The structure of an Hsp90-immunophilin complex reveals cochaperone recognition of the client maturation state.

RESOURCE AVAILABILITY

Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Daniel R. Sourthworth ( daniel.southworth@ucsf.edu ).

Materials Availability

Requests for resources and reagents should be directed to Daniel R. Sourthworth ( daniel.southworth@ucsf.edu ).

Data and code availability

Cryo-EM densities have been deposited at the Electron Microscopy Data Bank under accession codes EMD: 23213 (Hsp90:FKBP51:p23 closed-state complex) and EMD: 23214 (Hsp90:p23 closed-state complex). Atomic coordinates have been deposited at the Protein Data Bank under accession codes PDB: 7L7I (Hsp90:FKBP51:p23 closed-state complex) and PDB: 7L7J (Hsp90:p23 closed-state complex). This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Bacterial strains BL21-AI ™ One Shot ™ Chemically Competent

E. coli or BL21 Star ™ (DE3)pLysS One Shot ™ Chemically Competent E. coli from Invitrogen were used to express recombinant proteins. METHOD DETAILS Protein expression and purification NEB Q5 Site-directed Mutagenesis was used to introduce mutations (Deletion 401–457 (ΔH7e), N402K, M412K, and F413K) into the open reading frame of FKBP51 in pET151 vector with a 6x His-tag and TEV cleavage site (Invitrogen). Hsp90, p23, FKBP51 and its mutants were purified as previously described with minor modifications ( Southworth and Agard, 2008 , 2011 ). E. coli BL21 cells were transformed with pET151-Hsp90α, pET151-p23, and pET151-FKBP51 plasmids, respectively. The cells were grown in Terrific Broth (TB) media at 37°C and induced with 1 mM IPTG at OD 600 = 0.8 – 1.0 for overnight at 18°C after cooling on ice. The cells were lysed by either sonicator or Emulsiflex in the lysis buffer (20 mM Tris pH 8.0, 500 mM KCl, 6 mM β-mercaptoethanol, 3 mM imidazole, and 10% glycerol with EDTA-free protease inhibitor cocktail (Roche). The lysed cells were centrifuged with 40,000 rpm at 4°C for 45 min. The supernatant was incubated with HisPur Ni-NTA resin (Thermo Scientific) at 4°C for 1 hour. The resins were applied to a benchtop Ni-NTA column and bound proteins were eluted with an elution buffer (20 mM Tris pH 8.0, 100 mM KCl, 6 mM β-mercaptoethanol, and 300 mM imidazole) by gravity force. The eluted proteins were applied to ion exchange column (Mono Q 5/50 GL, GE Healthcare) and eluted with a gradient of ion exchange buffer A and B (A: 20 mM Tris pH 8.0, 20 mM KCl, and 6 mM β-mercaptoethanol; B: 20 mM Tris pH 8.0, 500 mM KCl, and 6 mM β-mercaptoethanol). The fraction containing proteins were dialyzed with a dialysis buffer (20 mM Tris pH 8.0, 100 mM KCl, 6 mM BME, and 10% glycerol) containing TEV protease to cleave the His tag at 4°C overnight. Dialyzed samples were applied to a size exclusion column (HiLoad Superdex-200 16/600 column, GE Healthcare) with a gel filtration buffer (20 mM Tris pH 7.5, 250 mM KCl, 6 mM β-mercaptoethanol). Protein fractions were collected and dialyzed with a storage buffer (20 mM HEPES pH 7.5, 50 mM KCl, 6 mM β-mercaptoethanol, and 10 % glycerol) at 4°C overnight. Dialyzed proteins were concentrated and frozen with liquid nitrogen for storage at −80°C. Purity of proteins was verified by SDS-PAGE.

Native gel electrophoresis

The formation of open (apo) and closed (ATP-bound) states of Hsp90 in different conditions (salt, temperature, and nucleotide as described in Figure 1A , S1A and S1B ) were tested using native gel electrophoresis. 2 uM of Hsp90 was incubated with a buffer (20 mM HEPES pH 7.5, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in addition to the conditions described for 2 hours. Hsp90 was then mixed with 2x Native PAGE sample buffer (62.5 mM Tris-HCl, pH 6.8, 40% glycerol, and 0.01% bromophenol blue) and run on a native gel (handcasting Tris polyacrylamide gel without SDS) in a Native PAGE buffer (25 mM Tris pH 8.3 and 0.192 M glycine). The gel was stained with Coomassie Blue (Bio-Rad). SEC-MALS and size exclusion chromatography 20 uM of Hsp90, FKBP51, and p23 were incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the absence or presence of 2 mM AMPPNP at 37°C for 2 hours before injection into a size exclusion column (Shodex KW-804) connected to an in-line DAWN HELEOS MALS and Optilab rEX differential refractive index detectors (Wyatt Technology Corporation) with the running buffer (20 mM HEPES pH 7.5, 150 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 4°C. Molecular weight of proteins and protein complexes were analyzed by the ASTRA V software package (Wyatt Technology Corporation). Superose 6 Increase 3.2/300 column (GE Healthcare Life Sciences) with the running buffer (20 mM HEPES pH 7.5, 150 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 4°C was used to analyze elution fractions by SDS-PAGE.

Fluorescence polarization assays

For the competition experiments, 40 nM of the C-terminal Hsp90α peptide tracer (FAM-DDTSRMEEVD) ( Assimon et al., 2015 ) was incubated with 5 μM of FKBP51 wildtype or ΔH7e in the tracer buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , and 0.002% Triton X-100). Hsp90 was incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the presence or absence of 2 mM AMPPNP at 37C for 2 hours to form the closed or open state, respectively. Then, two-fold serial dilution of Hsp90 was prepared in the protein buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , and 2 mM β-mercaptoethanol). 9 μl of Hsp90 serial dilution was mixed with 9 μl of the FKBP51 and tracer solution, and then the mixture was incubated at room temperature for 15 min in the dark before fluorescence polarization measurements were taken on a SpectraMax M5 plate reader. Raw polarization values (mP) were background subtracted, normalized to no-competitor control, and plotted relative to log 10 (Hsp90) (M). Curves were fit in GraphPad Prism 9.0 to the model [inhibitor] versus normalized response. IC 50 values were calculated based on following equation: Y = 100 / ( 1 + X / IC 50 ) K i values were calculated as previously described ( Nikolovska-Coleska et al., 2004 ) using the equation: K i = [ I ] 50 / ( [ L ] 50 / K d + [ P ] 0 / K d + 1 ) To determine K d of FKBP51 wildtype and ΔH7e for the Hsp90 tracer, two-fold serial dilution of FKBP51 wildtype or ΔH7e from 20 μM to 0.625 μM were incubated with 20 nM of the Hsp90 tracer in the fluorescent polarization buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , 0.001% Triton X-100, and 1 mM β-mercaptoethanol) for 15 min at room temperature in the dark. Raw polarization values (mP) were background subtracted and plotted relative to [FKBP51] (μM). Curves were fit in GraphPad Prism 9.0 using the model [agonist] versus response. K d was extrapolated from half-maximal effective concentration (EC 50 ) using the equation: Y =Bottom+ ( Top − Bottom ) / ( 1+10 ( ( logEC 50 − X ) × HillSlope ) ) Bpa (p-bnezoyl-L-phenylalanine) crosslinking E. coli cells were transformed with plasmids containing a TAG stop codon at specific sites (E23, S27, K29, D72, N74, and E75) of FKBP51 to incorporate Bpa by nonsense-suppression during translation in addition to Bpa-tRNA vector (pSUPT/BpF), encoding tRNA and tRNA synthetase, as describe previously ( Lennon et al., 2012 ; Ryu and Schultz, 2006 ). The cells containing the plasmids were grown in the TB media supplemented with 1 mM of Bpa at 37°C until OD 600 reached about 0.8–1. The cells were induced with 1mM of IPTG and 200 mg/L arabinose overnight at 18°C. The cells were lysed with sonicator with lysis buffer (20 mM HEPES pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 6 mM β-mercaptoethanol, and 10 mM imidazole) and incubated with HisPur Ni-NTA resin (Thermo Scientific) for 1 hour at 4°C. Bound proteins were eluted with elution buffer (20 mM HEPES pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 6 mM β-mercaptoethanol, and 300 mM imidazole) and used for crosslinking. The open and closed states of Hsp90 were formed by incubation with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the absence or presence of 2 mM AMPPNP at 37°C for 2 hours. FKBP51 were mixed with Hsp90 in the open and closed states, respectively, and exposed to UV lights for 1 hours. Crosslinking was analyzed by SDS-PAGE.

Electron Microscopy

(EM) data collection and processing For negative-stain EM Hsp90 was incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the presence or absence of 2 mM AMPPNP at 37°C to form the open and closed states, respectively. Hsp90 sampled at different incubation time points (0.5, 1, and 2 hours) was stained with uranyl formate for 30 seconds on carbon coated 400 mesh copper grids. The samples were imaged using a FEI Tecnai 10 operated at 100 keV. Micrograph images were recorded using a 4k × 4k CCD camera (Gatan). For cryo-EM Hsp90α, FKBP51, and p23 were incubated with binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 37°C for 2 hours. After incubation the sample was diluted and crosslinked with 0.01% glutaraldehyde for 30 min in 100 mM KCl. The crosslinked sample was applied on a glow-discharged holey carbon grid (R 1.2/1.3, Quantifoil) after quenching with 20 mM Tris (pH 7.5) and blotted with Whatman filter paper before plunge freezing in liquid ethane using a vitrobot (ThermoFisher Scientific). The sample was then imaged on a Titan Krios TEM (ThermoFisher Scientific) operated at 300 keV equipped with a Gatan BioQuantum imaging energy filter using a 20-eV energy slit (Gatan). Micrograph images were recorded on a K2 Summit and K3 direct electron detector (Gatan) at a magnification of x61,425 and x58,600 which corresponds to a physical pixel sizes of 0.814 and 0.834 Å/pixel, respectively. Serial-EM was used to collect data with a defocus range of −2.0 to −1.0 um at a dose rate of 6 (K2) and 8 (K3) e − /pix/s for total 8 and 6 seconds exposures at 0.05 s/frame in 120 frames resulting in total doses of 70 and 66 e − /Å 2 , respectively. Movies of images were corrected for drift using MotionCor2 ( Zheng et al., 2017 ) and were Fourier cropped by a factor of 2. Micrographs were rescaled to be combined with other datasets using relion_image_handler as described previously ( Scheres, 2012 ; Wilkinson et al., 2019 ). A total of ~22,000 micrographs were collected from three different dataset. Individual micrographs were manually curated using Scipion ( de la Rosa-Trevin et al., 2016 ) to exclude poor quality micrographs and CTF was estimated using Patch CTF in cryoSPARC2 ( Punjani et al., 2017 ). cryoSPARC2 v3.1.0 with standard parameters was used throughout the data processing. About 100 particles were picked to generate templates which were then used for automated particle picking for each dataset. The first round of reference-free 2D classification was conducted with the particles which were autopicked and extracted from each dataset separately. The 2D class averages with high resolution features were selected, removing noisy or indiscernible class averages, and combined with the other data sets for the second round of reference-free 2D classification. Three ab-initio models were generated with selected particles (~576,000 from about 3.6 million total picked particles) and used for subsequent 3D classification. For Hsp90:FKBP51:p23, initial 3D heterogeneous refinement was performed with three different classes ( Figure S3B ). Two classes (total of ~337,000 particles) containing distinct density for FKBP51 were combined for another round of 3D heterogeneous refinement with four different classes. Finally, non-uniform refinement was performed for a class (~122,000 particles) containing discrete density for FKPB51 and p23. A class (~295,000 particles) only containing p23 after the first round of 3D classification was used for non-uniform refinement to obtain the Hsp90:p23 reconstruction. The final resolutions for Hsp90:FKBP51:p23 and Hsp90:p23 were 3.3Å and 3.1Å, respectively as determined by the “Gold Standard” Fourier Shell Correlation method ( Henderson et al., 2012 ).

Molecular modeling

To build a model structure of the Hsp90:FKBP51:p23 complex, initial model of Hsp90α was generated by SWISS-MODEL ( Waterhouse et al., 2018 ) using Hsp90β structure (PDB: 5FWK) ( Verba et al., 2016 ) as a template. Hsp90, FKBP51 (PDB: 5NJX) ( Kumar et al., 2017 ), and p23 (PDB: 1EJF) ( Weaver et al., 2000 ) were docked into the Hsp90:FKBP51:p23 density using the UCSF Chimera ( Pettersen et al., 2004 ) fit-in-map function. AMPPNP were generated by Phenix eLBOW ( Afonine et al., 2018 ; Liebschner et al., 2019 ) and placed into the NBD of Hsp90 based on the crystal structures of mitochondrial Hsp90 in the closed state (PDB: 4IPE) ( Lavery et al., 2014 ). Initially Rosetta comparative modeling (RosettaCM) ( Song et al., 2013 ) was used to model Hsp90, FKBP51, and p23 according to the density. Homology model structures of Hsp90 (PDB: 5FWK, 2CG9, 2O1U, 4IPE, 2IOQ, and 4IVG), FKBP51 (PDB: 5OMP, 1P5Q, 2IF4, 4LAY, 5MGX, and 3O5D), and p23 (PDB: 1EJF, 2KMW, 2CG9, 1RL1, 2O30, and 2XCM) were determined by HHpred respectively and used to constrain model refinement in RosettaCM with template_weitght=0 and the initial model with template_weight=1. The lowest energy models were examined, and the models that best fit into the density were chosen. Rosetta enumerative sampling (RosettaES) ( Frenz et al., 2017 ) was used to build some regions of Hsp90 de novo. The Hsp90 residues 122–133, 175–183, 345–360, and 617–622 were deleted and RosettaES was run with beam width of 32, respectively. The resulting models of Hps90, FKBP51, and p23 were refined by Rosetta Relax with the Hsp90:FKBP51:p23 and Hsp90:p23 density maps, respectively. The final models were manually inspected with minor modification using ISOLDE ( Croll, 2018 ) in ChimeraX ( Pettersen et al., 2021 ) and Coot ( Casanal et al., 2020 ).

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification and statistical analysis are described in the figure legends and METHOD DETAILS when applicable. Error bars represent standard deviation. * p < 0.5, ns = no significance.

Materials Availability

Requests for resources and reagents should be directed to Daniel R. Sourthworth ( daniel.southworth@ucsf.edu ).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Bacterial strains BL21-AI ™ One Shot ™ Chemically Competent

E. coli or BL21 Star ™ (DE3)pLysS One Shot ™ Chemically Competent E. coli from Invitrogen were used to express recombinant proteins.

METHOD DETAILS Protein expression and purification NEB Q5 Site-directed Mutagenesis was used to introduce mutations (Deletion 401–457 (ΔH7e), N402K, M412K, and F413K) into the open reading frame of FKBP51 in pET151 vector with a 6x His-tag and TEV cleavage site (Invitrogen). Hsp90, p23, FKBP51 and its mutants were purified as previously described with minor modifications ( Southworth and Agard, 2008 , 2011 ). E. coli BL21 cells were transformed with pET151-Hsp90α, pET151-p23, and pET151-FKBP51 plasmids, respectively. The cells were grown in Terrific Broth (TB) media at 37°C and induced with 1 mM IPTG at OD 600 = 0.8 – 1.0 for overnight at 18°C after cooling on ice. The cells were lysed by either sonicator or Emulsiflex in the lysis buffer (20 mM Tris pH 8.0, 500 mM KCl, 6 mM β-mercaptoethanol, 3 mM imidazole, and 10% glycerol with EDTA-free protease inhibitor cocktail (Roche). The lysed cells were centrifuged with 40,000 rpm at 4°C for 45 min. The supernatant was incubated with HisPur Ni-NTA resin (Thermo Scientific) at 4°C for 1 hour. The resins were applied to a benchtop Ni-NTA column and bound proteins were eluted with an elution buffer (20 mM Tris pH 8.0, 100 mM KCl, 6 mM β-mercaptoethanol, and 300 mM imidazole) by gravity force. The eluted proteins were applied to ion exchange column (Mono Q 5/50 GL, GE Healthcare) and eluted with a gradient of ion exchange buffer A and B (A: 20 mM Tris pH 8.0, 20 mM KCl, and 6 mM β-mercaptoethanol; B: 20 mM Tris pH 8.0, 500 mM KCl, and 6 mM β-mercaptoethanol). The fraction containing proteins were dialyzed with a dialysis buffer (20 mM Tris pH 8.0, 100 mM KCl, 6 mM BME, and 10% glycerol) containing TEV protease to cleave the His tag at 4°C overnight. Dialyzed samples were applied to a size exclusion column (HiLoad Superdex-200 16/600 column, GE Healthcare) with a gel filtration buffer (20 mM Tris pH 7.5, 250 mM KCl, 6 mM β-mercaptoethanol). Protein fractions were collected and dialyzed with a storage buffer (20 mM HEPES pH 7.5, 50 mM KCl, 6 mM β-mercaptoethanol, and 10 % glycerol) at 4°C overnight. Dialyzed proteins were concentrated and frozen with liquid nitrogen for storage at −80°C. Purity of proteins was verified by SDS-PAGE.

Native gel electrophoresis

The formation of open (apo) and closed (ATP-bound) states of Hsp90 in different conditions (salt, temperature, and nucleotide as described in Figure 1A , S1A and S1B ) were tested using native gel electrophoresis. 2 uM of Hsp90 was incubated with a buffer (20 mM HEPES pH 7.5, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in addition to the conditions described for 2 hours. Hsp90 was then mixed with 2x Native PAGE sample buffer (62.5 mM Tris-HCl, pH 6.8, 40% glycerol, and 0.01% bromophenol blue) and run on a native gel (handcasting Tris polyacrylamide gel without SDS) in a Native PAGE buffer (25 mM Tris pH 8.3 and 0.192 M glycine). The gel was stained with Coomassie Blue (Bio-Rad). SEC-MALS and size exclusion chromatography 20 uM of Hsp90, FKBP51, and p23 were incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the absence or presence of 2 mM AMPPNP at 37°C for 2 hours before injection into a size exclusion column (Shodex KW-804) connected to an in-line DAWN HELEOS MALS and Optilab rEX differential refractive index detectors (Wyatt Technology Corporation) with the running buffer (20 mM HEPES pH 7.5, 150 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 4°C. Molecular weight of proteins and protein complexes were analyzed by the ASTRA V software package (Wyatt Technology Corporation). Superose 6 Increase 3.2/300 column (GE Healthcare Life Sciences) with the running buffer (20 mM HEPES pH 7.5, 150 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 4°C was used to analyze elution fractions by SDS-PAGE.

Fluorescence polarization assays

For the competition experiments, 40 nM of the C-terminal Hsp90α peptide tracer (FAM-DDTSRMEEVD) ( Assimon et al., 2015 ) was incubated with 5 μM of FKBP51 wildtype or ΔH7e in the tracer buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , and 0.002% Triton X-100). Hsp90 was incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the presence or absence of 2 mM AMPPNP at 37C for 2 hours to form the closed or open state, respectively. Then, two-fold serial dilution of Hsp90 was prepared in the protein buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , and 2 mM β-mercaptoethanol). 9 μl of Hsp90 serial dilution was mixed with 9 μl of the FKBP51 and tracer solution, and then the mixture was incubated at room temperature for 15 min in the dark before fluorescence polarization measurements were taken on a SpectraMax M5 plate reader. Raw polarization values (mP) were background subtracted, normalized to no-competitor control, and plotted relative to log 10 (Hsp90) (M). Curves were fit in GraphPad Prism 9.0 to the model [inhibitor] versus normalized response. IC 50 values were calculated based on following equation: Y = 100 / ( 1 + X / IC 50 ) K i values were calculated as previously described ( Nikolovska-Coleska et al., 2004 ) using the equation: K i = [ I ] 50 / ( [ L ] 50 / K d + [ P ] 0 / K d + 1 ) To determine K d of FKBP51 wildtype and ΔH7e for the Hsp90 tracer, two-fold serial dilution of FKBP51 wildtype or ΔH7e from 20 μM to 0.625 μM were incubated with 20 nM of the Hsp90 tracer in the fluorescent polarization buffer (20 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl 2 , 0.001% Triton X-100, and 1 mM β-mercaptoethanol) for 15 min at room temperature in the dark. Raw polarization values (mP) were background subtracted and plotted relative to [FKBP51] (μM). Curves were fit in GraphPad Prism 9.0 using the model [agonist] versus response. K d was extrapolated from half-maximal effective concentration (EC 50 ) using the equation: Y =Bottom+ ( Top − Bottom ) / ( 1+10 ( ( logEC 50 − X ) × HillSlope ) ) Bpa (p-bnezoyl-L-phenylalanine) crosslinking E. coli cells were transformed with plasmids containing a TAG stop codon at specific sites (E23, S27, K29, D72, N74, and E75) of FKBP51 to incorporate Bpa by nonsense-suppression during translation in addition to Bpa-tRNA vector (pSUPT/BpF), encoding tRNA and tRNA synthetase, as describe previously ( Lennon et al., 2012 ; Ryu and Schultz, 2006 ). The cells containing the plasmids were grown in the TB media supplemented with 1 mM of Bpa at 37°C until OD 600 reached about 0.8–1. The cells were induced with 1mM of IPTG and 200 mg/L arabinose overnight at 18°C. The cells were lysed with sonicator with lysis buffer (20 mM HEPES pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 6 mM β-mercaptoethanol, and 10 mM imidazole) and incubated with HisPur Ni-NTA resin (Thermo Scientific) for 1 hour at 4°C. Bound proteins were eluted with elution buffer (20 mM HEPES pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 6 mM β-mercaptoethanol, and 300 mM imidazole) and used for crosslinking. The open and closed states of Hsp90 were formed by incubation with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the absence or presence of 2 mM AMPPNP at 37°C for 2 hours. FKBP51 were mixed with Hsp90 in the open and closed states, respectively, and exposed to UV lights for 1 hours. Crosslinking was analyzed by SDS-PAGE.

Electron Microscopy

(EM) data collection and processing For negative-stain EM Hsp90 was incubated with the binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) in the presence or absence of 2 mM AMPPNP at 37°C to form the open and closed states, respectively. Hsp90 sampled at different incubation time points (0.5, 1, and 2 hours) was stained with uranyl formate for 30 seconds on carbon coated 400 mesh copper grids. The samples were imaged using a FEI Tecnai 10 operated at 100 keV. Micrograph images were recorded using a 4k × 4k CCD camera (Gatan). For cryo-EM Hsp90α, FKBP51, and p23 were incubated with binding buffer (20 mM HEPES pH 7.5, 500 mM KCl, 5 mM MgCl 2 , and 6 mM β-mercaptoethanol) at 37°C for 2 hours. After incubation the sample was diluted and crosslinked with 0.01% glutaraldehyde for 30 min in 100 mM KCl. The crosslinked sample was applied on a glow-discharged holey carbon grid (R 1.2/1.3, Quantifoil) after quenching with 20 mM Tris (pH 7.5) and blotted with Whatman filter paper before plunge freezing in liquid ethane using a vitrobot (ThermoFisher Scientific). The sample was then imaged on a Titan Krios TEM (ThermoFisher Scientific) operated at 300 keV equipped with a Gatan BioQuantum imaging energy filter using a 20-eV energy slit (Gatan). Micrograph images were recorded on a K2 Summit and K3 direct electron detector (Gatan) at a magnification of x61,425 and x58,600 which corresponds to a physical pixel sizes of 0.814 and 0.834 Å/pixel, respectively. Serial-EM was used to collect data with a defocus range of −2.0 to −1.0 um at a dose rate of 6 (K2) and 8 (K3) e − /pix/s for total 8 and 6 seconds exposures at 0.05 s/frame in 120 frames resulting in total doses of 70 and 66 e − /Å 2 , respectively. Movies of images were corrected for drift using MotionCor2 ( Zheng et al., 2017 ) and were Fourier cropped by a factor of 2. Micrographs were rescaled to be combined with other datasets using relion_image_handler as described previously ( Scheres, 2012 ; Wilkinson et al., 2019 ). A total of ~22,000 micrographs were collected from three different dataset. Individual micrographs were manually curated using Scipion ( de la Rosa-Trevin et al., 2016 ) to exclude poor quality micrographs and CTF was estimated using Patch CTF in cryoSPARC2 ( Punjani et al., 2017 ). cryoSPARC2 v3.1.0 with standard parameters was used throughout the data processing. About 100 particles were picked to generate templates which were then used for automated particle picking for each dataset. The first round of reference-free 2D classification was conducted with the particles which were autopicked and extracted from each dataset separately. The 2D class averages with high resolution features were selected, removing noisy or indiscernible class averages, and combined with the other data sets for the second round of reference-free 2D classification. Three ab-initio models were generated with selected particles (~576,000 from about 3.6 million total picked particles) and used for subsequent 3D classification. For Hsp90:FKBP51:p23, initial 3D heterogeneous refinement was performed with three different classes ( Figure S3B ). Two classes (total of ~337,000 particles) containing distinct density for FKBP51 were combined for another round of 3D heterogeneous refinement with four different classes. Finally, non-uniform refinement was performed for a class (~122,000 particles) containing discrete density for FKPB51 and p23. A class (~295,000 particles) only containing p23 after the first round of 3D classification was used for non-uniform refinement to obtain the Hsp90:p23 reconstruction. The final resolutions for Hsp90:FKBP51:p23 and Hsp90:p23 were 3.3Å and 3.1Å, respectively as determined by the “Gold Standard” Fourier Shell Correlation method ( Henderson et al., 2012 ).

Molecular modeling

To build a model structure of the Hsp90:FKBP51:p23 complex, initial model of Hsp90α was generated by SWISS-MODEL ( Waterhouse et al., 2018 ) using Hsp90β structure (PDB: 5FWK) ( Verba et al., 2016 ) as a template. Hsp90, FKBP51 (PDB: 5NJX) ( Kumar et al., 2017 ), and p23 (PDB: 1EJF) ( Weaver et al., 2000 ) were docked into the Hsp90:FKBP51:p23 density using the UCSF Chimera ( Pettersen et al., 2004 ) fit-in-map function. AMPPNP were generated by Phenix eLBOW ( Afonine et al., 2018 ; Liebschner et al., 2019 ) and placed into the NBD of Hsp90 based on the crystal structures of mitochondrial Hsp90 in the closed state (PDB: 4IPE) ( Lavery et al., 2014 ). Initially Rosetta comparative modeling (RosettaCM) ( Song et al., 2013 ) was used to model Hsp90, FKBP51, and p23 according to the density. Homology model structures of Hsp90 (PDB: 5FWK, 2CG9, 2O1U, 4IPE, 2IOQ, and 4IVG), FKBP51 (PDB: 5OMP, 1P5Q, 2IF4, 4LAY, 5MGX, and 3O5D), and p23 (PDB: 1EJF, 2KMW, 2CG9, 1RL1, 2O30, and 2XCM) were determined by HHpred respectively and used to constrain model refinement in RosettaCM with template_weitght=0 and the initial model with template_weight=1. The lowest energy models were examined, and the models that best fit into the density were chosen. Rosetta enumerative sampling (RosettaES) ( Frenz et al., 2017 ) was used to build some regions of Hsp90 de novo. The Hsp90 residues 122–133, 175–183, 345–360, and 617–622 were deleted and RosettaES was run with beam width of 32, respectively. The resulting models of Hps90, FKBP51, and p23 were refined by Rosetta Relax with the Hsp90:FKBP51:p23 and Hsp90:p23 density maps, respectively. The final models were manually inspected with minor modification using ISOLDE ( Croll, 2018 ) in ChimeraX ( Pettersen et al., 2021 ) and Coot ( Casanal et al., 2020 ).

Supplementary Material 1 3 Video S1. Overall structure of Hsp90:FKBP51:p23 complex, Related to Figure 3 . 4 Video S2. Specific interaction of FKBP51 with Hsp90 in the closed state recognized by the H7e, Related to Figure 5 .