Complete structure of the core signalling unit of the E. coli chemosensory array in an optimised minicell strain

Motile bacteria sense chemical gradients with transmembrane receptors organised in supramolecular signalling arrays.1,2 Understanding stimulus detection and transmission at the molecular level requires precise structural characterisation of the array building block known as a core signalling unit (CSU). Here we introduce a novel E. coli strain that forms small minicells possessing extended and highly ordered chemosensory arrays. We provide a three-dimensional (3D) map of a complete CSU at ~16 Å resolution by cryo-electron tomography (cryo-ET) and subtomogram averaging. This map, combined with previously determined high resolution structures and molecular dynamics simulations, yields an atomistic model of the membrane-bound CSU and enables spatial localisation of its signalling domains. Our work thus offers a solid structural basis for interpretation of existing data and design of new experiments to elucidate signalling mechanisms within the CSU and larger array.


solid structural basis for interpretation of existing data and design of new experiments 30
to elucidate signalling mechanisms within the CSU and larger array. 31 Bacteria survive and proliferate by sensing changes in their environment and 32 responding through metabolic adaptation or locomotion. Chemotactic bacteria, for example, 33 monitor attractant and repellent concentration gradients and promote movement towards 34 favorable niches. The chemosensory arrays that mediate this behavior are assembled from 35 CSUs that contain two trimers of receptor dimers (ToDs) interconnected at their cytoplasmic 36 tips by a dimeric histidine autokinase CheA and two copies of a coupling protein CheW, 37 which links CheA activity to receptor control (Fig. 1a, b, c). Chemoeffector binding to the 38 periplasmic domain of the receptors triggers a signalling cascade of intracellular 39 phosphorylation events which ultimately regulate the direction of the cell's flagellar motors. 40 To allow appropriate sensory responses over a wide range of chemoeffector concentrations, 41 receptor sensitivity is continuously tuned through the reversible methylation of receptors, 42 known as methyl-accepting chemotaxis proteins or MCPs. E. coli has four canonical MCPs 43 that share a common functional architecture (Fig. 1a); the two numerically predominant ones 44 are Tar (aspartate and maltose sensor) and Tsr (serine and autoinducer 2 sensor). Changes 45 in ligand occupancy of their periplasmic ligand-binding domains trigger conformational 46 rearrangements that propagate through the inner membrane to the HAMP domain, a 47 signalling element found in most microbial chemoreceptors and sensory kinases. 3 The HAMP 48 domain couples extracellular input to intracellular output by relaying stimulus information 49 through an extended methylation helix (MH) bundle, via a flexible region and glycine hinge, 50 to the signalling tips where it is further transmitted to CheA and CheW to affect kinase 51 activity. CheA functions as a homodimer with each monomer containing five domains 52 (referred to as P1-P5) connected by flexible linkers (Fig. 1b). 53 The mechanistic details of sensory signal transmission within the CSU remain 54 mysterious, in part because there is as yet no experimentally determined structure, even at 55 low resolution, for a complete CSU. On one hand, cryo-ET structures of complexes 56 composed of receptor cytoplasmic domains, CheA and CheW that are assembled on lipid 57 monolayers provide insights into CSU structure, 4 but lack the ligand-binding, transmembrane 58 up-regulation of chemotaxis and, presumably, flagellar genes, we found that WM4196 116 performed quite poorly in soft agar chemotaxis tests compared to RP437 derivatives UU3118 117 and UU3120 ( Supplementary Fig. 2). The observed migration difficulties of WM4196 could 118 be due to the fact that receptor arrays reside mainly at the poles of cells. Because WM4196 119 mother cells efficiently bud minicells at their poles, the minicells probably contain most of the 120 receptor arrays, leaving the mother cells with relatively few. However, since basal bodies 121 form at random locations around the cell, 25 (Fig. 3c). The minicell responses 132 exhibited hallmarks of sensory adaptation (response decay and activity overshoot; Fig. 3b) 133 consistent with a normal complement of the CheR and CheB receptor-modifying enzymes. 134 In summary, the receptor arrays in WM4196 minicells function comparably to those in RP437 135 cells. Therefore, we set out to calculate a 3D cryo-ET map of the CSU in WM4196 minicells. 136 In our cryo-ET reconstructions of individual minicells, the chemosensory array 137 followed the curved surface of the inner membrane, providing the ensemble of views 138 required for an isotropic 3D reconstruction by subtomogram averaging (see Methods, 139 Supplementary Fig. 3, Supplementary Fig. 4). Subtomogram averaging in Dynamo 28-30 140 resulted in a ~16 Å resolution 3D density map of a hexameric arrangement of three CSUs 141 ( Fig. 1c, Supplementary Fig. 3), from which we extract one for modelling and interpretation 142 ( Fig. 4, Supplementary Fig. 3). The most striking difference between our map and previously 143 reported structures 4-8 is the complete and continuous receptor density, showcasing the entire assembly from the periplasmic ligand-binding domain to CheA and CheW at the cytoplasmic 145 signalling tip (Fig. 4a, Supplementary Fig. 3). Indeed, densities corresponding to the 146 periplasmic domain, TM four-helix bundle, and HAMP domain were not resolved in previous 147 array studies, which had been attributed to inherent flexibility of the CSU. 3 Although our 148 current map probably derives from a mixture of receptors with a range of adaptational 149 modifications and signalling states, both the HAMP and MH bundles seem to be relatively 150 static. Thus, our 3D reconstruction demonstrates that complete CSUs are amenable to 151 visualisation by cryo-ET. The WM4196 minicell system, therefore, represents a valuable tool 152 for elucidating structure-function relationships in CSUs, for instance through the imaging of 153 arrays in different mutationally-imposed signalling states. 154 To enable spatial localisation of each signalling component and their individual 155 domains in the 3D density map, we constructed an atomistic model of the complete E. coli 156 CSU, using the density map to explicitly refine model tertiary structure (see Methods, Fig.  157 4a). The resulting molecular model provides several new structural insights throughout the 158 CSU into MCP structure and organization. In particular, visualisation of the receptor ligand-159 binding domains allows us to describe for the first time the periplasmic organisation of the 160 MCPs (Fig. 4b). Intriguingly, we observe in this region that receptors within a given ToD are 161 situated as close or closer to those from neighboring ToDs than those in the same ToD ( Fig.  162 4b, Supplementary Fig. 4), suggesting that minor diffusion within the membrane could give 163 rise to ToD interactions both within and between CSUs. 31 The impact of such interactions on 164 signalling and cooperativity has yet to be systematically explored and should now be 165 amenable to mutational and cross-linking analyses using WM4196-derived strains. 166 Furthermore, previously documented kinks between the HAMP and the MH bundles as well 167 as at the glycine hinge 32 are clearly not required to avoid structural clashes between 168 neighbouring MCPs (Fig. 4a), although gradual bending in these areas could play a role in 169 transitions between signalling states. 8 Overall, the symmetry axes of the ToDs are separated 170 by 7.4 nm, suggesting an array lattice constant of ~12.8 nm which is consistent with inter-171 particle distances in the tomograms. (Supplementary Fig. 5).
In the present model, the CSU baseplate, which comprises densities corresponding 173 to the MCP signalling tip, the CheA kinase dimer, and the CheW coupling protein (Fig. 1b) UU3120 is a derivative of UU3118 that carries the pflhDC(-10)g5a allele of WM4196 introduced 227 by two-step insertion/replacement. 228

Measuring cellular levels of chemotaxis proteins
Plasmid pRR48 42 was introduced into RP437, WM4196, and UU3111 to express the β-230 lactamase (Bla) protein (an internal reference for standardizing the relative levels of other cell 231 proteins). Cells were grown at 37°C in L broth (10 g/L tryptone, 5 g/L yeast extract, 5 g/L 232 NaCl) containing 25 µg/ml ampicillin and harvested at mid-exponential phase (OD600 = 0. ml of cell culture were centrifuged at ~8600 X g for 25' at 4°C. All subsequent centrifugation 247 steps were also carried out at 4°C. The supernatant was transferred to fresh tubes and 248 centrifuged at ~14,500 X g for 25'. The sample pellets were pooled and resuspended in ~1 ml 249 of motility buffer (see above) and carried through a second round of differential centrifugation 250 (9000 X g for 10'; 21000 X g for 20'). The final minicell pellet was resuspended in ~100 µl 251 motility buffer and applied to a polylysine-coated cover slip for the FRET assay. 252

WM4196 minicell purification for cryo-ET imaging 253
WM4196 minicells were grown in L broth supplemented with 34 µg mL -1 chloramphenicol for 254 12 hours. This culture was used to inoculate larger L broth cultures (without antibiotics) to an 255 OD600 value of 0.075. Bacteria were grown at 37 °C for 4h (the culture was grown to an OD600 256 value of 0.5 before being left to grow for a further 2h. Final OD600 value was 1.75). The culture 257 was centrifuged at 8683 g for 20 minutes at 4 °C, the supernatant carefully transferred to 258 another centrifuge tube and centrifuged again at 8683 g for 20 minutes at 4 °C. The resulting 259 supernatant was transferred to another centrifuge tube and centrifuged at 41500 g for 20 260 minutes at 4 °C. This time, the supernatant was discarded and the pellet gently resuspended 261 in the residual supernatant from the centrifuge tube. This suspension was then centrifuged at 262 5500 g for 5 minutes at 4 °C, and the supernatant transferred to another centrifuge tube to be 263 centrifuged at 16900 g for 15 minutes at 4 °C. The resulting minicell pellet was resuspended 264 in LB and the minicell suspension was placed at 4 °C.  -4, -6, -8, -10, 12, 14, 16, 18, 20, -12, -14, -16, -18, -20 etc). 46 Images at each tilt-angle were 279 acquired as movies comprising 5 frames. Target total dose for tilt-series acquisition was 60 e -280 Å -2 s -1 . 281

Pre-Processing 283
Frames of movies corresponding to each tilt-angle were aligned and motion-corrected using 284 MotionCor2 47 to mitigate the deleterious effects of beam-induced sample motion. Tilt-series 285 were assembled using the newstack command from IMOD. 48 286

Tomographic Reconstruction
Tilt-series were aligned, binned by two and tomograms were reconstructed by weighted back-288 projection using the IMOD software package. 48 Six tomograms of WM4196 minicells were 289 selected for further processing. 290

Initial Reference Generation 291
Particles were identified and picked using the Dynamo software package. 28-30 Sub-volumes 292 with a side length of 573 Å were extracted and initial orientations for each particle were 293 estimated by modelling a surface following the curvature of the chemosensory array in the 294 tomogram and imparting an orientation onto each particle which corresponded to the normal 295 to the modelled surface at the point closest to the particle. These orientations were further 296 refined manually, and these coarsely-oriented particles were then averaged to produce an 297 initial reference. The particles were subsequently locally aligned, constraining the angular 298 search to a 60° cone around the initial estimate for the orientation, and averaged to produce 299 initial references. 300

Particle Picking 301
The following analysis was performed in Dynamo. 29 Surfaces were modelled following the 302 curvature of the chemoreceptor arrays visible inside the minicells and a set of initial positions 303 and orientations was generated from this surface with an average distance of 30 Å between 304 each position. Sub-volumes with a side length of 573 Å were extracted at each of these 305 positions and each was aligned to the initial reference with an allowed translational freedom of 306 60 Å in each of the x-, y-and z-dimensions. Analysis of the post-alignment positions revealed 307 an ordered hexagonal array. Duplicate particles (defined as particles within 10 Å of each other) 308 were collapsed into one final position. The particle positions were then cleaned by selecting 309 only those which had greater than three nearest-neighbours at the expected distance of 310 120±20 Å. 311

Subtomogram Averaging 312
Alignment and averaging of the particles was performed in the Dynamo software package. 28 313 An iterative global alignment of the particle positions and orientations was performed starting 314 from the initial reference already generated, for which alignments were performed inside a 315 mask encompassing multiple core-signalling units in the chemosensory array and the inner membrane of the WM4196(DE3) minicell. These positions and orientations were then further 317 locally refined inside a mask containing only three core-signalling units, without the 318 membrane 10 , to produce a final reconstruction. 319

Post Processing 320
Two separate half-maps were generated from groups of particles coming from different 321 tomograms to allow estimation of the resolution of the final reconstruction. The fourier shell 322 correlation of these maps inside a mask, containing three core-signalling units and no 323 membrane, drops below 0.143 at a spatial frequency corresponding to a resolution of 16 Å. 324 The maps were subsequently subject to localised resolution estimation in RELION 49 with a 325 sampling rate of 20 Å and locally filtered to the estimated resolution of the map. The local-326 resolution filtered map was aligned to a C2-symmetric reference centered on one core-327 signalling unit and then itself symmetrised around the C2-axis to give a map centered on one 328 core-signalling unit. Masks for FSC calculation and map visualisation were calculated using a 329 combination of Chimera 50 , Dynamo 28 and RELION. 49