Similar protocols

Protocol publication

[…] For LdcI-LARA complex formation, 0.6 mg/ml of LdcI was mixed with 1 mg/ml of LARA for 15 min at room temperature in a buffer containing 50 mM Mes pH 6.5, 100 mM NaCl, 0.2 mM PLP, 1 mM DTT and 0.01% glutaraldehyde (glutarahaldehyde initially buffed in 800 mM Mes pH 6.5). The LdcI:LARA ratio was 1:10 (7.3 µM of LdcI for 73 µM of LARA). The glutaraldehyde cross-linking reaction was stopped by adding a final concentration of 62 mM TrisHCl pH 7. 4 µl of sample was applied to glow-discharged quantifoil grids (Quantifoil Micro Tools GmbH, Germany) 400 mesh 2/1, excess solution was blotted during 2 s with a Vitrobot (FEI) and the grid frozen in liquid ethane (). Data collection was performed on a FEI Polara microscope operated at 300 kV under low dose conditions. CryoEM micrographs were collected on a CCD Ultrascan Gatan USC 4000 (4 k × 4 k) using FEI EPU automatization software at a magnification of 102.413×, giving a pixel size of 1.464 Å. The contrast transfer function (CTF) for each micrograph was determined with CTFFIND3 () and corrected by phase flipping using bctf (). Defocus ranged between 1.5 and 2.7 µm.For initial 3D model determination, a subset of 11,102 particles was picked manually using EMAN boxer routine (). Band-pass filtered particles were centered against a rotationally averaged total sum and classified using multivariate statistical analysis (MSA) as implemented in IMAGIC (). A subset of ∼10 class averages was selected (based on the visual match between the class average and the individual particles) as references for multi-reference alignment (MRA). After three rounds of MRA/MSA, a 3D model was calculated by angular reconstitution. As two and fivefold symmetries were clearly visible in class averages and eigenimages, D5 symmetry was imposed for 3D calculations. Particle orientations were refined by multiple cycles of MRA, MSA and angular reconstitution, gradually incorporating more particles. The resulting initial model of LdcI-LARA (∼18 Å resolution) was used for particle picking using an automated particle picking procedure based on the Fast Projection Matching algorithm ().The resulting dataset of 26,165 LdcI-LARA particles was then used to refine the initial model by projection matching (AP SH command in SPIDER, D5 symmetry imposed) (; ). When the alignments had stabilized, more than 95% of the images aligned to the same references in consecutive rounds of alignment. The final 3D reconstruction comprised 23,540 particles. The resolution was estimated based on the gold-standard FSC = 0.143 criterion () by dividing the data in two independent halves and refining them iteratively against the angular reconstitution model low pass filtered to 50 Å resolution. The resolution estimated based on the unmasked FSC curve was 8.8 Å, whereas the FSC curve obtained with soft shaped masks (dilated 5 pixels and with a fall–off profile of a cosine half-bell of 4 pixel width) yielded the resolution of 7.5 Å (). The final map was sharpened with EMBfactor () using calculated B-factor of 463 Å2. [...] Conditions of optimal LdcI–RavA complex formation were initially tested by negative stain EM on a JEOL 1200 EX microscope. Different pHs (ranging from 6 to 8), salt concentrations (ranging from 100 to 300 mM), protein ratios (ranging from 1:1.5 to 1:3 LdcI monomer:RavA monomer), ADP concentrations (from 0 to 5 mM) were tested. No cross-linking agent was used to stabilize this high affinity complex. The most promising conditions were further analyzed by cryoEM on a FEI CM200 microscope. The final optimal condition chosen for data collection was: 1.26 mg/ml of LdcI mixed with 0.94 mg/ml of RavA for 15 min at room temperature in a buffer containing 25 mM Mes pH 6.5, 200 mM NaCl, 3 mM ADP, 0.8 mM PLP and 1 mM DTT. The resulting ratio of LdcI:RavA was 1:2 (in the complex, the ratio is LdcI–RavA ratio is 1:1.5, so there was 1.3 molar excess of RavA). Quantifoil grids were flash frozen in liquid ethane () using Vitrobot and a blotting time of 2 to 3 s.Initial 3D model was determined using cryo-electron tomography (cryoET) and subtomogram averaging as follows. Tomograms were recorded on a FEI Polara microscope between −65 and 65° with 2° angular step on a CCD Ultrascan Gatan USC 4000 at a nominal magnification of 51160× giving a pixel size of 2.932 Å. The total electron dose for each tomogram was around 40 electrons/Å2. The tomograms were aligned with the IMOD suite using 5 nm gold fiducial beads (BBInternational, UK) for frames alignment. 95 subvolumes were extracted in IMOD and aligned using PEET (). The subvolume average was consistent with the previously determined negative stain EM map of LdcI–RavA (; ) and clearly displayed a D5 symmetry.A high resolution dataset was then collected on a FEI Polara microscope operated at 300 kV under low dose conditions. Micrographs were recorded on Kodak SO-163 film at 59,000 magnification, with defocus ranging from 1.3 to 3.3 μm. Films were digitized on a Zeiss scanner (Photoscan) at a step size of 7 µm giving a pixel size of 1.186 Å. After CTF determination and correction (performed as for the LdcI-LARA dataset), particles were automatically picked using the Fast Projection Matching algorithm with projections of the cryoET reconstruction as a template. The resulting dataset of 35,443 particles was then used to refine the initial cryoET model by projection matching in SPIDER with D5 symmetry imposed (; ). 60% of the best correlating particles were used to refine the 3D reconstruction. When the alignments had stabilized, more than 95% of the images aligned to the same references in consecutive rounds of alignment. The final map comprised 21,265 particles with an even view distribution around the equatorial axis. The resolution was estimated based on the gold-standard FSC = 0.143 criterion () by dividing the data in two independent halves and refining them iteratively against the cryoET model low pass filtered to 50 Å resolution. The resolution estimated based on the unmasked FSC curve was 14 Å, whereas the FSC curve obtained with soft shaped masks (dilated 5 pixels and with a fall–off profile of a cosine half-bell of 4 pixel width) yielded the resolution of 11 Å (). [...] For LdcI-LARA fitting, LdcI decamer crystal structure (pdb 3N75) was fitted into the LdcI-LARA cryoEM map using the fit-in-map module of UCSF Chimera (), unambiguously identifying extra densities corresponding to LARA domains. The density of the LARA domain (residues 361–424, pdb 3NBX) was extracted and the LARA atomic model reliably rigidly fitted using SITUS (). Subtle domain movements being noted in the LdcI structure as compared to the crystal structure (), the LdcI density was extracted and used for flexible fitting with Flex-EM (). LdcI crystal structure was divided into three rigid bodies: the wing domain (residues 1–128), an intermediate domain (residues 131–142, 145–382, 386–397, 402–495, 527–562) and a C-terminal domain (residues 501–521 and 566–710), leaving inter-domains flexible. The Cα RMSD between the initial pdb of the LdcI monomer model and the final Flex-EM model is 0.65 Å. The cross correlation values between the cryoEM map and either the rigidly fitted monomer or the final Flex-EM model are 0.75 and 0.84 respectively. The FSC curves between the cryoEM map and either the rigid fit or the Flex-EM model give an FSC = 0.5 at 11 Å resolution and 8.3 Å resolution respectively ().For LdcI–RavA fitting, superposition of LdcI–RavA and LdcI-LARA maps (), demonstrated that LdcI and LARA positions were conserved in both maps (cross-correlation of 0.86 between the two maps filtered at 11 Å resolution). In order to take advantage of the higher resolution of the LdcI-LARA map, the LdcI-LARA fit was kept fixed for the LdcI–RavA fitting. Thus, only the AAA+ ATPase domain and the triple helical domains of RavA (pdb 3NBX) remained to be fitted. The initial rigid fit of these domains using UCSF Chimera fit-in-map module indicated that the AAA+ ATPase domain position was conserved compared to the isolated RavA structure, maintaining the integrity of the ATPase active sites. The rigid fit also revealed that RavA binding to LdcI required movements of the triple-helix domains to bring the LARA domains in their LdcI-interacting position, whereas the RavA legs involved in RavA–RavA equatorial interactions barely moved in comparison to the isolated RavA (). In order to propose a pseudoatomic model of RavA in its LdcI-binding state, RavA flexible fitting was thus performed using Flex-EM. The ATPase domains (residues 1–268), the triple helical bundles (residues 274–329, 444–496) and the LARA domains (residues 361–424) were considered as rigid, leaving inter-domains flexible. The LARA domains were given as initial position their position in LdcI-LARA maps, and the initial positions of the AAA+ ATPase domains were those of the rigid body fit. 25 cycles of Flex-EM led to improved position of triple-helix domains leaving the AAA+ ATPase and LARA domain fit quasi unchanged. The versatile 330–360 loop of the LARA domain was omitted from the models because it clearly undergoes major rearrangements that cannot be reasonably modeled based on the current quality of the EM maps.The FSC curves between the cryoEM map of the LdcI–RavA complex and either the rigid fit of the LdcI decamer and the RavA hexamer, or the Flex-EM model obtained as described, give an FSC = 0.5 at 26 Å resolution and 20 Å resolution respectively (). Interestingly, the FSC curve between the LdcI-LARA part of the LdcI–RavA map and its Flex-EM model has an FSC = 0.5 at 11 Å resolution, whereas the FSC curve between the equatorial region of the map (corresponding to the AAA+ domain core and the triple helical domains of RavA) and its Flex-EM model has an FSC = 0.5 at 20 Å resolution. This difference in the local quality of the Flex-EM model can be explained by (i) the better quality of the Flex-EM model of the LdcI-LARA region (because this part of the model is based on the higher resolution LdcI-LARA map), and (ii) the better definition of the LdcI-LARA region of the LdcI–RavA map in comparison to the equatorial region corresponding to the AAA+ domain core and the triple helical domains of RavA. […]

Pipeline specifications

Software tools CTFFIND, EMAN, IMAGIC, SPIDER, IMOD, UCSF Chimera, Situs, Flex-em
Applications cryo-EM, Protein structure analysis
Organisms Escherichia coli