In Situ Structure of an Intact Lipopolysaccharide-Bound Bacterial Surface Layer

Abstract

Most bacterial and all archaeal cells are encapsulated by a paracrystalline, protective, and cell-shape-determining proteinaceous surface layer (S-layer). On Gram-negative bacteria, S-layers are anchored to cells via lipopolysaccharide. Here, we report an electron cryomicroscopy structure of the Caulobacter crescentus S-layer bound to the O-antigen of lipopolysaccharide. Using native mass spectrometry and molecular dynamics simulations, we deduce the length of the O-antigen on cells and show how lipopolysaccharide binding and S-layer assembly is regulated by calcium. Finally, we present a near-atomic resolution in situ structure of the complete S-layer using cellular electron cryotomography, showing S-layer arrangement at the tip of the O-antigen. A complete atomic structure of the S-layer shows the power of cellular tomography for in situ structural biology and sheds light on a very abundant class of self-assembling molecules with important roles in prokaryotic physiology with marked potential for synthetic biology and surface-display applications.

Keywords: LPS; S-layer; bacteria; cryo-EM; cryo-ET; in situ structural biology; lipopolysaccharide; sub-tomogram averaging; surface layer; tomography.

Graphical abstract

Figure S1

Biochemical Reconstitution of the RsaA_NTD:PS Complex, Related to Figure 1 (A) Cryo-EM image of purified monomeric RsaA_NTD. (B) Image of reconstituted oligomeric RsaA_NTD:crude LPS aggregate complex. (C) Cryo-EM image of purified RsaA_NTD:PS complex. (D) Gel-filtration profiles of monomeric RsaA_NTD (red), RsaA_NTD + crude LPS (green) and RsaA_NTD + PS (blue) corresponding to images in (A–C). (E) SDS-PAGE of purified crude LPS stained with Pro-Q Emerald 300 (yellow) overlaid with the same gel stained with Coomassie brilliant blue G-250 (blue). (F) SDS-PAGE analysis of purified and mass-spectrometry verified RsaA_NTD:PS sample stained with Coomassie brilliant blue G-250 (black).

Figure 1

Cryo-EM Reconstruction of the RsaA_NTD:PS Complex at 3.7 Å Resolution (A) Cryo-EM image of the purified complex. Inset: class averages with the spiral-like nature of the complex highlighted (see Figure S1). (B) Density map of the complex (contour level on the lower left of panel). Different subunits are shown in different shades of gray, and density corresponding to the O-antigen of the LPS is orange (see Video S1). (C) Regions of the map along with the built atomic model showing resolved secondary structure elements and side-chain fits. Due to the α-helical nature of the RsaA_NTD, the fit of the model to the cryo-EM map is exceptional (see Figure S2). (D) The refined atomic model of a single RsaA_NTD subunit from the complex shown as a ribbon diagram. A stick representation of the main chain of the O-antigen is shown within the cryo-EM density. O-antigen chain is continuous along the spiral, denoted by asterisks (^∗). (E) Surface representation of a single RsaA_NTD subunit showing O-antigen binding residues in magenta. (F) Close up of two Ca²⁺ ion binding sites in relation to the O-antigen binding pocket.

Figure S2

Single-Particle Cryo-EM Reconstruction of the RsaA_NTD:PS Complex, Related to Figure 1 (A) Fourier shell correlation (FSC) curve of two random half sets of the final reconstructed RsaA_NTD:PS map shows better than 3.7 Å resolution according to the gold standard criterion of 0.143. (B) Local resolution differences plotted on the cryo-EM density of the RsaA_NTD:PS complex. (C) Examples of the de novo built atomic model fitted into the density contoured at 6 σ away from the mean.

Figure 2

Deducing Protein:Sugar Stoichiometry and Ca²⁺-Dependent Assembly Using High-Resolution Native MS (A) Native mass spectrum of purified RsaA_NTD:PS complexes show populations of monomer and dimer, both associated with LPS (average mass 10313 Da) and oligomers (20-mer and 21-mer) bound to one unit of PS and six units of LPS. Inset: High energy MS/MS of the oligomeric RsaA_NTD performed by isolating the peak at ~12404 m/z and dissociating at a voltage of 220V applied to the higher-energy collisional dissociation (HCD) cell. The inset spectrum (blue background) shows stripped oligomers generated by loss of single subunits from the parent complex (21-mer→20-mer→19-mer) and (20-mer→19-mer→18-mer) allowing us to conclude that the original oligomer consists predominantly of a 21-mer and a 20-mer with one unit of PS and six units of LPS each (see Table 1), although other LPS or PS hydrolysis products may also be present. (B) Mass spectrum of the RsaA_NTD:PS sample after Ca²⁺ removal shows presence of RsaA_NTD monomers, dimers, and tetramers only (see also Figure S3).

Figure S3

Investigation of the Effect of Ca²⁺ on RsaA_NTD Oligomerisation, Related to Figure 2 (A) Mass spectrum of the RsaA_NTD:PS sample after Ca²⁺ removal shows presence of RsaA_NTD monomers, dimers, and tetramers only. Panel same as Figure 2B, shown here for clarity. (B) Mass spectrum of the above sample following incubation with 1 mM calcium acetate indicates that Ca²⁺ ions stimulate the formation of oligomers (RsaA_NTD:PS complex). (C) After a second Ca²⁺ removal, the complex falls apart into RsaA_NTD monomers and dimers. (D) Top view of the RsaA_NTD:RsaA_NTD interaction interface in the cryo-EM structure shown as ribbon diagram. A single α-helix of one RsaA_NTD subunit forms the interaction interface with the next RsaA_NTD subunit. This interaction is duplicated around the spiral or in the native S-layer hexamer, likely giving large net stabilization. (E) A 90° rotated side view of (D) along the axis of the RsaA_NTD:PS spiral is shown. (F) Close-up view of (E) highlighting key residues at the RsaA_NTD:RsaA_NTD interaction interface. The interface is stabilized by an ionic interaction between Asp₃₀ of one RsaA_NTD monomer with Lys₁₄₂ of another RsaA_NTD molecule.

Figure 3

Probing RsaA_NTD Binding to the O-Antigen of LPS Using MD Simulations (A) MD simulation of RsaA_NTD bound to the O-antigen with no branching sugar moieties. (B) Simulation of RsaA_NTD bound to the O-antigen with 3-O-methyl-glucose (Glc) moieties at positions 3 and 9. (C) Simulation of RsaA_NTD bound to the O-antigen with Glc moieties at positions 6 and 12. (D) Plot of RMSF of the O-antigen atoms in the MD simulation presented in (A). (E) Plot of RMSF of the O-antigen atoms in the MD simulation presented in (B). (F) Plot of RMSF of the O-antigen atoms in the MD simulation presented in (C). (G–I) Cryo-EM density at different isosurface contour levels showing density for the branching sugar moieties at every third position (see Figure S4)

Figure S4

MD Simulations of RsaA_NTD Binding to O-antigen, Related to Figure 3 (A) Plot of amino acid residue and O-antigen (heavy-atom) interactions within 4 Å over the course of three 100 ns simulations. The protein-sugar interactions are normalized to 1 (brown), where 0 (white) relates to no contacts. (B) Interaction plot of protein residues and O-antigen with branching Glc moieties at the positions 3 and 9. (C) Protein and O-antigen interactions with branching Glc moieties at the positions 6 and 12. (D–F) Protein and O-antigen interactions from (A–C) are plotted on the ribbon diagram of the protein on a blue to red scale and on the O-antigen on a gray to purple scale. (G–I) Root mean square fluctuations (RMSF) of the O-antigen are displayed on the O-antigen stick diagram on a gray to red color scale (corresponding representation to data in Figures 3D–3F). (J–L) Plot of RMSF of the RsaA_NTD residues (see Figures 3A–3C), showing stabilization of the Ca²⁺ binding loop (residues 77–100) by the branching Glc moieties over the course of three 100 ns simulations (α-helical residues in blue background). (M) Interaction of protein residues with Ca²⁺ ions during MD simulations shown on a blue to red scale. All Ca²⁺ ions are stabilized by two aspartic acid residues and backbone carbonyl oxygens in the simulations, as well as in our cryo-EM structure.

Figure 4

Binding of RsaA_NTD to Cellular LPS Occurs along the Entire Length of the O-Antigen (A) Cryo-ET slice through a cellular stalk of CB15N C. crescentus. An assembled S-layer is observed, with RsaA_NTD and RsaA_CTD layers ~180 Å and ~230 Å away from the OM, respectively. (B) Subtomogram averaging of the stalk shows clear densities for the OM, RsaA_NTD, and RsaA_CTD. (C) Slice through a stalk lacking RsaA (ΔrsaA). (D) Corresponding subtomogram average from ΔrsaA stalks. (E) Slice through a ΔrsaA cellular stalk with exogenous RsaA_NTD added. (F) Corresponding subtomogram average shows three density layers on the outside of the OM. Inset: Side view of the RsaA_NTD:PS cryo-EM structure also shows three layers of protein bound to the O-antigen PS with the same spacing (see Figure S5).

Figure S5

Probing RsaA Binding to Cellular LPS, Related to Figure 4 (A) Cryo-ET slice through a cellular stalk of C. crescentus. (B) Slice through a cell stalk lacking RsaA (ΔrsaA). (C) Slice through a ΔrsaA cell stalk with exogenous full-length RsaA added. Decoration of the LPS in three layers is observed (as in Figures 4E and 4F); however, an additional fourth density layer is observed at the same distance from the OM as the native S-layer RsaA_CTD. This suggests that RsaA molecules bound to the tip of O-antigen form at least a partial outer S-layer lattice by oligomerization of RsaA_CTD. (D) Normalized density profiles through subtomogram averages of (A–C) aligned to the OM showing that exogenous added full-length RsaA binds to the entire length of the O-antigen, while forming a partial outer S-layer lattice. (E–G) Corresponding sub-tomogram averages of (A–C) (Figures S5A and S5B; E–G are the same as Figures 4A–4D, shown here for clarity). (H) Cryo-EM image of a ΔrsaA cell with exogenous RsaA_NTD added together with EGTA. Chelation of Ca²⁺ by EGTA prevents S-layer assembly at the cell surface. (I) Cryo-ET slice through the top of a cell stalk of C. crescentus showing a normal, hexagonal S-layer. (J) Slice though the top of a cell stalk lacking RsaA (ΔrsaA). (K) Slice through the top of a ΔrsaA cell stalk with exogenous full-length RsaA added showing irregularly arranged spiral-like structures (black arrow). (L) Slice through the top of a ΔrsaA cell stalk with exogenous RsaA_NTD added showing irregularly arranged spiral-like structures (arrow) with characteristics similar to the RsaA_NTD:PS complex.

Figure S6

Structure of the Native C. crescentus S-Layer Determined by Subtomogram Averaging, Related to Figure 5 (A) Fourier shell correlation (FSC) curve of two half sets of the final reconstructed native RsaA S-layer at a 4.82 Å resolution according to the 0.143 criterion. (B) Local resolution differences plotted on the cryo-ET density showing resolution anisotropy between RsaA_NTD and RsaA_CTD. (C) Atomic models docked into the cryo-ET density (gray) (contour levels on lower left side of panel). The RsaA_CTD X-ray structure (PDB: ID 5N8P) fits the central pore region exceptionally well. (D) A side view cross-section of the isosurface (gray) is shown with the docked RsaA_CTD X-ray structure (red) and the RsaA_NTD cryo-EM structure (blue). RsaA_CTD is connected to RsaA_NTD by a small linker region. (E) A single monomer of the native RsaA S-layer is shown as top view as in Figure 5B. (F) Close up view of the connecting region as shown in (D) highlights the exceptional model fit of both domains. The C terminus of the solved cryo-EM structure (Pro₂₄₃) is ~19 Å away from the N terminus (Gly₂₄₉) of the X-ray structure (red). The linker region consisting of five residues is poorly resolved indicating flexibility.

Figure 5

In Situ Cryo-ET of the Native C. crescentus S-Layer at 4.8 Å Resolution (A) Subtomogram averaging at 4.8 Å resolution of the native S-layer from cell stalks. RsaA_CTD X-ray structure (PDB: 5N8P, red ribbon) and RsaA_NTD cryo-EM structure (blue ribbon) are docked into the density (contour level shown on lower left of each panel, see Figure S6). (B) Top View of a Single Hexameric Unit (C) Six subunits of the cryo-EM structure of RsaA_NTD docked into the cryo-ET map. LPS densities highlighted with asterisks (^∗). (D) A closeup of a single RsaA_NTD α-helix showing some resolved bulky side chains in the cryo-ET map. (E) A ribbon diagram of one RsaA_NTD subunit overlaid on a slice through the cryo-ET map. A clear density for the O-antigen is observed at the same relative location as in the cryo-EM RsaA_NTD:PS structure. (F) A side view of a single hexamer is shown relative to the OM of the cell. Densities of O-antigen bound to RsaA extend downward to the OM (black outline density). Positions of RsaA_NTD:PS density layers as seen in Figure 4F are highlighted with blue arrows (see Video S3). (G) A closeup view of the O-antigen binding pocket resolved in the cryo-ET map. (H) Cellular structural biology from cells to atoms. Tomographic slice of a C. crescentus cell. Copies of the 4.8 Å cryo-ET structure are overlaid on the tomographic slice at their refined cellular locations. Atomic structures determined by X-ray crystallography (RsaA_CTD 2.7 Å) and cryo-EM (RsaA_NTD 3.7 Å) are docked into the cryo-ET map (see Video S5).

Figure S7

Overall S-Layer Arrangement on the Flat Cell Body and Highly Curved Cell Stalk Is the Same, Related to Figure 6 (A) A cryo-ET slice through the side of a C. crescentus cell body (protein density black in all raw cryo-ET slices). The OM is clearly decorated with a S-layer made up of RsaA_NTD and RsaA_CTD layers (marked). (B) Cryo-ET slice through the tip of a C. crescentus cell stalk. The highly curved OM is covered by a S-layer, consisting of the same RsaA_NTD and RsaA_CTD layers, with same ultrastructural morphology as the S-layer on the cell body. (C) Cryo-ET slice through the top surface of a C. crescentus cell body. The near hexagonal planar arrangement of the S-layer with a hexamer:hexamer distance of 220 Å is seen, as shown previously (Bharat et al., 2017) and confirmed in this study. (D) Cryo-ET slice through the top of a C. crescentus cell stalk. Although the S-layer lattice is highly curved around the stalk, the pseudo-hexagonal arrangement of the S-layer with a hexamer:hexamer distance of 220 Å is observed, same as the cell body. (E) Despite considerably increased specimen thickness, we performed subtomogram averaging of the outer surface of the C. crescentus cell body, which shows clear densities for the OM, RsaA_NTD, and RsaA_CTD (protein density white in all averages). The distance between the OM and the RsaA_CTD layer is ~230 Å. (F) Subtomogram averaging of the cell stalk (as shown in Figure 4B) shows a highly curved OM, surrounded by a S-layer. The distance between the OM and the RsaA_CTD layer is ~230 Å, same as the cell body, indicating the length of the LPS underneath the S-layer is the same between the stalk and the cell body. (G) The dimeric RsaA_CTD interface observed in the outer S-layer lattice in flat planar sheets, solved by X-ray crystallography (PDB: 5N8P). (H) Two copies of RsaA_CTD were fitted separately into a subtomogram averaging map produced from curved cell stalks with a large box size to visualize the hexamer:hexamer interfaces. The fit shows a mismatch between the X-ray structure and the subtomogram averaging map, suggesting rearrangement of residues at the dimeric interface concurrent with lattice curvature.

Figure 6

Schematic Model of C. crescentus S-Layer Assembly on the Cell Surface (A) Density of the cryo-EM structure of the isolated RsaA_NTD:PS complex contoured at 3 σ away from the mean. The reconstruction of the entire spiral is shown, no mask applied (3.9 Å resolution map). RsaA_NTD subunits are shown in different shades of gray (additionally numbered), and density corresponding to the O-antigen of the LPS is shown in orange. Asterisk (^∗) denotes density for LPS extending at lower contour levels only in one direction. (B) Subtomogram averaging of the sample with exogenous added full-length RsaA to cells lacking native S-layer (ΔrsaA) demonstrates that full-length RsaA can bind along the entire length of the O-antigen and can form a partly assembled outer lattice (panel same as Figure S5G). We expect that only RsaA molecules at the tip of the O-antigen are able to partially assemble the outer lattice because of steric hindrance by a mesh of LPS in the layers below. (C) Using a combined structural approach, X-ray crystallography (RsaA_CTD, PDB: 5N8P, red), cryo-EM (RsaA_NTD, blue), cryo-ET (subtomogram average, gray), and native MS (O-antigen, orange), we report a model of a full bacterial S-layer bound to LPS. (D) Schematic model of C. crescentus S-layer assembly. RsaA is secreted to the extracellular milieu, where RsaA binds to Ca²⁺ and LPS. This binding has been observed in our cryo-EM structure of the RsaA_NTD:PS complex (Figure 1) and verified by native MS (Figure 2). Next, RsaA is guided on LPS molecules by binding to the O-antigen along multiple sites, as observed in our cryo-EM structure (Figure 1), confirmed by MD simulations (Figures 3 and S4) as well as by in situ experiments showing binding of RsaA along the entire length of the LPS O-antigen (Figure 4). RsaA molecules are unable to assemble into an S-layer lattice near the OM (Figure S5), likely due to steric hindrance by a meshwork of LPS molecules. At the tips of the LPS O-antigen, whose length we accurately estimated using native MS (Figure 2), RsaA molecules can bind with a pre-existing S-layer to complete gaps in the lattice via oligomerization through RsaA_CTD (B and Bharat et al., 2017).