3.1. Structure determination
Full-length SmLip consists of 176 residues. A truncated version of SmLip consisting of an N-terminal hexahistidine tag plus a TEV protease recognition site followed by residues Ala30–Asp176 was obtained in recombinant form and purified. The N-terminal 29 amino acids, which include the lipidation signal peptide and the first four residues of the mature protein, have been omitted. This sample gave monoclinic crystals. The asymmetric unit consists of four polypeptide chains, labelled A–D, with an estimated solvent content of 45% and a V
M of 2.27 Å3 Da−1.
Medium-resolution diffraction data were recorded in-house and the anomalous scattering information was used in a SAD approach to phasing. 13 potential iodide positions were identified and produced a figure of merit of 0.43 to 2.35 Å resolution. Subsequently, 12 of these positions were confirmed by refinement with this data set. The initial model constructed in RESOLVE consisted of 293 residues, with a correlation coefficient of 0.55 and R
work and R
free values of 46% and 49%, respectively. The first round of model building in Coot extended this to 467 residues, with a correlation coefficient of 0.72 and R
work and R
free values of 33% and 37%, respectively. At this point the high-resolution synchrotron data (1.92 Å resolution) became available and were used to continue the analysis. The refinement proceeded with the release of NCS restraints and the incorporation of water molecules, an Na+ ion, ethylene glycol and a number of side chains with dual rotamer conformations. This data set was derived from crystals grown in the presence of chloride instead of iodide. However, we did not assign any chloride ions to the structure, noting that typical water molecules occupy the previously identified iodide-binding sites. The refinement was terminated when there were no significant changes in R
work and R
free and inspection of the difference density map suggested that no further corrections or additions were justified. Several dual rotamers are incorporated into the model. Disorder was evident at several positions, for example the N-terminus, where it was not possible to interpret diffuse and weak electron density. Consequently, several residues are absent from the model. Molecule A consists of residues 32–173; molecule B of residues 33–142 and 147–176; molecule C of residues 34–50, 53–143 and 147–175; and molecule D of residues 33–50 and 55–175. The geometry of the model is acceptable (Table 1).
3.2. Self-association and localization in vivo
Previous work on SciN, the Lip homologue from enteroaggregative E. coli
, showed that the protein is localized in the outer membrane, facing the periplasm (Aschtgen et al.
). Examination of the amino-acid sequence of the N-terminus of Sm
Lip predicts that this is also an outer-membrane lipoprotein. The LipoP
1.0 algorithm (Juncker et al.
) predicts that Sm
Lip has a lipoprotein signal peptide and that signal peptidase II cleavage occurs between Gly25 and Cys26, with the cysteine subsequently being lipidated. Additionally, the residue at the +2 position following cleavage is Met27 (i.e.
it is not an aspartate, which directs retention in the inner membrane); therefore, Sm
Lip should proceed to the outer membrane via
the Lol system (Bos et al.
In order to investigate whether Sm
Lip undergoes self-interaction, the bacterial two-hybrid system (Karimova et al.
) was utilized in E. coli
. This assay involves reconstitution of adenylate cyclase activity from two non-interacting cyclase fragments, called T18 and T25, from Bordetella pertussis
. The presence of cyclic AMP activates the transcription of maltose and lactose catabolic operons by E. coli
. This can be detected by direct measurement of β-galactosidase activity or by using the observation that bacteria capable of fermenting maltose acidify the medium and thus generate a red colour on MacConkey–maltose indicator plates.
SmLip was introduced as both bait and prey by encoding on plasmids pUT18 and pT25, and a strong positive result was observed (Fig. 1). Mature SmLip (lacking the N-terminal signal peptide) was used for this experiment, firstly to correspond to the form of SmLip for which the structure was solved and secondly to ensure that both partners were localized together in the cytoplasm after fusion with T18 or T25. This positive result indicates that Lip does indeed self-associate within the cell and that neither localization in the outer membrane nor other components of the type VI secretion apparatus are required for self-interaction. We note, however, that this system is unable to distinguish between dimerization or higher order oligomerization.
Figure 1 Detection of Lip–Lip self-interaction. The bacterial two-hybrid system was used to detect an in vivo interaction between Lip (minus signal peptide) fused to T25 (pSC080) and Lip (minus signal peptide) fused to T18 (pSC072). The empty vectors pUT18 (more ...)
As a control for any propensity of SmLip to form nonspecific interactions, in addition to the lack of interaction with the T18 and T25 proteins demonstrated in Fig. 1 we tested whether SmLip gave a positive bacterial two-hybrid result with several cytoplasmic components of the T6SS (with which, as a periplasmic protein, it should not interact). SmLip gave a negative result (indistinguishable from the T25 + T18 negative control) when tested against the proteins VipB, TssK and TssL (data not shown).
In order to confirm the localization of the native Lip protein in S. marcescens, we utilized an anti-Lip polyclonal antibody to probe each of the major cellular fractions in this organism. As shown in Fig. 2, native SmLip is found exclusively in the membrane fraction, confirming the predicted localization of the protein and the functionality of the signal peptide.
Figure 2 Cellular localization of native SmLip in S. marcescens. Anti-Lip immunoblot of whole cells or cellular fractions prepared from wild-type S. marcescens strain Db10 or the Δlip mutant SJC10 (WC, whole cell; Peri, periplasm; Sph, sphaeroplast; Cyto, (more ...)
3.3. Overall structure
Lip polypeptide can be classified as a new member of the transthyretin-like superfamily and a detailed comparison will be given below. The protein displays a compact globular structure dominated by an eight-stranded β-sandwich (Fig. 3; Supplementary Fig. S11
). The order of the strands is 8–7–1–4 and 6–5–2–3. There are three short α-helical segments and three 310
-helix turns. The four Sm
Lip polypeptide chains in the asymmetric unit are similar, with the root-mean-square (r.m.s.) deviations between superimposed Cα
atoms ranging from 1.3 Å (monomers A
) to 0.8 Å (monomers A
) with an average value of 0.95 Å.
Figure 3 The secondary structure and fold of SmLip. β-Strands are shown as blue arrows and α-helices and 310-turns as red and green ribbons, respectively. The N- and C-terminal residues are labelled and the orientation of the protein with respect (more ...)
Although a set of core conserved proteins are encoded by the T6SS gene clusters in different Gram-negative bacteria (data not shown), there is a large degree of variation in the amino-acid sequences of these proteins. Lip and its orthologues, for example, are relatively poorly conserved. Excluding the signal peptide and lipobox motif (Fig. 4), SmLip shares only about 20% sequence identity with SciN, the homologue from enteroaggregative E. coli. This increases to near 40% in comparison with the homologue from the P. aeruginosa HSI-1 T6SS. Sequence conservation is noted in loop 1, near α1 and α2, in loop 2 and in the loop 4–β6 region (Fig. 4).
Figure 4 The primary and secondary structure of SmLip and sequence alignment with two homologues. S. marcescens Lip is aligned with the homologous proteins from enteroaggregative E. coli (GenBank CBG37366.1) and P. aeruginosa (NCBI Reference Sequence NP_248770.1, (more ...)
An alignment of Sm
Lip with eight orthologues (Supplementary Fig. S21
) reinforces the observation of a low level of sequence identity for this protein. Excluding two residues in the lipobox motif, only six residues are strictly conserved: Asn48, Leu99-X
-Pro101-Gly102, Gly120 and Ala124. All six residues appear to contribute to the conformation of specific parts of the fold (data not shown). The side chain of Asn48 accepts a hydrogen bond from the main-chain amide of Gln126, helping to define the conformation of loop 4. The Leu99-X
-Pro101-Gly102 segment defines the structure of the turn after β3 leading into loop 2. Gly120 and Ala124 occur in β5 and contribute hydrogen bonds to form interactions with β2 and β6 on either side. An increase in size of the side chain at either of these positions would be likely to be disruptive to the formation of this β-sheet, which forms one side of the structure. There is no obvious hydrophobic, basic or acidic surface feature on Sm
Lip that is likely to be conserved within the Lip proteins since the few conserved residues are mainly buried.
The information provided in §
3.2 identifies that the N-terminus of the structure is placed close to the outer membrane, hence the assignment of the orientation of Sm
Lip with respect to the outer membrane (Fig. 3). By extension, we note that the areas of Sm
Lip in which sequence conservation is observed mainly appear to contribute to stabilizing parts of the structure that jut out into the periplasm. They may therefore serve to define the structure of parts of Lip that are responsible for interaction with other molecules in the periplasm.
3.4. The tetramer is likely to be a crystallographic artefact
Gel-filtration data acquired during purification indicated that SmLip is a monomer in solution (data not shown). In contrast, the bacterial two-hybrid data revealed a propensity for self-interaction and the asymmetric unit is a tetramer displaying 222 point-group symmetry (Fig. 5). The accessible surface area (ASA) of the SmLip polypeptide averages out at approximately 8350 Å2; the range is from 8200 Å2 for molecule D to 8510 Å2 for molecule A. Each molecule in the asymmetric unit interacts with two of the other three and two types of protein–protein interface are formed between molecules A–B and C–D (interface I) and between molecules A–C and B–D (interface II). The type I interface, which is larger, covers an area that is approximately 13% of the ASA of the SmLip molecule. Such coverage certainly indicates potential for a biologically relevant dimer. This interface is primarily formed by the antiparallel alignment of two β7 strands. Three aromatic residues, Phe147, Trp151 and Phe153, contribute van der Waals interactions to the association and, by virtue of their relative bulk, also to the ASA (data not shown). The type II interface covers about 6.5% of the ASA of a molecule, a level typical of the values observed simply owing to molecular packing in a crystal lattice. This interface is formed by the antiparallel alignment of two β4 strands. The areas of SmLip involved in forming a tetramer are not conserved in the homologues from E. coli or P. aeruginosa (Fig. 4) and it is unlikely that such a tetramer is a generic feature of this lipoprotein.
Figure 5 The asymmetric unit. The four molecules that constitute the asymmetric unit are shown in different colours using the secondary-structure assignment given in Fig. 1 and labelled. The two types of protein–protein interface are labelled, (more ...)
The spatial placement of the N-terminal residues in the asymmetric unit is such that it is unlikely that an oligomeric assembly could form when the protein is anchored in the membrane by the lipidated Cys26 at the N-terminus. The N-termini of molecules A and D are on the same side of the tetrameric assembly but are opposite to those of molecules B and C. As explained, there are no direct interactions formed between molecules A and D or molecules B and C. That the bacterial two-hybrid experiments reveal a propensity for self-interaction of the truncated protein in the cytoplasm is in one sense consistent with the crystal structure of the truncated version of SmLip, which shows a tetrameric assembly containing a plausible dimer. On the other hand, the structure of the tetramer is incompatible with dimeric or tetrameric structures if the N-terminus is membrane-bound. These observations may be a result of the different concentrations and experimental conditions used. We suggest that SmLip is a membrane-bound monomer but displays a propensity to interact with itself.
A reviewer commented on the possibility that the SmLip tetramer might represent an inactive or alternative state of the protein. This is an intriguing suggestion and raises questions about how conversion to an active form might occur and how the T6SS itself is regulated. We have no data to address this issue and further studies would be required to investigate such a possibility.
3.5. Comparisons with structural homologues
A search for structural neighbours in the Protein Data Bank using the PDBeFold
) and ProFunc
servers (Laskowski et al.
) gives a Z
score of 6.1 with sea bream transthyretin (Eneqvist et al.
; PDB entry 1sn0
). This matched 84 residues with an r.m.s.d. of 2.7 Å. The β-sheet structures align well (Supplementary Fig. S31
). The r.m.s.d. and relatively low Z
score reflect the low sequence identity shared between the two proteins of approximately 7%. Nevertheless, the structural relationship is clear and Sm
Lip can be classed as a new member of the transthyretin-like protein family. The only other member of this protein family is 5-hydroxyisourate hydrolase (EC 126.96.36.199; Hennebry et al.
), an enzyme that is found only in prokaryotes, leading to the conclusion that this represents an example of divergent evolution (Hennebry, 2009
). The sequence identity shared between this hydrolase and Sm
Lip is only 6%, but the similarity in fold is evident (data not shown). We carried out further comparisons seeking to inform on Lip function.
Transthyretin binds the hormone thyroxine, self-interacts to form a tetramer and also forms a complex with retinol-binding protein (Blake et al.
; Wojtczak et al.
; Monaco et al.
; Zanotti et al.
). In common with transthyretin, Sm
Lip forms a tetrameric assembly. However, the Sm
Lip oligomer is distinct and an overlay of one Sm
Lip polypeptide with a subunit from transthyretin does not produce an overlap of any of the other polypeptides (data not shown).
Transthyretin forms a dimer by antiparallel self-association of the β6 and β8 strands, creating a curved eight-stranded β-sheet (Blake et al.
). The binding of the hormone thyroxine occurs at the tetramer interface created by the convex surfaces of two of these eight-stranded β-sheets as the protein assembles as a dimer of dimers. The thyroxine-binding residues in transthyretin are not conserved in Sm
Lip and an overlay of an Sm
Lip polypeptide and transthyretin subunit places the ligand-binding site on the surface of the former (Supplementary Fig. S31
). It is unlikely that Sm
Lip acts to bind hydrophobic ligands of the type that transthyretin can bind.
Transthyretin associates with retinol-binding protein using residues in three turns: two from one subunit that link β1 to β2 and β4 to β5, and one from another subunit that links β1 to β2 (Monaco et al.
). These parts of the transthyretin structure correspond to loops 1 and 3 of Sm
Lip. Loop 1 is directed out from the globular fold into the periplasmic space; it is placed to interact with physiological partners and may represent a binding site for other proteins/molecules.
In a recent study of Klebsiella pneumoniae
5-hydroxyisourate hydrolase, the residues important for catalytic function were confirmed as His7, Arg41, His92 and Ser108, which together with Tyr105 form a polar and symmetric active site at a dimer interface (French & Ealick, 2011
). A structure-based sequence alignment matches four of these catalytic residues (with the exception being Ser108) to Asp42, Gly105, His161 and Val172, respectively, in Sm
Lip. The polypeptides do not overlay in the vicinity of Ser108 (data not shown) and it is unlikely that Lip has any hydrolase activity.
The biological role of SmLip or its orthologues in the T6SS has yet to be unambiguously defined. Structural comparisons appear to rule out, rather than assign, a function. The propensity to self-associate using parts of the SmLip structure that will be exposed in the periplasm suggests that this protein, exploiting the lipid anchor, helps to bind and position different components of the secretion apparatus at the outer membrane. Future experiments, aided by the structural model, can address this hypothesis.