A coiled-coil domain acts as a molecular ruler to regulate O-antigen chain length in lipopolysaccharide.

Long-chain bacterial polysaccharides have important roles in pathogenicity. In Escherichia coli O9a, a model for ABC transporter-dependent polysaccharide assembly, a large extracellular carbohydrate with a narrow size distribution is polymerized from monosaccharides by a complex of two proteins, WbdA (polymerase) and WbdD (terminating protein). Combining crystallography and small-angle X-ray scattering, we found that the C-terminal domain of WbdD contains an extended coiled-coil that physically separates WbdA from the catalytic domain of WbdD. The effects of insertions and deletions in the coiled-coil region were analyzed in vivo, revealing that polymer size is controlled by varying the length of the coiled-coil domain. Thus, the coiled-coil domain of WbdD functions as a molecular ruler that, along with WbdA:WbdD stoichiometry, controls the chain length of a model bacterial polysaccharide.

Introduction

Lipopolysaccharide (LPS) is a major constituent of the outer membrane of Gram-negative bacteria. In most cases, its presence is essential for cell viability and in addition LPS plays important roles in pathogenicity. LPS is composed of a conserved membrane anchor molecule, lipid A (endotoxin), and a short conserved core oligosaccharide that links lipid A to an immunogenic O-antigen polysaccharide of variable length [1,2]. The carbohydrate composition and size of the O antigen is hyper-variable with more than 180 variants described in E. coli strains alone [3]. The O-antigen is among the first molecules encountered by the host during infection and it has been shown to be vital in protecting pathogenic bacteria from the host-immune response by avoiding complement mediated killing [4]. A fascinating structural feature of most O antigens is that their chain lengths fall within defined ranges (termed the ‘modal distribution’) that are O-serotype dependent (Figure 1A). Such exquisite control of polymerization is a recurring theme in biology (e.g. phage tails or the length of injection needles in type-III-secretion-systems[5,6]) but its origin is generally poorly understood at a structural and molecular level, even where the identities of the protein components from the assembly systems are known.

Figure 1

A) Silver stained SDS-PAGE of wild type E. coli O9 LPS, illustrating the modal size distribution with an average length of 14 repeating units (RU). The data are reproduced from a previous study[16]. B) Schematic view of the biosynthesis of O9a polymer. The polymer contains a repeating tetrasaccharide (RU) and is built as an undecaprenol diphosphate (Und-PP)-linked intermediate by 3 mannosyltransferases, including the polymerizing enzyme WbdA. The N-terminal domain of WbdA catalyzes formation of the α-(1→2) linkage and the C-terminal domain is predicted to form the α-(1→3) linkages. Polymerisation is terminated by WbdD which caps the polymer by phosphorylating then methylating the terminal mannose and the resulting molecule is exported by the ABC transporter before ligation to lipid A-core.

There are three recognized bacterial polysaccharide biosynthesis pathways that produce O-antigen glycans (the Wzy dependent-, the ATP-binding cassette (ABC) transporter dependent- and the synthase dependent pathway) [1]. The components that determine the modal distributions of O-antigen glycans vary in these biosynthesis pathways but none of the processes are fully understood at a molecular level. This study focuses the assembly of the serotype O9a antigen from on E. coli, a well-characterized prototype of the ABC transporter-dependent pathway.

The O9a glycan chain is synthesized in the cytoplasm using an undecaprenyl diphosphate carrier molecule. The chain is extended by addition of mannose residues to the non-reducing terminus and glycan chain length is terminated by the addition of a phosphomethyl moiety that blocks further chain extension [7] (Figure 1B). Elongation of the polymannose O9a-antigen is performed by WbdA, whose two glycosyltransferase (GT) domains cooperate in a distributive reaction mechanism to produce a glycan composed of alternating pairs of α-(1→2)- and α-(1→3)-linked mannoses in a tetrasaccharide repeat unit [8,9] (Figure 1B). The N-terminal domain of WbdA is predicted to catalyse formation of the α-(1→2) linkage and the C-terminal domain the α-(1→3) linkage; both domains belong to the glycosyl transferase family 4 [9]. The overall structure of WbdA and the positioning of the two catalytic sites relative to one another have not been reported. Chain termination is catalyzed by WbdD, (a kinase and methyltranferase) that adds a terminal phosphomethyl moiety at the O3 position of the terminal mannose residue halting further chain extension [10-12] (Figure 1B). The terminating phosphomethyl modification is recognized by a specific carbohydrate binding module forming part of the ABC transporter, giving an elegant quality control method to ensure only (completed) glycans of the desired length are exported and assembled into LPS [13,14]. WbdD is a membrane-associated protein and its interaction with WbdA is essential for the ability of the otherwise soluble polymerase to act on the membrane-embedded undecaprenyl lipid-linked acceptor [11]. Full-length WbdD (708 residues) is membrane bound and forms aggregates, which has precluded in vitro studies of the complex. Only WbdD constructs truncated at residue 556 (WbdD1-556) or 459(WbdD1-459), thus lacking both WbdA and membrane interacting regions, have been crystallized[15]. The structure of WbdD1-556 revealed a trimer, in which each monomer is comprized of a methyltransferase domain, followed by kinase domain followed by a 12 residue helix; residues beyond 470 were disorderd and thus not seen in the electron density[15]. The short 12 residue helix forms a trimeric bundle with the same helix in two other monomers around a three fold axis, giving the overall structure the appearance of an umbrella[15]. WbdD1-459 lacks the helix and exists as a monomer.

Mathematical modelling has been used in an effort to understand the molecular basis for the observed modality of the O9a glycan [16]. This resulted in a “variable geometry model”, in which the stoichiometry of the WbdA:WbdD complex is a crucial factor regulating modal length and size distribution of the polymer. Perturbing expression of WbdA or WbdD to alter the stoichiometry altered chain lengths and distribution in a manner predicted by the model [16]. However, in addition to complex stoichiometry, the mathematical model requires a measuring element (hereafter referred to as a molecular ruler). In this study, we set to define the molecular ruler. We demonstrate that the C-terminus of WbdD forms an extended coiled-coil and that this structural feature acts as a central element of the required molecular ruler that together with complex stoichiometry defines the size of the O9a glycan. This is the first complete molecular and structural description of such a polymerisation system.

Results

Extended X-ray structure of the C-terminal helix-bundle WbdD1-556

As we had reported previously, WbdD truncated beyond residue 459 (WbdD1-459) gave a monomeric protein that crystallized well but possessed severely reduced kinase activity. The latter was unexpected, since both catalytic domains were left intact [17]. A second truncated construct, WbdD1-556, produced active trimeric protein that gave cubic crystals which were refractory to optimization and diffracted to resolutions between 4 and 8 Å. We reported that a dehydration of these crystals gave a 2.2 Å resolution structure with a monomer in the asymmetric unit. In this crystal structure only a poorly defined C-terminal poly-alanine helix (that forms a helical bundle) comprising residues 459 to 473 (approximately) was built, precluding analysis of the C-terminal domain[15].

For this work, we evaluated 10 datasets obtained from non-dehydrated crystals on the hypothesis[15] that the dehydration itself was responsible for disordering of the C-terminus. One WbdD1-556 crystal diffracted to 3.9 Å and the C-terminal domain up to residue 505 was clearly visible in the electron density, following deformable elastic network (DEN) refinement [18]. Only 51 residues remained disordered in the resulting structure (Table 1). The newly traced residues adopted a helical arrangement generating an extended C-terminal helix to which sequence was assigned (Figure 2A, B). The extended helix forms a coiled-coil arrangement via the crystallographic three-fold axis. Furthermore, this coiled-coil domain interacts with the kinase active site, resulting in the ordering of two previously disordered active site loops (αQ-αR, residues 398-416 and α16-βQ, residues 377-381, the “activation” loop). These two loops interact with each other via a π-stacking interaction between W408 and W382 (Figure 2B). We had previously obtained weak but interpretable density for two mannose residues at the kinase active site[15], this newly ordered structure completes the kinase catalytic site (Figure 2B), supplies residues to recognize the substrate (consistent with site directed mutagenesis)[15], and rationalizes the role of the C-terminus in kinase activity[15].

Table 1

Data collection and refinement statistics

	WbdD1-556, PDB-ID: 4UW0
Data collection [1]
Space group	I23
Cell dimensions
a, b, c (Å)	181.3, 181.3, 181.3
α, β, γ (°)	90.0, 90.0, 90.0
Resolution (Å)	128.2-3.87 (4.0-3.87) *
R merge	0.07 (0.59)
I / σI	11.6 (2.3)
Completeness (%)	98.6 (99.3)
Redundancy	4.0 (3.7)
Refinement
Resolution (Å)	128.2-3.87 (4.0-3.87)
No. reflections	8823
Rwork / Rfree	0.27 / 0.34
No. atoms
Protein	4047
Ligand/ion	27
Water	0
B-factors (Å2)
Protein	171.5
Ligand/ion	126.8
Water	n/a
r.m.s. deviations
Bond lengths (Å)	0.007
Bond angles (°)	1.301

Values in parentheses are for highest-resolution shell.

The reported data were collected from a single crystal.

Figure 2

A) 3.9 Å X-ray structure of WbdD1-556, the trimer is shown in cartoon representation. One monomer is colored with a gradient running from blue (N-terminus) to pink (C-terminus). Spheres represent the nucleotides in the catalytic domains. The trimer axis is drawn as a red line. Arrows mark the start- and end-point of the coiled-coil. B) Interaction of the coiled-coil domain with the active site of WbdD (stereo pair). The color scheme is identical to C. Selected residues are numbered and shown as sticks. The cofactor ATP is shown as ball-and-stick model. The mannose (purple) is from a superimposed model of the WbdD1-459 structure [15].

The structure of the C-terminal coiled-coil domain in WbdD

Small-angle X-ray scattering (SAXS) data were collected for the WbdD1-459, WbdD1-556 and WbdD1-600 constructs (Figure 3A). WbdD1-600 behaved as an aggregated assembly in gel filtration and its SAXS data were uninterpretable. The scattering from WbdD1-459 yielded a molecular mass (MM) estimate of 52±5 kDa (Table 2) in agreement with the calculated molecular mass (MM) of the monomer (49kDa). The reconstructed ab initio model of WbdD1-459 fits the SAXS data with discrepancy χ = 0.9 and matches well with the crystallographic monomer[15] (Supplementary Figure 1).

Figure 3

A) Fits of different structural models (panels B,C) against the experimental WbdD1-556 SAXS data (open circles with error bars representing standard deviations computed from propagated Poisson counting statistics) with the residuals of the fits shown below. B) Cartoon model of the new low-resolution structure with extended coiled-coil (same as in Figure 2A). C) Ab initio SAXS envelope of WbdD1-556 (cyan) superposed with a model of WbdD 1-556 (purple) that was built as described in the main text and refined against the SAXS data.

Table 2

SAXS Data Collection and Scattering Derived Parameters.

Data collection parameters	WbdD1-459	WbdD1-556
Instrument	X33 (DORIS)	X33 (DORIS)
Beam geometry
Wavelength (Å)	1.5	1.5
q-range (Å−1)	0.01-0.60	0.01-0.60
Exposure time (s)	15	15
Concentration range (mg ml−1)	1 - 10	1 – 5
Temperature (K)	283	283
Structural parameters *
I(0) (arbitrary units) (from P(r))	47.5 ± 0.5	144.9 ± 0.5
Rg (Å) (from P(r))	32 ± 3	53 ± 5
I(0) (arbitrary units) (from Guinier)	47.4 ± 0.5	144.0 ± 0.5
Rg (Å) (from Guinier)	31 ± 3	52 ± 5
Dmax (Å)	100 ± 5	170 ± 10
Porod volume (103 Å3)	90 ± 10	380 ± 40
Dry volume calculated from sequence (103Å3)	59.2	76.6
Molecular mass determination *
MMPOROD (from Porod volume, kDa)	53 ± 5	220 ± 30
Contrast (Δρ × 1010 cm−2)	3.047	3.047
MMsaxs (from I(0), kDa)	52 ± 5	165±20
Calculated monomeric MM from sequence (kDa)	48.9	63.3

Reported for infinite dilution of concentration series measurements

The MM estimate of WbdD1-556 from SAXS (MM = 165±20 kDa, Table 2) indicates a trimeric assembly of this construct, in agreement with gel filtration and crystallography[15]. The ab initio WbdD1-556 model (imposing P3 symmetry) reveals a mushroom-like particle that resembles the crystal structure (Figure 3B,C). The trimeric arrangement of the catalytic domains of WbdD1-556 corresponds to the head of the mushroom and the coiled-coil partly matches its stalk. In the ab initio model, the stalk is clearly longer than that in the crystal structure (Figures 3B, C), indicating that the coiled-coil extends further than is currently resolved by crystallography. The theoretical scattering, computed from the crystallographic model at 3.9 Å where residues from 506 to 556 are missing, provides a poor fit to the SAXS data from WbdD1-556 with χ = 4.8 (Figure 3A, green curve).

Sequence analysis using the COILS algorithm [19] predicts that residues 460-600 of WbdD (and 460-556 of WbdD1-556) form a coiled-coil structure (Supplementary Figure 2). When the new structure (Figure 3B) was analysed by the CCCP server [20], residues 462-505 gave an average coiled-coil radius of 8.5 Å with a 1.42 Å rise-per-residue of the coiled-coil axis (Supplementary Table 1, TWISTER [21] an alternative program gave similar values, radius: 8.4 Å, rise per residue: 1.42 Å/residue). Using these experimental parameters, we built a model of the complete coiled-coil of WbdD1-556 using the CCBuilder server [22], leading to a coiled-coil of ~135 Å in total length. The hybrid structure (X-ray structure of N-terminal domains plus model of coiled-coil) has an improved fit to the SAXS scattering (χ = 3.3 vs 4.8) compared to just the residues of the X-ray model. Rigid body refinement with SASREF was employed and the angle of catalytic domains relative to coiled-coil was allowed to move [23]. Precedent for such movement was seen in the various crystal structures[15,17] which have different angles with a more closed cap seen at higher resolution consistent with tighter packing. The angle between the catalytic domains and the coiled-coil in the SASREF model opened by ~5° resulting in a χ value of 1.29 (Figure 3A). Figure 3C shows that this refined rigid-body molecular model of WbdD1-556 fits very well the shape of the ab initio SAXS model (which is bulkier as the solvation shell is visible in SAXS). Circular dichroism analysis using the CONTINLL algorithm[24] (via Dichroweb [25]) on spectra collected on B23 at Diamond of WbdD1-459 and WbdD1-556 confirms a detectable increase in helical content as expected for the addition of a coiled-coil (Supplementary Figure S3). The good fit between experimental and model structures encouraged us to extrapolate our model to include the additional 44 residues (predicted to be coiled-coil). By this approach the coiled-coil of full-length WbdD is predicted to be around 200 Å.

Insertions and deletions in the coiled-coil affect polysaccharide length

The length of the coiled-coil was experimentally modulated by insertion or deletion of residues within this structure. To minimize stoichiometry changes in the WbdA:WbdD complex[16] resulting from mutations in WbdD perturbing its expression level, an experimental system was used in which modified His6-WbdD proteins were expressed from the same transcript (and promoter) as WbdA-Flag in an E. coli wbdD wbdA deletion mutant (CWG917). This experimental test system gives a polymer with a modal length of 11 repeat units; slightly shorter than the 14 observed in the native O9a product. Four consecutive heptads were deleted within the coiled-coil region of WbdD (Fig 3A, Δ(GHIJ)), amino acids 558-585). Expression of the Δ(GHIJ) derivative in CWG917 resulted in LPS with a shorter average O-PS chain length compared to the modal length obtained with the wild-type His6-WbdD (Fig 3B and C). Deletion of an additional heptad (Fig 3A, Δ(GHIJK), amino acids 558-592) caused a further reduction by one repeat unit in the O-PS (Fig 3, B and C). A His6-WbdD derivative containing a coiled-coil domain extended by insertion of two additional blocks of heptads CDEF (i.e. amino acids 466-507) designated (CDEF)2 (Fig 3A) produced a polymer with a longer average O-PS chain length (Fig 3 B and C). A longer coiled-coil insertion (ABCDEF)2 was also constructed, the O-PS synthesized was longer than the native product but not within error longer than that observed from the (CDEF)2 derivative, indicating there is a limit to the extension possible. Since full length protein was refractory to structural analysis, we made two insertion and two deletion variants of WbdD1-556 in order to gain some structural insight into the effect of the changes in the coiled-coil region. CD spectroscopy (Supplementary Figure S3) was used to characterize constructs WbdD1-459, WbdD1-556, WbdD1-556(CDEF)2, WbdD1-556(ABCDEF)2, WbdD1-556(ΔGHIJK) and WbdD1-556(ΔGHIJ). Although the presence of aggregates (which hinders precise concentration estimation) means we treat the numbers with caution, we do observe a correlation between predicted and observed changes in helical content (Supplementary Figure S3). SAXS data on these constructs also showed that solutions of these proteins contain aggregates that render detailed analysis impossible (as seen for WbdD1-600).

Previous work has shown that the chain-length of the O9a antigen can be manipulated by altering the levels of expression of either WbdA or WbdD [16]. To determine whether altered stoichiometry could explain the results with the coiled-coil variants, the relative amounts of His6-WbdD and WbdA-Flag were quantitated from the whole-cell lysates by Western immunoblotting using anti-His and anti-Flag antibodies [16]. The WbdD:WbdA ratios for the strains containing His6-WbdD(ΔGHIJK) and (ΔGHIJ) were approximately 1.3 and 1.7-fold lower, respectively than the ratio observed for the wild type His6-WbdD. However, these changes in stoichiometry would tend to generate a longer polymer, rather than the observed shorter products [16]. The WbdD(CDEF)2:WbdA ratio was 1.5-fold higher than that that observed with wild type His6-WbdD which would tend to yield a shorter polymer, opposite to the observed products [16]. Thus, in each of these cases, the change in stoichiometry would operate counter to the experimental observation and we can conclude that changes in polymer length are due to the change in the protein structure, rather than altered stoichiometry. However, the ratio of WbdD(ABCDEF)2:WbdA was 2.5-fold lower than that of the wild type and thus possible that the increase in O-PS chain length in this mutant was due to decreased expression of WbdD(ABCDEF)2 relative to WbdA. To determine whether a 2.5-fold decrease in WbdD expression could affect the chain length to the extent observed for WbdD(ABCDEF)2, the expression level of plasmid-encoded wild type His6-WbdD was titrated relative to constitutively expressed endogenous WbdA. The LPS chain-length was examined by SDS-PAGE and the His6-WbdD levels were quantitated by Western immunoblotting. From this analysis it was determined that a 2.5-fold decrease in WbdD(ABCDEF)2 relative to WbdA was insufficient to cause the observed O-PS chain length increase (Supplementary Figure S4). As a final control for mutant function, we confirmed that dramatic changes in stoichiometry in complexes containing the WbdD constructs with altered coiled-coils reproduced the size-distribution shifts (for one insertion and one deletion) observed for the native protein [16] (Supplementary Figure S5).

Discussion

The concept that biology can control the length of polymeric structures by means of a ‘molecular ruler’ dates back to the observation that the lengths of bacteriophage tail structures are narrowly distributed [26]. Since then, molecular rulers have been identified in other polymerization systems [27]. In principle, a molecular ruler is a protein, or a domain of a protein-protein complex, whose size defines the length of a polymeric structure by terminating polymerisation when the nascent polymer reaches the requisite length [27]. Well studied examples include the phage-λ tail system, which involves two proteins gpH (the ruler) and gpU (the terminator)[5], flagellar hooks in Salmonella [28,29] using FliK and YscP and type-three secretion system needles in Yersinia [6,30-32]. Although these systems have been thoroughly characterized at a genetic level, detailed molecular descriptions of the ruler have been lacking. Polymerisation control but of a different sort is also seen in the Cas10•Csn ribonucleoprotein complex from the CRISPR system[33], where multiple copies of Csm3 protein bind the nascent oligonucleotide and each additional copy of Csm3 results in a 6-nucleotide increment.

An explicit factor in the mathematical model [16] developed to explain chain length regulation in E. coli O9a LPS system is the presence of a molecular ruler separating WbdA and WbdD. The rigid coiled-coil domain of WbdD is the obvious candidate for this ruler. The coiled-coil is at the C-terminus and effectively separates the kinase (capping reaction) from WbdA (the polymerase), which interacts with the C-terminus of WbdD [12]. WbdA has two active sites at which the undecaprenol diphosphate anchored polymer is elongated by addition of carbohydrate monomers. Coiled-coils have been shown to act as rigid spacers in other non-polymerization based systems: e.g. tetherin [34] or type-I restriction modification enzymes [35,36]. The 92 kDa gpH ruler protein from the phage-λ ruler system [5] has a predicted coiled-coil [19], although the coiled-coil has not been experimentally investigated.

The GLYCAM server (www.glycam.org) calculates the length of a repeat unit of the O9a E. coli carbohydrate polymer to be ~15 Å, meaning 14 repeating units will be around 210 Å (Figure 4A). This is close to the 200 Å length of the coiled-coil in WbdD calculated from our crystallographic and modelling data. We suggest that the polymer grows at WbdA by switching between the two mannosyltransferase active sites in this two-domain protein [9] until long enough to reach and be capped by the kinase domain of WbdD (which is held by the coiled-coil domain at around 200 Å from the WbdA) (Figure 5). Direct experimental assessment of a reconstituted WbdD: WbdA complex has not been possible as purified full length WbdD aggregates. Instead we tested our hypothesis by engineering WbdD with coiled-coil domains with shorter and longer lengths. We deleted 28 and 35 residues (multiples of 7 were used to conserve the heptad register), corresponding to 40 Å (2 repeat units) and 50 Å (3 repeat units) respectively (Figure 4A). In vivo data shows the resulting polymers produced in these mutants are indeed progressively shorter (Figure 4 B,C). Insertions of 56 residues (80 Å increase; five repeats) and 84 residues (120 Å increase; nine repeats) (Figure 4A) both give O-polysaccharides with chain lengths longer than that of the wild type. Taken together the in vitro data establish the direct relationship between coiled-coil domain length and LPS polymer length required for a molecular ruler (Figure 4B,C, 5, S5). The close correlation between a physical length and sequence of coiled-coil structures may make them particularly suited as molecular rulers for biological systems more generally.

Figure 4

A) Comparison of O9a LPS size distribution with lengths of the coiled-coil domains of different WbdD constructs in the experimental system. B) Silver stained SDS-PAGE of the LPS from WbdD wild type and the coiled-coil domain mutants shown in A). C) Intensity profiles of the gels shown in B). The profiles were calculated with ImageJ[42]. The maximum intensity band in each trace is marked by an asterisk.

Figure 5

Proposed model for elongation and termination of the O9a antigen. The membrane-attached polymer is extended (shown as translucent spheres, coloring as in Fig. 1B) by 3 copies of the multidomain WbdA polymerase (green ring). WbdA forms a complex with full length trimeric WbdD (which is anchored to the membrane, purple); note that the stoichiometry of the complex also influences the chain length[16]. As the polymer grows in length it will become long enough to reach the kinase active site of WbdD where polymerisation is terminated. The coiled-coil domain of WbdD separates the kinase from WbdA in space and thus is the essential element in the molecular ruler that regulates chain length of O-PS.

The bifunctional kinase-methyltransferase WbdD component is identical in E. coli serotypes O9 and O9a, the difference in O antigens arises to due to changes in linkage geometry catalysed by the differing WbdAs[37]. O antigens and biosynthesis systems similar to E. coli O9 are also found in Klebsiella pneumoniae [37] and Hafnia alvei [38]. A polymannose O antigen with a trisaccharide repeat unit is produced by E. coli serotype O8 (identical to K. pneumoniae O5). It possesses a WbdA protein with 3 mannosyltransferase domains [9] and the cognate WbdD protein only possesses the methyltransferase activity but the coiled-coil domain is conserved (Supplementary Figure 6) [10]. Consequently we suggest a similar ruler mechanism is operative in each of these polymannose O antigen systems. An ABC transporter-dependent process is also used by Geobacillus stearothermophilus, where the S-layer glycoprotein carries O-linked poly-D-rhamnose glycans capped with methyl groups[39]. Interestingly the organization of the system differs, as there is no separate terminating methyltransferase protein. Instead the methyltransferase is part of a multidomain protein that includes one of the glycosyltransferases required for chain extension. Importantly, the methyltransferase and glycosyltransferase domains are separated by a region of predicted coiled-coil structure (Supplementary Figure S6) and we predict this region may function as the ruler. An analogous situation may occur in biosynthesis of the O12 antigen in K. pneumoniae O12 and a related O-antigen structure in Raoutella terrigena [40,41], where terminal capping is achieved by addition of a β-linked 3-deoxy-D-manno-oct-2-ulosonic acid (Kdo) residue. In this multidomain protein, the putative Kdo transferase domain is separated from two chain extension glycosyltransferase domains by a region of coiled-coil structure (Supplementary Figure S6).

To our knowledge, the coiled-coil domain of WbdD represents the first molecular ruler characterized at a molecular level and we have shown that the ruler can be engineered by genetic modification to produce both longer and shorter carbohydrate polymers. The ease of the genetic manipulation that can be used to tune the property of polymer may be another positive facet of coiled-coils as molecular rulers in nature but maybe also for chemical applications.

Online Methods

Protein expression & production

The procedures used for the cloning, expression and purification of WbdD1-600, WbdD1-556 and WbdD1-459 have been described previously[15,17].

Crystallography

WbdD1-556 was crystallized as previously reported[17]. Before flash cooling in liquid nitrogen, crystals were cryo-protected by transfer into a solution consisting of 1 μl ADP (5 mM), 0.5 μl magnesium chloride (100 mM), 3 μl glycerol and 6 μl mother liquor (0.13 M lithium sulfate; 0.10 M Tris-HCl pH 8.52; 1.18 M ammonium sulfate). A 3.9 Å diffraction data set was collected (λ = 0.9173 Å) using synchrotron beamline IO4-1 at the Diamond synchrotron (Didcot, UK). The data set was processed using XIA2 [43] and the structure was solved by molecular replacement (PHASER, [44]), using the WbdD1-459 structure as search model (PDB-ID: 4AZW) [15]. The program gave a clear solution with a translation function Z-score of 30.9. Using deformable elastic network refinement [45] and jellybody refinement as implemented in REFMAC [46], the molecular replacement solution could be refined to R/Rfree-factors of 26.8/34.1 and a model for the C-terminal helix could be build up to residue 498. The geometry of the model was analysed with MOLPROBITY [47] and was placed overall in the 91st centile (Molprobity Score: 2.90/ 91st percentile; Molprobity Clashscore: 30.1/ 83rd percentile). Ramachandran statistics were also calculated with MOLPROBITY (favoured: 83.0 %, outliers: 3.6 %). The data collection and refinement statistics are listed in Table 1, the coordinates are available from the RCSB with code 4uw0.

SAXS

SAXS measurements were carried out at beamline X33 (EMBL-DESY, Hamburg) at the DORIS III storage ring using a Pilatus 1M detector (Dectris) [48]. Purified WbdD constructs (1-459 (C380S mutant), 1-556 and 1-600) were measured at four different concentrations (1, 2, 5 and 10 mg/ml) in 20 mM BisTris pH 7, 50 mM NaCl, 5 mM DTT. For each measurement eight 15 s frames were collected and averaged. No substantial changes of the scattering intensity were detected during the successive exposures indicating the absence of radiation damage. At the sample-detector distance of 2.7 m and a wavelength of λ = 1.5 Å the momentum transfer range of 0.01 < s < 0.6 Å−1 was covered (s = 4π sinθ/λ, where 2θ is the scattering angle). The program PRIMUS [49]. was used to correct the data for buffer contribution, scale for solute concentration, and extrapolate the data to infinite solute dilution [49]. The statistics of the SAXS data collection are presented in Table 2. Data were reduced by automated radial averaging [50]. The radius of gyration Rg and the forward scattering intensity I(0), were determined using Guinier analysis [51] and the indirect Fourier transformation using GNOM [52]. The latter program was employed to evaluate the maximum particle dimension D and the pair-distance distribution function p(r). The molecular mass (MM) of the protein constructs was calculated by comparison of the extrapolated forward scattering with that from a reference bovine serum albumin (BSA) sample (MM = 66 kDa). The excluded volume of the hydrated protein V was obtained with DATPOROD [53] providing an independent estimate of the molecular mass (MM). For globular proteins, hydrated protein volumes in Å3 are about 1.7 times the molecular masses in Da.

SAXS modelling

Theoretical scattering profiles from the high-resolution models were calculated by CRYSOL [54]. Ab initio models of the WbdD1-459 mutant and WbdD1-556 constructs were obtained from the scattering data using the bead-modeling program DAMMIN [55]. An average of 10 independent reconstructions was used to generate a representative model by SUPCOMB [56] and DAMAVER [57]. Rigid-body modeling against the SAXS data was done using SASREF [23] to search for the optimum configuration of the N-terminal and C-terminal domains to minimize the discrepancy χ between the experimental and computed SAXS data. The complete coiled-coil of WbdD1-556 built using CCBuilder server [22] was modelled as a single entity.

CD measurements

CD experiments were performed using a nitrogen-flushed Module B end-station spectrophotometer at B23 Synchrotron Radiation CD Beamline at the Diamond Light Source, Oxfordshire, UK [58,5960]. Samples were typically prepared in 30 mM NaPi and 50mM NaF, pH 7.0. Measurements were carried out at 20 °C with sample concentration of 0.3mg/ml with 0.02cm pathlength cell using average amino acid molecular weight of 113. The data was processed using B23 CDApps (http://www.diamond.ac.uk/Beamlines/Soft-Condensed-Matter/B23/manual/Beamline-software.html).

Site-directed mutagenesis of WbdD and in vivo O-PS chain length analysis

The coiled-coil domain insertion ((CDEF)2, (ABCDEF)2) and deletion (Δ(GHIJK), Δ(GHIJ)) mutants were constructed by PCR based mutagenesis [61] using the primers described in Supplementary Table 2 with pWQ470 (His6-WbdD) [11] as the template. To generate mutant plasmids encoding His6-WbdDΔ(GHIJK) and Δ(GHIJ) two different forward primers (DELGJSPEF and DELGKSPEF, respectively), were used together with the same reverse primer (DELGKCOMR). Plasmids encoding the insertion mutants His6-WbdD(CDEF)2 and His6-WbdD(ABCDEF)2 were generated using the primer pairs, INS2XEFR/INS2XEFF, INSCDEFR/INSCDEFF and INSATOFR/ INSATOFF, respectively. The resulting mutants were sequenced and protein expression of these mutants was verified by western immunoblotting.

Wild-type and mutant His6-WbdD constructs were cloned immediately upstream of a derivative of the wbdA gene (coding for a WbdA-Flag fusion protein) creating the same organization that occurs in the chromosomal locus and were expressed under the control of the L-arabinose-inducible PBAD promoter. The His6-WbdD and WbdA-Flag fusion proteins were expressed contiguously to provide consistent stoichiometry of the overexpressed proteins across individual experiments. Pwo DNA polymerase was used to amplify the wbdA gene from genomic E. coli O9a DNA from strain CWG28 [62], using the primers, WbdABgFw and WbdARv. The WbdARv primer contained sequence coding for a C-terminal Flag epitope tag. EcoRI and BglII restriction endonuclease sites (underlined) were incorporated into the forward primer and used with the blunt end formed at the 3′ end of the PCR product for subsequent ligation reactions. The PCR product was digested with EcoRI and then ligated into EcoRI-SmaI-digested pWQ589 [8] to form pWQ457. Genes encoding the His6-WbdD derivatives were amplified by PCR using primers, WBBDDHF1 and WbdDBg708. The amplification product was digested with EcoRI-BglII and ligated into EcoRI-BglII-cut pWQ457 to give, pWQ458 (wildtype His-WbdD), pWQ468 (Δ(GHIJK)), pWQ459 (Δ(GHIJ)), pWQ460 ((CDEF)2), and pWQ469 ((ABCDEF)2), respectively (Supplementary Table 2). To overexpress His6-WbdD and the mutant derivatives together with WbdA-Flag, an E. coli O9a wbdD wbdA deletion mutant (CWG917)[11] was transformed with the relevant plasmid and grown for 16 h at 37° C in 5 mL of LB broth containing kanamycin (Km, 50 μg/mL) and glucose (0.4% w/v). The cultures were then diluted 1:100 into 5 mL of LB broth containing Km (50 μg/mL), D-mannose (0.1% w/v), and L-arabinose (0.01% w/v) and grown at 37° C until A600nm values of 0.4-0.7 were reached. Three biological replicates were grown for each transformed strain. A volume equivalent to 1 A600nm unit of each culture was collected by centrifugation at 12,000 × g for 1 min. Cells were lysed in 0.1 mL of SDS-PAGE loading buffer. LPS was analysed by treating one half of each sample with proteinase K [63] followed by SDS-PAGE in Tris-glycine buffer (Laemmli) and visualization by silver staining [64].

Proteins were analysed by Western immunoblotting using the remaining half of the lysed samples. Following SDS-PAGE, proteins were transferred to nitrocellulose membranes (Protran, PerkinElmer). Duplicate membranes were incubated with either anti-Penta-His antibody (Qiagen cat no 34660; used at a 1:1000 dilution)), or with anti-Flag M2 antibody (Sigma cat no F1804; used at a 1:1000 dilution) for detection of His6-WbdD derivatives and WbdA-Flag, respectively. Horseradish peroxidase-conjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories cat no 115-036-003; used at a 1:3000 dilution) was used together with Luminata Classico Western HRP Substrate (Millipore) for protein detection. Validation and specificity information for the antibodies is found in the manufacturers literature. Immunoblot data was collected with a BioRad Chemidoc XRS system and Image processing was performed with ImageJ [42] to quantify the proteins by densitometry. For each mutant derivative, the mean pixel density ratio derived from the anti-His and the anti-Flag signals, respectively, was compared to the mean ratio derived from strains expressing the wildtype WbdD. To control for differences due to Western blot exposure times, wildtype and mutant signals were only compared among samples on the same blot.