Co-chaperone Regulation of Conformational Switching in the Hsp90 ATPase Cycle*

ATP hydrolysis by the Hsp90 molecular chaperone requires a connected set of conformational switches triggered by ATP binding to the N-terminal domain in the Hsp90 dimer. Central to this is a segment of the structure, which closes like a “lid” over bound ATP, promoting N-terminal dimerization and assembly of a competent active site. Hsp90 mutants that influence these conformational switches have strong effects on ATPase activity. ATPase activity is specifically regulated by Hsp90 co-chaperones, which directly influence the conformational switches. Here we have analyzed the effect of Hsp90 mutations on binding (using isothermal titration calorimetry and difference circular dichroism) and ATPase regulation by the co-chap-erones Aha1, Sti1 (Hop), and Sba1 (p23). The ability of Sti1 to bind Hsp90 and arrest its ATPase activity was not affected by any of the mutants screened. Sba1 bound in the presence of AMPPNP to wild-type and ATPase hyperactive mutants with similar affinity but only very weakly to hypoactive Isothermal Titration Calorimetry and K d Determinations— Heats of interaction were measured on a MSC system (Microcal) with a cell volume of 1.458 ml. For Sti1 interactions, 10 aliquots of 27 (cid:3) l of 90 (cid:3) M Hsp90, T22I, T101I, or A107N were injected into 6 (cid:3) M Sti1, or 43.8 (cid:3) M F349A was injected into 4 (cid:3) M Sti1 at 30 °C in 40 m M Tris, pH 8.0, containing 1 m M EDTA and 5 m M NaCl. For Aha1 interactions, 15 aliquots of 20 (cid:3) l of 220 (cid:3) M Aha1 were injected into 22 (cid:3) M Hsp90, T22I, T101I, or A107N, or 128.5 (cid:3) M Aha1 was injected into 12.85 (cid:3) M F349A, or 324 (cid:3) M NAha1 was injected into 32.4 (cid:3) M Hsp90 at 30 °C in 40 m M Tris, pH 8.0, containing 1 m M EDTA and 5 m M NaCl. For Sba1 interactions, 15 aliquots of 20 (cid:3) l of 150 (cid:3) M Sba1 were injected into 30 (cid:3) M Hsp90 N-terminal domain, T22I, or F349A mutant, or 140.7 (cid:3) M Sba1 was injected into 28.14 (cid:3) M Hsp90, T101I, or A107N at 30 °C in 40 m M Tris, pH 8.0, containing 1 m M EDTA, 5 m M NaCl, 7 m M MgCl 2 , and 5 m M AMPPNP. Sba1 experiments without magnesium-AMPPNP were per- formed by injecting 15 (cid:3) l aliquots of 150 (cid:3) M Sba1 into 30 (cid:3) M Hsp90 or A107N. Heats of dilution were determined in a separate experiment by diluting protein into buffer, and the corrected Hsp90 and Sba1 in the absence of AMPPNP so that the obs spectrum cannot be reproduced exactly. However, the results indicate that the proteins interact to form a complex (Sba1-AMPPNP- Hsp90). B , near-UV CD spectra for a Cp50-Hsp90-AMPPNP-Sba1 (50, 100, and 25 (cid:3) M , respectively) protein mixture ( obs ). The simulated spectra for when one component remains unbound are represented by the curves a , Sba1 unbound, and b , Cp50 unbound. Curve c represents no interaction between all the components, and curve d represents interactions between all the components (Sba1-AMPPNP-Hsp90-Cp50 complex). sim represents the simulated spectrum that matches the observed spectrum most closely and suggests that Sba1 and Cp50 compete for binding to Hsp90. sim assumes the formation of a protein mixture consisting of 20% Hsp90-AMPPNP-Cp50, 40% Hsp90-Cp50, and 40% Hsp90-AMPPNP. Within this mixture 40% of the AMPPNP, 40% Cp50, and 60% Sba1 remained unbound.

The molecular chaperone Hsp90 is responsible for the in vivo activation or maturation of specific client proteins (reviewed in Refs. [1][2][3][4]. Crucial to such activation is the essential ATPase activity of Hsp90 (5), which drives a conformational cycle involving transient association of the N-terminal nucleotide-binding domains within the Hsp90 dimer (6). A variety of studies have shown that Hsp90 is structurally and biochemically related to DNA-gyrase B and MutL (7)(8)(9)(10). Structures of co-crystals of GyrB (11) and MutL (12) with the non-hydrolyzable ATP analogue AMPPNP, 1 identify a segment of the commonly conserved N-terminal nucleotide-binding domain, that acts as an "ATP lid," closing over the mouth of the nucleotide-binding pocket when ATP is present. Hsp90 also possesses a potential lid segment (Gly 100 -Gly 121 in yeast Hsp90), which is larger than the equivalents in GyrB and MutL, and unlike those has a fully ordered open conformation in the absence of nucleotide. Although the structure of Hsp90 with bound ATP (or AMPNP) has not yet been directly observed, biochemical and mutagenesis studies (6) suggest that, as with GyrB and MutL, the lid segment in Hsp90 closes over bound ATP facilitating N-terminal dimerization (6,13) and docking with the middle segment of the chaperone (10) to form an active ATPase conformation.
A number of mutations isolated either through genetic screens (14) or deliberately engineered, have been found to affect the ATPase activity of Hsp90 in ways consistent with the proposed ATP lid mechanism (6). Thus the T101I mutation within the lid segment appears to stabilize the ATP lid in the open conformation seen in crystal structures of the isolated N-terminal domain (9,15,16), substantially decreasing ATPase activity. A107N, also within the lid segment, probably stabilizes the closed conformation greatly enhancing ATP turnover and N-terminal dimerization. T22I, which also enhances ATPase activity, probably does not affect lid closure directly but favors consequent association of the N-terminal domains. Mutations in residues outside the N-terminal nucleotide-binding domain, in particular Phe 349 in the middle segment, have also been found to have significant effects on ATPase activity (10). How the F349A mutation causes a substantial drop in catalytic activity is not certain, but it is likely that Phe 349 contributes to the interface between the middle segment and the exposed hydrophobic face of the ATP lid in its closed state.
The inherent ATPase activity of Hsp90 is also greatly influenced by some of the co-chaperones with which it cooperates in the activation of client proteins in vivo. Aha1 (and the related Hch1) is an activator of the ATPase of Hsp90 able to stimulate the inherent activity of yeast Hsp90 by ϳ12-fold and human Hsp90␤ by ϳ50-fold (17). Biochemical studies have shown that Aha1 binds to the middle region of Hsp90 (17,18), and recent structural studies of the Aha1-Hsp90 core complex suggest that the co-chaperone promotes a conformational switch in the middle segment catalytic loop (370 -390) of Hsp90 that releases the catalytic Arg 380 and facilitates its interaction with ATP in the N-terminal nucleotide-binding domain (19). The co-chaperones Sti1 (Hop in mammals) and p50 cdc37 , both of which are involved in the recruitment of client proteins to the Hsp90 system, are able to arrest the ATPase cycle of Hsp90 to facilitate client protein loading (20,21). Recent structural studies have shown that p50 cdc37 achieves this arrest by binding to surfaces of the Hsp90 N-domain implicated in ATP-dependent N-terminal dimerization and association with the middle segment, fixing the ATP lid in an open conformation and preventing a transactivating interaction of the N-domains (22). The mechanism of ATPase arrest by Sti1 is not yet fully understood but is also likely to involve an interaction with the N-terminal domains of the Hsp90 dimer (20,23). Sba1 (p23 in mammals), which binds preferentially to Hsp90 in its ATP-bound state (6,8,13,24), is also able to slow the ATPase cycle (17) presumably by stabilizing the N terminally dimerized conformation. However, as Sba1 binding is dependent on prior binding of ATP, ATPase activity can never be totally inhibited. Where and how Sba1 binds specifically to the N terminally dimerized Hsp90 is still unknown. To gain further insight into the molecular mechanism of Hsp90, we have now analyzed the effect of various Hsp90 mutations on the ability of co-chaperones to bind to Hsp90 and regulate its ATPase activity.

Sti1 Binding and ATPase Regulation of Hsp90
Mutants-Sti1 is a potent inhibitor of the Hsp90 ATPase activity, but the precise mechanism by which it achieves this is unknown (20,23). The stoichiometry of the Hsp90-Sti1 interaction as determined by isothermal titration calorimetry (ITC) is most probably 1:1 (Fig. 1), indicating that one Sti1 dimer interacts with one dimeric Hsp90 molecule, which is consistent with previous observations (20). The K d for the binding of Sti1 to wild-type Hsp90 was determined by ITC as 0.24 M (Fig. 1A), consistent with previous results (20). K d values determined with Hsp90 mutants (T22I, K d ϭ 0.18 M; T101I, K d ϭ 0.18 M; A107N, K d ϭ 0.41 M; F349A, K d ϭ 0.13 M) suggest that none of the mutations significantly modify the binding of Sti1 (Fig. 1).
Although the binding of Sti1 was unaffected, its ability to inhibit the ATPase activity of the Hsp90 mutants might still be compromised. ATPase assays showed that Sti1 inhibited the ATPase activity of the T22I and A107N mutants to the same degree as wild-type ( Fig. 2A). Sti1 inhibition assays with the Hsp90 T101I and F349A mutants were not possible because of their inherently low ATPase activity.
Although the primary binding site for Sti1 maps to the extreme C terminus of Hsp90, the perturbation of the CD spectrum of Hsp90-bound geldanamycin by Sti1 clearly indicates that ATPase arrest by Sti1 involves a direct interaction with FIG. 2. ATPase activity of Hsp90 tsmutants and their regulation by cochaperones. Sti1 (A) and Sba1 (B) inhibition of the ATPase activity of Hsp90 ts-mutants showing that their ability to inhibit the A107N and T22I mutants is unaffected. C, activation of Hsp90 ts-mutants by Aha1. D, activation of the same mutants by Aha1 plotted as the -fold activation. The results show that Aha1 activation of the T101I mutant is normal, whereas the A107N mutation significantly reduces the requirement for Aha1, that the T22I mutant is less responsive to Aha1, and finally that Aha1 effectively suppresses the F349A mutation. the N-terminal domain (20) but without preventing nucleotide binding (23). That the A107N mutation does not affect Sti1 binding to Hsp90, nor the ability of Sti1 to arrest the ATPase when bound, suggests that the mechanism of ATPase arrest by Sti1 is distinct from that of p50/Cdc37, which binds specifically to the open conformation of the ATP lid in the N-domain and prevents its closure (22). Furthermore, the lack of the effect of the F349A mutation on Sti1 binding also suggests that arrest by Sti1-mediated ATPase arrest does not involve interference with communications between the N-terminal domain and the middle segment of Hsp90. It is likely that structural studies will be required to fully understand the inhibitory function of Sti1.
Sba1 Binding and ATPase Regulation of Hsp90 Mutants-In the presence of saturating AMPPNP, the binding of Sba1 to mutants previously shown to favor an association of the Nterminal domains (T22I, K d ϭ 2.91 M; A107N, K d ϭ 3.69 M) was comparable to that with wild-type Hsp90 (K d ϭ 1.75 M) (Fig. 3). As binding to the T22I mutant evolved relatively small heats and therefore gave a less reliable ITC, its affinity for Sba1 was also measured by difference CD, (K d ϭ 1.5 M), and found to be similar to that of wild-type Hsp90 (ITC, K d ϭ 1.75 M; CD, K d ϭ 1.16 M) (Figs. 3, A and B and 4, A-E). Unexpectedly, in all cases the observed stoichiometry for the interaction of Sba1 with Hsp90, in both ITC and CD measurements, was close to 1:2 (Sba1:Hsp90) (Figs. 3 and 4, A-C). As previously seen for wild-type Hsp90 (17), the ATPase activity of the T22I and A107N mutants could be partially inhibited by Sba1 (Fig. 2B) (15), so that their lower affinity for Sba1 is not because of a defect in ATP binding itself but rather because of an impaired ability to adopt a stable N terminally dimerized conformation on the binding of ATP to which Sba1 then binds. ATPase assays with the F349A and T101I mutants were not conducted because of their inherently low ATPase activities and lack of interaction with Sba1.
The ATP dependence of Sba1/p23 binding to Hsp90 is well known (25, 26) but has not previously been quantified. When AMPPNP was omitted, Sba1 binding to wild-type Hsp90 was ϳ70-fold weaker (K d ϭ 120 M in ITC). Tighter binding was observed with the A107N mutant (K d ϭ 19.2 M) (Fig. 5), which displays an enhanced N-terminal dimerization (6), but this was still significantly weaker than the affinity for wild-type or A107N Hsp90s in the presence of AMPPNP. These results are fully consistent with previous suggestions that Sba1 (and p23) binding to Hsp90 is not dependent on ATP binding to Hsp90 per se but is dependent on the N-terminal association that ATP binding promotes (13,15). Most surprising however was the stoichiometry observed for the Sba1-Hsp90 interaction in both ITC and ⌬CD measurements, which showed that a single Sba1 molecule binds to an Hsp90 dimer. Monomer-dimer interactions are rare, but not unknown (27, 28), and usually involve a quasi-symmetric bridging interaction in which equivalent sites in the dimer interact with the monomeric ligand but in nonidentical ways. In the case of Hsp90 and Sba1 the sites involved are yet to be determined. However, such an interaction would clearly stabilize Hsp90-client protein complexes by promoting the N-terminal dimeric state. The monomeric state of Sba1 is consistent with human p23 studies (29).
Aha1 Binding and ATPase Activation of Hsp90 Mutants-The stress-regulated co-chaperone Aha1 has previously been shown to be a strong activator of the ATPase activity of Hsp90 (17). The K d for the binding of Aha1 to wild-type Hsp90 measured here (0.65 M) (Fig. 6) was consistent with previous esti-  (Fig. 6, A-E). The N-terminal domain of Aha1 also bound to Hsp90 (K d ϭ 1.75 M) with an affinity similar to the intact Aha1 (Fig. 6, A and F) indicating that NAha1 is primarily responsible for binding to Hsp90. In all cases Aha1 bound to Hsp90 with a 1:1 stoichiometry consistent with the structural studies of the Hsp90-Aha1 core complex (19).
Although the binding of Aha1 was not affected by the Hsp90 mutations, the degree to which Aha1 was able to stimulate their inherent ATPase activity varied significantly (Fig. 2, C  and D). For the T101I Hsp90 mutant, which has a much lower inherent ATPase than the wild-type does, the Aha1-stimulated ATPase activity was significantly lower in absolute value than the Aha1-stimulated activity of the wild-type, but the -fold activation achieved was very similar. We have previously suggested that the decreased activity of the T101I mutant results from an increased stability of the open conformation of the ATP lid, in which Thr 101 is packed on the underside of the lid in a hydrophobic environment. The essentially wild-type -fold activation achieved by Aha1 binding to T101I suggests that Aha1 exerts its stimulatory effect at a point in the ATPase after lid closure has occurred.
The two previously characterized ATPase-activating mutations T22I and A107N (6) were both susceptible to activation by Aha1, but the combined influence of the stimulatory mutation and co-chaperone binding was relatively weak without any strong synergy. As with the deactivating T101I, the relative lack of effect of the T22I and A107N mutations on Aha1 activation is again consistent with Aha1 acting at a point in the ATPase pathway after lid closure and N-terminal domain association.
In contrast, the effect of Aha1 on the virtually ATPase-dead F349A mutant was dramatic, effectively suppressing the effect of the mutation. Unlike the other residues for which mutants have been assessed, Phe 349 is not in the N-terminal nucleotidebinding domain itself but is part of a hydrophobic patch exposed on the surface of the first ␣-␤-␣ domain of the middle segment of Hsp90 (10). Modeling of the ATP-bound closed dimeric conformation of Hsp90 suggests that this hydrophobic patch forms a key part of the complex-transient interface formed between the N-terminal domain with its bound ATP and the middle segment catalytic loop. By analogy with other GHKL ATPases such as GyrB and MutL (11,12), the assembly of this interface is essential to allow correct positioning of the middle segment catalytic residue Arg 308 , which interacts with the ␥-phosphate of the nucleotide to promote the key catalytic step of the ATPase reaction, hydrolysis of the ␤-␥ phosphodiester bond. Mutation of Phe 389 would be expected to affect this interface, preventing the formation of the catalytically active conformation of Arg 308 , with consequently severe impact on ATP hydrolysis, as observed. The ability of Aha1 to suppress the effect of the F389A mutation, suggests that it acts to facilitate formation of the catalytic active conformation of Arg 308 .
Insight into the mechanism by which Aha1 achieves this has come from the recent determination of the structure of a complex between the N-terminal domain of Aha1 (equivalent to the whole of Hch1) and the middle segment of Hsp90 (19). The binding of NAha1 elicits a substantial remodeling of the middle segment catalytic loop in Hsp90 (Pro 375 -Ile 388 ) so that the cat-  ϭ 120 Ϯ 10.1 M). B, near-UV CD difference spectra for the Hsp90-Sba1 (100 and 50 M, respectively) interaction without magnesium-AMPPNP. Obs, the observed spectrum for an Sba1 and Hsp90 protein mixture; sim, the simulated spectrum without interaction between Sba1 and Hsp90 showing that the spectrum differs to that observed. This is consistent with molecular interaction between Sba1 and Hsp90. C, ITC of Sba1 injected into the A107N ts-mutant of Hsp90. The K d of this interaction was measured at 19.2 Ϯ 2.5 M indicating that the binding of Sba1 is favored by the A107N mutation over that with Hsp90 (K d ϭ 120 Ϯ. alytic Arg 308 is released from a retracted conformation observed in the structure of the unliganded middle segment. Compatibility of Co-chaperones in Complex with Hsp90 -The number of different Hsp90 co-chaperones identified offers the theoretical possibility of a very wide range of different Hsp90-co-chaperone complexes. However it is clear that some co-chaperones only bind to specific conformational states of Hsp90 (8, 13, 20 -22, 24), whereas others have common binding sites and are therefore mutually exclusive (30). The protein kinase specific co-chaperone p50 cdc37 had previously been thought to bind to Hsp90 at a site overlapping but not identical with the tetratricopeptide repeat-domain binding site in the C terminus (30,31). However subsequent structural studies (22) have shown that p50 cdc37 binds to the N-terminal nucleotidebinding domain, reinforcing an ATPase-arrested conformation of Hsp90 (21). To gain some further insight into co-chaperone compatibility we have used difference circular dichroism to analyze formation of Hsp90-based complexes in the presence of p50 cdc37 and other co-chaperones.
When Hsp90, Aha1, and the C-terminal Hsp90-binding region of p50 cdc37 (Cp50) were simultaneously present, the observed difference CD spectrum could not be simulated by any combination of observed spectra for the components and their pairwise complexes in which one component remained unbound. The observed spectrum was, however, satisfactorily simulated by linear combination of the observed spectra for Hsp90-Aha1 and Hsp90-Cp50 pairwise complexes (Fig. 7, A  and B). This is most simply explained by formation of an Hsp90-bridged three-way complex in which Cp50 and Aha1 make no direct mutual interaction. Consistent with this, the order of addition of Cp50 and Aha1 had no affect on the nature of the complex formed (Fig. 7B). The structurally distinct binding sites on Hsp90 for these co-chaperones would allow for such a three-way complex. p50/Cdc37 is responsible for the recruitment of kinases into the Hsp90 complex, whereas deletion of Aha1 can prevent the activation of v-Src protein kinase (17). Thus it is possible that in vivo the ATPase activity of Hsp90 in kinase complexes, which is down-regulated by p50/Cdc37, may be posed for activation by Aha1 upon displacement of p50/ Cdc37. Interestingly we also observed that Cp50 could form a complex with Hsp90-AMPPNP (Fig. 7C).
We next investigated the ability of Sba1, AMPPNP, and Cp50 to bind simultaneously to Hsp90. As with the Cp50-Hsp90-Aha1 mixture (see above) the observed ⌬CD spectrum is also similar to that with intact Aha1 (K d ϭ 0.65 Ϯ 0.06 M) and indicates that the N-terminal domain of Aha1 is primarily responsible for binding to Hsp90. The stoichiometry of the interaction is most probably 1:1 (Hsp90:Aha1) as determined previously (17).
for a mixture of Sba1, AMPPNP, and Hsp90 could not be simulated by any combination of the observed spectra for the components and their pairwise complexes in which one component remained unbound, indicative of the formation of a threeway Hsp90-AMPPNP-Sba1 complex (Fig. 8A). Although the observed spectrum was most closely approximated by a combination of the observed spectra for Hsp90-AMPPNP and Hsp90-Sba1 mixtures, it could not be reproduced exactly, as there is little complex formation between Hsp90 and Sba1 in the absence of nucleotide.
When Cp50 was added to the system, the observed near-UV CD spectrum for the Cp50, Sba1, Hsp90, and AMPPNP mix-FIG. 7. Analysis of Hsp90, Cp50, and Aha1 interactions by near-UV CD. A, near-UV CD spectra. obs1, Hsp90-Cp50 (50 M); obs2, Aha1-Cp50 (50 M); obs3, Hsp90-Aha1 (50 M). Simulated spectra for the protein mixtures is also shown in A. sim1, Hsp90-Cp50; sim2, Aha1-Cp50; sim3, Hsp90-Aha1. Simulated spectra are derived from the spectra from individual components without interaction. Where simulated spectra are dissimilar to the observed spectra (sim1 and obs1, and sim3 and obs3) this indicates that the proteins interact to form complexes (Hsp90-Cp50 and Hsp90-Aha1). obs2 and sim2 are similar indicating that Aha1 and Cp50 do not interact. B, near-UV CD spectrum for Hsp90-Cp50-Aha1 (50 M) protein mixtures. Spectra could be superimposed irrespective of the order of addition of the components to Hsp90. obs1, Cp50 added after Aha1 addition; obs2, Aha1 added after Cp50 addition. The simulated spectra for when one component remains unbound are represented by the curves a, Cp50 unbound, and b, aha1 unbound. Curve c represents no interaction between all the components, and curve sim represents interactions between Hsp90 and Cp50 as well as Hsp90 and Aha1 and was simulated assuming the formation of an Aha1-Hsp90-Cp50 complex with 10% unbound Aha1 (Aha1 added after Cp50), and matches the observed (obs) spectrum for the protein mixture. C, near-UV CD spectra for an Hsp90-AMPPNP-Cp50 (50 M for each protein and 100 M AMPPNP) protein mixture (obs). The simulated spectra for when one component remains unbound are represented by the curves a, Cp50 unbound, and b, AMPPNP unbound. Curve c represents interactions between Hsp90 and Cp50 as well as Hsp90 and AMPPNP and was simulated assuming the formation of an AMPPNP-Hsp90-Cp50 complex. sim represents interactions between Hsp90 and Cp50 as well as Hsp90 and AMPPNP and was simulated assuming the formation of two types of complexes, namely 90% AMP-PNP-Hsp90-Cp50 and 10% Hsp90-Cp50 and unbound AMPPNP, and matches the observed (obs) spectrum for the protein mixture. The results are consistent with the formation of a three-way complex of these components and shows that AMPPNP and Cp50 can bind simultaneously to Hsp90.  a and b, respectively. sim represents all components bound (Sba1-AMPPNP-Hsp90 complex) and most closely resembles the obs spectrum. None of the spectra match the observed (obs) spectrum, because there is very little interaction between Hsp90 and Sba1 in the absence of AMPPNP so that the obs spectrum cannot be reproduced exactly. However, the results indicate that the proteins interact to form a complex (Sba1-AMPPNP-Hsp90). B, near-UV CD spectra for a Cp50-Hsp90-AMPPNP-Sba1 (50, 100, and 25 M, respectively) protein mixture (obs). The simulated spectra for when one component remains unbound are represented by the curves a, Sba1 unbound, and b, Cp50 unbound. Curve c represents no interaction between all the components, and curve d represents interactions between all the components (Sba1-AMPPNP-Hsp90-Cp50 complex). sim represents the simulated spectrum that matches the observed spectrum most closely and suggests that Sba1 and Cp50 compete for binding to Hsp90. sim assumes the formation of a protein mixture consisting of 20% Hsp90-AMPPNP-Cp50, 40% Hsp90-Cp50, and 40% Hsp90-AMPPNP. Within this mixture 40% of the AMPPNP, 40% Cp50, and 60% Sba1 remained unbound. ture could not be simulated by assuming an Sba1-AMPPNP-Hsp90-Cp50 complex but could be most closely simulated by a combination of observed spectra for three distinct types of complex, Cp50-Hsp90, Cp50-AMPPNP-Hsp90, and Sba1-AMP-PNP-Hsp90 (Fig. 8B) all at specific concentrations within the protein mixture. Our results also suggested that the binding of Cp50 and p23/Sba1 to Hsp90 is mutually exclusive and is difficult to reconcile with the observation suggesting the formation of a p23/Sba1-Hsp90-p50/Cdc37 complex (32).
Conformational Switches-The pattern of effects of mutations and co-chaperones on the ATPase cycle of Hsp90, suggests that this activity is not a simple enzymatic reaction in which a distinct rate-limiting step can be defined. Rather it is a complicated process involving coupled conformational switches in several regions of the structure, which together conspire to assemble an active site competent for ATP hydrolysis. This complex series of conformational changes is regulated by the binding of co-chaperones.
The presence of the first of these conformational switches is elaborately demonstrated by the binding characteristics of Sba1 to the Hsp90 ts-mutants. Disruption in Sba1 binding was seen with the T101I and F349A mutants (Figs. 3 and 4), which promote the open ATP lid state, whereas those favoring the closed state had no effect. The results support the idea that a conformational switch exists that promotes ATP lid closure leading to N-terminal dimerization to which Sba1 can then bind. However, how this ATP lid is triggered remains to be seen. Our results suggest that because Aha1 activates the T101I mutant (open state) normally, it is unlikely that Aha1 triggers closure of the lid, but rather helps to stabilize it in the closed state. This idea is supported by the observation that the A107N mutation, which favors a stable N-terminal dimerized state (closed ATP lid), largely bypasses the requirement for Aha1 activation (Fig. 6, C and D), whereas the F349A mutation, which destabilizes the closed state, is highly dependent on Aha1 for its activation.
As with Aha1, NAha1 activates the ATPase activity of Hsp90, albeit to a lesser degree than the intact protein (17), but nonetheless it stimulates ATP turnover. Our structure did not show whether NAha1 could stabilize N-terminal dimerization, although the bound NAha1 may be close enough to the Nterminal domains that it might interact with them. It is also not unreasonable that the catalytic loop released from the middle domain of Hsp90 may itself interact with the N-terminal domains as well as the bound ATP and that these interactions might help to stabilize N-terminal dimerization. However, what is apparent is that the rate-limiting step of ATP turnover by Hsp90 is simply not ATP hydrolysis (33) but appears to consist of a complex restructuring of Hsp90 involving ATP lid closure, interaction of the ATP lid with the middle domain of Hsp90, a second molecular switch that releases the catalytic loop, and finally N-terminal dimerization. Because of this complex restructuring of Hsp90 its ATPase activity can be influenced by mutations that appear unrelated, such as A107N and F349A, and regulated by co-chaperones that bind different regions of Hsp90, such as Aha1, p23/Sba1, and Cp50, but nonetheless act by affecting the same overall mechanism. Hsp90 ATPase activity can also be influenced by the disruption of C-terminal dimerization, which drastically reduces the ATPase activity of Hsp90 (6). This suggests that the full ATPase activity of Hsp90 is only attained when the two halves of Hsp90 cooperate by N-terminal dimerization and in so doing help to stabilize the formation of the catalytically active unit.
In conclusion, Aha1 and Sba1 seem to regulate Hsp90 by mechanisms directly involving the conformation of the ATP lid, and therefore N-terminal dimerization, whereas Sti1 inhibition does not. The results we have presented so far support the ATP lid model and suggest that the formation of an N-terminally dimerized catalytically active molecule involves several conformational switches including the interaction of the N-terminal domains with the middle-domains of Hsp90, especially around F349, and release of the catalytic loop for interaction with the bound ATP (Fig. 9). FIG. 9. Model of the ATP lid mechanism and location of the ts-mutants that affect the hydrolysis of ATP by Hsp90. In the inactive state the ATP lid of Hsp90 is in the open state and this allows the binding of ATP. Following ATP binding the ATP lid traps the bound ATP. The lid is stabilized by interaction with its own N-terminal domain and by interaction with a hydrophobic patch, around F349, in the middle domain of Hsp90 and finally by dimerization of the N-terminal domains. The T101I mutation (•) stabilizes the lid in an open conformation by interacting with hydrophobic residues, whereas the A107N mutation (f) promotes a closed state by interacting with a hydrophilic patch in the N-terminal domain. The T22I mutation (ࡗ) promotes N-terminal dimerization by acting at a point directly involving association of the N-terminal domains. R represents Arg 380 from the catalytic loop of the middle domain. Finally F349A destabilizes closer of the ATP lid probably by weakening the interactions of the ATP lid with the hydrophobic patch around this position. Aha1 probably acts to stabilize the formation of this catalytic unit, because it can effectively suppress the affects of the F349A mutation, whereas A107N effectively bypasses the Aha1 requirement.