Cryo-EM reveals how Hsp90 and FKBP immunophilins co-regulate the glucocorticoid receptor

Hsp90 is an essential molecular chaperone responsible for the folding and activation of hundreds of ‘client’ proteins, including the glucocorticoid receptor (GR). Previously, we revealed that Hsp70 and Hsp90 remodel the conformation of GR to regulate ligand binding, aided by co-chaperones. In vivo, the co-chaperones FKBP51 and FKBP52 antagonistically regulate GR activity, but a molecular understanding is lacking. Here we present a 3.01 Å cryogenic electron microscopy structure of the human GR:Hsp90:FKBP52 complex, revealing how FKBP52 integrates into the GR chaperone cycle and directly binds to the active client, potentiating GR activity in vitro and in vivo. We also present a 3.23 Å cryogenic electron microscopy structure of the human GR:Hsp90:FKBP51 complex, revealing how FKBP51 competes with FKBP52 for GR:Hsp90 binding and demonstrating how FKBP51 can act as a potent antagonist to FKBP52. Altogether, we demonstrate how FKBP51 and FKBP52 integrate into the GR chaperone cycle to advance GR to the next stage of maturation.

Hsp90 is required for the functional maturation of 10% of the eukaryotic proteome 1 .Hsp90 'clients' are enriched in signaling proteins, such as steroid hormone receptors (SHRs), making Hsp90 an important clinical target [2][3][4] .SHRs, including the glucocorticoid receptor (GR), are hormone-regulated transcription factors that depend on Hsp90 for function throughout their lifetimes [5][6][7][8][9][10] .We previously established in vitro reconstitution of the 'GR chaperone cycle', revealing that GR ligand binding is inactivated by Hsp70 and reactivated by Hsp90  (ref.11).In the GR chaperone cycle, understood in atomic detail through cryogenic electron microscopy (cryo-EM), GR ligand binding is regulated by a cycle of three distinct chaperone complexes 12,13 .In this chaperone cycle, GR is first inhibited by Hsp70 and Hsp40, then loaded onto Hsp90:Hop (Hsp70/Hsp90 organizing protein co-chaperone) forming an inactive 'GR-loading complex' (GR:Hsp70:Hsp90:Hop) 12 .Upon ATP hydrolysis by Hsp90, Hsp70 and Hop release, and p23 binds to form an active 'GR-maturation complex' (GR:Hsp90:p23), restoring GR ligand binding with enhanced affinity 13 .Cryo-EM structures of the GR-loading complex and GR-maturation complex revealed that Hsp70 and Hsp90 locally unfold and refold the GR ligand binding domain (GR LBD ) to directly regulate ligand binding.
In vivo, additional Hsp90 co-chaperones are found associated with the GR chaperone cycle 6,14 , including the large immunophilins, FKBP51 and FKBP52 (ref.6).FKBP51 and FKBP52 are peptidyl proline isomerases (PPIases) that contain an N-terminal FK1 domain with PPIase activity, an enzymatically dead FK2 domain, and a C-terminal tetratricopeptide repeat (TPR) domain, which binds the EEVD motifs at the C-termini of Hsp90 and Hsp70 (refs.15-18).Additionally, the TPR domain contains a helical extension at the C-terminus (H7e), which binds the C-terminal domain (CTD) closed dimer interface of Hsp90 (refs.19,20).Although FKBP51 and FKBP52 are 70% similar in sequence, these co-chaperones have antagonistic functional effects on GR in vivo 21 .FKBP51 inhibits GR ligand binding, nuclear translocation and transcriptional activity, while FKBP52 potentiates each of these fundamental GR activities [22][23][24][25][26][27][28][29][30][31][32] .FKBP51 and FKBP52 are also known to regulate the other SHRs 21,33 .Due to the critical importance of steroid hormone signaling in the cell, altered expression of FKBP51 or FKBP52 is associated with https://doi.org/10.1038/s41594-023-01128-y was stabilized with sodium molybdate, further purified and lightly crosslinked (Extended Data Fig. 1d,e).A 3.01 Å cryo-EM reconstruction of the GR:Hsp90:FKBP52 complex was obtained (Fig. 1a,b, Table and Extended Data Figs.1f and 2a,b), revealing a fully closed Hsp90 dimer (Hsp90A and Hsp90B) with a single GR and a single FKBP52, both occupying the same side of Hsp90 (Fig. 1a,b; for a discussion of Hsp90 nucleotide state, see Extended Data Fig. 3a and Methods).Despite using a multi-domain GR construct, only GR LBD was visible on the map.
In interface 2, Hsp90B MEEVD binds in the FKBP52 TPR helical bundle (approximately 779 Å 2 BSA) (Fig. 1g and Extended Data Fig. 4b), with multiple hydrogen bonds, salt bridges and hydrophobic interactions, analogous to FKBP51:Hsp90 MEEVD structures 18,19 .However, the MEEVD peptide binds in an opposite orientation relative to the FKBP52:Hsp90 MEEVD crystal structure 17 , which may have been incorrectly modeled, as suggested 18,43 .Interface 3 is composed of FKBP52 Helix5/6 in the TPR domain binding to Hsp90B CTD , stabilized by multiple hydrogen bonds (approximately 193 Å 2 BSA) (Extended Data Fig. 4c), also observed in the FKBP51:Hsp90 cryo-EM structure 19 .While the interactions between FKBP52 TPR/H7e :Hsp90 are conserved in the FKBP51:Hsp90 structure, the positions of the FKBP52 FK1 and FK2 domains are notably altered (Extended Data Fig. 4d), owing to the presence of GR, as discussed below.

FKBP52 directly binds GR, which is critical for GR function
Unexpectedly, FKBP52 directly and extensively interacts with GR, with all three FKBP52 domains wrapping around GR, cradling the folded, ligand-bound receptor near the GR ligand-binding pocket (Fig. 2a).The tertiary structure within each FKBP52 domain closely matches isolated domains from FKBP52 crystal structures; however, the interdomain angles are significantly different (Extended Data Fig. 4d), probably owing to the extensive interaction with GR.There are three major interfaces between FKBP52 and GR (Fig. 2b-d): (1) FKBP52 FK1 :GR, (2) FKBP52 FK2 :GR and (3) FKBP52 FK2/TPR-linker :GR Helix12 .
In interface 1, FKBP52 FK1 interacts with a large surface on GR, canonically used for GR dimer formation, consisting of the post-Helix 1 strand (Helix 1-3 loop), Helix 5 and β1,2 (approximately 280 Å 2 BSA) (Fig. 2b).Three-dimensional variability analysis in CryoSparc revealed that the interaction between FKBP52 FK1 and GR is highly dynamic, even as the other FKBP52 domains (FK2 and TPR) remain stably associated with GR (Supplementary Movies 1 and 2).At the FK1:GR interface, GR Y545 on the post-Helix 1 strand interacts with a hydrophobic surface formed by the FKBP52 [81][82][83][84][85][86][87][88] loop and forms a hydrogen bond with FKBP52 Y113 .Supporting this interaction, residues of GR post-Helix1 (GR [544][545][546] ) have previously been implicated in FKBP51/52-dependent regulation of GR activity 44,45 .In addition, the FKBP52 proline-rich loop (β4-β5 loop or 80s loop) contacts GR Helix5/β1,2 .Three-dimensional variability analysis in CryoSparc revealed that the proline-rich loop positioning is flexible, deviating from the position in the crystal structure (Protein Data Bank (PDB) ID: 4LAV) (ref.46) and adopting different interfaces with GR (Supplementary Movies 3 and 4).In the consensus 3D refinement map, the proline-rich loop adopts a position similar to the crystal structure, and FKBP52 A116,S118,P119 interact with GR Helix5/β1,2 .The FKBP52 P119L mutation has been shown to reduce GR and AR activation in vivo, while FKBP52 A116V has been shown to increase AR activation in vivo 29 .We also demonstrate that the FKBP52 S118A mutation significantly reduces FKBP52-dependent GR potentiation in vivo (Fig. 2e), further demonstrating the functional significance of this interaction site.In addition, S118 has been identified as a phosphorylation site on FKBP52, but not FKBP51 (qPTM database 47 ) (possibly due to the unique adjacent proline, FKBP52 P119 , which could recruit proline-directed kinases).Phosphorylation at FKBP52 S118 may help promote the interaction between the proline-rich loop and GR, which could also explain the large effect of the FKBP52 S118A mutation in vivo.While FKBP52 FK1 is known to have PPIase enzymatic activity, GR is not bound in the PPIase active site and, accordingly, no GR prolines were found to have been isomerized compared to other GR structures (PDB IDs: 1M2Z (ref.48) and 7KRJ (ref.13)).Consistent with this, mutation of GR prolines does not disrupt FKBP52-dependent regulation of GR 45 .Additionally, mutations that disrupt PPIase activity do not affect FKBP52-dependent GR potentiation in vivo 29 .Conversely, PPIase inhibitors have been shown to block the FKBP52-dependent potentiation of GR in vivo 23 .This can now be understood, as docking of PPIase inhibitors (FK506 and rapamycin) into the PPIase active site demonstrates that the inhibitors would sterically block the FKBP52 FK1 :GR interface (Extended Data Fig. 4e), as previously hypothesized 23,29 .
Interface 2 is composed of the FKBP52 FK2 Y161 sticking into a shallow hydrophobic pocket formed by GR Helix3 , GR Helix11-12 loop (GR T561, M565,E748 ), and a hydrogen bond between the FKBP52 backbone and GR E748 (approximately 125 Å 2 BSA) (Fig. 2c).Supporting this interaction, we show that the FKBP52 Y161D mutation significantly reduces FKBP52-dependent GR potentiation in vivo, demonstrating the importance of this interaction (Fig. 2e).In interface 3, the solvent exposed, conserved FKBP52 W259 on the FK2-TPR linker makes electrostatic and hydrophobic interactions with GR Helix12 (approximately 235 Å 2 BSA) (Fig. 2d), which adopts the canonical agonist-bound position even in the absence of a stabilizing co-activator peptide interaction 48 (Extended Data Fig. 3b).We show that the corresponding FKBP52 W259D mutation significantly reduces FKBP52-dependent GR potentiation in vivo, demonstrating the functional importance of this single residue (Fig. 2e).Interestingly, FKBP52 W259 is also conserved in the FKBP-like co-chaperone XAP2 and a recent structure reveals XAP2 engages with an Hsp90-client using the analogous XAP2 W168 , suggesting this residue is critical more broadly for FKBP co-chaperone:client engagement 42 .At interface 3, FKBP52 K254,E257,Y302,Y303 make further polar interactions between the FK2-TPR linker and GR Helix12 (Fig. 2d).While a significant portion of the GR Helix12 co-activator binding site is available in the FKBP52bound GR, the N-terminus of a co-activator peptide would sterically clash with FKBP52 TPR based on the GR:co-activator peptide structure 48 (Extended Data Fig. 5b).Thus, co-activator binding in the nucleus could help release GR from its complex with Hsp90:FKBP52.We also find that the residues at the FKBP52:GR interfaces are conserved across metazoans (Fig. 2f,g) and have been identified as sites that crosslink to GR in vivo (FKBP52 Y159, W257 ) (ref.49), in agreement with our results that single point mutations at each of the three FKBP52:GR interfaces has a significant effect on GR function in vivo.Based on the observation that FKBP52 TPR is sufficient to bind Hsp90 and Hsp90:SHR complexes 19,20 , the FKBP52 single point mutants probably do not disrupt FKBP52 binding to the GR:Hsp90 complex, but specifically disrupt the GR:FKBP52 interaction and prevent stabilization of GR LBD by FKBP52.

FKBP52 advances GR to the next stage of maturation
We previously described another GR:chaperone complex, the GRmaturation complex (GR:Hsp90:p23) (ref.13), which also contains a closed Hsp90 dimer and a folded, ligand-bound GR (Fig. 3a).However, in the GR:Hsp90:FKBP52 complex, GR is rotated by approximately 45° relative to the GR-maturation complex (Fig. 3a).Hsp90A Src-loop interacts with GR pre-Helix1 in the maturation complex, but flips out to stabilize the rotated GR position in the GR:Hsp90:FKBP52 complex by interacting with the GR hydrophobic-patch (Fig. 3b,c).In both complexes, GR pre-Helix1 is threaded through the Hsp90 lumen ; however, in the GR:Hsp90:FKBP52 complex, GR has translocated through the Hsp90 lumen by two residues, positioning prolines (GR P522,P526 ) in the hydrophobic pockets of the lumen rather than leucines (GR L525,L528 ) (Fig. 3d).This translocation positions GR LBD further from Hsp90, probably allowing enough space for the observed GR rotation.Despite the translocation and rotation of GR, Hsp90 uses the same surfaces to bind GR (Hsp90B amphi-α , Hsp90A Src-loop and Hsp90A W320 ); however, the contact surfaces on GR are different.The rotation of GR may facilitate LBD dimerization, which is on pathway to activation.In the GR-maturation complex, dimerization of the LBD clashes with Hsp90 CTD ; however, due to the rotation of the LBD in the GR:Hsp90:FKBP52 complex, LBD dimerization would now be sterically permitted after FKBP52 release (Extended Data Fig. 5c).

FKBP52 competes with p23 through allostery
Surprisingly, FKBP52 competes with p23 to bind the GR:Hsp90 complex, although there is no direct steric conflict between FKBP52 and p23 binding (Fig. 3a).During 3D classification on the cryo-EM dataset, GR:Hsp90:p23 complexes were observed at low abundance (~74,000 particles); however, the GR:Hsp90:FKBP52 complexes showed no apparent p23 density (Extended Data Fig. 2a,b), despite p23 being present at high concentration in the reconstitution.Furthermore, FKBP52 was only found associated with the rotated GR position, while the GR position in the p23-containing classes was only consistent with the GR-maturation complex.Thus, FKBP52 appears to specifically bind the rotated GR position, which is not compatible with p23 binding.This is consistent with mass spectrometry studies, demonstrating FKBP52 competes with p23 to form a stable GR:Hsp90:FKBP52 complex 50 .In the rotated GR position, Hsp90A Src-loop flips out of the Hsp90 lumen to bind the GR hydrophobic-patch , which was previously engaged by the p23 tail-helix (Fig. 3a-c).Thus, rotation of GR dictates the accessibility of the GR hydrophobic-patch to either Hsp90 or p23.FKBP52 stabilizes the rotated position of GR and therefore favors GR binding to Hsp90A Src-loop over p23.

FKBP52 potentiates GR ligand binding in vitro
To quantitatively assess the functional significance of FKBP52 on GR activation, we added FKBP52 to the in vitro reconstituted GR chaperone cycle, using the GR DBD-LBD construct (residues 418-777 containing the F602S solubilizing mutation) and monitored GR ligand binding, as previously described 11,13 .Addition of FKBP52 to the GR chaperone cycle resulted in the enhancement of GR ligand binding above the already enhanced GR + chaperones control reaction at equilibrium (Fig. 3e), strongly suggesting FKBP52 potentiates the GR ligand binding affinity beyond the minimal chaperone mixture, consistent with reports in vivo 23 .We hypothesized that FKBP52 functions in a similar manner to the p23 tail-helix in stabilizing the ligand-bound GR.As previously Article https://doi.org/10.1038/s41594-023-01128-ydescribed, removal of the p23 tail-helix (p23Δhelix) resulted in a decrease in GR ligand binding activity in the GR chaperone system 13 ; however, addition of FKBP52 to the reaction fully rescued GR ligand binding in the p23Δhelix background (Fig. 3e).Thus, FKBP52 can functionally replace the p23 tail-helix , probably by also directly stabilizing the ligand-bound GR.Given that FKBP52 and the p23 tail-helix bind different locations on GR, the mechanisms of stabilization may be unique.Additionally, in the p23Δhelix background, FKBP52 potentiated ligand binding to a greater extent than in the wild-type p23 background.We hypothesize that removing p23 tail-helix alleviates the competition between p23 and FKBP52, allowing p23 to remain bound to the GR:Hsp90:FKBP52 complex.Given that p23 is known to stabilize the closed Hsp90 conformation 13,51 , the enhanced ligand binding in the p23Δhelix background may be due to stabilization of closed Hsp90 by p23Δhelix.Interestingly, FKBP52 also affected GR ligand binding independent of Hsp90, with addition of FKBP52 to GR resulting in enhanced ligand binding, probably due to an Hsp90-independent chaperoning effect 15,52 (Extended Data Fig. 5d).

Upon Hsp90 closure, FKBP52 can functionally replace p23
Given that FKBP52 can functionally replace p23 tail-helix , we wondered whether FKBP52 could also functionally replace p23 altogether.p23 is known to stabilize Hsp90 NTD closure through the globular p23 domain 13,51 in addition to stabilizing the ligand-bound GR through the p23 tail-helix 13 .Omitting p23 from the GR chaperone cycle drastically reduces GR ligand binding, as previously described 11,13 .The addition of FKBP52 in place of p23 results in a small increase in ligand binding but does not fully rescue ligand binding activity (Fig. 3f).We reasoned this

Article
https://doi.org/10.1038/s41594-023-01128-ycould be due to the inability of FKBP52 to sufficiently stabilize Hsp90 closure, as previously suggested 53 .Therefore, we added molybdate to these reactions, which stabilizes Hsp90 NTD closure by acting as a γ-phosphate analog in the Hsp90 NTD ATP-binding site 13,54 .Addition of molybdate to the reaction lacking p23 resulted in a small increase in GR ligand binding but did not fully rescue ligand binding activity.However, addition of molybdate to the reactions containing FKBP52 without p23 resulted in a full reactivation of ligand binding and even potentiated ligand binding over the control GR + chaperones reaction (Fig. 3f), much like with p23Δhelix.Thus, FKBP52 is able to functionally replace p23 if Hsp90 NTD closure is stabilized.Taken together, these results suggest FKBP52 can stabilize the ligandbound GR, like p23, but cannot stabilize the closed Hsp90 NTD conformation, which requires p23.

GR:Hsp90:FKBP51 structure determination
In vivo, the interplay between FKBP52 and the highly similar FKBP51 have profound implications for GR activity.FKBP51 is functionally antagonistic to FKBP52-dependent potentiation of GR in vivo; thus, the relative ratios of FKBP51 and FKBP52 dictate GR activity levels 23,28,55 .To understand how FKBP51 antagonizes FKBP52, we prepared the human GR:Hsp90:FKBP51 complex by in vitro reconstitution of the GR chaperone cycle with FKBP51 (Extended Data Fig. 1d-f).
We obtained a 3.23 Å cryo-EM reconstruction of GR:Hsp90:FKBP51 (Fig. 4a,b, Table 1 and Extended Data Fig. 6a,b).Contrary to our expectations, the FKBP51-containing structure appears nearly identical to the FKBP52-containing structure.The GR:Hsp90:FKBP51 structure reveals a fully closed Hsp90 dimer complexed with a single GR and a single FKBP51, which occupy the same side of Hsp90 (Fig. 4a,b; for a discussion of Hsp90 nucleotide, see Extended Data Fig. 7a and Methods).As with the FKBP52 dataset, GR:Hsp90:p23 complexes were also observed during 3D classification and the GR:Hsp90:FKBP51 complexes showed no apparent p23 density (Extended Data Fig. 6a,b).The FKBP51:Hsp90 interactions are analogous to the FKBP52:Hsp90 interactions, including Hsp90B MEEVD :FKBP TPR and Hsp90 CTD :FKBP H7e , also seen in the Hsp90:FKBP51:p23 structure 19 , but distinct from a previous nuclear magnetic resonance model 56 (Fig. 4b and Extended Data Fig. 7b-d).The GR:Hsp90 interfaces are nearly identical when comparing the FKBP51 and FKBP52-containing complexes, including Hsp90 Src-loop :GR hydrophobic-patch and Hsp90 lumen :GR pre-Helix1 (Fig. 4b and Extended Data Fig. 7e-g).FKBP51 also directly binds GR in an analogous manner to FKBP52 (Fig. 4c-e).FKBP51 binds the folded, ligand-bound, rotated GR using the same three major interfaces (1) FKBP51 FK1 :GR, (2) FKBP51 FK2 :GR Helix3 and (3) FKBP51 FK2/TPR-linker :GR Helix12 .The GR:FKBP52 interaction residues are largely conserved for GR:FKBP51 (Fig. 2f).As with the FKBP52-containing structure, no GR prolines appear to be isomerized and the PPIase inhibitors rapamycin and FK506 sterically clash with the GR backbone.Interestingly, the small FKBP51-specific inhibitor, SAFit2 (PDB ID: 6TXX) 35,57 , does not clash with the GR backbone and may be accommodated with only side chain rotations, consistent with in vivo data 49 (Extended Data Fig. 7h).Furthermore, the FKBP51 FK1 domain and FK1 proline-rich loop are highly dynamic, as revealed by CryoSparc 3D variability analysis, analogous to the FKBP52-containing structure (Supplementary Movies 5 and 6).However, in the GR:Hsp90:FKBP51 complex, the FK1 domain contacts GR at a different angle relative to the GR:Hsp90:FKBP52 complex.Thus, the FK1:GR interface is distinct between the two complexes, specifically at the functionally important, but divergent, residue 119 in the proline-rich loop (FKBP51 L119 and FKBP52 P119 ) (Figs. 2b and 4c) 29 , which we investigated further below.

FKBP51/52 functional difference is dependent on residue 119
To quantitatively assess the functional effect of FKBP51 on GR in vitro, we added FKBP51 to the GR chaperone cycle and measured ligand binding activity.FKBP51 had no effect on the GR equilibrium value (Extended Data Fig. 8a), unlike FKBP52 which potentiated GR ligand binding.However, we found FKBP51 can functionally replace p23 tail-helix (or p23, if molybdate is added), just as we observed with FKBP52 (Fig. 4f and Extended Data Fig. 8b).However, FKBP51 does not potentiate GR ligand binding in any of these conditions, unlike FKBP52, recapitulating in vivo findings 23,29 .
The residues responsible for the functional difference between FKBP51 and FKBP52 in vivo are in the FK1 domain proline-rich loop, specifically the divergent residue 119 (FKBP51 L119 and FKBP52 P119 ) 29 .To assess whether this residue is responsible for the functional difference between FKBP51 and FKBP52 in vitro, we swapped residue 119 (FKBP51 L119P and FKBP52 P119L ) and added these mutants to the in vitro reconstituted GR chaperone cycle.We then measured ligand binding activity in the p23Δhelix background, where the largest potentiation due to FKBP52 was observed.Surprisingly, the residue 119 swapped mutants almost fully reversed the effects of FKBP51 and FKBP52 on GR: FKBP51 L119P potentiated GR ligand binding over the GR + chaperones control reaction, while FKBP52 P119L showed significantly less potentiation of ligand binding compared to wild-type FKBP52 (Fig. 4f).These results are consistent with the effects of the FKBP51/52 residue 119 swapped mutants in vivo 29 .Thus, residue 119 on the proline-rich loop provides a critical functional difference between FKBP51 and FKBP52 on GR activity in vitro and in vivo, probably driven via the differential positioning of the loop seen in our structures.

Discussion
We present two cryo-EM structures, demonstrating how the co-chaperones FKBP51 and FKBP52 bind to an Hsp90-client complex.The 3.01 Å human GR:Hsp90:FKBP52 structure reveals that FKBP52 directly and extensively binds the client using three distinct interfaces that stabilize the folded, ligand-bound conformation of GR.We show that FKBP52 enhances GR ligand binding in vitro and that each of the three observed GR:FKBP52 interfaces is critical for FKBP52dependent potentiation in vivo.We also provide a 3.23 Å human GR:Hsp90:FKBP51 structure, unexpectedly demonstrating FKBP51 binds to the GR:Hsp90 complex similarly to FKBP52, providing a molecular explanation for the functional antagonism between FKBP51 and FKBP52.Our structures contribute to an emerging theme in which Hsp90 co-chaperones bind to distinct Hsp90 conformations, while simultaneously stabilizing specific client conformations to regulate client activity 12,13,41,42,54 .
A recent study 49 using in vivo chemical crosslinking validates our structures remarkably well, recapitulating all three major GR:FKBP51/52 contacts as well as the FKBP-mediated rotated GR position.Given that the in vivo crosslinking between GR and FKBP51/52 was performed in the absence of ligand, together our findings demonstrate that FKBP51 and FKBP52 bind apo GR LBD in a similar, if not identical manner to the ligand-bound GR LBD observed here in our structures.In vivo, ligand addition dissociates GR:Hsp90:FKBP51/52 complexes 49 probably due to the rapid ligand-dependent nuclear translocation of GR 22,58 .While our high-resolution reconstructions unambiguously contain ligand, apo GR:Hsp90:FKBP51/52 complexes probably also exist in our dataset, but are less well ordered (consistent with the GR-maturation complex 13 ).In addition, that study provided further in vivo validation of our structural models by demonstrating FK506, but not SAFit2, inhibits FKBP51-dependent regulation of GR in vivo (Extended Data Figs.4e and 7h) and that the FKBP51/52 FK1 domain is dynamically associated with GR (Supplementary Movies 1-6).Altogether, these studies complement each other extraordinarily well, demonstrating direct association of FKBP51 and FKBP52 with GR LBD in vivo and in vitro at single-residue resolution.
Unexpectedly, our structures also demonstrate that FKBP51 and FKBP52 compete with p23 to bind the GR:Hsp90 complex through an allosteric mechanism.Previous reports showed FKBP51 Article https://doi.org/10.1038/s41594-023-01128-yand p23 could simultaneously bind the closed Hsp90 in the absence of client 19 .We demonstrate that the position of the client can dictate which co-chaperone is bound, with the FKBPs and p23 binding to distinct GR surfaces accessible in distinct GR orientations.FKBP51 and FKBP52 stabilize a rotated position of GR relative to the GR-maturation complex, which may facilitate post-translational modifications, interactor binding or GR dimerization, as suggested previously 59 , raising the possibility that the FKBPs promote the next step in maturation.Although the FKBPs directly contact GR, they do not appear to isomerize GR prolines or engage GR NLS1 (nuclear localization signal 1) (GR 467-505 ) (ref.60) to regulate GR activity, as previously hypothesized 8,[61][62][63] .
While FKBP51 binds GR:Hsp90 similarly to FKBP52, we find that, unlike FKBP52, FKBP51 does not enhance GR ligand binding in vitro, consistent with in vivo studies 23,27,29 .Importantly, we find proline 119 on FKBP52 is critical for enhancement of ligand binding in vitro, also consistent with in vivo studies 29 .Proline 119 on FKBP52 was found to decrease dynamics of the proline-rich loop (80s loop, β4-β5 loop) by nuclear magnetic resonance, relative to leucine 119 on FKBP51 (ref.64).Three-dimensional variability analysis of our structures demonstrates the FKBP51/52 proline-rich loop dynamically interacts with GR and differences in dynamics may dictate the specificity and/or stability of the interaction, leading to distinct regulation of GR activity by the FKBPs.
Based on our structures of the GR:Hsp90:FKBP51 and GR:Hsp90:FKBP52 complexes, we propose additional steps in the GR chaperone cycle accounting for FKBP51/52 incorporation and subsequent regulation of GR activity (Fig. 5).In the cytosol, GR cycles between Hsp70 and Hsp90, which locally unfold and refold GR to directly control ligand binding [11][12][13] .Once the folded GR reaches the GR-maturation complex (GR:Hsp90:p23), either FKBP51 or FKBP52 binds the complex and competes with p23 to advance GR to the next

Article
https://doi.org/10.1038/s41594-023-01128-ystage of maturation.Given that the folded GR is strongly stabilized and tightly associated with Hsp90 and the FKBPs, we suggest that it is unlikely that ligand binding/unbinding happens in the context of FKBP-bound complexes.Instead, we propose that ligand binds before the formation of either the GR-maturation complex or the GR:Hsp90:FKBP complexes, and that unbinding mostly occurs by recycling GR back to Hsp70, as previously described [11][12][13] .The functional outcome for GR is dictated by which FKBP binds.FKBP52 stabilizes ligand-bound GR, resulting in enhanced ligand affinity, and facilitates rapid GR nuclear translocation on dynein 22,24,25,65 , allowing GR to proceed with dimerization and transcription activation.In contrast, FKBP51 keeps GR sequestered in the cytosol, enabling GR to recycle back into the chaperone cycle, inhibiting GR translocation and transcription activation.Interestingly, the expression of FKBP51, but not FKBP52, is upregulated by GR (also PR and AR), leading to a short negative feedback loop, which may help dampen chronic GR activation and signaling 27,[66][67][68][69][70] .Thus, the relative concentrations of FKBP51 and FKBP52 in the cell dictate the level of GR activity in vivo 23,28,55 .
Beyond GR, FKBP51/52 are known to regulate the entire SHR class.Given the sequence and structural conservation of SHR LBD at the FKBP binding sites, we propose FKBP51/52 engage with all SHRs in a similar manner (Extended Data Fig. 9a,b).Thus, FKBP51/52 can fine-tune the activity of these critical and clinically important signaling molecules and allow for crosstalk between the hormone signaling pathways.Altogether, we demonstrate how Hsp90 provides a platform for the FKBP co-chaperones to engage Hsp90 clients after folding and promote the next step of client maturation, providing a critical layer of functional regulation.(orange) binds GR and locally unfolds GR to inhibit cortisol binding, stabilizing GR in a partially unfolded, apo state.Hsp70 transfers the partially unfolded GR to Hsp90 (light and dark blue):Hop (pink) to form the GR-loading complex 12 , in which GR is stabilized in a partially unfolded, apo state.Cortisol (pink), which enters the cell through diffusion, binds to GR during the transition from the GR-loading complex to the GR-maturation complex when Hsp90 refolds the GR to a native conformation 13 .In the GR-maturation complex, the cortisol-bound, folded GR is stabilized by Hsp90 and p23 (green) and is protected from Hsp70 re-binding.Depending on the relative concentrations of the FKBPs, either FKBP51 (purple) or FKBP52 (teal) can bind the GR:Hsp90:p23 complex, competing with p23, and stabilizing the rotated position of GR.FKBP51 sequesters GR:Hsp90 in the cytosol until ATP hydrolysis on Hsp90 allows release of GR back to the chaperone cycle.In contrast, FKBP52 promotes rapid nuclear translocation of GR:Hsp90 (refs.22,24,25,65).Once in the nucleus, the cortisol-bound GR can dimerize, nucleate the assembly of transcriptional regulatory complexes, and regulate transcription, including activating expression of FKBP51, leading to a negative feedback loop that regulates GR activity in the cell 27,[66][67][68][69][70] . https://doi.org/10.1038/s41594-023-01128-y
For dataset IV and GR ligand binding assays, apo GR DBD-LBD was expressed and purified as in a similar manner as described above; however, after Ni-NTA purification, the protein was dialyzed overnight in a buffer containing 30 mM Tris pH 7.5, 150 mM KCl, 10% glycerol and 2 mM DTT.Protein was then purified by hydrophobic interaction chromatography on a HiScreen Butyl-S FF (4.7 mL) column (Cytiva Life Sciences) to remove degradation products.First, solid KCl was slowly added to the protein solution at 4 °C to a final concentration of 2 M KCl.Then the protein was injected onto the hydrophobic interaction chromatography column and eluted over a gradient of 2 M-0 M KCl over 10 column volumes in a buffer containing 30 mM Tris pH 7.5, 10% glycerol and 2 mM DTT.Then the protein was further purified by size exclusion in 30 mM HEPES pH 8, 150 mM KCl, 10% glycerol and 0.5 mM tris(2-carboxyethyl) phosphine using a HiLoad 16/60 Superdex 200 (GE Healthcare).Protein was then dialyzed for 3 days, with fresh buffer each day, in a buffer containing 30 mM HEPES pH 8, 150 mM KCl, 10% glycerol and 2 mM DTT.Protein was concentrated, flash frozen and stored at −80 °C.

GR:Hsp90:FKBP complex sample preparation
The GR chaperone cycle was reconstituted in vitro with purified components as previously described 11 with addition of 15 μM FKBP51 or FKBP52.After 60 min at room temperature, another 15 μM FKBP51 or FKBP52 was added, along with 15 μM Bag-1 and 20 mM sodium molybdate.The reaction was incubated at room temperature for another 30 min and then a pulldown on MBP-GR DBD-LBD was performed as previously described.The elution was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (4-12% acrylamide gel) (Extended Data Fig. 1d).The elution was concentrated and purified by size exclusion using a Superdex 200 Increase 3.2/300 (Cytiva Life Sciences), and fractions were analyzed by SDS-PAGE (4-12% acrylamide gel) (Extended Data Fig. 1e).Fractions containing the full complex were concentrated to ~2 μM and crosslinked with 0.02% glutaraldehyde for 20 min at room temperature (Extended Data Fig. 1d).Then 2.5 μl of sample was applied to glow-discharged QUANTIFOIL R1.2/1.3, 400-mesh, copper holey carbon grid (Quantifoil Micro Tools GmbH) and plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI) with a blotting time of 12-16 s, blotting force 3, at 10 °C, and with 100% humidity.

Cryo-EM data acquisition
All data were acquired using SerialEM software v.4.0 (ref.73) and collected on an FEI Titan Krios electron microscope (Thermo Fisher Scientific) using a K3 direct electron camera (Gatan) and equipped with a Bioquantum energy filter (Gatan) set to a slit width of 20 eV (example micrographs in Extended Data Fig. 1f).For additional parameters, see Table 1.All datasets were acquired using fringe-free imaging (FFI) and multi-hole targeting using image shift in which three micrographs were collected per hole.Two small datasets on the GR:Hsp90:FKBP52 and GR:Hsp90:FKBP51 complex were collected before the larger datasets and used for initial model generation (described below).
Datasets I-IV were motion corrected using UCSF MotionCor2 and analyzed with RELION v.3.1.0.Motion-corrected images with dose weighting were used for CTF estimation using CTFFIND v.4.1 and reference-based picking was done in RELION using the corresponding references from the smaller datasets described above.The processing scheme for datasets I-IV are depicted in Extended Data Figs.2a and 6a.After initial rounds of 3D classification with symmetry C1, the GR:Hsp90:FKBP52 datasets (I and II) were combined and GR:Hsp90:FKBP51 datasets (III and IV) were combined.
For the GR:Hsp90:FKBP52 combined dataset, a particle stack of ~496,000 particles was obtained representing a GR:Hsp90:FKBP52 reconstruction at nominal resolution 3.82 Å.This stack was then subjected to 3D classification without alignment and subsequent 3D refinement on the best classes (~307,000 particles), which yielded the best overall consensus reconstruction at a nominal resolution of 3.56 Å.Additionally, to improve the resolution of the GR:FKBP52 region, the ~496,000 particle stack was subjected to signal subtraction of the Hsp90 region.Focused refinement on the Hsp90-subtracted particle stack was then performed (initial angular sampling 1.8°, initial offset range 3 pixels, initial offset step 0.75 pixels, local searches from auto-sampling 1.8°) using a mask including GR and FKBP52 only.
https://doi.org/10.1038/s41594-023-01128-yFocused classification without alignment was then performed using a mask including GR and FKBP52 only.Focused refinement on the best class (~107,000 particles) was then performed (initial angular sampling 0.9°, initial offset range 3 pixels, initial offset step 0.75 pixels, local searches from auto-sampling 0.9°) using focused refinement using a mask including GR and FKBP52 only, which yielded a GR:FKBP52 reconstruction with a nominal resolution of 4.31 Å.
For the GR:Hsp90:FKBP51 combined dataset, a particle stack of ~500,000 particles was obtained, representing a GR:Hsp90:FKBP51 reconstruction at a nominal resolution of 4.05 Å.This stack was then subjected to 3D classification without alignment and subsequent 3D refinement on the best classes (~172,000 particles) to obtain the best overall consensus reconstruction at a nominal resolution of 4.18 Å.Additionally, to improve the resolution of the GR:FKBP51 region, the ~500,000 particle stack was subjected to signal subtraction of the Hsp90 region.Focused refinement on the Hsp90-subtracted particle stack was then performed (initial angular sampling 0.9°, initial offset range 3 pixels, initial offset step 1 pixels, local searches from auto-sampling 0.9°) using a mask including GR and FKBP51 only.Focused classification without alignment was then performed using a mask including GR and FKBP51 only.Focused refinement on the best class (~120,000 particles) was then performed (initial angular sampling 0.9°, initial offset range 3 pixels, initial offset step 1 pixels, local searches from auto-sampling 0.9°) using focused refinement using a mask including GR and FKBP51 only, which yielded a GR:FKBP51 reconstruction with a nominal resolution of 4.31 Å.
Per-particle CTF, beam-tilt refinement, trefoil and fourth order aberration refinement, and astigmatism were estimated for both the consensus reconstructions and GR:FKBP focused reconstructions in RELION.The corrected particle stacks were then imported to CryoSparc (v3.3.2), and 2D classification was performed to clean up the particle stacks.The consensus reconstructions were subjected to non-uniform refinement with an envelope mask and a mask including Hsp90 only.The GR:FKBP focused reconstructions were subjected to local refinement with a mask including GR and FKBP only.
All final reconstructions were post-processed in CryoSparc in which the nominal resolution was determined by the gold standard Fourier shell correlation (FSC) using the 0.143 criterion (Extended Data Figs.2a and 6a).Maps were sharpened in CryoSparc and filtered to their estimated resolution.A composite map for both GR:Hsp90:FKBP51 and GR:Hsp90:FKBP52 was generated by combining the overall consensus refinement map with the GR:FKBP focused refinement map using vop maximum in Chimera.Note that the composite maps were only used for presentation in Figs.1a and 4a, but not used in atomic model building or refinement.
For both the GR:Hsp90:FKBP51 and GR:Hsp90:FKBP52 complexes, no ligand-free GR complexes were identified during image analysis, despite many rounds of focused classification on GR at various stages of data processing.Only classes with clear ligand density in the GR ligand binding pocket were obtained, suggesting ligand-free GR is either too dynamic or quickly released from the complex, consistent with findings during processing of the GR-maturation complex 13 .
Note that the nucleotide density in the Hsp90 NTD pockets were modeled as ATP in both the GR:Hsp90:FKBP51 and GR:Hsp90:FKBP52 atomic models; however, we cannot unambiguously determine whether the density corresponds to ATP or ADP:molybdate due to the difficulty in assigning the γ-phosphate density as molybdate.Our previous work demonstrated that Hsp90 needs to hydrolyze at least one ATP to reach the GR-maturation complex 11 , strongly suggesting the nucleotide is in a hydrolyzed ADP:molybdate state.Furthermore, the γ-phosphate density is relatively strong compared to the rest of the map (Extended Data Figs.3a and 7a), suggesting the presence of molybdate in at least some population of particles.However, Hsp90 may be in a hemi-hydrolyzed state with one protomer bound to ATP and one protomer bound to ADP:molybdate, which we cannot unambiguously determine from the reconstruction.Therefore, we have modeled ATP into the density, which is consistent with our treatment of the GR-maturation complex 13 .Models were refined using Rosetta v.3.11throughout.Following the split map approach 79 to prevent and monitor overfitting, the Rosetta iterative backbone rebuilding procedure was used to refine models against one of the half maps obtained from RELION, with the other half map only used for validations.Structurally uncharacterized regions, including the FKBP52 TPR:Hsp90 CTD interaction, the FKBP51:HSP90 CTD interaction, and the Hsp90 lumen:GR pre-Helix 1 interaction, were built de novo into consensus maps or focused maps using RosettaCM 80 .The final refinement statistics are provided in Table 1.

Fluorescence polarization assays
Fluorescence polarization of fluorescent dexamethasone (F-dex) (Thermo Fisher) was measured on a CLARIOstar Plus microplate reader (BMG LabTech) with excitation/emission wavelengths of 485/538 nm, and temperature control set at 25 °C.Buffer conditions were 50 mM HEPES pH 8, 100 mM KCl, 2 mM DTT.For equilibrium ligand binding in Figs.3e,f and 4f, and Extended Data Fig. 8a,b, proteins were pre-equilibrated together at room temperature for 60 min before F-dex addition.Proteins and reagents were added at the following concentration: 10 nM F-dex, 100 nM GR DBD-LBD, 2 μM Hsp40, 2 μM Bag-1, 15 μM Hsp70, 15 μM Hsp90, 15 μM Hop, 15 μM p23 or p23Δhelix, 15 μM FKBP or FKBP mutants, 5 mM ATP/MgCl 2 and 20 mM sodium molybdate where indicated.Note that the dissociation constant (K D ) between GR and F-dex is ~150 nM (ref.11).Ligand binding was initiated with 10 nM F-dex, and association was measured until reaching equilibrium.The plotted equilibrium values in Figs.3e,f  For equilibrium ligand binding in Extended Data Fig. 5d, proteins were pre-equilibrated together at room temperature for 30 min before F-dex addition.Proteins and reagents were added at the following concentration: 10 nM F-dex, 100 nM GR and 15 μM FKBP51 or FKBP52.Ligand binding was initiated with 10 nM F-dex, and association was measured until reaching equilibrium.The plotted data points for each reaction represent three biologically independent samples.GR ligand binding behavior was affected by buffer conditions; therefore, reactions were always normalized such that each reaction had equivalent amounts of buffer reagents.

In vivo GR activity assays
Relating to Fig. 2e, the effect of overexpression of wild-type FKBP52 on GR activity was determined as previously described 23 .GR activity was measured in the wild-type Saccharomyces cerevisiae strain ( JJ762) expressing wild-type, human GR on a single-copy plasmid (p414GPD-GR) and the GRE-lacZ reporter plasmid pUCDSS-26X.Wild-type or mutant FKBP52 was expressed in the pRS423GPD plasmid.Cells were grown at 30 °C with shaking overnight in selective media, diluted tenfold and grown to OD 600 0.4-0.5.Cultures were split in two, and one set was induced with ligand (50 nM deoxycorticosterone) (Sigma) for 1 h.The β-galactosidase (β-gal) activity of paired samples in the presence and absence of hormone was measured as described using the yeast β-gal assay kit from Thermo Fisher Scientific (catalog number #75768).Assays contained triplicate samples and were conducted at least twice with each mutant.A representative assay is shown.
Fold GR activity was determined by the increase in normalized β-gal activity in the hormone treated sample relative to the untreated paired sample.Relative GR activation was calculated by normalizing the fold GR activity of each sample to the average fold GR activity of strain JJ762 expressing p423GPD (e.v.).The fold increase in GR activities compared to the e.v.control is shown (mean ± s.d.).

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Extended Data Fig. 1 | Sample Preparation.a, The in vitro reconstituted GR chaperone cycle.On the left, GR is active and able to bind ligand.Hsp70, aided by the co-chaperone Hsp40, engages GR and inhibits ligand binding.Hsp70 loads apo GR onto Hsp90 and Hop, which forms the 'GR-loading complex' (PDB ID 7KW7).Hsp70 and Hop are released, Hsp90 hydrolyzes ATP to fully close, and the co-chaperone p23 binds, forming the 'GR-maturation complex' (PDB ID 7KRJ).GR binds ligand in the transition from the GR-loading complex to the GR-maturation complex.In the maturation complex, GR is in a fully folded, native conformation and bound to ligand.Upon Hsp90 re-opening, GR is released from the complex to return to the cycle.b, Domain organization of the proteins in the GR:Hsp90:FKBP complexes and p23.c, Structural motifs of Hsp90, GR, and FKBP51/52.d, Coomassie-stained SDS-PAGE (4-12% acrylamide gel) with MBP-GR pulldown elutions from the in vitro reconstituted GR chaperone cycle.Lane 1-elution from MBP-GR pulldown for FKBP51-containing reaction; Lane 2sample from Lane 1 after size exclusion chromatography (SEC) (e) and chemical crosslinking with 0.02% glutaraldehyde; Lane 3-elution from MBP-GR pulldown for FKBP52-containing reaction; Lane 4-sample from Lane 3 after SEC (e) and chemical crosslinking with 0.02% glutaraldehyde.This experiment was repeated 7 independent times with similar results.e, Size exclusion chromatography (SEC) profile of the elution from the MBP-GR pulldown.The green trace represents the SEC profile from the reconstituted GR chaperone cycle with FKBP51, while the blue trace represents the SEC profile from the reconstituted GR chaperone cycle with FKBP52.mAU = milli-absorbance units.Coomassie-stained SDS-PAGE (4-12% acrylamide gel) of the fractions from SEC corresponding to the GR:Hsp90:FKBP52 sample (top) or the GR:Hsp90:FKBP51 sample (bottom).Colors indicate which gel lanes correspond to specific regions of the SEC profile.Sample fractions from the region highlighted in purple were collected and used for cryo-EM data collection.This experiment was repeated 11 independent times with similar results.f, Representative electron micrograph for the cryo-EM dataset of the GR:Hsp90:FKBP52 complex (left) (−1.48 μm defocus) and GR:Hsp90:FKBP51 complex (right) (−2.23 μm defocus).A total of 11,162 and 26,413 micrographs were obtained, respectively.
Extended Data Fig. 2 | Cryo-EM Data Analysis for the GR:Hsp90:FKBP52 Complex.a, Cryo-EM data processing procedure for the GR:Hsp90:FKBP52 complex performed in RELION and CryoSparc.Gold-standard Fourier shell correlation (GSFSC) curves of the final 3D reconstructions, including the focused maps and the consensus map, are shown (bottom).The blue lines intercept the y-axis at an FSC value of 0.143.Angular distribution of particles and local resolution are shown for the consensus map (bottom, middle).b, Map-to-model FSC curves between the GR:Hsp90:FKBP52 atomic model and the consensus GR:Hsp90:FKBP52 map, along with different views of the model within the map.The black dotted line intercepts the y-axis at an FSC value of 0.5.
Extended Data Fig. 3 | Hsp90:GR Interfaces in the GR:Hsp90:FKBP52 Complex.a, Hsp90:GR:FKBP52 complex map density with atomic model showing ATP-magnesium density in both Hsp90 protomers (Hsp90A/B).Bottom images show increased contour level on the map density to indicate that the ATP γ-phosphate position has much stronger density relative to the α and β-phosphates, likely corresponding to molybdate, which may act as a γ-phosphate analog (see Methods).b, Atomic model of GR from the GR:Hsp90:FKBP52 complex (yellow) aligned with GR from the crystal structure (PDB ID 1M2Z) (light pink) with co-activator peptide NCoA2 (purple) and GR from the GR-maturation complex structure (PDB ID 7KRJ).GR Helix 12 is indicated.c-f, Atomic model of GR:Hsp90:FKBP52 complex with Hsp90A (dark blue), Hsp90B (light blue), GR (yellow).Side chains in contact between GR and Hsp90 are shown, along with hydrogen bonds (dashed pink lines).c,Interface 1 of the GR:Hsp90 interaction depicting the GR hydrophobic patch (GR Helices 9 and 10) interacting with the Hsp90A Src loop (Hsp90 345-360 ), Hsp90A W320 , and Hsp90A NTD/MD helices.d, Interface 2 of the GR:Hsp90 interaction depicting the GR pre-Helix 1 strand and Helix 1 packing up against the Hsp90B amphipathic α-helices.e, Interface 2 of the GR:Hsp90 interaction depicting GR in surface representation colored by hydrophobicity (green = polar, brown = nonpolar) with Hsp90B Y627 sticking into the BF3 druggable hydrophobic pocket.f, Interface 3 of the GR:Hsp90 interaction depicting the GR pre-Helix 1 strand threading through the Hsp90 lumen between Hsp90A and Hsp90B.
Extended Data Fig. 4 | Hsp90:FKBP52 Interfaces in the GR:Hsp90:FKBP52 Complex.Atomic model of the GR:Hsp90:FKBP52 complex with Hsp90A (dark blue), Hsp90B (light blue), GR (yellow), and FKBP52 (teal).Side chains in contact between Hsp90 and FKBP52 are shown, along with hydrogen bonds (dashed pink lines).a, Interface 1 of the Hsp90:FKBP52 interaction depicting the FKBP52 TPR H7e binding to the Hsp90A/B CTD dimer interface.The helix of FKBP52 H7e breaks to fit into the cleft formed by the Hsp90 CTDs.b, Interface 2 of the Hsp90:FKBP52 interaction depicting the Hsp90B MEEVD motif binding the FKBP52 TPR helical bundle.c, Interface 3 of the Hsp90:FKBP52 interaction depicting the FKBP52 TPR Helices 5 and 6 binding to the Hsp90B CTD.d, FKBP52 (teal) from the GR:Hsp90:FKBP52 atomic model aligned with the cryo-EM structure of FKBP51 (light blue) (PDB ID 7L7I) (top) and crystal structures of FKBP52 (PDB ID 1P5Q, 1Q1C) (bottom) showing the difference in interdomain angles.1P5Q contains the FKBP52 FK1 and FK2 domain, while 1Q1C contains the FKBP52 FK2 and TPR domains.e, The GR:Hsp90:FKBP52 atomic model with FKBP52 (teal), GR (yellow), and dexamethasone (pink) with proline-isomerase inhibitors, rapamycin (brown) or FK506 (orange), docked into the atomic model to indicate the steric clash with GR.Rapamycin was docked in based on the FKBP52:rapamycin crystal structure (PDB ID 4DRJ) and FK506 was docked in based on the FKBP52:FK506:FRB crystal structure (PDB ID 4LAX).
Extended Data Fig. 5 | Analysis of the GR:Hsp90:FKBP52 Structure.a, Expression of human FKBP52 or FKBP52 mutants in wild-type yeast strain JJ762 assayed by immunoblot with a monoclonal antibody specific for FKBP52.A monoclonal antibody against PGK1 was used as a loading control.The asterisk marks an unknown protein that cross reacts with the anti-FKBP52 antibody.This experiment was performed one time.b, Atomic model of the GR:Hsp90:FKBP52 complex shown in surface representation.FKBP52 (teal), GR (yellow).The NCoA2 (nuclear coactivator 2) co-activator peptide is docked in based on the GR:NCoA2 crystal structure (PDB ID 1M2Z).While most of the coactivator peptide binding is sterically permitted, the N-terminus of NCoA2 clashes with the FKBP52 TPR domain (red circle).c, Atomic models of the GRmaturation complex (GR:Hsp90:p23) (left) and the GR:Hsp90:FKBP52 complex without FKBP52 (right) depicting that GR LBD dimerization is permitted once FKBP52 is released.Hsp90A (dark blue, surface representation), Hsp90B (light blue, surface representation), GR (yellow, surface representation), p23 (green, surface representation).In both complexes, the GR LBD dimerization site is accessible, however; binding of the second GR LBD (light pink) to the GR-maturation complex clashes with the Hsp90B CTD, shown with a red circle (left).Binding of the second GR LBD (light pink) to the GR:Hsp90:FKBP52 complex (right).Docking of the dimerized GR LBD is based on the GR LBD dimer crystal structure (PDB ID 1M2Z).d, Equilibrium binding of 10 nM fluorescent dexamethasone to 100 nM GR DBD-LBD with addition of 15 μM FKBP51 ('51') or FKBP52 ('52') measured by fluorescence polarization (mean ± SD). n = 3 biologically independent samples per condition.Fluorescence polarization values are baseline subtracted in accordance with the measured fluorescent dexamethasone baseline polarization value.Statistical significance was evaluated by an ordinary one-way ANOVA (F (2,6) = 1414, p < 0.0001) with post-hoc Tukey's multiple comparisons test.P-values: p(GR vs. GR + 51) < 0.0001, p(GR vs. GR + 52) < 0.0001, p(GR + 51 vs. GR + 52) = 0.004.(n.s.P > 0.05; * P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.001; **** P ≤ 0.0001).Source data Extended Data Fig. 6 | Cryo-EM Data Analysis for the GR:Hsp90:FKBP51 Complex.a, Cryo-EM data processing procedure for the GR:Hsp90:FKBP51 complex performed in RELION and CryoSparc.Gold-standard Fourier shell correlation (GSFSC) curves of the final 3D reconstructions, including the focused maps and the consensus map, are shown (bottom).The blue lines intercept the y-axis at an FSC value of 0.143.Angular distribution of particles and local resolution are shown for the consensus map (bottom, middle).b, Map-to-model FSC curves between the GR:Hsp90:FKBP51 atomic model and the consensus GR:Hsp90:FKBP51 map, along with different views of the model within the map.The black dotted line intercepts the y-axis at an FSC value of 0.5.
Extended Data Fig. 7 | Interfaces in the GR:Hsp90:FKBP51 Complex.Atomic model of the GR:Hsp90:FKBP51 complex with Hsp90A (dark blue), Hsp90B (light blue), GR (yellow), and FKBP51 (purple).Side chains in contact between Hsp90 and FKBP51 or Hsp90 and GR are shown, along with hydrogen bonds (dashed pink lines).a, Hsp90:GR:FKBP51 complex map density with atomic model showing ATP-magnesium density in both Hsp90 protomers (Hsp90A/B).Bottom images show increased contour level on the map density to indicate that the ATP γ-phosphate position has much stronger density relative to the α and β-phosphates, likely corresponding to molybdate, which may act as a γ-phosphate analog (see Methods).b, Interface 1 of the Hsp90:FKBP51 interaction depicting the FKBP51 TPR H7e binding to the Hsp90A/B CTD dimer interface.The helix of FKBP51 H7e breaks to fit into the cleft formed by the Hsp90 CTDs.c, Interface 2 of the Hsp90:FKBP51 interaction depicting the Hsp90B MEEVD motif binding the FKBP51 TPR helical bundle.d, Interface 3 of the Hsp90:FKBP51 interaction depicting the FKBP51 TPR Helices 5 and 6 binding to the Hsp90B CTD.e, Interface 1 of the GR:Hsp90 interaction depicting the GR hydrophobic patch (GR Helices 9 and 10) interacting with the Hsp90A Src loop (Hsp90 345-360 ), Hsp90A W320 , and Hsp90A NTD/MD helices.f, Interface 2 of the GR:Hsp90 interaction depicting GR pre-Helix 1 strand and Helix 1 packing up against the Hsp90B amphipathic α-helices.g, Interface 3 of the GR:Hsp90 interaction depicting the GR pre-Helix 1 strand threading through the Hsp90 lumen between Hsp90A and Hsp90B.h, The GR:Hsp90:FKBP51 atomic model with FKBP51 (purple), GR (yellow), and dexamethasone (pink) with prolineisomerase inhibitors, rapamycin (brown) or FK506 (orange), docked into the atomic model to indicate the steric clash with GR.Rapamycin was docked in based on the FKBP51:rapamycin crystal structure (PDB ID 4DRI) and FK506 was docked in based on the FKBP51:FK506:FRB crystal structure (PDB ID 3O5R).The FKBP51-specific inhibitor SAFit2 was docked into the atomic model to indicate there is no steric clash with GR at the backbone level (although some side chains clash).SAFit2 was docked in based on the FKBP51:SAFit2 crystal structure (PDB ID 6TXX).Extended Data Fig. 9 | Modeling FKBP Binding to All Five Steroid Hormone Receptors (SHRs).a, FKBP52 bound to the glucocorticoid receptor (GR, yellow), progesterone receptor (PR, tan), mineralocorticoid receptor (MR, orange), estrogen receptor (ERα, pink), or androgen receptor (AR, green) based on the structure of the GR:Hsp90:FKBP52 complex.Due to the structural conservation of the LBDs across the five SHRs, all SHRs fit well with FKBP52 using the GR:Hsp90:FKBP52 atomic model, with no backbone clashes between the SHRs and FKBP52.The PDB IDs used to dock in the SHRs are as follows: PR (1A28), MR (2AA7), ERα (1ERE), AR (1T7R).b, Sequence conservation across human steroid hormone receptors (GR, MR, ERα, ERβ, AR) plotted onto the GR structure from the GR:Hsp90:FKBP52 atomic model.Residues are colored from most variable (cyan) to most conserved (maroon).Residues that interact with FKBP51/52 are shown as spheres.

Fig. 1 |
Fig. 1 | Architecture of the GR:Hsp90:FKBP52 complex.a, Composite cryo-EM map of the GR:Hsp90:FKBP52 complex.Hsp90A, dark blue; Hsp90B, light blue; GR, yellow; FKBP52, teal.Color scheme is maintained throughout.b, Atomic model in cartoon representation with boxes corresponding to the interfaces shown in detail in c-g.c, Interface 1 of the Hsp90:GR interaction, depicting the Hsp90A Src loop (Hsp90A 345-360 ) interacting with the GR hydrophobic patch.GR is in surface representation.d, Interface 2 of the Hsp90:GR interaction, depicting GR Helix 1 (GR 532-539 ) packing against the entrance to the Hsp90 lumen.

Fig. 5 |
Fig.5| Mechanism of GR regulation by FKBP51 and FKBP52 during the GR chaperone cycle.Schematic depicting how the FKBP co-chaperones integrate with the GR chaperone cycle and how this cycle may take place within a cellular context.Starting on the top left, GR (yellow, cartoon representation) is in dynamic equilibrium between cortisol-bound and unbound (apo) states.Hsp70 (orange) binds GR and locally unfolds GR to inhibit cortisol binding, stabilizing GR in a partially unfolded, apo state.Hsp70 transfers the partially unfolded GR to Hsp90 (light and dark blue):Hop (pink) to form the GR-loading complex12 , in which GR is stabilized in a partially unfolded, apo state.Cortisol (pink), which enters the cell through diffusion, binds to GR during the transition from the GR-loading complex to the GR-maturation complex when Hsp90 refolds the GR to a native conformation13 .In the GR-maturation complex, the cortisol-bound, folded GR is stabilized by Hsp90 and p23 (green) and is protected from Hsp70 re-binding.Depending on the relative concentrations of the FKBPs, either FKBP51 (purple) or FKBP52 (teal) can bind the GR:Hsp90:p23 complex, competing with p23, and stabilizing the rotated position of GR.FKBP51 sequesters GR:Hsp90 in the cytosol until ATP hydrolysis on Hsp90 allows release of GR back to the chaperone cycle.In contrast, FKBP52 promotes rapid nuclear translocation of GR:Hsp90 (refs.22,24,25,65).Once in the nucleus, the cortisol-bound GR can dimerize, nucleate the assembly of transcriptional regulatory complexes, and regulate transcription, including activating expression of FKBP51, leading to a negative feedback loop that regulates GR activity in the cell27,66-70 .