Search tips
Search criteria 


Logo of hhmipaabout author manuscriptssubmit a manuscriptHHMI Howard Hughes Medical Institute; Author Manuscript; Accepted for publication in peer reviewed journal
J Mol Biol. Author manuscript; available in PMC 2010 January 10.
Published in final edited form as:
PMCID: PMC2735500

pH dependent conformational changes in the bacterial Hsp90 reveal a Grp94-like conformation at pH6 that is highly active for suppression of citrate synthase aggregation


The molecular chaperone Hsp90 depends upon large conformational rearrangements for its function. One driving force for these rearrangements is the intrinsic ATPase activity of Hsp90, as seen with other chaperones. However, unlike other chaperones, structural and kinetic studies have shown that the ATPase cycle of Hsp90 is not conformationally deterministic. That is, rather than dictating the conformational state, ATP binding and hydrolysis shifts the equilibrium between a pre-existing set of conformational states in an organism-dependent manner. While many conformations of Hsp90 have been described, little is known about how they relate to chaperone function. In this study, we show that the conformational equilibrium of the bacterial Hsp90 (HtpG) can be shifted with pH. Using small angle x-ray scattering we identify a two-state pH-dependent conformational equilibrium for apo HtpG. Our structural modeling reveals that this equilibrium is between the previously observed extended state and a second state that is strikingly similar to the recently solved Grp94 crystal structure. In the presence of the non-hydrolyzable AMPPNP, a third state, identical to the solved `ATP' structure from yeast Hsp90, is populated. Electron microscopy confirmed the observed conformational equilibria. We also identify key histidine residues that control this pH-dependent equilibrium, and by mutagenesis we successfully modulate the conformational equilibrium at neutral pH. Using these mutations, we show that the Grp94-like state provides stronger aggregation protection compared to the extended apo conformation, in the context of a citrate synthase aggregation assay. These studies provide a more detailed view of HtpG's conformational dynamics and provide the first linkage between a specific conformation and chaperone function.

Keywords: Hsp90, molecular chaperone, SAXS, conformational dynamics, pH-dependence


Hsp90 is a member of the diverse class of proteins known as molecular chaperones. Molecular chaperones are required in order to maintain the correctly folded state of other proteins in the cell. This is generally accomplished by the recognition of hydrophobic surfaces on the client protein and then repeated rounds of ATP hydrolysis-dependent conformational changes by the chaperone, facilitating substrate folding.1 Two well-studied members of this class, Hsp70 (DnaK) and Hsp60 (GroEL), interact with nascent polypeptide chains and promote their folding.2; 3 In addition to the stress response role in preventing aggregation,4 Hsp90 under unstressed conditions appears to act at later stages of the protein folding pathway and interacts with client proteins in their near-native states. Hsp90 induces subtle conformational changes that promote the binding and release of ligands or the interaction with partner proteins.5; 6; 7; 8; 9; 10; 11 Hsp90's client proteins include serine/threonine and tyrosine kinases, steroid hormone nuclear receptors, transcription factors, and tumor suppressor proteins.9; 12; 13; 14 Many of Hsp90's client proteins, including Cdk4, c-src and v-src, are oncogenic or otherwise required for cell proliferation, making Hsp90 an attractive target for anti-cancer therapeutics.15 Inhibition of Hsp90 with small molecules such as geldanamycin and its derivatives has been shown to be antitumorigenic, and several of these compounds are currently in clinical trials.16; 17; 18; 19

While the molecular mechanisms of Hsp70 and Hsp60 are well characterized, the conformational changes and determinants of substrate recognition associated with Hsp90 function remain unclear. For Hsp70 and Hsp60, client protein binding stabilizes the unstructured states; only upon release (either into solution or the lumen of the Hsp60 cavity) can folding occur. Thus in the bound state, the client protein adapts its conformation to the binding surfaces for these chaperones.20 By contrast, Hsp90 substrates have already achieved a partially folded or almost fully folded conformation before they interact with Hsp905 suggesting that Hsp90 must adapt its conformation to match each substrate or that different conformations recognize different substrates. This issue is particularly pronounced for the bacterial Hsp90, HtpG, for which there are no known cochaperones to improve the efficiency of client protein recruitment.

Hsp90 exists as an obligate dimer and the crystal structure of the nucleotide-free (apo) state reveals exposed hydrophobic surfaces on each of the three domains within a protomer (N-terminal (NTD), middle (MD), C-terminal (CTD)) that likely serve for client protein binding.21 Other crystal structures show that binding of nucleotide to the NTD induces remarkably distinct conformational states that decrease (the non-hydrolyzable ATP-analog, AMPPNP)22 or completely bury (ADP)21 the exposed hydrophobic patches. These conformational changes represent unexpectedly large changes where the NTD moves by approximately 50Å between each state (Figure 1). These rearrangements presumably act to alter client protein structure and to modulate client protein binding. Although localized changes in the active site lid23; 24 (residues 100-126), and the src loop25 (residues 281-296) occur upon nucleotide binding, these vastly different conformational states are well described by rigid body rearrangements of the domains about flexible linkers. A crystal structure of the full-length ER homolog, Grp94, has recently been solved26 revealing a state that appears to be intermediate between the apo open state and the closed ATP state (Figure 1). This structure further highlights the conformational flexibility of Hsp90s.

Figure 1
Previously solved structures demonstrating the conformational flexibility of Hsp90. Apo and ADP forms were crystallized using HtpG,21 ATP was crystallized in yeast,22 Grp94 is the crystal structure of the canine mitochondrial homolog,26 and the SAXS model ...

While the crystal structures provide important and essential information into the structure of Hsp90, they present only a static picture of Hsp90. It is, however, known that Hsp90 is a highly dynamic molecule and therefore static representations are insufficient for gaining a complete understanding of the molecular mechanism of Hsp90. In order to examine the solution structure of Hsp90 and its flexibility, we have employed small angle x-ray scattering (SAXS). SAXS defines the three-dimensional shape of a protein by providing low-resolution (10-30Å) data for the solution state of the protein. SAXS can also elucidate the conformations of multiple populations in solution.27; 28 Our previous SAXS study of the bacterial homolog, HtpG,27 revealed that the dominant solution conformation in the absence of nucleotide is a more open and extended structure (extended state) than that observed in the crystal structures (Figure 1). The SAXS data was highly reproducible making it possible to determine even subtle rearrangements in domain structure. It was also demonstrated that upon the addition of AMPPNP, HtpG did not shift completely to the previously crystallized `ATP' state (closed state). Instead, an equilibrium between the two states was established.27 Kinetic studies29; 30 EM single particle studies of the yeast and human proteins,31 and FRET studies32; 33 further demonstrated the generality of complex conformational equilibria for all Hsp90s. While conformational equilibria appear to be a conserved aspect of Hsp90's function, the positioning of these equilibria and the response to nucleotide appear to be extremely species specific. In humans, the equilibrium is heavily biased towards the extended state even in the presence of ATP,31 whereas in yeast, the equilibrium is shifted predominantly towards the closed state with the addition of nucleotide.31; 32 Notably the prokaryotic Hsp90 seems poised to equally sample the extended and closed states in the presence of ATP.27; 31

As described here, HtpG displays an even broader range of conformational states than previously anticipated. By varying the pH and examining the solution conformation with SAXS, we show that at low pH a previously unknown HtpG conformation, closely resembling the reported crystal structure of Grp94, dominates in solution. In the absence of nucleotide, a two-state equilibrium, with a pH midpoint around pH 7.2, exists between the extended state and the Grp94-like conformation described here. Through mutagenesis of a single histidine residue we can bias the conformation of HtpG independent of pH providing powerful new tools for examining the functional properties of each conformation. Importantly, we also demonstrate a functional difference between these two apo conformations where the Grp94-like conformation is the active state in a citrate synthase chaperoning assay. These results provide the first direct evidence for distinct functional behaviors of different apo conformations.


pH causes the solution structure of HtpG to shift from an extended state to a more compact state

Our previous SAXS study of HtpG revealed a more open and extended structure in solution than observed in the crystal.27 This model was determined at pH 9 given that both the SAXS data and EM studies had suggested that HtpG was more homogenous at pH 9. However at physiological pH, both SAXS and EM studies indicate that HtpG is conformationally heterogeneous.27; 31 These results suggest that varying the pH could be an effective and important tool for investigating the conformational dynamics of HtpG in solution.

Indeed, SAXS measurements show dramatic changes between pH 6 and 9. The scattering intensity, I(Q), was measured for scattering angles (Q=4πsinθ/λ) from 0.01 to 0.3Å-1 (Figure 2(a)). The scattering curves for each pH represent the average of multiple curves (See Materials and Methods) and the differences seen between the pH conditions are very robust especially for low Q where the largest differences are seen between the different pH conditions (average error of 2% for Q=0.06-0.12). The I(Q) data was then converted to the interatomic distance distribution function, P(r), by Fourier transform (Figure 2(b)). The P(r) function provides a distribution of the interatomic distances found within the molecule giving information on shape as well as maximum end-to-end distance and provides a particularly sensitive indicator of conformational changes. Both the I(Q) and the P(r) functions for HtpG at the different pHs demonstrate a clear transition to a more compact state as the pH of the solution is lowered. HtpG at pH 6 has a radius of gyration, Rg, of 46.3Å as compared to an Rg of 55.7Å for HtpG at pH 9. To clearly analyze the pH dependence of the conformational transition, P(r) at 54Å can be plotted versus pH (Figure 2(c)). The transition is well fit by a two-state pH titration curve resulting in a pKa of 7.2. The resulting titration curve is the same when you use the P(r) values at 50Å or 126Å (data not shown). That the data is best fit with a two state model indicates that two distinct conformations are present in solution.

Figure 2
Scattering data for apo HtpG shows a more compact conformation at low pH. (a) Averaged and scaled solvent-subtracted scattering curves (I(Q)) at varied pH (pH 6, purple; pH 6.5, gold; pH 7, green; pH 7.5, red; pH 8, blue; pH 9, grey). (b) Interatomic ...

The pKa of 7.2 immediately suggests the involvement of histidine in the conformational change. Fitting the titration data with a multiple binding site model shows that 1-2 histidines are required for the transition between the two populations (Supplemental Figure S1). There are 14 histidines present in one monomer of HtpG. To narrow down the list of candidates, we focused on the domain interfaces. As was previously noted, the differences in the various conformations of Hsp90 result from large domain rearrangements, where the origins of these motions lie at the NM interface (between the NTD and MD) or at the MC interface (between the MD and CTD). There are 8 histidines clustered at the NM interface and 3 clustered at the MC interface. To isolate which interface is relevant for the pH dependent conformational changes, the effect of pH on a monomeric truncation mutant containing only the NTD and MD was analyzed by SAXS. Scattering data was collected for the truncation mutant (NMHtpG, residues 1-495) for Q = 0.009-0.3 at pH 6, 7, 8, and 9. No significant differences were seen for the scattering data at different pHs (Supplemental Figure S2). The slight differences seen in the pH 6 curve (purple) are due to aggregation in the sample and do not result from actual conformational differences in the monomer. These results indicate that the pH dependent conformational changes are localized to the MC domain interface.

Modeling of the scattering data reveals the presence of a second Grp94-like population

From the P(r) data collected in this study and the pH titration analysis, it appears that a two state conformational equilibrium exists with one state, at high pH, represented by the previously described extended conformation (Figure 3(a)). If this hypothesis is true, the SAXS data collected at the other pH values should be fit by a linear combination of the previous model and a new model that represents the low pH state. To maximally use the information from all of the SAXS data, we decided to simultaneously fit all of the pH data as a linear combination of two conformations. To do so, we modified a previously developed rigid body modeling program PRFIT, that was used to fit the SAXS data of HtpG at high pH.27 In the modified version, rigid body refinement was performed on the low pH conformation while the extended state (the previously determined pH 9 state) was held fixed.

Figure 3
Modeling of the apo HtpG SAXS data reveals a second population of a more compact conformation. (a) Surface representations of the low pH state and the previously determined extended conformation.27 (b) Representative fit of the model (blue) to the experimental ...

In order to model the low pH state, we treated the NM domains as rigid bodies that pivot around the MC interfaces given that the NM domain does not exhibit pH-dependent conformational changes. We also maintained two-fold dimer symmetry. Residue 500 on the flexible linker between the MD and CTD was defined as the hinge point, and conformational space was systematically searched at defined angles about a general rotation axis as described in our previous work.27 We found that -70 to 70 degrees sufficiently covered the conformational space accessible to the HtpG dimer. At each refinement step, the best scoring linear combination for each pH condition was calculated, and the combination with the best overall R-factor for the six pH conditions was chosen as the best model (Figure 3(a) and (b)).

The P(r) data for each of the six pHs is very well fit by a linear combination of the extended, high pH state and the more closed compact low pH state (Table 1, and Supplemental Figure S3). The low pH model was also confirmed using negative stain EM (Figure 3(c) and (d)). EM images were collected for HtpG at pH 5.5 and pH 7.5. From examination of the 2-D images a significantly larger population of more compact particles is clearly present at low pH compared to neutral pH. High pH (pH 9) was also tested and extended particles predominated (data not shown). In order to better compare our compact, low pH SAXS model to the EM data we collected ~1,000 single particles and generated reference free class averages. Class averages were not generated for the neutral pH particle given the heterogeneity of the particles. Reference free alignments of the more compact particles were then compared to 2-D projections of the low pH SAXS model. Remarkably, when the averages are aligned to 2-D projections of our SAXS model the conformations appear nearly identical, confirming the structure of this compact state and establishing the accuracy of our SAXS modeling methods (Figure 3(d)).

Table 1
Populations of extended and low pH (Grp94-like) state for WT HtpG at varied pH

The low pH SAXS model is surprisingly reminiscent of the Grp94 crystal structure (Figure 4). While the Grp94 crystal structure and the low pH model (Grp94-like state) have similar open angles, the Grp94-like state has a more pronounced U shape while the Grp94 crystal structure more closely resembles a V shape. This difference is a result of rotations of both the MD and NTD of the Grp94-like state as depicted in Figure 4(b). While the two structures are similar the combinations with the low pH model fit the data best (Figure 4(c)). Linear combinations of the Grp94 crystal structure and the extended state gave higher R-factors overall (pH 6, 6.14; pH 6.5, 4.74; pH 7, 2.62; pH 7.5, 2.34; pH 8, 4.16) than mixtures with the low-pH model (pH 6, 5.52; pH 6.5, 2.57; pH 7, 3.17; pH 7.5, 2.07; pH 8, 2.8).

Figure 4
The low pH, Grp94-like, HtpG state closely resembles the Grp94 crystal structure. (a) Top-down and side views of the Grp94 crystal structure and the Grp94-like HtpG model. (b) Cartoon representation of the Grp94-like model depicting the rotations necessary ...

The AMPPNP-bound conformation of HtpG is dependent on pH

From previous SAXS and EM studies it is clear that nucleotide binding also significantly influences the conformational equilibrium of HtpG. The addition of AMPPNP at pH 8 shifts the conformation of HtpG from the apo state to an approximately 40:60 equilibrium between the apo state and the closed ATP state identified crystallographically.27; 31 Given the importance of pH in influencing the conformational equilibrium for apo HtpG, we sought to assess the effect of pH on the conformational equilibrium of the AMPPNP-bound state.

Scattering data was collected from pH 6 to 9 for HtpG in the presence of saturating levels of AMPPNP. As can be seen in Figure 5(a), there also exists a pH dependent conformational shift in the presence of nucleotide. Contrasting with the apo data, we were unable to fit the AMPPNP data with a simple two state titration model, suggesting that a third conformation may exist. To better define the source of the conformational equilibrium in the presence of AMPNP we examined the structure of the NM domain. SAXS data for NMHtpG in the presence of AMPPNP supports the presence of the AMPPNP-bound crystal state as an additional conformation in solution. At low pH, the NM domain scattering matches the single NTD-MD conformation found in the low and high pH apo states, while at high pH, the NMHtpG data is indicates that the structure of the AMPPNP-bound NM domain matches that seen in the full-length yeast crystal structure. The best fit was achieved with a mixture of monomeric and dimeric forms of the AMPPNP-bound NM domain (R-factor 3.5% with 74% dimer and 26% monomer, Supplemental Figure S4). A DAMMIN34 reconstruction also indicated the presence of the AMPPNP dimer at high pH (data not shown). This is the first observation of an NM dimer in solution and may help explain the residual ATPase activity for the isolated NM domain. To test for the yeast crystal state in full-length HtpG, each pH state was fit with a linear combination of three conformations, the two apo states (extended and Grp94-like) described above and the ATP state (closed) as seen in the yeast crystal structure (Table 2, Supplemental Figure S5). The three conformations were sufficient to provide good fits to the AMPPNP data, suggesting that at physiologic pH in the presence of AMPPNP, HtpG is in equilibrium between these three states.

Figure 5
The conformation of AMPPNP-bound HtpG is also affected by pH. (a) P(r) functions for HtpG in the presence of 10mM AMPPNP (pH 6, purple; pH 6.5, gold; pH 7, green; pH 7.5, red; pH 8, blue; pH 9, grey). (b) Relative populations of the closed state (pink), ...
Table 2
Populations of extended, Grp94-like, and closed state for HtpG:AMPPNP at varied pH

In the presence of AMPPNP, the amount of each conformation varies significantly with pH as seen in Figure 5(b). The dominant states are represented by the Grp94-like state and the closed state. Interestingly, the Grp94-like state is greatly stabilized relative to the extended state by the presence of AMPPNP especially around pH 7.5 (Figure 5(b) and (c)). The lack of the closed state at low pH is not, however, indicative of differential AMPPNP binding. Using a standard tryptophan fluorescence assay35 we confirmed that the binding of AMPPNP is independent of pH (data not shown). This leads to the question of what the roles of each of these conformations are in the functional cycle of HtpG.

Mutations at H446 shift the conformation of HtpG independently of pH

Given that the titration analysis of the apo conformational change suggests the involvement of 1-2 histidine residues, it is reasonable to assume that a histidine could be mutated to shift the conformation of apo HtpG regardless of pH. As previously discussed, histidines (H446, H530, H574) located at the MC interface (Figure 6(a)), are the most likely candidates for the responsible titratable group. We systematically mutated these groups to lysines to mimic the protonated state of the histidine expected at low pH. To insure that the mutations did not disrupt the dimerization interface, the molecular weight of each mutant was determined using multiangle light scattering. Both the H446K and H530K mutants formed dimers, whereas the H574K mutant was a monomer in solution (data not shown). Residue H574 was therefore not further pursued. The mutants were also examined by CD spectroscopy and no differences were seen between H446K, H530K and wild-type (WT) HtpG (data not shown) indicating that the overall secondary structure is unchanged.

Figure 6
H446 is primarily responsible for the observed pH dependent conformational equilibrium. (a) Cartoon depicting the location of histidine residues at the MC interface. The histidines are represented as red spheres with the NTD in blue, the MD in green, ...

The SAXS profiles for the H446K and H530K mutations were initially collected at pH 6 and 9, and the resulting P(r)s were compared to the P(r)s for the wild-type protein. The H446K mutation significantly shifts the conformational equilibrium towards the low pH state at both pH 6 and pH 9 (Figure 6(b)). This shift is consistent with the pH-dependence seen in the wild-type protein. The H530K mutation, on the other hand, shifts the conformational equilibrium towards the high pH state (Supplemental Figure 6(a)). We also examined H530K in the context of a non-titratable, yet polar, mutation of H446 (H530K/H446Q). This showed a complete shift to the high-pH extended state (Supplemental Figure 6(c)).

Because the H446K mutation is consistent with the dominant pH effect seen for the wild-type protein while H530K behaves in an opposite manner, we focused on residue 446 for further analysis. Modeling of the H446K scattering data shows that the solution state is a combination of ~80% of the low pH state and ~20% of the high pH state regardless of pH (Figure 7(b)). Residue 446 was then mutated to glutamic acid (H446E) and the conformation shifted completely to the extended state as predicted (Figure 6(c) and 7(b)). In the presence of AMPPNP, both H446K and H446E are still able to populate the closed ATP state (data not shown), suggesting that both the extended and Grp94-like states are able to convert to the closed state. Both mutations are also able to turn over ATP in an ATPase assay, however their overall activity is ~1/3 that of wild-type HtpG (Figure 6(d)).

Figure 7
H446K shifts the conformation to the Grp94-like state at neutral pH whereas H446E shifts the population to the extended conformation. (a) Single particle negative stain EM of H446K at pH 7.5 shows more compact particles that are identical to the 2-D projections ...

H446Q was made as a non-titratable mutation that should be able to act as either a hydrogen bond donor or acceptor. As such, we expected it to remove any pH dependence resulting from residue 446. Surprisingly, this mutant appears to be almost identical to the WT protein in terms of its pH dependence (Supplemental Figure 6(b)) suggesting that another residue must be involved (Figure 7(b)). Since, as discussed above, mutation of the only other relevant histidine residue (H530) produced an effect opposite to that observed with the wild type protein, this result appears paradoxical. Examination of the relevant crystal structures shows that H530 resides in a highly polar/charged environment in both the extended and Grp94-like states; hence the H530K mutation could result in interactions with neighboring groups that are quite distinct from those that interact with the wild-type histidine. Thus while H530 undoubtedly contributes to the overall pH effects seen in the wild-type protein, the complexity of the local environment makes it difficult to probe its contributions in a straightforward manner.

By contrast, our SAXS data indicate that H446E and H446K shift the conformation in a way consistent with the wild-type protein behavior. Given the potential importance of these mutations for testing the functional role of different HtpG conformations without having to shift the pH from neutrality, we sought confirmation of the SAXS results by negative-stain electron microscopy. As expected, the H446K mutant HtpG particles incubated at pH 7.5 were predominantly in a more compact state identical to wild type HtpG at low pH. Conversely, H446E particles at pH 7.5 were more extended and similar to wild type HtpG at high pH. Reference free alignments of the H446K particles match remarkably well to 2-D projections of the Grp94-like model, confirming that the compact structure seen for H446K is the same structure as seen in the WT HtpG at low pH (Figure 7(a)).

The different conformations have distinct functional roles in suppressing the aggregation of citrate synthase

Since it has been postulated that client proteins first bind to the apo state, of key importance is determining whether or not the two conformations are functionally distinct. To test this, we used the well-established citrate synthase (CS) aggregation assay. It has been shown that at 43°C CS aggregates in solution, and that this aggregation can be suppressed by adding Hsp90.36 This suppression of aggregation does not depend on nucleotide.4; 37

The CS assay emphasizes the importance of mutations that alter the conformation of Hsp90 while maintaining a constant pH. Changes in the pH of the system cause the aggregation behavior of CS to vary greatly (data not shown). The H446K and H446E mutations allow us to assess the role of Hsp90 conformation in the aggregation assay in the absence of any extraneous pH effects. As expected, the addition of HtpG at pH 7.5 greatly reduces CS aggregation (Figure 9). We tested both the H446K (Grp94-like conformation) and the H446E (extended conformation) mutants at pH 7.5 for their ability to prevent CS aggregation. As shown in Figure 9, H446K, the Grp94-like conformation, is considerably more effective than the WT protein in preventing the aggregation of CS whereas, H446E, the extended conformation, has little effect on the aggregation of CS. Moreover, WT activity corresponds to the fraction of the low pH state normally present at pH 7.5. Wild-type HtpG shows a slight aggregation over 18 minutes at 43°C but this aggregation is negligible compared to CS. Both of the mutants displayed no aggregation after 18 minutes at 43°C (Supplemental Figure S7). This data indicates that there are significant functional differences between the two conformations of HtpG.


It is becoming more and more apparent that structural dynamics play an important role in the function of Hsp90, and understanding the conformational ensemble is imperative for better defining its molecular mechanism. Large conformational changes occur throughout the ATPase hydrolysis cycle, and previous structural studies, as well as kinetic studies have shown that the conformational state is not rigorously determined by the bound nucleotide, but instead that nucleotide binding results in a conformational equilibrium between apo and nucleotide states.27; 29; 31 Conformational equilibria have been shown as functionally important in other systems as well. For example, NMR studies have revealed that the proto-oncogene Ras exists in a conformational equilibrium that can be shifted with point mutations. The equilibrium shift then promotes interactions with effector proteins leading to an increase in oncogenic potential.38 In another example, single molecule FRET studies revealed that even in the absence of substrate, adenylate kinase exists in both the open and closed state. Notably, the closed state, previously thought to exist only in the presence of substrate, is favored.39 These examples also support the realization that the functional pathway of Hsp90 must be considered as a stochastic process rather than as a deterministic ATP machine as has been observed for other chaperones.

Here, both SAXS and single particle EM have revealed an unanticipated solution conformation of apo HtpG that exists in a pH-dependent two-state conformational equilibrium with the previously described extended apo state. Interestingly, the apparent pKa for the conformational transition between the extended state and the Grp94-like state is 7.2 suggesting that under physiological conditions the system is optimized for maximum levels of both conformations. By using a linear combination of two structures together with rigid body modeling, we were able to dissect out the two populations of HtpG and to discover that the low pH conformation of HtpG is remarkably similar to the recent Grp94 crystal structure. While the exact rotation of the middle and N-terminal domains differs between our low pH structure and the Grp94 crystal structure, the overall architecture is surprisingly similar. Although our SAXS data is very well fit with a two state model, it is likely that a third state, the closed ATP state, is present as well. Recent single particle EM studies have directly observed the presence of the closed state under apo conditions.31 The closed state has also been identified under apo conditions for yeast by single-molecule FRET, and analysis of the energy barriers suggests that all states observed in the presence of nucleotide are accessible in the absence of nucleotide.32 If the closed state represents less than 5-10% of the total population, its effect on the SAXS modeling would be negligible. Therefore, the apo state of HtpG is most likely represented by the extended state, the Grp94-like state, and a small amount of the closed state. The conformational equilibrium of HtpG persists in the presence of saturating AMPPNP with the closed state now being significantly populated. The AMPPNP-bound equilibrium can also be shifted with pH with the Grp94-like conformation most populated at low pH and with the closed state most populated at high pH.

Given that the Grp94-like state has both conformational features of the extended (NTD-MD angle) and closed states (MD-CTD angle), it is possible that the low pH Grp94-like state (G) represents an intermediate step along the pathway proceeding from the extended model (E) to the closed model (C) (1). Alternatively, the Grp94-like state could be off-pathway (2) or the Grp94-

equation image

equation image

equation image

like and extended states could be combined in a three-way pathway with the closed state (3). Typically, kinetics measurements are used to determine the connectivity of a particular pathway. However it seemed that similar information should be extractible from the pH equilibrium data. We know from the mutant HtpG SAXS data in the presence of AMPPNP that both the extended and the Gpr94-like states can convert to the closed state suggesting that pathway (3) is the relevant pathway. Given that the only structural changes required for the transformation from the extended to the Grp94-like state occurs at the MC interface, we assumed that this pH-dependent equilibrium is unaffected by the addition of AMPPNP. With this assumption, we modeled the pathways using standard kinetic equations and confirmed that only a model allowing all three states to interconvert (3) can recapitulate the steady-state levels observed in the presence of AMPPNP. The modeling indicates that the conversion from the extended to Grp94-like state and the extended to closed state both have a similar pKa, whereas the conversion from the Grp94-like state to the closed state is independent of pH. This suggests that the main role of pH is to shift the equilibrium for the MC domain conformation. Because the Grp94-like state is stabilized in the presence of AMPPNP, the dominant pathway from the extended state to the ATP-state most likely goes through the Grp94-like state, despite the fact that all three species can interconvert. A recent hydrogen exchange study of HtpG further supports this idea with the demonstration of an intermediate state between the `relaxed' apo state and the `tense' ATP state.30

Using the histidine mutations H446K and H446E, we probed the functional role of the two apo conformations. Using a citrate synthase aggregation assay, we demonstrate that, at least for this substrate protein, the different apo conformations of HtpG are functionally quite distinct. The Grp94-like conformation is very efficient at preventing aggregation whereas the extended conformation has little or no effect. This result was surprising, in that we anticipated the extended conformation to be better able to accommodate the large (50kD), dimeric citrate synthase. The wild type protein showed an intermediate ability to suppress aggregation as expected for being a nearly 1:1 mixture of the two conformations. Since chaperone binding via exposed hydrophobic residues on citrate synthase is likely to be the main requirement for blocking aggregation, the observed behavior suggests that the Grp94-like conformation presents hydrophobic surfaces that are better aligned to those exposed on citrate synthase. Previous work has shown that Hsp90 interacts with structured unfolding intermediates of CS,4 and that this interaction only required the MC domain of Hsp90 and was independent of nucleotide.37 This correlates well with the results presented here showing that a change in only the MC orientation of HtpG affects the interaction with CS.

One possible mechanism for the interaction is that a locally unfolded region of CS prone to aggregation could interact deep in the cleft of the Grp94-like structure binding to the hydrophobic surfaces on both MD arms of HtpG and the amphipathic α-helices in the CTD. The close juxtaposition of MD and CTD hydrophobic surfaces would not occur in the extended HtpG state. In the wild-type protein where the two conformations are in equilibrium, the citrate synthase would be able to bind to only half of the HtpG present accounting for the intermediate level of activity that is seen as compared to the two mutants. When the population is already shifted in favor of the Grp94-like structure (H446K), binding and aggregation prevention are most efficient. By contrast, the H446E mutation is unable to prevent citrate synthase aggregation because it is unable to convert to the Grp94-like state even in the presence of citrate synthase. The inability of CS to shift the conformational equilibrium of HtpG also suggests that the interaction is weak and transitory. It is also possible that the CS/HtpG interaction is mediated by local changes in HtpG that occur between the two apo states. In order to distinguish these and other possible modes of interaction, more detailed structural studies of the interaction of HtpG with CS will be required.

In order to better understand the molecular mechanism for the conformational switch, we examined the residues contacting H446 in both the extended high pH model and the Grp94-like low pH model. Unfortunately, no obvious contacts were being made that would account for the pH dependence of the conformational change. Comparison of the analogous residue to H446 in the Grp94 structure (F554) provides some insight into the molecular basis of the conformational switch. In Grp94, F554 is nicely packed in a hydrophobic pocket comprised of residues that are conserved in HtpG as hydrophobic. The one non-conserved residue contacting F554 is a serine (S725) in Grp94 and a glutamate (E604) in HtpG. This residue could explain the preference of HtpG for the Grp94-like state only when the histidine is either protonated at low pH or replaced with a lysine. Both of these would be able to form a favorable electrostatic interaction with E604 whereas the H446E mutation would be expected to clash with E604 shifting the population towards the extended state as is observed.

The use of histidines to modulate the conformation of Hsp90 appears to be a uniquely bacterial trait. The histidines involved in HtpG are not conserved in higher organisms. The similarity of the functionally active apo HtpG state to the Grp94 structure, however, suggests that Grp94 may utilize the same structural pathway as other Hsp90 homologs, and that the different structural states are conserved from bacteria to higher eukaryotes. Given the evidence that the extended, ATP and ADP states are structurally conserved from bacteria to higher eukaryotes, it is probable that the intermediate Grp94-like conformation found here is also important in higher eukaryotes. Recent FRET studies have shown that yeast Hsp90 has at least two intermediate states between the already described open apo state and closed ATP state.32; 33 One possibility is that one of the intermediate states in yeast is related to the Grp94-like state described here, but this hypothesis will require further characterization of the yeast solution structure.

Our previous work27; 31 and the FRET studies demonstrate that nucleotide shifts the conformational equilibrium as opposed to explicitly determining the conformational state of Hsp90. An important difference between eukaryotes and prokaryotes may be the other factors that determine the conformational equilibrium. One hypothesis is that the equilibrium in eukaryotes has evolved to be controlled by client/cochaperone binding instead of pH. Bacteria may rely more heavily on the finely tuned conformational equilibrium to optimize the levels of both conformations because of the lack of cochaperones. It is also possible that the HtpG pH equilibrium acts as a sensor to tune chaperone response in times of metabolic stress. Identification of HtpG client proteins will assist in the development of in vivo HtpG functional assays and provide information as to the types of client proteins that interact with each conformational state.

The functional differences of the different apo states observed here may also be relevant to other homologs. The question remains as to the generalizability of the preferential binding of the Grp94-like structure to client proteins. One possibility is that different substrates are recognized by different conformations of Hsp90, and this structural diversity is therefore one mechanism by which Hsp90 recognizes such a broad set of substrate proteins. All of these questions will require further studies of the yeast and human proteins and their interactions with client proteins. The results discussed here with HtpG provide critical insights into the mechanism of substrate activation and the differences in regulation between different species, and these results are an important step in better understanding the chaperone cycle of Hsp90.

Materials and Methods

Protein Expression and Purification

Full-length HtpG (residues 1-624) was cloned into pET29b (Invitrogen) and the HtpG truncation mutant, NMHtpG (residues 1-495) was cloned into pET151 using the TOPO cloning system (Invitrogen). Point mutations were made using PCR and 30-35 nucleotide primers encoding for the mutated amino acid. Parental DNA was digested with Dpn1 after the completed reaction, and the resulting plasmid was transformed into TOP10 (Invitrogen) cells. All mutations were confirmed by sequencing of the appropriate regions. All constructs were expressed in E. coli using BL21(DE3) cells (Invitrogen). The full-length WT protein was purified by successive passage over DE52 (Whatman), monoQ and gel filtration (Amersham Biosciences) columns with a final buffer of 20mM Tris pH 7.5, 100mM NaCl. NMHtpG and the point mutations were first passed over Ni-NTA agarose and the collected peak was cleaved with TEV overnight. The cleaved protein was then purified by gel filtration (Amersham Biosciences) into a final buffer of 20 mM Tris pH 7.5 and 100 mM NaCl. All proteins ran as a single monodisperse peak on gel filtration. CD spectroscopy of the proteins at 20uM in 50mM Tris pH 8, 50 mM KCl, and 10 mM MgCl2 showed no difference in 2° structure between the different constructs. The purified proteins were collected and concentrated using 10 kD molecular weight cutoff Amicon Ultra spin concentrators (Millipore). Using a desalting column (GE Healthcare) the samples were exchanged into 50 mM buffer (MES pH 6, Bis-tris pH 6.5, or Tris pH 7, 7.5, 8, or 9), 50 mM KCl, 10 mM MgCl2, and 1 mM DTT. Using Millipore Ultrafree®-0.5 Centrifugal Filter Units, samples were concentrated to between 10 and 30mg/ml as estimated by absorbance at 280 nm. The A260 nm/A280 nm ratio was measured as < 0.6 giving confidence that no nucleotide was present in the purified sample. Samples were split into two aliquots with 10 mM AMPPNP added to one aliquot. Samples were then flash frozen in liquid nitrogen and stored at -80°C.

SAXS data collection

Initial SAXS measurements were collected at beamline 4-2 at the Stanford Synchrotron Radiation Laboratory (SSRL) and at the BioCAT beamline (18-ID) at the Advanced Photon Source (APS). Data reported here was collected at SSRL beamline 4-2 and at the SIBYLS beamline (12.3.1) at the Advanced Light Source (ALS). To minimize aggregation, samples were spun in a table top microcentrifuge for 5 minutes immediately before data collection. Data was then collected at 25° C at 2.5, 5 and 10mg/ml. At the ALS, samples were exposed for 7 and 70 seconds at a detector distance of 1.6m. At SSRL, samples were exposed for five 30 second exposures at a detector distance of 2.5m. The data was collected on a Mar165CCD detector, and the detector channels were converted to Q = 4πsinθ/λ, where 2θ is the scattering angle and λ is the wavelength, using a silver behenate sample as a calibration standard. The data was circularly averaged over the detector and normalized by the incident beam intensity.

SAXS data analysis

The raw scattering data, I(Q), were scaled and the buffers were subtracted. Variations were seen in the amplitude of the high Q data for data sets that should be identical. To correct for this variation, data from multiple concentrations were linearly extrapolated to the buffer background signal. The experimental buffers were then scaled to match the calculated buffer background signal. This correction was only necessary for precise comparison to molecular models and did not affect the overall shape of either the scattering curves or the interatomic distance profiles (P(r)). Individual scattering curves for a given nucleotide condition collected at different concentrations and over different scattering angle ranges were scaled and merged to yield a single low-noise average scattering curve. The radii of gyration (Rg) were initially calculated from the Guinier plot as implemented in the program PRIMUS.40 The interatomic distance distribution functions (P(r)) were then calculated using the program GNOM.41 Dmax was determined by constraining rmax to zero and then varying rmax between 150 and 200Å. Dmax was chosen as the rmax where the P(r) smoothly approached zero based upon perceptual criteria. The overall shape of the P(r) function was very robust for Dmax ± 10Å. The molecular modeling of the scattering data was then done using the in house software PRFIT as previously described.27

Electron Microscopy

HtpG protein was negatively stained with uranyl format on thin carbon layered (40-50 Å thick) 400 mesh copper grids (Pelco) as described.31; 42 Prior to staining, wildtype and H446K HtpG (200 nM) were incubated at 37° for 10 minutes in either 20 mM Tris pH 7.5 or 20 mM MES pH 5.5 and 50 mM KCl. Samples were imaged using a Tecnai G2 Spirit TEM (FEI) operated at 120 keV. Micrograph images were recorded using a 4k × 4k CCD camera (Gatan) at 68,000x magnification with 2.21 Å pixel size.

Two-dimensional single particle analysis was carried out using the EMAN software package.43 Reference-free class averages were determined using starnrclasses from a set of 811 and 1109 single particles (collected from 9 and 20 micrographs, respectively) for wildtype (pH 5.5, Figure 3D) and H446K (pH 7.5, Figure 7A) respectively. This method involves a k-means clustering analysis where the particles are separated into a given number of classes and no other information, such as the SAXS models, were used for alignment or as a reference to generate the image. This was critical for generating unbiased averages that were not influenced by known structures. Between 1 and 5% of the total particles collected were aligned in each of the average images. This was expected given the apparent random orientation of the particles on the grid. The class averages (27 and 22, respectively) were then low-pass filtered (25 Å) and aligned and rotated (using classesbymra and align2d) to 2-D projections of the low-pH HtpG SAXS model for qualitative comparison.

Citrate Synthase Assay

Citrate synthase aggregation was monitored with a Jobin Yvon FluoroMax-3 fluorescence spectrophotomer with a temperature controlled jacket and automatic stirring. The excitation and emission wavelengths were both set to 500 nm with slit widths of 2 nm. Citrate synthase was purchased from Sigma and prepared for thermal aggregation as described previously.36 HtpG and variants at 300 nM dimer were incubated for 5 minutes at 43 °C, 40 mM HEPES pH 7.5 before rapidly diluting CS 1:100 to a final concentration of 75 nM dimer and monitoring the resulting light scattering.

Figure 8
H446K, the Grp94-like state, most efficiently blocks the aggregation of citrate synthase (CS). CS was incubated at 43°C by itself (red) or in the presence of WT or mutant HtpG (CS with WT HtpG, green; CS with H446K, black; CS with H446E, blue), ...

Supplementary Material


Funding for this project was provided by the Howard Hughes Medical Institute and a UC Discovery Grant (bio03-10401/Agard). K.A.K. was supported by a NSDEG graduate fellowship. D.R.S. was supported by an American Cancer Society postdoctoral fellowship. T.O.S was supported by a Damon Runyon Cancer Research Foundation fellowship. We wish to thank the staff at the Advanced Light Source SIBYLS beamline 12.3.1 (especially G. Hura), the Stanford Synchrotron Radiation Laboratory beamline 4-2 (especially H. Tsuruta), and the Advanced Photon Source Biocat beamline 18-ID. A special thanks to U. Boettcher, C. Cunningham, L. Lavery, and N. Salimi for many helpful discussions and comments on the manuscript.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


1. Young JC, Agashe VR, Siegers K, Hartl FU. Pathways of chaperone-mediated protein folding in the cytosol. Nat. Rev. Mol. Cell. Biol. 2004;5:781–91. [PubMed]
2. Weissman JS, Hohl CM, Kovalenko O, Kashi Y, Chen S, Braig K, Saibil HR, Fenton WA, Horwich AL. Mechanism of GroEL action: productive release of polypeptide from a sequestered position under GroES. Cell. 1995;83:577–87. [PubMed]
3. Kurt N, Rajagopalan S, Cavagnero S. Effect of hsp70 chaperone on the folding and misfolding of polypeptides modeling an elongating protein chain. J. Mol. Biol. 2006;355:809–20. [PMC free article] [PubMed]
4. Jakob U, Lilie H, Meyer I, Buchner J. Transient interaction of Hsp90 with early unfolding intermediates of citrate synthase. Implications for heat shock in vivo. J. Biol. Chem. 1995;270:7288–94. [PubMed]
5. McLaughlin SH, Smith HW, Jackson SE. Stimulation of the weak ATPase activity of human hsp90 by a client protein. J. Mol. Biol. 2002;315:787–98. [PubMed]
6. Freeman BC, Yamamoto KR. Disassembly of transcriptional regulatory complexes by molecular chaperones. Science. 2002;296:2232–5. [PubMed]
7. Picard D. Heat-shock protein 90, a chaperone for folding and regulation. Cell. Mol. Life Sci. 2002;59:1640–8. [PubMed]
8. Pratt WB, Toft DO. Regulation of signaling protein function and trafficking by the Hsp90/Hsp70-based chaperone machinery. Exp. Biol. Med. (Maywood) 2003;228:111–33. [PubMed]
9. Richter K, Buchner J. Hsp90: chaperoning signal transduction. J. Cell. Physiol. 2001;188:281–90. [PubMed]
10. Young JC, Moarefi I, Hartl FU. Hsp90: a specialized but essential protein-folding tool. J. Cell Biol. 2001;154:267–73. [PMC free article] [PubMed]
11. Zhao R, Davey M, Hsu YC, Kaplanek P, Tong A, Parsons AB, Krogan N, Cagney G, Mai D, Greenblatt J, Boone C, Emili A, Houry WA. Navigating the chaperone network: an integrative map of physical and genetic interactions mediated by the Hsp90 chaperone. Cell. 2005;120:715–27. [PubMed]
12. Caplan AJ, Mandal AK, Theodoraki MA. Molecular chaperones and protein kinase quality control. Trends Cell Biol. 2007;17:87–92. [PubMed]
13. Pearl LH, Prodromou C. Structure and mechanism of the Hsp90 molecular chaperone machinery. Annu. Rev. Biochem. 2006;75:271–94. [PubMed]
14. Picard D. Chaperoning steroid hormone action. Trends Endocrinol. Metab. 2006;17:229–35. [PubMed]
15. Neckers L. Heat shock protein 90: the cancer chaperone. J. Biosci. 2007;32:517–30. [PubMed]
16. Chiosis G, Lucas B, Huezo H, Solit D, Basso A, Rosen N. Development of purine-scaffold small molecule inhibitors of Hsp90. Curr. Cancer Drug Targets. 2003;3:371–6. [PubMed]
17. Neckers L, Ivy SP. Heat shock protein 90. Curr. Opin. Oncol. 2003;15:419–24. [PubMed]
18. Workman P. Combinatorial attack on multistep oncogenesis by inhibiting the Hsp90 molecular chaperone. Cancer Letters. 2004;206:149–57. [PubMed]
19. Solit DB, Osman I, Polsky D, Panageas KS, Daud A, Goydos JS, Teitcher J, Wolchok JD, Germino FJ, Krown SE, Coit D, Rosen N, Chapman PB. Phase II trial of 17-allylamino-17-demethoxygeldanamycin in patients with metastatic melanoma. Clin. Cancer Res. 2008;14:8302–7. [PMC free article] [PubMed]
20. Bukau B, Horwich AL. The Hsp70 and Hsp60 Chaperone Machines. Current Biology. 2002;92:351–66. [PubMed]
21. Shiau AK, Harris SF, Southworth DR, Agard DA. Structural analysis of E. coli Hsp90 reveals dramatic nucleotide-dependent conformational rearrangements. Cell. 2006;127:329–40. [PubMed]
22. Ali MM, Roe SM, Vaughan CK, Meyer P, Panaretou B, Piper PW, Prodromou C, Pearl LH. Crystal structure of an Hsp90-nucleotide-p23/Sba1 closed chaperone complex. Nature. 2006;440:1013–7. [PubMed]
23. Prodromou C, Roe SM, O'Brien R, Ladbury JE, Piper PW, Pearl LH. Identification and structural characterization of the ATP/ADP-binding site in the Hsp90 molecular chaperone. Cell. 1997;90:65–75. [PubMed]
24. Stebbins CE, Russo AA, Schneider C, Rosen N, Hartl FU, Pavletich NP. Crystal structure of an Hsp90-geldanamycin complex: targeting of a protein chaperone by an antitumor agent. Cell. 1997;89:239–50. [PubMed]
25. Meyer P, Prodromou C, Hu B, Vaughan C, Roe SM, Panaretou B, Piper PW, Pearl LH. Structural and functional analysis of the middle segment of hsp90: implications for ATP hydrolysis and client protein and cochaperone interactions. Mol. Cell. 2003;11:647–58. [PubMed]
26. Dollins DE, Warren JJ, Immormino RM, Gewirth DT. Structures of GRP94-nucleotide complexes reveal mechanistic differences between the Hsp90 chaperones. Mol. Cell. 2007;28:41–56. [PMC free article] [PubMed]
27. Krukenberg KA, Förster F, Rice LM, Sali A, Agard DA. Multiple conformations of E. coli Hsp90 in solution: insights into the conformational dynamics of Hsp90. Structure. 2008;16:755–65. [PMC free article] [PubMed]
28. Putnam CD, Hammel M, Hura GL, Tainer JA. X-ray solution scattering (SAXS) combined with crystallography and computation: defining accurate macromolecular structures, conformations and assemblies in solution. Quarterly Reviews of Biophysics. 2007;40:191–285. [PubMed]
29. Frey S, Leskovar A, Reinstein J, Buchner J. The ATPase cycle of the endoplasmic chaperone Grp94. J. Biol. Chem. 2007;282:35612–20. [PubMed]
30. Graf C, Stankiewicz M, Kramer G, Mayer MP. Spatially and kinetically resolved changes in the conformational dynamics of the Hsp90 chaperone machine. EMBO J. 2009;28:602–13. [PubMed]
31. Southworth DR, Agard DA. Species-dependent ensembles of conserved conformational States define the Hsp90 chaperone ATPase cycle. Mol. Cell. 2008;32:631–40. [PMC free article] [PubMed]
32. Mickler M, Hessling M, Ratzke C, Buchner J, Hugel T. The large conformational changes of Hsp90 are only weakly coupled to ATP hydrolysis. Nat. Struct. Mol. Biol. 2009;16:281–6. [PubMed]
33. Hessling M, Richter K, Buchner J. Dissection of the ATP-induced conformational cycle of the molecular chaperone Hsp90. Nat. Struct. Mol. Biol. 2009;16:287–93. [PubMed]
34. Svergun DI. Restoring low resolution structure of biological macromolecules from solution scattering using simulated annealing. Biophys. J. 1999;76:2879–86. [PubMed]
35. Csermely P, Kajtar J, Hollosi M, Jalsovszky G, Holly S, Kahn CR, Gergely P, Jr., Soti C, Mihaly K, Somogyi J. ATP induces a conformational change of the 90-kDa heat shock protein (hsp90) J Biol. Chem. 1993;268:1901–7. [PubMed]
36. Buchner J, Grallert H, Jakob U. Analysis of chaperone function using citrate synthase as nonnative substrate protein. Methods Enzymol. 1998;290:323–38. [PubMed]
37. Scheibel T, Weikl T, Buchner J. Two chaperone sites in Hsp90 differing in substrate specificity and ATP dependence. Proc. Natl. Acad. Sci. USA. 1998;95:1495–9. [PubMed]
38. Spoerner M, Wittinghofer A, Kalbitzer HR. Perturbation of the conformational equilibria in Ras by selective mutations as studied by 31P NMR spectroscopy. FEBS Letters. 2004;578:305–10. [PubMed]
39. Hanson JA, Duderstadt K, Watkins LP, Bhattacharyya S, Brokaw J, Chu JW, Yang H. Illuminating the mechanistic roles of enzyme conformational dynamics. Proc. Natl. Acad. Sci. 2007;104:18055–60. [PubMed]
40. Konarev PV, Volkov VV, Sokolova AV, Koch MHJ, al. e. PRIMUS: a Windows PC-based system for small-angle scattering data analysis. J. Appl. Crystallogr. 2003;36:1277–82.
41. Svergun DI. Determination of the regularization parameter in indirect-transform methods using perceptual criteria. J. Appl. Crystallogr. 1992;25:495–503.
42. Ohi M, Li Y, Cheng Y, Walz T. Negative Staining and Image Classification - Powerful Tools in Modern Electron Microscopy. Biol. Proceed. Online. 2004;6:23–34. [PMC free article] [PubMed]
43. Ludtke SJ, Baldwin PR, Chiu W. EMAN: semiautomated software for high-resolution single-particle reconstructions. J. Struct. Biol. 1999;128:82–97. [PubMed]
44. Kleywegt GJ. Use of non-crystallographic symmetry in protein structure refinement. Acta Crystallogr. D Biol. Crystallogr. 1996;52:842–57. [PubMed]
45. Cunningham CN, Krukenberg KA, Agard DA. Intra- and Intermonomer Interactions Are Required to Synergistically Facilitate ATP Hydrolysis in Hsp90. J. Biol. Chem. 2008;283:21170–8. [PMC free article] [PubMed]