2016-07-13 REL 2016-02-18 PDBe PDBe 2016-03-09 2016-06-29 Verba KA, Wang RYR, Arakawa A, Liu Y, Shirouzu M, Yokoyama S, Agard DA Hsp90, Cdc37, Cdk4, chaperone, kinase, unfolding 5fwm Verba KA, Wang RYR, Arakawa A, Liu Y, Shirouzu M, Yokoyama S, Agard DA Atomic structure of Hsp90-Cdc37-Cdk4 reveals that Hsp90 traps and stabilizes an unfolded kinase SCIENCE 352 1542 1547 2016 27339980 doi:10.1126/science.aaf5023

emd_3342.tif

emd_3342_fsc.xml 3 Complex of Human Hsp90 beta, human Cdc37 and human Cdk4 One Hsp90 homodimer binds to one Cdc37 and one Cdk4 0.245

All three proteins were co-expressed in Sf9 cells.

As cloned, verified by SDS-PAGE 0.245 protein Heat Shock Protein HSP 90 beta Hsp90 0.083 Homo sapiens Human 2 true Dimer P08238 GO:0000052 IPR001404 cytoplasm pFastBacHT Spodoptera frugiperda protein Hsp90 co-chaperone Cdc37 Cdc37 0.0445 true Homo sapiens 1 Human GO:0000002 Q16543 throughout Spodoptera frugiperda pFastBacHT protein Cyclin-dependent kinase 4 Cdk4 0.0337 Homo sapiens 1 Human true P11802 IPR000719 GO:0000038 throughout pFastBacHT Spodoptera frugiperda Single blot from 4 to 6 seconds, at 20C ETHANE 90 FEI VITROBOT MARK III 95 At high mag via FT. FIELD EMISSION GUN 44 BRIGHT FIELD 1400 3800 OTHER GATAN K2 SUMMIT (4k x 4k) 2.7 FEI TITAN KRIOS 25-NOV-2014 FEI TITAN KRIOS AUTOGRID HOLDER 300 22500 3718 8

38 frames, 7.6 seconds total exposure

Residues 5-85 (N-lobe) of Cdk4 from 3G33 were fit in Chimera as rigid body into EMD-3342. Such fit N-lobe of Cdk4 was further tweaked in Coot in the context of 5FWK to join the Cdk4 chain and minimize clashes for each of the maps. To relieve atomic clashes or bond length/angle distortions at the linker regions, these models were subjected to "Cartesian space relax" protocol within corresponding density maps using Rosetta. Final models were selected using the combined score of Rosetta all-atom physically-realistic score and electron density score.

flexible REAL cross correlation of fit into the density with Rosetta force field score Chimera, Rosetta 130 5FWK A B E K 130 REAL cross correlation of fit into the density with Rosetta force field score flexible Chimera, Rosetta 3G33 particle 0.27 Glow discharged for 30 sec, C-flat 400 mesh 1.2/1.3 thick carbon grids (Protochips)

20mM Tris-HCl (pH 7.5), 150 mM NaCl, 10 mM KCl, 10 mM MgCl2, 20 mM Na2MoO4, 2mM DTT, 0.085mM DDM

7.5 singleParticle Relion 8 FSC 0.143, gold-standard 1 C1 61981

Image stacks were corrected for motion and summed as described previously, resulting in binned sums (1.315A/pix). For particle picking the images were binned to 5.2A/pix and Gaussian bandpass filtered between 15A and 500A using EMAN2. SamViewer template based picking was then used to pick particles from all the micrographs, followed by manual review of all the picks. After such procedure 802877 particles were picked in total and extracted from images binned to 2.6A/pix. CTFFIND4 was used to estimate defocus parameters for all the images. Relion 1.4 was used for all the following steps unless noted otherwise. Reference free 2D classification into 300 classes for 75 iterations was performed followed by manual examination of the resulting class averages. Low resolution/signal to noise/feature class averages and contributing particles were discarded, resulting in 670000 particles left. The resulting particles were 3D classified into 4 classes resulting in two classes having high-resolution features (390000 particles). At this stage particles were extracted from 1.315A/pix micrographs and all the following processing was done with these particles. Using 3D Auto-refine in Relion 1.4, a reconstruction was obtained from 390000 particles resulting from 3D classification above (using highest resolution 3D class as initial model, low pass filtered to 20A). Using the resulting parameters, the particles were further drift corrected per particle and dose weighted using the Particle Polishing feature. The B-factor weighing curve was fit by a polynomial (with a rationale that such a curve should be smooth) and used to generate new weighting parameters for Particle Polishing, with which 390000 particles were then polished. All further data processing was done using the polished particles. Re-refinement of the 390000 particles after polishing yielded the map at about 4A resolution (determined using gold standard FSC in the PostProcessing tab). The 390000 particles were then 3D classified into four different classes without particle re-alignment, using the alignment parameters from 4A reconstruction and subtraction procedure described previously. Particles contributing to each of the four classes were grouped and a full 3D refinement with a spherical 200A mask was performed with each of the four groups of particles using the same initial model, low pass filtered to 20A. This is one of the resulting reconstructions. For best visualization, reconstruction was low pass filtered to 7A and sharpened with a Bfactor of -130.