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

Design and genetic characterisation of the WM4196 strain Strain WM4196 is a distant derivative of a minicell mutant first described over 50 years ago. This prototype strain, containing the minD allele, a dnaB ts allele, as well as several auxotrophic markers, was converted to dnaB + and prototrophy by conjugation with a prototrophic Hfr donor strain 29 . The F′ was then cured to make strain x1411, which was obtained from the E. coli Genetic Stock Center, Yale University (CGSC#6397). The mreB -A125V allele 28 was further subsequently introduced into x1411 by cotransduction with a tightly linked yhdE::cat chromosomal allele to create the WM4196 strain.

DNA sequencing and strain constructions WM4196 chromosomal

DNA was PCR-amplified and sequenced with primer pairs specific for the minCDE , flgM , and flhDC loci. Strain UU3118 (RP437 mreB-A125V minD-G262D ) was constructed by (a) introducing the mreB allele from WM4196 into RP437 by phage P1-mediated cotransduction with the linked ∆yhdE::cat allele; and (b) introducing the minD-G262D mutation with homologous recombination-mediated insertion and replacement steps 47 . Strain UU3120 is a derivative of UU3118 that carries the pflhDC(-10)g5a allele of WM4196 introduced by two-step insertion/replacement. Measuring cellular levels of chemotaxis proteins Plasmid pRR48 48 was introduced into RP437, WM4196, and UU3111 to express the β-lactamase (Bla) protein (an internal reference for standardizing the relative levels of other cell proteins). Cells were grown at 37 °C in L broth (10 g l −1 tryptone, 5 g l −1 yeast extract, 5 g l −1 NaCl) containing 25 µg ml −1 ampicillin and harvested at mid-exponential phase (OD 600 = 0.5–0.6). Approximately equal numbers of cells (based on culture OD) were used to prepare lysates. Cells were pelleted by centrifugation, washed once with motility buffer (10 mM K-PO 4 [pH 7.0], 0.1 mM K-EDTA) and lysed in SDS–PAGE buffer 49 . Samples were analysed in 11% polyacrylamide–SDS gels and MCP, CheA and Bla bands were detected by immunoblotting with a mix of three polyclonal rabbit antisera. Bands were visualized with a Cy5-labeled goat anti-rabbit antibody and quantified with a fluorescence imager. The uncropped and unprocessed scan of the gel is shown in Supplementary Fig. 2 and in the Source Data file to Fig. 3a . In vivo FRET kinase assay This method is described in detail in previous publications 32 , 50 . For the present study, we introduced into WM4196 plasmid pVS88 33 , which expresses fusion proteins CheZ-CFP (FRET donor) and CheY-YFP (FRET acceptor) under IPTG-inducible control. The plasmid-carrying strain was grown at 37 °C to OD 600 = 0.65–0.7 in L broth (see above) containing 25 µg ml −1 ampicillin and 1 mM IPTG. (The high IPTG concentration compensates for elevated levels of untagged CheY and CheZ proteins expressed from the WM4196 chromosomal genes.) 300 ml of cell culture were centrifuged at ~8600 × g for 25 min at 4 °C. All subsequent centrifugation steps were also carried out at 4 °C. The supernatant was transferred to fresh tubes and centrifuged at ~14,500 × g for 25 min. The sample pellets were pooled and resuspended in ~1 ml of motility buffer (see above) and carried through a second round of differential centrifugation (9000 × g for 10 min; 21,000 × g for 20 min). The final minicell pellet was resuspended in ~100 µl motility buffer and applied to a polylysine-coated cover slip for the FRET assay.

WM4196 minicell purification for cryo-ET imaging

WM4196 minicells were grown at 37 °C in L broth supplemented with 34 µg ml −1 chloramphenicol for 12 h. Small volumes of this culture were used to inoculate larger volumes of L broth media (without antibiotics) to an initial OD 600 value of 0.075. These larger volume cultures were grown at 37 °C for 4 h, the final OD 600 value was 1.75. The culture was centrifuged at 8683 × g for 20 min at 4 °C, the supernatant carefully transferred to another centrifuge tube and centrifuged again at 8683 × g for 20 min at 4 °C. The resulting supernatant was transferred to another centrifuge tube and centrifuged at 41,500 × g for 20 min at 4 °C. This time, the supernatant was discarded and the pellet gently resuspended in the residual supernatant from the centrifuge tube. This suspension was then centrifuged at 5500 × g for 5 min at 4 °C, and the supernatant transferred to another centrifuge tube to be centrifuged at 16,900 × g for 15 min at 4 °C. The resulting minicell pellet was resuspended in LB and the minicell suspension was placed at 4 °C.

Cryo-EM imaging and cryo-ET data acquisition

Minicell solution was supplemented with 10 nm colloidal protein A-gold particles (Cell Microscopy Core, University Medical Center, Utrecht, The Netherlands). R 2/1 or R 3.5/1 on 300 mesh Cu/Rh grids (QUANTIFOIL) were glow discharged for 45 s at 25 mA. 3 μl of the sample was applied to grids and plunge frozen in liquid ethane using a Vitrobot Mark IV (Thermo-Fisher Scientific). Grids were stored under liquid nitrogen until imaging. Screening of the grids was performed on an FEI F20 microscope at 200 keV. Tilt-series acquisition was performed on a Titan Krios transmission electron microscope operated at 300 keV, through a Gatan Quantum energy filter and Volta phase plate onto a K2 Summit direct electron detector, using SerialEM software 51 . For both search and navigational purposes, low-magnification montages were acquired. For tilt-series acquisition, a magnification which produced a calibrated pixel size of 2.24 Å was selected. Data were collected at tilt-angles between −60° and +60° in 2° increments in a grouped dose-symmetric tilt scheme (0, +2, +4, +6, +8, +10, −2, −4, −6, −8, −10, 12, 14, 16, 18, 20, −12, −14, −16, −18, −20, etc.) 52 . Images at each tilt-angle were acquired as movies comprising five frames. Target total dose for tilt-series acquisition was 60 e − Å −2 .

Image processing and pre-processing

Frames of movies corresponding to images at each tilt-angle in each tilt-series were aligned and motion-corrected using MotionCor2 53 to mitigate the deleterious effects of beam-induced sample motion. Tilt-series were assembled using the newstack command from IMOD 54 .

Tomographic reconstruction

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

Initial reference generation

Particles were identified and picked using the Dynamo software package 34 – 36 . Sub-volumes with a side length of 573 Å were extracted and initial orientations for each particle were estimated by modelling a surface following the curvature of the chemosensory array in the tomogram and imparting an orientation onto each particle, which corresponded to the normal to the modelled surface at the point closest to the particle. These orientations were further refined manually, and these coarsely oriented particles were then averaged to produce an initial reference. The particles were subsequently locally aligned, constraining the angular search to a 60° cone around the initial estimate for the orientation, and averaged to produce initial references.

Particle picking

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

Subtomogram averaging

Alignment and averaging of the particles was performed in the Dynamo software package. An iterative global alignment of the particle positions and orientations was performed starting from the initial reference already generated, for which alignments were performed inside a mask encompassing multiple CSUs in the chemosensory array and the inner membrane of the WM4196(DE3) minicell. These positions and orientations were then further locally refined inside a mask containing only three CSUs, without the membrane 15 , to produce a final reconstruction resulting from 681 subtomograms derived from six tomograms.

Post processing

Two separate half-maps were generated from groups of particles coming from different tomograms to allow estimation of the resolution of the final reconstruction. The Fourier shell correlation (FSC) of these maps inside a mask, containing three CSUs and no membrane, drops below 0.5 and 0.143 at spatial frequencies corresponding to resolution estimates of 18 and 16 Å, respectively (Supplementary Fig. 5 ). The maps were subsequently subjected to localised resolution estimation in RELION 55 with a sampling rate of 20 Å and locally filtered to the estimated resolution of the map. A histogram of the per-voxel local resolution estimations for regions inside the FSC mask shows that most areas of the map fall in the range 15–25 Å (Supplementary Figs. 5 and 6 , Fig. 4c ). The local-resolution-filtered map was aligned to a C2-symmetric reference centred on one CSU and then itself symmetrised around the C2-axis to give a final map centred on one CSU. Separate masks for FSC calculation and map visualisation were calculated using a combination of Chimera 56 , Dynamo and RELION.

Molecular modelling and simulations

Atomic coordinates for E. coli CheA.P3.P4.P5 and CheW were derived from existing T. maritima crystal structures 57 , 58 using template-based homology modelling. An all-atom model of the E. coli Tsr (residues 1–518) was constructed from existing crystal structures of the ligand binding and cytoplasmic domains as well as a HAMP homology model based on a crystal structure of A. fulgidus HAMP. The transmembrane four-helix bundle was modelled using existing cross-linking data between the individual helices. As the long, unstructured C-terminus of Tsr (residues 519–551) was not resolved in the cryo-ET map, we did not include these residues in the model. The intact Tsr model was then embedded in a 3:1 POPE:POPG lipid bilayer and equilibrated for 500 ns using molecular dynamics simulation, providing an initial structure for further conformational refinement via MDFF. All-atom molecular dynamics simulations were carried out using NAMD 2.13 59 (and the CHARMM36 force field 60 ). MDFF simulations were performed in the NVT ensemble at 310 K in explicit solvent. A scaling factor of 0.1 was used to couple backbone atoms to the MDFF potential. To prevent the loss of secondary structure as well as the formation of cis-peptide bonds and chirality errors, additional harmonic restraints were applied to the protein backbone during the fitting simulations. Reporting summary Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Supplementary information Supplementary information Reporting Summary Peer Review File Supplementary Movie 1 Description of Additional Supplementary Files