. ). However, such a mechanism is unlikely for three reasons. First, as described above, TolAIII and Pal do not interact in vitro. Second, a direct TolAIII-Pal interaction obviates the need for TolB in the periplasm yet deletion of tolB results in a classic tol phenotype. Indeed, mutations in TolB tend to be the most deleterious of all tol mutations in E. coli (Szczepaniak et al. ). Third, the TolAIII-TolB complex is clearly the focal point of the force that is generated by the PMF-linked TolQ-TolR stator complex (Szczepaniak et al. ). For TolB, a soluble protein, to be the target of force transduction in the periplasm only makes biological sense when viewed in the context of TolB's association with Pal in the outer membrane. This in turn implies that force-mediated dissociation of TolB-Pal in vivo must occur via a ternary TolA-TolB-Pal complex so that Pal can (re)bind PG.

32Which brings us to the recent studies of (Petiti et al. ) and (Szczepaniak et al. ). Both studies demonstrate that a major physiological role of the entire Tol-Pal assembly is the PMF-driven accumulation of Pal at division sites, where its binding of the cell wall helps invaginate the outer membrane and prevent blebbing. In addition, Szczepaniak et al. exploited a novel mathematical approach, developed by Seán Murray (Max Planck, Marburg), called SpatialFRAP in order to dissect the underlying mobilisation-and-capture mechanism (Szczepaniak et al. ). SpatialFRAP was used to extract effective diffusion coefficients (Deff) from fluorescence recovery after photobleaching (FRAP) data for Pal-mCherry expressed from the chromosomal locus in E. coli. This development was important because the diffusion of Pal varies both spatially and temporally during the E. coli cell cycle. Consequently, FRAP curves do not plateau and so standard FRAP analyses cannot be used to determine diffusion coefficients. Employing SpatialFRAP in conjunction with engineered strains Szczepaniak et al. uncovered the definitive characteristics of Pal mobility and the role of Tol proteins (Szczepaniak et al. ). First, the mobility of Pal in the outer membrane of non-dividing cells is very slow (effective diffusion coefficient, Deff ∼10−4 µm2.s−1) due to binding of the PG. Second, the onset of division leads to an acceleration in Pal mobility throughout the cell except at the division site where instead Pal molecules accumulate and mobility is similar to that in non-dividing cells. Third, all components of the Tol system and the PMF are required for these combined effects. A particularly remarkable aspect of this mechanism is the action-at-a-distance on Pal mobility when the divisome is formed. How do Pal molecules far from the divisome have their outer membrane mobility enhanced while those at the divisome do not and how is cellular energy expended to achieve these joint outcomes?

33The answer as we currently understand it is comprised of four elements, two reasonably well-understood and two hypothetical (Fig. 4). The well-established elements are: (1) the recruitment of PMF-linked TolQ-TolR and TolA to the divisome, albeit the mechanism is still not known (Gerding et al.; Petiti et al. ), and, (2) TolB's inhibition of PG binding by Pal (Bonsor et al. ), which likely increases Pal mobility in the outer membrane and the chances of a diffusing TolB-Pal complex being captured by TolQ-TolR-TolA at the divisome. The two elements for which there is as yet no direct evidence are: (1) active dissociation of TolB-Pal complexes by PMF-linked TolQ-TolR-TolA, and (2) translocation of dissociated TolB molecules through holes in the PG layer by TolA, the same holes TolA itself would have used to reach the outer membrane in the first place (Fig. 4). We argue that it is this spatial separation of TolB molecules (those actively dissociated by TolQ-TolR-TolA from those remaining bound to Pal in the outer membrane) by the intervening PG layer that explains action-at-a-distance on Pal mobility. Because of this spatial separation, dissociated TolB molecules can only diffuse between the inner membrane and the PG until a hole is found through which they can again reach the outer membrane to rebind Pal (Fig. 4). We note that this is not the first model to suggest the importance of PG pores for spatial separation of periplasmic proteins. Regulation of peptidoglycan synthesis, for example, involves outer membrane lipoproteins reaching through holes in PG to interact with inner membrane proteins, as in the case of LpoB and PBP1B (Egan and Vollmer; Turner et al.; Egan et al. ). Sacculi are known to contain pores as large as 5–16 nm (Turner et al. ), which presumably also reflects the situation at the septum where largescale remodelling takes place during cell division. To conclude, TolB serves as a PMF-recycled catalyst of Pal mobility, mobilising Pal molecules anywhere in the cell except at the divisome where Pal is kept free of TolB through the localised action of TolQ-TolR-TolA.

34A major change in Pal mobility ensues in non-dividing cells when the TolQ-TolR-TolA complex is no longer confined to the divisome (Szczepaniak et al. ). Now, a TolB-Pal complex anywhere in the cell can be captured by diffusing TolQ-TolR-TolA, releasing Pal to bind the cell wall. The net result is that in non-dividing cells Pal is predominantly bound to the PG because the small number of TolB molecules (present at ∼10% the levels of Pal) that could increase its diffusion are prevented from doing so. As a result, Pal's lateral diffusion in the outer membrane slows. A potential consequence of TolB-Pal complexes being continually captured by TolQ-TolR-TolA in non-dividing cells is the redistribution of Pal in the cell envelope, which is further addressed below.

35How is the mobilisation-and-capture of Pal described above linked to force generation by TolQ-TolR-TolA? We speculate that extension-retraction of TolA may be coincident with the flow of protons through the stator and the (as yet unresolved) structural changes in TolQ that cause unfurling of the strand-swapped TolR dimer so that it can extend and bind PG (Fig. 4). In this TolR-PG-anchored state, TolA extends through the periplasm, possibly also interacting with PG-bound TolR (a similar interaction has been suggested to occur between ExbD and TonB; (Ollis and Postle )), to capture TolB from a TolB-Pal complex in the outer membrane. Reversal of these steps, for example through the loss of protonation, would result in both TolR and TolA, the latter now bound to TolB, returning to the inner membrane in their retracted states. We suggest that the TolQ-TolR-TolA complex may be continuously going through this cycle in response to the PMF.

36The mechanism we propose for the Tol-Pal system raises many questions that are also pertinent for other PMF-driven nanomachines in the bacterial cell envelope. Does the movement of protons through these conserved complexes transduce force to their specific partners, the flagellum, TonB, TolA, by similar mechanisms? In the case of the bacterial flagellum, many MotA-MotB stators engage with the flagellum and even exchange during active rotation of the flagellum (Leake et al.; Reid et al.; Brenzinger et al. ). How many stators are involved in driving the motion of TonB and TolA? Alternatively, can more than one TonB/TolA engage with a single stator complex? What are the structural transitions experienced by TonB and TolA and how are these coupled to the unplugging of TBDTs and the dissociation of TolB-Pal complexes, respectively, in the outer membrane? Can the plug domains of TBDTs bound to TonB be brought through the PG layer as we have postulated for TolB bound to TolA?

37Recent work has revealed that once localised to the divisome, the Tol-Pal assembly has a broader role within the cell envelope beyond stabilising the connection between the outer membrane and cell wall. Tol-Pal is also involved in remodelling the PG at division sites. One of these roles involves cpoB, the terminal gene in the tol-pal operon. CpoB (coordinator of PG synthesis and outer membrane constriction associated with PBP1B, formerly known as YbgF) is a 28 kDa periplasmic protein that has long been an enigma. Although widely conserved in bacteria, deletion of cpoB does not generate a characteristic tol-pal phenotype but does sensitise cells to certain β-lactam antibiotics, such as cefsulodin, which target penicillin binding protein 1B (PBP1B). Krachler et al. demonstrated that CpoB has an elongated oligomeric structure, composed of a trimeric coiled-coil attached to a three-repeat tetratricopeptide repeat (TPR) domain, and that this structure is disrupted when the TPRs of CpoB associate with TolAII, generating a heterodimeric CpoB-TolA complex (Krachler, Sharma and Kleanthous; Krachler et al. ). Subsequent studies by Gray et al. showed that CpoB is an important regulator of PBP1B transpeptidase activity and that this regulation is further moderated by PMF-linked TolQ-TolR-TolA (Gray et al. ).

38PBP1B is an inner membrane bifunctional PG synthase with both glycosyltransferase and transpeptidase activity. These activities are stimulated by the outer membrane lipoprotein LpoB, resulting in PBP1B producing hyper-crosslinked PG. The TPR domain of CpoB associates with PBP1B to block LpoB-mediated activation of PG crosslinking thereby generating fewer peptide crosslinks within the PG. TolA, which also binds to PBP1B, reverses the inhibitory effect of CpoB on PBP1B transpeptidase activity, reinstating hyper-crosslinked PG. It is not clear what the oligomeric status of CpoB is when bound to PBP1B nor if the same (or different) regions of the CpoB TPR domain that bind TolA also bind PBP1B. Importantly, however, TolA needs to be coupled to the PMF, which implies that in order for CpoB's inhibitory effect on PBP1B-LpoB transpeptidase activity to be reversed TolAII must extend through the PG layer. Hence, not only is the PMF-linked Tol-Pal system involved in loading division septa with Pal that bind non-crosslinked stem peptides within PG, but it also regulates the degree of peptide crosslinking at these sites by modulating CpoB's influence on PBP1B-LpoB transpeptidase activity.

39In the latter stages of bacterial cell division glycan strands connecting daughter cells need to be cleanly cut. Two recent studies point to Tol-Pal being involved in this process. During daughter cell separation crosslinks connecting glycan strands are cut by amidases and endopeptidases. Their action is tightly controlled by specific activators, NlpI (Banzhaf et al. ) and NlpD and EnvC (Uehara et al. ). Tol-Pal exerts a degree of control over amidase activity through NlpD. Although no direct interactions between Tol proteins and NlpD have been described, cells deficient in envC and tol genes display the same growth defects as cells lacking both amidase regulators (Tsang, Yakhnina and Bernhardt ) suggesting Tol-Pal may be involved.

40One of the phenotypic outcomes of tol-pal mutations is cell chaining (Fung, MacAlister and Rothfield; Fung et al.; Gerding et al. ). Given the importance of Tol-Pal for invaginating the outer membrane this phenotype has always been interpreted as demonstrating an outer membrane defect in tol-pal mutants. Yakhnina and Bernhardt reported recently that this is not the case (Yakhnina and Bernhardt ). Instead, Tol-Pal is needed for efficient processing of septal PG. They found that sacculi generated from tol-pal mutants are also chained suggesting that the Tol-Pal system plays a role in promoting the cleavage of PG-linked daughter cells. Yakhnina and Bernhardt conducted a phenotypic suppressor screen to identify cell components involved in this activity (Yakhnina and Bernhardt ). They identified a number of suppressors in the protease Prc which, together with its partner protein NlpI, hydrolyses the cell wall endopeptidase MepS. Subsequent multicopy suppressor analysis identified other PG hydrolase targets of Prc, including a novel amidase christened DigH, the overexpression of which complemented the cell chaining phenotype of tol-pal mutants. These authors also showed that DigH is recruited to the divisome independent of Tol-Pal and that it preferentially cleaves glycan chains lacking stem peptides. These observations may help explain why in some species of bacteria (for example Chlamydia spp) a lytic transglycosylase is associated with the tol-pal operon that could serve a similar role in cleaving glycans connecting daughter cells. How Tol-Pal promotes efficient septal PG hydrolysis via DigH and other lytic transglycosylases and whether this requires the PMF remains to be established.

41E. coli tol-pal mutants accumulate phospholipids in their outer membranes, similar to bam and lptD mutants (Shrivastava, Jiang and Chng ). However, unlike bam mutants, tol-pal mutants seem to have impaired retrograde phospholipid transfer to the inner membrane and retain phospholipids in the outer leaflet of the outer membrane (Shrivastava, Jiang and Chng; Shrivastava and Chng ). Unidirectional transport of phospholipids in and out of the outer membrane is mediated by the Mla system, mutations in which affect the lipid asymmetry of the outer membrane (Malinverni and Silhavy ). Overexpression of mla in tol-pal cells partially recovers outer membrane asymmetry (Shrivastava, Jiang and Chng ). It has been suggested that since the same phospholipid-retaining phenotype is observed in conditions where tol cells are able to divide normally (Gerding et al. ) then this phenotype does not stem from cell septation problems (Shrivastava, Jiang and Chng ). However, it is possible that phospholipids are retained in the outer leaflet of the outer membrane to compensate for the loss of the Tol-Pal system. Another Tol-Pal connection to phospholipid biosynthesis is the tol-pal operon gene ybgC, which encodes a thiol diesterase (Gully and Bouveret ). In S. Typhimurium, ybgC mutants accumulate phosphatidylglycerol and phosphatidylethanolamine in the outer membrane, similar to tol-pal mutants (Masilamani, Cian and Dalebroux ). There is increasing evidence that MCE transporter proteins such as LetB (Isom et al. ) form protein tunnels that act as conduits for phospholipids to the outer membrane, further suggesting that any involvement in phospholipid trafficking by Tol-Pal is indirectly linked to its outer membrane stabilising role.

42The Tol-Pal system has been implicated in the polar localization of several inner membrane proteins with consequent impact on bacterial development and behaviour. In C. crescentus the system is required for polar localization of TipN, which regulates cell asymmetry and polar development in the organism (Yeh et al. ). Tol-Pal is required for cell motility in both P. putida and E. coli (Llamas, Ramos and Rodriguez-Herva; Youderian et al.; Gao, Meng and Gao ). In E. coli this has been shown to be due to recruitment of chemoreceptor clusters to cell poles (Santos et al.; Neeli-Venkata et al. ). Although co-immunoprecipitation analyses in these studies show that Tol-Pal proteins associate with the proteins being localized to the poles it remains to be established if this is due to direct interactions with Tol-Pal proteins or an indirect result of Tol-Pal activity; for example, the accumulation of Pal at new poles following the completion of cell division. It is also not known if Tol-Pal coupling to the PMF is required for polar localization of these systems.

43Why do bacteria need an energised system to stabilise the connection between the outer membrane and the underlying cell wall when other PG binding proteins exist in the outer membrane that could conceivably carry out the same stabilising function? In E. coli, the two other main PG binding proteins in the outer membrane are OmpA and Braun's lipoprotein, Lpp. OmpA, which has a similar abundance to Pal in the outer membrane (∼105 copies), is composed of an integral outer membrane β-barrel and a periplasmic PG binding domain similar to that of Pal. The biogenesis of outer membrane proteins can occur everywhere except the poles in E. coli (Rassam et al. ) and so OmpA could contribute to outer membrane stabilisation at the divisome. However, only those molecules inserted close to the divisome would be useful in this regard since OmpA cannot diffuse laterally in the outer membrane (Verhoeven, Dogterom and den Blaauwen ).

44Lpp, one of the most abundant proteins in bacteria (∼106 copies), is covalently cross-linked by a suite of transpeptidases to the same mDAP side-chain to which Pal binds non-covalently (Asmar and Collet ). lpp mutants, which also have destabilised outer membranes, can be rescued by overexpressing pal but pal mutants are not similarly rescued by lpp overexpression (Cascales et al. ). This observation demonstrates that Lpp cannot compensate for the loss of Pal's outer membrane stabilising function at division sites whereas Pal can compensate for the loss of Lpp crosslinks, although it has been reported that tol mutants have less Lpp bound to PG (Weigand and Rothfield ). Since Pal actively relocates to mid-cell en masse through the action of the Tol-Pal system and its binding is mutually exclusive with the covalent attachment of Lpp we speculate that it may have a role in modulating Lpp crosslinking to PG at division sites. When viewed in the context of Tol-Pal's regulation of PG peptide crosslinking density at division sites (via CpoB) and its involvement (direct or indirect) in the cleavage of glycan chains, these observations all point to Tol-Pal being able to coordinate outer membrane invagination with separation of daughter cells. Moreover, the status of this coordinating role is communicated to the FtsZ constriction ring in the cytoplasm since a delay in the recruitment of Tol-Pal to the divisome delays closure of the Z-ring (Rassam et al. ).

45Why is the tol-pal operon essential in some Gram-negative bacteria but not others? In some instances, the answer may lie in the lack of redundancy in systems that stabilise the outer membrane; C. crescentus, for example, where tol-pal is essential, lacks lpp (Yeh et al. ). Environmental factors could also contribute to tol-pal’s importance especially if these place additional stresses on the outer membrane; for example, tol-pal is essential for the infection of hosts by many pathogens (Bowe et al.; Fortney et al.; Dubuisson et al.; Abdelhamed et al.; Masilamani, Cian and Dalebroux; Hirakawa et al. ). A major factor in the successful exploitation of diverse ecological niches by Gram-negative bacteria is the presence of O-antigen in the outer membrane which brings additional stability to the membrane. Species where tol-pal is essential such as P. aeruginosa present O-linked sugars on the surface (Rivera et al. ) whereas E. coli K-12, where tol-pal is not essential, does not produce O-antigen. Might the presence of O-antigen on the bacterial surface require cells to have tol-pal? The study of Gaspar et al. suggests this might be the case (Gaspar et al. ). This study asked two related questions: Is tol-pal essential in wild-type E. coli O7 antigen expressing strains and what happens when O7 antigen expression is introduced into E. coli K-12 cells which otherwise does not make O7? They found that tol-pal genes could not be deleted from E. coli O7 and that O7 expression in E. coli K-12 was significantly reduced if these strains also carried tol-pal deletions. tol-pal essentiality may therefore stem from problems invaginating the outer membranes of soon-to-be-daughter cells if these are inter-digitated due to the presence of O-antigen. Such inter-digitation might require an energised system to invaginate the outer membrane and so separate cells, whereas this requirement might be relaxed in the absence of O-antigen. Consistent with this idea, coarse-grained molecular dynamics simulations of asymmetric outer membrane models indicate that the strong cohesive interactions of tightly-packed O-antigen in smooth LPS make the membrane much more resistant to mechanical deformation compared to rough LPS that lacks O-antigen (Jefferies, Shearer and Khalid ). There are exceptions however that contradict this idea. Salmonella enterica serovar Typhimurium is able to produce full-length LPS in a tol-pal background (Prouty, Van Velkinburgh and Gunn ) albeit these mutants decrease the LPS content of the outer leaflet by retaining phospholipid (Masilamani, Cian and Dalebroux ).

46The TolQ-TolR-TolA complex is released from the divisome when septation is complete, leaving the proteins free to diffuse in the inner membrane. Pal that had accumulated at the divisome is polar in daughter cells redistributes before the next division (Szczepaniak and Kleanthous, unpublished observations). This redistribution is also likely to be dependent on TolQ-TolR-TolA and the PMF since Pal diffusion is too slow otherwise. If this is the case, this would imply that even when diffusing in the inner membrane the TolQ-TolR-TolA complex uses the PMF to scan the outer membrane for TolB-Pal complexes on which to pull (as postulated in Fig. 4). Hence, the Tol-Pal system may have another outer membrane stabilising role in bacteria beyond that at the divisome, as a de facto outer membrane surveillance system, using the PMF to redistribute Pal connections to the PG. Such an activity would be advantageous for a Gram-negative bacterium since it could help maintain outer membrane stability in the event of damage, for example, by antimicrobial peptides. An outer membrane surveillance role might be a contributory factor in the pleiotropic instability phenotype typically associated with tol-pal mutations, such as the production of OMVs. OMVs mediate macromolecule transfer between bacterial cells and are implicated in biofilm formation and pathogenesis, but how their production is regulated is poorly understood (Schwechheimer and Kuehn ). Tol-Pal has long been associated with OMV production since tol-pal mutations hyper-vesiculate, particularly at division sites, but it has been unclear if this activity is regulated in any way. The mobilisation-and-capture mechanism uncovered for Tol-Pal could be amenable to regulated production of OMVs through, for example, the modulation of TolB interactions with Pal and/or TolA.

47Fig. 3 shows the modes of binding of Ff phage coat protein g3p (a) and colicin A (b) with TolAIII, and colicin E9 with the β-propeller domain of TolB (c). The small N1 domain of g3p interacts with β3 of TolAIII through β-strand augmentation (Lubkowski et al. ), but adopts an anti-parallel orientation compared to the parallel orientation seen in the endogenous TolAIII-TolB complex (Szczepaniak et al. ) (compare with Fig. 1D). Colicin A on the other hand binds through an alternative β-strand augmentation site, interacting with β2 on the opposite side of the β-sheet (Fig. 3B) (Li et al. ). Bacteriocins such as colicins E2-E9 target TolB using an intrinsically disordered protein epitope that is part of the bacteriocin's translocation domain (residues 32–47) (Bonsor et al. ). Colicin E9 mimics interactions of Pal with TolB but without inducing the structural changes in TolB that normally diminish its binding to TolA (Loftus et al.; Bonsor et al. ). This stealth mechanism enables the surface-bound colicin E9 to connect itself to the PMF, which is required for the early stages of import (Rassam et al. ).