Electron Paramagnetic Resonance (EPR) Spectroscopy Reveals Nucleotide- and Partner-Dependent Conformational Ensembles in Hsp90

Abstract

Heat shock protein 90 kDa (Hsp90) is an ATP-dependent molecular chaperone whose function relies on conformational rearrangements and interactions with cochaperones and clients. Binding of nucleotides, cochaperones, and clients leads to functional structural rearrangements, often associated with heterogeneous structural ensembles that are challenging to characterize using high-resolution structure determination methods alone. Electron paramagnetic resonance (EPR) spectroscopy has emerged as a powerful approach for tracking the conformational landscape of proteins. However, despite its great potential, its application to Hsp90 homologs has so far remained limited, and the insights gained have not yet been comprehensively reviewed. In this mini-review, we summarize EPR studies on cytosolic, mitochondrial, and endoplasmic reticulum Hsp90 with the aim of highlighting how EPR has advanced our understanding of nucleotide- and partner-dependent conformations. Specifically, we summarize how continuous wave EPR (CW-EPR), double electron–electron resonance (DEER), and electron nuclear double resonance (ENDOR) spectroscopies revealed conformational equilibria, symmetry breaking, dissociation constants, and the structural role of intrinsically disordered regions within Hsp90, while also discussing methodological advances and recent extensions to cellular contexts. Overall, these studies establish EPR spectroscopy as a valuable biophysical tool for resolving conformational ensembles and structural heterogeneity underlying Hsp90’s function, with general relevance to the structural biology, spectroscopy, and chaperone fields, and pave the way for future in vitro and in-cell investigations of this exciting molecular chaperone.

1. Hsp90A Dynamic ATP-Driven Molecular Machine

Ηeat shock protein 90 kDa (Hsp90) is a highly conserved molecular chaperone abundant among all organisms (yeast, bacteria, and humans) and critical in maintaining cellular proteostasis. It does so by folding, remodeling, and activating hundreds of substrate proteins called clients, contributing to 10% of the proteome. In some cases, Hsp90 holds proteins to protect them from aggregation and degradation. Hsp90 clients are involved in signaling, immunity, and metabolism and include many oncogenic proteins, whose stabilization promotes tumorigenesis. Therefore, Hsp90 is a major pharmacological target. , Hsp90 is an ATPase using Mg­(II) ion as an essential cofactor of ATP hydrolysis. Its function is finely tuned by cochaperones that modulate the rate of ATP hydrolysis, remodel Hsp90 in a conformation primed for client loading or release, or assist with loading the client.

Structurally, Hsp90 is a homodimer (∼180 kDa), with each monomer comprising three consecutive domains (Figure ): the N-terminal domain (NTD), the middle domain (MD), and the C-terminal domain (CTD). The NTD harbors the ATP-binding site, serves as the binding site for some cochaperones, and contains a flexible “ATP lid” segment that undergoes conformational rearrangement upon ATP binding. The MD is a critical structural element of ATP hydrolysis, and is the binding site of clients, although other domains might be involved as well. The CTDs form the primary dimerization interface with a dissociation constant, K _d, in the tens of nanomolar range. The three domains of Hsp90 are known to communicate allosterically along the ∼12 nm long dimer. −

1.

X-ray structures of Hsp90 homologs that are relevant to the discussion in this review. (A) yHsp90 in the closed conformation crystallized in the presence of the NTD cochaperone Sba1 (omitted here for clarity) and Mg­(II)/AMP-PNP (PDB entry 2CG9). The tail was not resolved in the X-ray structure; instead, it is added schematically here to help visualize spin labeling positions at the tail. (B) Mitochondrial Hsp90 (TRAP1) from Danio rerio in the closed conformation in the presence of Mg­(II)/AMP-PNP (PDB entry 4IPE). (C) Endoplasmic reticulum Hsp90 (Grp94) from Canis lupus familiaris in the closed conformation in the presence of Mg­(II)/AMP-PNP (PDB entry 5ULS). In each panel, one protomer is shown in gray, and the other is colored by domain, with the NTD in cyan, the MD in yellow, and the CTD in blue; the Mg­(II) ions and nucleotides are shown as magenta spheres and green sticks, respectively. The residues labeled and studied with EPR techniques and discussed herein are shown as red spheres. All panels were generated using PyMOL.

Across species, the core Hsp90 fold is conserved, but sequence features and regulation differ substantially. In bacteria, HtpG is the cytosolic Hsp90 homolog and is only about 50% similar to human Hsp90, whereas human (hHsp90) and yeast Hsp90 (yHsp90) have 61% sequence identity. yHsp90 contains two closely related cytosolic isoforms, Hsp82 and Hsc82, distinguished mainly by their regulation, with Hsp82 being heat-inducible and Hsc82 more constitutive. In human cells, the two cytosolic isoforms are Hsp90α and Hsp90β; the former is induced by heat shock and stress, while the latter is constitutively expressed. In eukaryotic cytosolic Hsp90, the CTD possesses a terminal disordered tail with a conserved MEEVD motif at its end, whereupon tetratricopeptide repeat domains (TPR) cochaperones bind. , By contrast, HtpG lacks this motif and the cochaperones associated with it. Moreover, Hsp90 features a flexible charged linker between the MD and NTD, which is important for cochaperone regulation. Interestingly, the length of the linker is highly variable, ranging from 10 residues in HtpG to 56 in yHsp90 and 63 in hHsp90.

Eukaryotic cells encode four major Hsp90 homologs: the two cytosolic forms mentioned above, one mitochondrial (TRAP1), and one that resides in the endoplasmic reticulum (ER, Grp94). TRAP1 is more closely related to the bacterial HtpG than to eukaryotic cytosolic Hsp90, as it lacks the charged linker and the MEEVD motif and does not bind standard cytosolic cochaperones. , TRAP1 also contains a mitochondrial targeting sequence and a unique NTD “strap” extension that stabilizes the closed state and modulates ATPase activity. Grp94 is regulated by an NTD pre-N extension that suppresses spontaneous nucleotide-driven closure. Grp94 also carries a KDEL-type ER retention signal, rather than a cochaperone recruitment tail.

During its functional cycle, Hsp90 undergoes several large and small conformational changes at the NTDs and MDs. These span from highly heterogeneous V-shaped ensembles, to closed conformations stabilized in the presence of nucleotides, , and cochaperones. , Importantly, nucleotide binding is not deterministic in the conformation Hsp90 adopts, with the system best described as being in an equilibrium between open ↔ closed conformations, dependent on nucleotide, cochaperone, and client binding. − The open V-shaped ensembles are occupied in the absence of nucleotide, i.e., in the apo state, , and are known to serve the recruitment of a diverse set of clients. However, these open conformations are also observed in the presence of bound nucleotides, as found from Förster resonance energy transfer (FRET), nuclear magnetic resonance (NMR), electron paramagnetic resonance (EPR), and electron microscopy/X-ray crystallography. Posthydrolysis, i.e., after ATP hydrolysis but before ADP release, the ADP-bound state adopts a more compact conformation compared to the ATP-bound state, as found from both high-resolution and frozen solution data, as well as from biochemical assays. Throughout the cycle, the CTDs remain dimerized independent of nucleotide or cochaperone binding, although single-molecule FRET (smFRET) data have suggested a CTD opening that is anticorrelated with respect to the NTD closure. Structural differences of Hsp90 are also evidenced between different homologs. ,

The binding of cochaperones occurs at different stages of Hsp90’s cycle and stabilizes specific conformations. For example, human Hop (analog Sti1 in yeast) binds and stabilizes open conformations, inhibiting ATPase activity, while TPR cochaperones bind at the CTDs, acting as structural scaffolds that recruit Hsp70 to facilitate client transfer. In contrast, human p23 (analog Sba1 in yeast) binds to closed NTD conformations and promotes maturation of the client. Structural characterization of ternary complexes of Hsp90/cochaperone/client is more challenging, and insights stem mostly from cryoEM data, which, although they provide binding sites and insights into the mechanism of action, they report only a snapshot of these interactions, lacking dynamical information. −

Collectively, the structural plasticity of Hsp90, along with the multiple nucleotide-dependent conformational equilibria and the formation of heterogeneous cochaperone and cochaperone/client-bound complexes, make the structural characterization of Hsp90 throughout its functional cycle challenging. High-resolution techniques such as X-ray crystallography and cryoEM tend to capture the most stable or highly populated conformational states of Hsp90, while transient and sparsely populated intermediates that underlie its conformational plasticity remain largely inaccessible. , Moreover, Hsp90 and its clients and cochaperones feature intrinsically flexible/disordered regions whose electron density cannot be easily resolved. ,,

For these reasons, spectroscopic techniques that can probe dynamics in solution, such as NMR, FRET, and EPR, and report directly on conformational ensembles and residue-level dynamics, are needed to elucidate Hsp90’s function. FRET reports on real-time conformational changes at the ensemble or single-molecule level and has been central in establishing open-closed equilibria in Hsp90. ,, FRET also reports distances between donor and acceptor labels, and given that FRET labels are bulky, the precision of the distance information may be compromised. NMR is pertinent for providing structure and residue-level information on dynamics and transient interactions, relying heavily on isotope labeling and being free of molecular labeling, but it is difficult to use in large proteins like Hsp90, making full-length (FL) protein studies challenging. ,, EPR spectroscopy can report on both local (residue) dynamics and nanometer-range distance distributions upon spin labeling, which can resolve heterogeneous conformational ensembles without limitations on the size of the protein and complements the other techniques. Although EPR is highly suitable for investigating systems with the characteristics of Hsp90, it has been applied to Hsp90 only sparingly in the past decade. In this mini-review, we present these studies to highlight the potential of EPR in Hsp90 research.

We first give a brief introduction to the EPR methods used in Hsp90 studies, then describe their application to specific Hsp90 homologs in solution, frozen solution, frozen cell lysates, and cells, in the presence of nucleotides, as well as their complexes with partner proteins. These studies provide quantitative information on local dynamics, conformational ensembles, conformational equilibria, and the structural changes induced by ATP binding and hydrolysis and partner binding. We also discuss how the observables obtained by EPR are linked to the mechanisms of Hsp90 function.

2. EPR Spectroscopic Techniques and Observables

2.1. Spin Labeling

EPR techniques require the presence of paramagnetic centers in the protein under study that serve as probes; these can be either endogenous, e.g., paramagnetic metal ions, or intrinsic radicals, or introduced via site-directed spin labeling (SDSL) approaches, most commonly on cysteine (Cys) residues. In cases where the protein has functional cysteines, alternative labeling approaches must be pursued. − The choice of the labeling sites is critical for EPR studies and depends on the specific structural questions. Also, the chosen labeling sites should not perturb correct protein folding and function; therefore, surface-exposed sites are preferred, and functional assays are typically required. Another important issue is the flexibility of the labels. If the label is too flexible, then it can contribute to the observed protein dynamics and obscure the intrinsic local protein dynamics, preventing reliable data interpretation. For intracellular studies, additional requirements apply: the spin label itself must be redox-stable in the reducing cellular environment, and the conjugation bond to the protein must resist cleavage by endogenous proteases or reductants.

Typically used EPR spin labels are in the form of nitroxide (referred to as NO herein) and trityl , radicals, and Gd­(III), and Cu­(II) chelates. Their employment depends on the biological question, the biological setting (in vitro or in cell), and available instrumentation. NOs are the gold standard for protein structural studies in vitro due to their small size, which causes minimal structural perturbation, as well as their commercial availability. Importantly, they are excellent reporters of protein mobility and solvent accessibility at room temperature and at commonly available X-band (∼9.5 GHz) spectrometers. Additionally, the analysis of EPR data is well established, making data interpretation straightforward. , A major drawback of the common NOs is their reduction in the intracellular environment, which is challenging for in-cell studies. Several new redox-stable NOs have been synthesized and applied for intracellular studies. , Trityl radicals provide high sensitivity due to their narrow EPR line width and are stable and redox resistant; therefore, they are pertinent for in-cell studies. , On the downside, trityls are bulkier than NOs and can perturb protein structure; they are also less sensitive to their environment compared to NOs. Additionally, their labeling chemistry is less mature compared to NO radicals, and their commercial availability is limited. Gd­(III) chelates are extremely stable and redox resistant. Therefore, when combined with a rigid and redox-resistant tether to the protein, they are ideal candidates for in-cell studies. They perform well at medium to high fields (Q- and W-band) for distance measurements due to their relatively prolonged phase memory times, giving access to long interspin distances. , However, the room-temperature Gd­(III) EPR spectrum is not informative in terms of local dynamics. Cu­(II) chelates are rigid spin labels that provide narrow distance distributions in EPR distance measurements, albeit at the cost of engineering metal-binding sites (e.g., the double-His motif). The binding affinity of Cu­(II) to the double-His motif is very high and has allowed measurements at the nanomolar regime , and in the cell with high sensitivity. Currently, the analysis of their EPR spectra is more demanding compared to that of NOs and trityls. In Hsp90 studies, Gd­(III) and NO spin labels have so far been used (see Figure D).

2.

EPR techniques applied to the study of Hsp90: (A) X-band CW-EPR spectra of a NO spin label with different rotational correlation times (indicated); the spectra were simulated with EasySpin. (B) DEER and (C) ENDOR pulse sequences. (D) NO and Gd­(III) spin labels used in the studies of Hsp90. (E) experimental setup for DEER measurements at W-band for (i) NO–NO, (ii) Gd­(III)–Gd­(III), Mn­(II)–Mn­(II), and Mn­(II)-NO pairs; reproduced with permission from ref copyright 2021 American Chemical Society. The blue arrows indicate the observed frequency, whereas the light blue areas indicate the position and bandwidth of the pump pulse(s). For NO–NO DEER, the observed pulse sequence is set near the maximum of the NO EPR spectrum, while the pump pulse is applied at a frequency offset that inverts a large portion of NO spins without perturbing the observed spins. In Gd­(III)–Gd­(III) DEER, the observed pulses are placed on the intense central transition, while pump pulses excite Gd­(III) spins on both sides of the peak. In Mn­(II)–Mn­(II) DEER, the observed pulses are set on one of the six hyperfine lines, and the pump pulse is applied on the neighboring maxima. For heterospin Mn­(II)-NO DEER, the observed pulse sequence is applied on a Mn­(II) hyperfine line, while the pump pulse is set on the maximum of the nitroxide spectrum to achieve maximum inversion efficiency and thus enhanced sensitivity.

2.2. CW-EPR

EPR is divided into two modalities: continuous-wave EPR (CW-EPR) and pulsed EPR. CW-EPR is typically used to record the EPR spectrum, where the sample is continuously irradiated with microwaves while varying the magnetic field. The room-temperature CW-EPR spectrum of a NO spin label contains information on local dynamics, the environment of the labeling site, transient interactions, and oxygen accessibility. It features the characteristic three-line spectrum arising from the hyperfine interaction (A) of the unpaired electron (S = 1/2) with the ¹⁴N (I = 1) nearby nuclear spin. The center of the triplet (given by the g-value) and the splitting of the triplet (A) are anisotropic; namely, they depend on the orientation of the magnetic field with respect to the nitroxide moiety. Accordingly, the shape, width, and relative intensities of these lines are sensitive to the mobility (rotational correlation time, τ_R) of the spin label on the nanosecond time scale. Rapid isotropic motion averages the anisotropy and produces three narrow lines of nearly equal intensity, whereas restricted motion leads to line broadening, anisotropic line shapes, and unequal peak intensities due to incomplete averaging of the magnetic interactions (simulated NO spectra with different τ_R are shown in Figure A). Thus, CW-EPR spectra encode information on the spin-label mobility stemming from backbone and side chain motion, as well as from rotational diffusion for small proteins. CW-EPR spectra of proteins often feature multiple components, reflecting spin labels experiencing distinct local environments, conformations, and transient interactions. Quantifying the proportion and rotational mobility of each component requires spectral simulations using user-friendly available software packages. ,,

2.3. Pulsed EPR–DEER

Pulsed-EPR is typically performed at cryogenic temperatures because of the need for sufficiently long relaxation times. The double electron–electron resonance (DEER or PELDOR) experiment (four-pulse pulse sequence in Figure B) is used for measuring the distance distribution between two spin labels. A pulse sequence, applied at the observer microwave frequency (MW₁) creates a refocused echo for spin population A. During the dipolar evolution period, τ₁, a π pulse applied at a second microwave frequency (MW₂) inverts (pumps) a second population of spins, B. If the B spins are coupled to the A spins, the pump π pulse changes the local magnetic field experienced by the A spins, resulting in a modulation of the echo intensity. By incrementing the position of the MW₂ π pulse and recording the echo intensity, a time-domain dipolar evolution trace is obtained. The frequency of these oscillations reflects the strength of the dipolar interaction between the spins, given by

$${\omega }_{\mathrm{A}\mathrm{B}}=\frac{(1-3\mathrm{c}\mathrm{o}{\mathrm{s}}^2\theta )}{4\pi \hslash r_{\mathrm{A}\mathrm{B}}^3}({\mu }_0{\beta }_{\mathrm{e}}^2{g}_{\mathrm{A}}{g}_{\mathrm{B}})$$

where θ is the angle between the vector r _AB connecting the two spins and the external magnetic field, g _A, g _B are the g-values of the two spins, μ₀ is the vacuum permeability, β _e the Bohr magneton, and ℏ the reduced Planck constant.

The extraction of distance(s) from the DEER trace can be achieved by employing some regularization method to account for the ill-posed nature of the inverse problem. More recently, machine-learning approaches have been integrated into data analysis to enhance robustness. , The typical accessible distance range of DEER spectroscopy is 1.5–8 nm. Other pulse sequences, such as relaxation-induced dipolar modulation enhancement (RIDME), which works particularly well for unlike spins, can in principle be applied, − but it has not yet been applied to Hsp90.

In DEER, unlike FRET, the B spin can be of the same type or different compared to the A spin, which facilitates sample preparation significantly. The choice of the observed MW₁ and pump MW₂ frequencies is dictated by the character of their EPR spectra and aims to maximize excitation and inversion efficiencies while minimizing overlap between observe and pump pulses to achieve the best sensitivity. In the case of Hsp90, DEER has been applied to NO–NO, Gd­(III)–Gd­(III), Mn­(II)–Mn­(II), and Mn­(II)-NO pairs, each requiring a different experimental setup, as shown in Figure E.

DEER distance distributions report on all conformational ensembles present at the time of sample freezing; therefore, DEER is well suited for differentiating well-ordered proteins, which give narrow distance distributions, from flexible, heterogeneous systems and their complexes that give broad distance distributions, as schematically illustrated in Figure . DEER also allows observation of less stable or sparsely populated conformations, as well as flexible regions, that are difficult to access using X-ray crystallography and cryoEM. , Specifically, changes in the shape, width, and modality of distance distributions report on population shifts, conformational heterogeneity, and structural rearrangements within the ensemble. Moreover, the DEER signal encodes information on the fraction of coupled spin-pairs from the modulation depth parameter, allowing measurement of the K _d of dimerization , or probing the multimerization degree. ,

3.

Schematic representation of how conformational heterogeneity is reflected in DEER distance distributions. (A) The presence of a single conformation gives rise to a single distance distribution, whereas (B) the presence of an ensemble of conformations produces multiple peaks or a broadened peak corresponding to different interspin distances.

2.4. Pulsed EPR–ENDOR

Apart from CW-EPR and DEER spectroscopies, which monitor solely electron spins, it is possible to probe electron–nuclear interactions directly using experiments that can excite nearby coupled NMR-active nuclei. For simplicity, let us consider a spin S = 1/2, coupled to a nucleus with I = 1/2; in this case, the EPR signal will split and form a doublet, the splitting of which is the hyperfine coupling, A. When A is small and not resolved in the EPR spectrum, it can be detected by electron–nuclear double resonance (ENDOR) experiments. A common ENDOR experiment, Davies ENDOR, is shown in Figure C. Here, the first MW pulse inverts selectively one of the doublet components, thus creating a nonequilibrium spin polarization. Then, a radiofrequency (RF) pulse, at v _RF, applied to one of the NMR transitions, induces a nuclear spin flip that changes the electron-spin population. The change in the spin population is detected by a Hahn echo sequence, and the echo intensity is monitored as a function of v _RF, yielding the ENDOR spectrum, from which A can be measured. For Hsp90, the Davies ENDOR experiment measured the hyperfine coupling of ³¹P of the nucleotide to the Mn­(II) metal cofactor at the ATPase site (vide infra).

3. EPR Studies on Hsp90 Family Homologs

3.1. Nucleotide- and Cochaperone-Dependent NTD Closed Conformations in yHsp90

The closed conformation of yHsp90 (Figure A) was postulated to be the one resolved in the presence of Mg­(II)/AMP-PNP (the nonhydrolyzable ATP analog) and the NTD cochaperone Sba1. However, negative-stain electron microscopy and X-ray studies supported two different closed nucleotide-dependent conformations for the E. coli homolog in the absence of cochaperone. To address whether such conformations are ubiquitous, EPR spectroscopy was employed on frozen solutions of yHsp90³⁷ and different nucleotide conditions were compared: the apo state (no nucleotide), the prehydrolysis state (mimicked with Mg­(II)/AMP-PNP or Mg­(II)/ATP short incubation time), the posthydrolysis state (mimicked with Mg­(II)/ADP or Mg­(II)/ATP prolonged incubation), and the high-energy transition state, HES (Mg­(II)/ATP with orthovanadate, Vi, which mimics the γ-phosphate of the transition state). Spin labeling with NO, 3-maleimido-proxyl (hereafter referred to as proxyl, Figure D), across all domains (residues D61C, A152C at the NTD, Q385C, E517C at the MD, and D560C, K637C at the CTD, Figure A) yielded CW-EPR spectra characteristic of two populations, with no visible spectral changes between apo and nucleotide-bound states (see, e.g., CW-EPR spectra of A152C-NO in Figure A, left). Additionally, DEER measurements on these constructs (setup shown in Figure E, (i)) gave broad heterogeneous interprotomer distance distributions without unambiguous changes upon nucleotide binding, which were attributed to the rotational motion of the NO label and the inherent structural flexibility of Hsp90. Unexpectedly, broad distance distributions were observed even for D560C-NO (Figure A and Figure B, top). To bypass the issue of spin label flexibility and improve structural resolution at the NTDs, the essential ATPase cofactor, Mg­(II) was replaced by paramagnetic Mn­(II) in wild-type (WT) yHsp90, serving as a single-atom spin label. The ATPase activity of Hsp90 was not affected by this exchange. Mn­(II), like Gd­(III), has high spin (S = 5/2) and therefore is best measured at high frequencies, where the EPR central transition, M _S = −1/2 to M _S=+1/2 is narrow. W-band (95 GHz) Mn­(II)–Mn­(II) DEER was carried out to determine the interprotomer distance (setup in Figure E, (iii)) in the pre- and posthydrolysis, as well as the HES state, revealing distinct conformations at the NTDs: one with a distance distribution with a maximum at ∼4.3 nm in the prehydrolysis state and a second, more flexible conformation at ∼3.8 nm in the posthydrolysis and HES states, termed “closed” and “compact” respectively. Concurrently, the ENDOR experiment (sequence in Figure C) confirmed the nucleotide hydrolysis state by monitoring the coordination of the Mn­(II) to nucleotide phosphates through changes in the ³¹P ENDOR spectrum in the pre- and posthydrolytic ATP states, as well as in the HES (Figure C). The ENDOR data established that the distance changes observed by DEER correlate directly with ATP hydrolysis, and not with ATP binding alone, incomplete hydrolysis, or sample heterogeneity. The EPR data are in good agreement with smFRET experiments, which reported on the closed state for the AMP-PNP-bound prehydrolysis state, but they did not resolve the compact conformation for the posthydrolysis state. One possibility for this discrepancy is the selectivity of Mn­(II)–Mn­(II) for the compact subpopulation or the lower FRET resolution. The approach of using Mn­(II) as a spin label is not unique to Hsp90 but can be extended to any protein with ATPase activity.

4.

CW-EPR, DEER, and ENDOR data on FL yHsp90. (A) X-band CW-EPR spectra at room temperature on the A152C-NO mutant in the apo (black), pre- (green), and posthydrolytic (blue) ATP states (left, unpublished data from the Goldfarb lab) and of MTSL-labeled cochaperone Sba1 (right, reproduced with permission from ref copyright 2021 American Chemical Society). (B) W-band NO–NO DEER at 25 K (top) and Gd­(III)–Gd­(III) at 10 K (bottom) on the D560C CTD mutant at different ATP hydrolysis states (as explained in A); adapted with permission from ref copyright 2020 National Academy of sciences. (C) W-band ENDOR spectra at 10 K of Mn­(II) ions at the ATPase sites under different hydrolysis conditions (indicated), monitoring the ³¹P nucleotide signal; 2 h and 5 s refer to the time after the addtion of the nucleotide at which the sample was frozen. Reproduced with permission from ref copyright 2020 National Academy of sciences. (D) W-band DEER data in the prehydrolysis state of ATP in the absence (blue) and presence (cyan) of the NTD cochaperone Sba1; DEER at 25 K on the A152C-NO mutant (top), Mn­(II)-NO DEER at 10 K between A152C-NO and the Mn­(II) ions at the ATPase sites (middle), and Mn­(II)–Mn­(II) DEER data between the ATPase sites; adapted with permission from ref copyright 2021 American Chemical Society. (E) Cartoon model summarizing the findings on the NTDs of yHsp90. Panels B and D show DEER time-domain data after background removal (left) and the corresponding distance distributions (right); the orange-shaded area in B corresponds to the modeled distance distribution and was produced using the MtsslWizard software, and the dashed line in D represents the modeled distances from the X-ray structure..

To evaluate whether the CTD structure is affected by the closure of the NTD’s in solution and how it compares with the X-ray crystallography-determined structure, DEER was applied to yHsp90 labeled in the CTD. Using rigid Gd­(III) labels (setup in Figure D, (ii)) at two positions (K637C and D560C, Figure A) it was found for both positions that the dimer structure is preserved with similar interprotomer distances across all hydrolysis states and in agreement with crystallographic data of the closed state (see an example in Figure B, bottom). Crucially, the Gd­(III)–Gd­(III) DEER data afforded narrower distance distributions compared to the respective NO–NO data (Figure B, top), due to the Gd­(III) label rigidity, thereby reflecting mostly intrinsic protein flexibility. The DEER data in frozen solutions did not track the anticorrelated CTD opening observed with FRET.

To study whether the closed structure of the Sba1-bound yHsp90 complex solved by X-ray crystallography (Figure A) persists in solution, DEER distance measurements using a combination of Mn­(II) and NO labeling were employed. First, Sba1 was labeled on its native Cys35 with the MTSL NO label (S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)­methylmethanesulfonothioate, Figure D) and the CW-EPR spectrum of NO-labeled Sba1 showed the presence of one population, which is immobile on the EPR time scale (Figure A, right). Distance measurements between this label and the Mn­(II) at the ATPase sites confirmed cochaperone binding at both pre- and posthydrolysis states, with distances matching the modeled ones from the X-ray structure. Then, to address whether Sba1 binding alters the global NTD conformation, residue A152C near the ATPase site at the NTD was labeled with the proxyl NO label (Figure D) and interprotomer distances between the two NTDs (setup in Figure D, (i)) were measured in the absence and presence of Sba1 in the prehydrolysis state (Figure D, top). Unfortunately, the resolution was too low to detect potential structural changes. To enhance structural resolution, the observed spin was set to be the Mn­(II) atom at the ATPase sites, while the pumped spin was set to be the proxyl spin label at the A152C site (setup in Figure E, (iv)). Mn­(II)-NO DEER measurements between Mn­(II) and A152C-NO reported a shift of the conformational ensembles from two populations in the absence of Sba1 to one in its presence (Figure D, middle), reporting conformational rearrangements within the Hsp90/Sba1 complex, previously not resolved with the NO–NO DEER. To detect the proximity of NTDs in the presence of Sba1, Mn­(II)–Mn­(II) DEER measurements (setup in Figure E, (iii)) were done. These measurements revealed a tightly closed conformation (termed “packed”) in the presence of bound nucleotides and Sba1, different from those observed before in the absence of the cochaperone. Lastly, the integrity of the CTD dimer in the presence of Sba1 was further evaluated with interprotomer CTD DEER measurements using Gd­(III) labels, showing distances similar to those in the absence of Sba1.

Overall, DEER spectroscopy structurally resolved three distinct NTD closed conformational ensembles of yHsp90 (see Figure E), establishing ATP hydrolysis, as well as Sba1 cochaperone binding in the presence of Mn­(II)/AMP-PNP and Mn­(II)/ADP, as driving forces for the conformational rearrangements within the closed NTD ensembles. Here, the use of Mn­(II) was instrumental in resolving subtle but functionally important states that are difficult to access with other techniques or with standard spin labeling. These results establish that cochaperone binding reshapes the NTD conformation in yHsp90, rather than simply stabilizing a preexisting nucleotide-bound conformation, while maintaining CTD dimerization. Moreover, DEER probed the CTD dimerization in the apo and nucleotide-bound states, as well as in the Sba1/Hsp90 complex, with distances matching the X-ray structure.

3.2. CTD Dimer in yHsp90 is Allosterically Modulated by the Other Domains and Its Own Intrinsically Disordered Tail

The functional cycle of Hsp90, beyond nucleotide- and cochaperone-dependent NTD rearrangements, also depends on the stability and structure of the CTD dimer interface. FL yHsp90 forms a very stable dimer with a K _d of 60 nM, as determined from size exclusion chromatography (SEC). However, it was unclear whether long-range allosteric communication between domains modulates CTD K _d.

To address this, FL and isolated CTD (IsoC) yHsp90 were investigated with Gd­(III)-based DEER spectroscopy. Measurements were performed for two labeling sites within the CTD (K637C and D560C; see positions in Figure A) at increasing protein concentrations in the apo state. Analysis of the concentration-dependent modulation depth values allowed determination of the apparent K _d of the dimer. The data revealed that the K _d of IsoC is higher than that of the FL protein, suggesting that the removal of the MD and NTD moderately destabilizes the CTD dimerization, and that long-range allosteric effects modulate the dimer’s K _d. The DEER data, which were obtained at cryogenic temperatures, were further supported by room-temperature microscale thermophoresis (MST) and native mass spectrometry (native-MS) data, where the D61C (Figure A) FL variant was also explored as a native-like reference, yielding a K _d of 100 nM via MST, consistent with SEC data on yHsp90.

A recent study focused on the CTD flexible, disordered tail, exploring its conformational ensemble and its effect on CTD dimerization. This tail is inaccessible to high-resolution static techniques. Gd­(III) DEER measurements at position K637C near the dimer interface (Figure A) revealed that distinct populations appeared depending on protein context, i.e., IsoC vs FL. In the IsoC, with DEER measurements on K637C–Gd­(III) two populations at 3.4 and 4.9 nm were evident, whereas in FL only one population was present at 3.6 nm (Figure A), suggesting structural heterogeneity for the IsoC, which is suppressed in the FL protein via interactions with the MD. Truncation of the tail for both FL and IsoC proteins yielded a different, more compact CTD conformation (Figure A), indicating that the tail modulates CTD conformation. Notably, binding of the TPR cochaperone Cpr6 reduced the second population in IsoC-K637C, while it did not influence the FL protein (Figure A). Interdomain distance measurements at positions further away from the dimerization interface (K627C and D560C, Figure A) showed no difference between IsoC and FL forms, indicating that the effects of the tail perturb the structure of the CTD dimer near the tail site only.

5.

(A) W-band Gd­(III)–Gd­(III) DEER distance distributions of IsoC (left) and FL (right) yHsp90 measured at 10 K on K637C–Gd­(III) in the absence and presence of the TPR cochaperone Cpr6 (molar ratio indicated), as well as after tail truncation. (B) Gd­(III)–Gd­(III) primary DEER data (left, dashed lines show the background decay) and corresponding distance distributions (right) of IsoC-K637C–Gd­(III) inside human HeLa cells and cell extracts (CE). Data in A and B are reproduced with permission from ref copyright 2026 American Chemical Society. (C) Cartoon models describing the findings on the CTD (K637C–Gd­(III)) and CTD tail (A687C–Gd­(III)) of yHsp90 with color coding as in Figure E.

To establish the biological relevance of the results, DEER measurements were also carried out under cellular conditions, that is, in yeast and human HeLa cell extracts, as well as in HeLa cells. For these experiments, the purified IsoC-K637C–Gd­(III) was either directly mixed with cell extracts or internalized inside HeLa cells via electroporation, followed by Gd­(III)–Gd­(III) DEER measurements. It was found that in both types of extracts, the percentage of the second population diminished, while it had completely vanished inside HeLa cells (Figure B). As in the solution case, no effect was observed for the FL form. These data suggest that endogenous TPR cochaperones modulate the conformational landscape near the dimer interface for the IsoC protein.

In the same study, the tail conformational ensemble was directly followed with Gd­(III)–Gd­(III) DEER distance measurements for a labeling site within the tail at a position located close to the folded part of the CTD (A687C, Figure A). For FL-A687C–Gd­(III), DEER gave an interspin distance shorter than that for IsoC, further supporting the idea that the tail engages in intramolecular interactions in the FL protein. Upon binding of the Cpr6 cochaperone, the IsoC-687-Gd­(III) distance shortened and became similar to that of the FL, consistent with Cpr6 causing a more compact tail conformation. Similar measurements on a position located toward the end of the tail and close to the MEEVD motif (A701C, Figure A) yielded distances beyond the accessible DEER range, supporting an extended conformation between these sites both in the absence and presence of Cpr6. Complementary CW-EPR measurements using NO labeling showed reduced spin label mobility in the FL protein vs the IsoC protein. Specifically, the CW-EPR spectra of FL687-NO and FL701-NO, featuring, in addition to a fast motion component, a second spectral component indicative of restrictive motion, support the presence of transient interactions of the tail with the folded domains (CTD and/or MD). The binding of Cpr6 further increased the percentage of the restricted motion component at both labeling sites. In contrast, for the analogous IsoC, the restrictive motion was not present. Complementary NMR measurements further provided direct evidence of transient intermolecular interactions of the tail with the folded part of the CTD.

These results establish that the intrinsically disordered CTD tail of yHsp90 is an active structural regulatory element that reshapes the conformation of the folded CTD without altering dimer stability. The findings on the CTD dimer interface and CTD tail conformational ensemble are summarized in Figure C.

3.3. Symmetry Breaking and Sequential ATP Hydrolysis in TRAP1

In contrast to cytosolic Hsp90, which adopts several closed conformations in a highly symmetric manner in the presence of nucleotides, studies on the mitochondrial Hsp90 homolog (TRAP1) have revealed an asymmetric closed-state dimer in the presence of nucleotides, proposed to be important for client loading. Asymmetry appeared at the MD-CTD interface, as detected by X-ray crystallography (Figure B), in the presence of nucleotide analogs (AMP-PNP, ADP-BeF₃ ^–, ADP-AlF₄ ^–), and was further confirmed in solution using small-angle X-ray scattering (SAXS) and DEER spectroscopy. In the latter, a double Cys mutant with the mutations introduced at the MD and CTD (K439C, D684C, Figure B) was labeled with NO (proxyl, Figure D) and DEER measurements were carried out on a mixture with a WT:mutant ratio of 10:1 to remove interfering contributions from protomer–protomer distances. The results revealed two distances at 2.2 and 4.1 nm in the heterodimers, confirming that the asymmetry is not a crystallization artifact but an inherent feature of TRAP1’s closed state.

Subsequent EPR studies focused on understanding the role of ATP hydrolysis in the conformational asymmetry of TRAP1. For this, the same double Cys mutant (K439C/D684C) was labeled with NO (4-maleimido-TEMPO, Figure D) and two distances at 2.2 and 4.1 nm were resolved for the Mg­(II)-free ATP-bound WT:mutant mixed dimer, corresponding to the buckled and straight protomer, respectively (Figure A). To mimic a hemihydrolyzed state (ATP/ADP), a covalently linked +/R402A heterodimer was generated, in which one protomer was the labeled WT and the other carried the R402A mutation that lacked the ability to coordinate the γ-phosphate of ATP (serving as an ADP-state mimic). When labels were introduced into the R402A protomer, DEER showed mostly the 4.1 nm distance, corresponding to the straight conformation. In contrast, when the labels were placed on the opposite (WT) protomer, the distance shifted to 2.2 nm, corresponding to the buckled conformation (Figure A).

6.

Q-band (34 GHz) DEER distance distributions (A) of the double Cys TRAP1 mutant K439-NO/D684-NO in the ATP/ATP and ATP/ADP states; reproduced with permission from ref under CC-BY 4.0 license: https://creativecommons.org/licenses/by/4.0/. Copyright 2017 eLife sciences publications, Ltd. (B) Cartoon model describing the findings of the asymmetry in TRAP1 in the spirit of ref (C) Q-band DEER time-domain data after background removal (left) and corresponding distance distributions (right) at 80 K of K161C-NO in the presence of the partner BiP in the apo state or in the presence of nucleotides (indicated); adapted with permission from ref copyright 2024 National Academy of sciences. The modeled distance distributions shown as black and gray shaded areas were produced using the MtsslWizard software and the closed and partially open X-ray structures of Grp94, respectively. (D) Cartoon model describing the findings on the NTDs of Grp94. Color coding in B and D is as in Figure E.

Overall, the DEER data in TRAP1 showed that ATP hydrolysis proceeds in a sequential and deterministic manner, in contrast to earlier models that assumed independent or weakly cooperative ATPase sites. , Specifically, the data revealed that in the ATP/ATP state, one protomer adopts a bent conformation, while the other adopts a straight conformation. After hydrolysis of the first ATP molecule at one ATPase site, the dimer remains in a closed state but undergoes a switch in conformation, such that the ATP-bound protomer becomes buckled and the ADP-bound protomer becomes straight (results are summarized in Figure B).

The changes in the symmetry of TRAP1’s closed state support the notion that the energy provided by ATP hydrolysis is converted to mechanical work that rearranges the chaperone’s internal architecture at the client-binding sites, making TRAP1 a unique system to understand how ATP hydrolysis relates to client remodeling. While crystallographic data revealed the asymmetric closed conformation, DEER demonstrated that this asymmetry persists in solution, confirming that it is an intrinsic property of the chaperone.

3.4. Partner-Induced Remodeling of the NTD in Grp94

Large-scale conformational changes of the NTD of Grp94, the endoplasmic reticulum Hsp90, were observed upon nucleotide and partner BiP (an Hsp70 homolog) binding. Bip chaperone interacts directly with Grp94, accelerating the ATP-driven NTD closure. Measurements were performed on NO (proxyl, Figure D) labeled variants at positions reporting on the NTD (K161C, K87C), the ATP lid region (K177C), and the MD (R395C), and in a construct lacking the pre-N-domain (Δ72) (labeling positions are shown in Figure C).

DEER distance distributions between K161C-NO in the apo state and in the presence of Mg­(II)/ADP, ATP revealed broad distance distributions between ∼2–6 nm, indicative of a highly flexible open NTD arrangement. In contrast, binding of Mg­(II)/AMP-PNP led to the appearance of a partially closed conformation with a distance of ∼2.2 nm. Comparison between Grp94 and Grp94-Δ72 showed enhanced sensitivity of Δ72 to nucleotide, allowing ATP and AMP-PNP to populate the closed conformation more easily, suggesting that the pre-N domain acts as a structural barrier for Grp94 to occupy the closed conformation.

Addition of the partner BiP induced an NTD conformational change, revealing a distance of ∼4.2 nm for K161C-NO, corresponding to a partially closed intermediate (Figure C), whereas addition of BiP+Mg­(II)/nucleotides (ATP, ADP) led to a fully closed, catalytically active NTD-dimerized state at ∼2.3 nm (Figure C). Complementary DEER data for K87C-NO confirmed the dimerization of the NTDs. Both K161C-NO and K87C-NO DEER data are in agreement with the modeled distance distribution from the X-ray structure closed NTD conformation.

Additional structural information was obtained from CW-EPR measurements on the K177C-NO mutant at the ATP lid region in the absence and presence of nucleotides and BiP. The spectra revealed the presence of two components: a mobile and a rigid component, interpreted by the authors as the lid being in the closed and open conformations, respectively. In the presence of Mg­(II)/nucleotides, an increase in the mobile component was found, whereas in the presence of both BiP and nucleotide, these effects were further enhanced, suggesting a cooperative stabilization of the closed conformation and supporting that changes in the local dynamics of the lid are coupled to large global NTD closure.

DEER measurements at the MDs (R395C-NO, Figure C) showed no changes across Mg­(II)/nucleotide and BiP bindings, with an inter-MD distance of 4.5 nm, indicating that the MD remains structurally unchanged. This suggests that Grp94’s function is governed by the NTD rearrangements and lid dynamics, and does not involve the MD.

Overall, these results support a model in which Grp94 relies on Mg­(II)/nucleotide and BiP assistance to remodel its conformational landscape toward a more compact NTD conformation. Here, the pre-N domain acts as a suppressor of NTD closure, while the partner BiP stabilizes partially and fully closed conformations required for chaperoning activity. The findings are summarized in Figure D.

4. Summary and Outlook

The studies presented herein establish EPR spectroscopy as a powerful, yet underutilized, biophysical tool in the structural and functional study of Hsp90 homologs. The inherent flexibility and concurrent conformational states that Hsp90 adopts require approaches capable of resolving heterogeneous ensembles. EPR has proven to be suitable for probing both local dynamics and long-range distance distributions in solution, frozen solution, cell extracts, and living cells. Importantly, methodological aspects like the substitution of the native Mg­(II) cofactor with paramagnetic Mn­(II) acting as a single-atom label at the ATPase sites , and the use of rigid Gd­(III) labels, as well as a Mn­(II)-nitroxide spin pair, significantly enhanced resolution and sensitivity compared to classical nitroxide spin labeling.

EPR ensemble data indicate that Hsp90 function is dictated by shifts in conformational populations, as well as distinct structural transitions. Specifically, EPR provided quantitative structural access to ensembles and their relative populations, offering complementary insights to those obtained by other common techniques. The summary of findings in terms of the conformational changes detected is given in Figures –. The technique also allowed quantification of the CTD dimerization constants, as well as probing conformations of intrinsically disordered regions within the chaperone (CTD tail) that so far had been ignored in Hsp90 structural studies. Importantly, EPR measurements were recently extended to cellular environments, opening new directions to studying the structural dynamics of Hsp90 in biologically relevant contexts. This was achieved by combining highly sensitive, high-field measurements with intracellularly stable Gd­(III) spin labels, enabling measurements in the low micromolar regime within cells.

However, several limitations should be considered when applying EPR to Hsp90. EPR structural studies provide only scarce structural information closely related to the labeling positions. While a number of pairs can be introduced, this cannot provide a full structure, but it is particularly useful for tracking conformational changes and the extent of conformational ensembles. Accordingly, EPR is best coupled with other structural information and potentially with AlphaFold predictions. As it relies mostly on SDSL, the introduction and removal of cysteine residues must be carefully designed and validated to ensure they do not perturb ATPase activity, dimerization, cochaperone, or client loading. This is particularly relevant for human Hsp90 due to the presence of multiple native cysteines that complicate selective labeling strategies, although efforts in this direction are starting to develop. Moreover, traditionally used NO labels possess flexible linkers that may lead to loss of information, as shown earlier in this review. In addition, the inherent flexibility of Hsp90 makes data interpretation even more challenging, often requiring the integration of complementary structural and biophysical methods. Finally, pulsed EPR experiments are usually performed at cryogenic temperatures, which might alter populations or dissociation constants, even though optimized freezing protocols are followed.

Up to date, several important biological questions on Hsp90 remain open. These involve, for example, intraprotomer structural ensembles, like those resolved for TRAP1, which have not been explored for other Hsp90 homologs. In addition, the study of ternary complexes of Hsp90/cochaperone/client in solution remains unexplored. Such studies would be essential for linking conformational landscapes and complex stoichiometries to functional outcomes. Another important direction would be to resolve the potential regulatory role of the disordered CTD tail. Such a study would require investigation of the effects of nucleotides, mutations, and cochaperone recruitment on the conformational landscape of the tail, its interaction with the other domains, and its coupling to ATP hydrolysis.

Ultimately, expanding EPR studies to cellular environments is the next natural step in understanding Hsp90 structure and function. It would be particularly interesting to monitor the NTD and MD conformations of mammalian and bacterial Hsp90 directly inside cells, given that in-cell EPR is now established for both cell lines. , Methodological developments in spin label chemistry, such as rigid Cu­(II)-doubleHis motifs, − promise improved distance resolution while circumventing the need for Cys-based labeling, which is a significant advantage for hHsp90 where native Cys are present. Ultimately, combining EPR data with other biophysical techniques like cryoEM, NMR, and biochemical assays would provide a more comprehensive, ensemble-based picture of Hsp90s function in vitro and ultimately in cells.

Biographies

Christini Fiotaki received her BSc degree in Chemistry from the University of Crete, Greece, in 2024 and is currently pursuing an MSc in Organic and Biological Chemistry at the same University. Her current research focuses on the function and interactions of the molecular chaperone Hsp90. Her broader research interests include structural biology, protein biochemistry, molecular chaperones, and phase separation in biological systems.

Daniella Goldfarb is Professor Emerita in the Department of Chemical and Biological Physics at the Weizmann Institute of Science. She received her PhD from the Weizmann Institute in 1984, where she studied the NMR of liquid crystals. She then carried out postdoctoral research at the University of Houston, Texas, focusing on the application of pulsed EPR spectroscopy to zeolites. She joined the Weizmann Institute as a principal investigator in 1987, was promoted to Full Professor in 1998, and became Professor Emerita in 2023. At the Weizmann Institute, she established a laboratory dedicated to electron paramagnetic resonance (EPR) spectroscopy, where she focused on the development of new experimental methodologies. These techniques were initially applied to the study of porous materials and later to investigations of protein structure and dynamics, both in solution and within cells

Angeliki Giannouli received her BSc and MSc degrees in Chemistry from the University of Ioannina, Greece, and her PhD in Chemistry from the University of St Andrews, UK. She then conducted postdoctoral research at the Weizmann Institute of Science, Israel, where she later served as a Staff Scientist. In 2025, she joined the Department of Chemistry at the University of Crete as an Assistant Professor and established her independent research laboratory. Her research lies at the interface of biophysics, structural biology, and chemical biology, with a particular focus on electron paramagnetic resonance (EPR) spectroscopy to investigate protein structure, dynamics, and interactions. Her research interests include molecular chaperones, enzyme mechanisms, biomolecular condensates, and the development of chemical and biophysical approaches for studying complex biological systems.

The open access publishing of this article is financially supported by HEAL-Link.

The authors declare no competing financial interest.