A chemical compound inhibiting the Aha1–Hsp90 chaperone complex

The eukaryotic Hsp90 chaperone machinery comprises many co-chaperones and regulates the conformation of hundreds of cytosolic client proteins. Therefore, it is not surprising that the Hsp90 machinery has become an attractive therapeutic target for diseases such as cancer. The compounds used so far to target this machinery affect the entire Hsp90 system. However, it would be desirable to achieve a more selective targeting of Hsp90–co-chaperone complexes. To test this concept, in this-proof-of-principle study, we screened for modulators of the interaction between Hsp90 and its co-chaperone Aha1, which accelerates the ATPase activity of Hsp90. A FRET-based assay that monitored Aha1 binding to Hsp90 enabled identification of several chemical compounds modulating the effect of Aha1 on Hsp90 activity. We found that one of these inhibitors can abrogate the Aha1-induced ATPase stimulation of Hsp90 without significantly affecting Hsp90 ATPase activity in the absence of Aha1. NMR spectroscopy revealed that this inhibitory compound binds the N-terminal domain of Hsp90 close to its ATP-binding site and overlapping with a transient Aha1-interaction site. We also noted that this inhibitor does not dissociate the Aha1–Hsp90 complex but prevents the specific interaction with the N-terminal domain of Hsp90 required for catalysis. In consequence, the inhibitor affected the activation and processing of Hsp90–Aha1-dependent client proteins in vivo. We conclude that it is possible to abrogate a specific co-chaperone function of Hsp90 without inhibiting the entire Hsp90 machinery. This concept may also hold true for other co-chaperones of Hsp90.

Heat shock protein 90 (Hsp90) 4 is one of the most abundant and evolutionary conserved molecular chaperones. Together with its co-chaperones, it plays a key role in stress and cell cycle control, as well as hormone signaling in the eukaryotic cytosol (1)(2)(3). Hsp90 is a flexible homodimer consisting of three domains per monomer. The N-terminal domain (NTD) contains the ATP-binding site, the middle domain (MD) has been implicated in client protein and co-chaperone binding, and the C-terminal domain is the site of dimerization (4 -7) and binds several TPR-containing co-chaperones (8 -10). Hsp90 undergoes large conformational changes from an open to an N-terminally closed state during its ATPase cycle (11)(12)(13)(14). These rate-limiting steps are modulated by co-chaperones (1,15,16). Among them, Aha1 (activator of the Hsp90 ATPase) is a potent accelerator of the Hsp90 ATPase (17,18).
Aha1 is a monomeric, two-domain protein that has two interaction sites with Hsp90: the Hsp90 MD binds to the Aha1 N-domain (19) and the Aha1 C-domain binds more dynamically to the N-domain of Hsp90 supporting the formation of the unfavorable and rate-limiting N-terminally closed state of Hsp90 (20,21). In the chaperone cycle, Aha1 interaction precedes the formation of the fully closed state (22). Specifically, Aha1 accelerates the association of the two N-terminal domains (20). The activation mechanism is asymmetric because one Aha1 per Hsp90 dimer is sufficient to fully stimulate the Hsp90 ATPase (20). Aha1 is involved in the interaction of Hsp90 with specific client proteins like protein kinases and steroid hormone receptors (17,23,24). Furthermore, Aha1 together with Hsp90 has been shown to play an important role in the quality control of mutants of the cystic fibrosis transmembrane conductance regulator (CFTR) (21,25,26). Mutations in the CFTR represent the primary cause of cystic fibrosis. In particular, deletion of Phe 508 in CFTR results in a loss of function and rapid degradation (25,(27)(28)(29). A knockdown of Aha1 led to increased levels of the CFTR ⌬F508 mutant (26), suggesting that inhibition of the Hsp90 -Aha1 interaction may be beneficial in this context.
Targeting the Hsp90 -Aha1 interaction by chemical compounds is therefore attractive as a potential therapeutic approach, and at the same time it could provide insight into the possibility of targeting specific Hsp90 -co-chaperone effects. Efforts to target the Hsp90 machinery by small molecules focused mainly on Hsp90. Promising Hsp90 inhibitors like the geldanamycin (GA) derivative 17-allylamino-17-demethoxygeldanamycin (17-AAG), IPI-504 (retaspimycin hydrochloride), and radicicol derivatives like NVP-AUY922 (luminespib) or STA-9090 (ganetespib) entered clinical trials (30 -33). All these inhibitors bind to the ATP-binding site of Hsp90. Alternatively, other interaction sites in Hsp90, like the C-terminal domain, have been tried to target by inhibitors. In this context, the amino-coumarin antibiotic novobiocin binds to the C-terminal domain of Hsp90 and has shown promising results in prostate cancer cells (33,34). In a previous proof of concept study, we showed that it is also possible to identify compounds that accelerate the ATPase cycle of Hsp90 (35), reminiscent of the function of the co-chaperone Aha1.
For the Aha1-Hsp90 interaction, it has been shown that it is in principle possible to interfere with complex formation by chemical compounds (36). Here we report in a proof-of-princi-ple study identifying modulators that specifically interfere with the catalytic function of the co-chaperone Aha1 but do not affect complex formation with Hsp90 in general. The characterization of the mode of action of the key compound reveals that by binding to the Hsp90 N-domain, the interaction important for accelerating the ATPase is blocked.

Identification of Hsp90 -Aha1 modulators
We monitored the interaction between Hsp90 and Aha1 by FRET (22) to screen a compound library of ϳ15,000 chemical compounds. Our FRET system consists of Hsp90 carrying a fluorescent acceptor dye (Atto 550) in the M-domain and of Aha1, which is randomly labeled with a donor dye (Alexa 488). Binding of Aha1 to Hsp90 results in an increase of the FRET A, Hsp90 -Aha1 FRET system with donor-labeled yAha1 and acceptor-labeled yHsp90 was used to screen a 15,000-compound library for the identification of specific modulators of the Hsp90 -Aha1 interaction. B, FRET kinetics of Hsp90 -Aha1 with active compounds. Compounds (500 g/ml; HAM-1 (green), HAM-2 (lilac), HAM-3 (dark blue), HAM-4 (pink), HAM-5 (blue), and HAM-6 (red)) were added to each 500 nM of Hsp90 and Aha1 and preincubated. The addition of 2 mM AMP-PNP allowed monitoring complex formation. DMSO (black) and GA (purple) served as negative and positive controls, respectively. C, structures of the HAMs.
Chemical compound inhibiting Aha1-Hsp90 chaperone complex efficiency (Fig. 1, A and B). Because Aha1 binds preferably to a closed conformation of Hsp90 that can be induced by the nonhydrolyzable ATP analog AMP-PNP (8,11,12), FRET kinetics were recorded after the addition of AMP-PNP to labeled Hsp90 and Aha1 (Fig. 1B). DMSO, added as a negative control, had no influence on the FRET kinetics. Because there are no specific Hsp90 -Aha1 modulators known so far, the Hsp90 inhibitor GA served as a positive control. GA binds to the nucleotidebinding pocket of Hsp90 and keeps Hp90 in an open conformation (37). Accordingly, the FRET kinetics were significantly influenced. Having established that the Hsp90 -Aha1 FRET system responds to perturbations of the interaction, we analyzed the effects of the library compounds. Representative kinetic traces for the screened compounds are depicted in Fig.  1B. Some compounds exhibited strong inhibitory effects on the Hsp90 -Aha1 interaction and significantly slowed down the FRET kinetics. We also found activators of the Hsp90 -Aha1 interplay as manifested by faster FRET kinetics. Compounds altering the FRET kinetics of the Hsp90 -Aha1 complex formation were rescreened twice to exclude false positive hits. We identified ϳ40 compounds as possible modulators of the Hsp90 -Aha1 interaction. Fig. 1C depicts the structures of the six most effective Hsp90 -Aha1 modulators (HAMs). The same library had been recently used to identify Hsp90 ATPase accelerators (35). Interestingly, there is no overlap of compounds identified in this study and the molecules uncovered by the Hsp90 -Aha1 FRET system. This further implies that the compounds identified are specific for modulating the interaction of Hsp90 with Aha1.

The modulators alter the stimulatory effect of Aha1 on the Hsp90 ATPase
The identified modulators were further studied for their effects on the Hsp90 ATPase in the presence of Aha1. Aha1 is a strong accelerator of the Hsp90 ATPase cycle (17,18,20). In the presence of the identified compounds (Fig. 1C), the stimulatory effect of Aha1 on the Hsp90 ATPase was altered ( Fig. 2A). Three compounds acted as inhibitors of the Aha1 stimulation on Hsp90, whereas three other compounds further stimulated catalysis. Compound HAM-1, the strongest inhibitor, is able to almost completely suppress the Aha1-mediated stimulation of Hsp90 (93 Ϯ 1% inhibition). The modulators HAM-2 and HAM-3 inhibit the Aha1 stimulation by 27 Ϯ 3 and 20 Ϯ 1%, respectively. The activating compounds HAM-4 and HAM-5 increased the ATPase by 1.3-and 1.5-fold (130 Ϯ 8 and 151 Ϯ 8% of the Hsp90 -Aha1 activity), respectively. HAM-6 did not have a significant effect on the ATPase.
The apparent binding constant (K D app ) of each modulator to the Hsp90 -Aha1 complex was determined by titration of the compounds in the ATPase assay. All compounds displayed affinities in the micromolar range (Table 1). HAM-1 has the highest affinity for Hsp90 -Aha1 with a K D app of 24 Ϯ 2 M. For

Chemical compound inhibiting Aha1-Hsp90 chaperone complex
HAM-2, a K D app of 40 Ϯ 5 M was determined, whereas the affinity for HAM-3 was found to be higher than 100 M (K D app Ͼ 100 M). The activators, HAM-4 and HAM-5, have K D app values of 163 Ϯ 42 and 145 Ϯ 70 M, respectively. It should be noted that because of the low affinity for HAM-4 and HAM-5, we could not reach saturation in the ATPase assays (supplemental Fig. S1). The assays described above do not tell us whether the modulators act on the Hsp90 -Aha1 complex or on Hsp90 alone. To clarify this, we performed ATPase assays in the absence of Aha1 ( Fig. 2B and Table 1). Here, the modulator HAM-1 did not have a significant influence on the Hsp90 ATPase activity without Aha1 (inhibition by HAM-1, 22 Ϯ 2%; activation by HAM-6, 26 Ϯ 3%). For HAM-2, surprisingly, we found that it activates the Hsp90 ATPase 2-fold when Aha1 is not present. Together with Aha1, HAM-2 acts as an inhibitor of the Hsp90 -Aha1 chaperone system (inhibition by 27 Ϯ 3%). This implies that HAM-2 is a unique type of modulator of the Hsp90 chaperone machinery. HAM-4 exhibited an almost 6-fold activation (565 Ϯ 8%) of the Hsp90 ATPase activity. When Aha1 is present, HAM-4 has a much weaker stimulatory effect on the Hsp90 ATPase (30 Ϯ 8% stimulation). HAM-5, screened as an activator of the Hsp90 -Aha1 complex, also activated the Hsp90 ATPase 2-fold (220 Ϯ 5%). These findings suggest that HAM-2, -4, and -5 might directly bind to Hsp90. Because HAM-1 and -6 do not significantly alter the Hsp90 ATPase activity, they may either affect Aha1 directly or the Aha1-Hsp90 interaction.
In the following, we focused on HAM-1, because this compound exhibited the most promising properties. Derivatives of HAM-1 (supplemental Fig. S2) were generated and characterized ( supplemental Figs. S3 and S6). However, none of these showed improved characteristics concerning the effects on Hsp90 -Aha1.
Because HAM-1 had a profound effect on the stimulation of the Hsp90 ATPase by Aha1, it could interfere with the binding of Aha1 to Hsp90. Analytical ultracentrifugation was used to analyze complex formation between Hsp90 and Aha1 in the presence of HAM-1 (Fig. 2C). Under all conditions tested (ϩHAM-1, ϩAMP-PNP, or ϩHAM-1/AMP-PNP), the Hsp90 -Aha1 complex was formed. Also, a titration of HAM-1 to Hsp90 and Aha1 did not reveal an influence on complex formation. Therefore, we conclude that HAM-1 does not abolish Aha1 association with Hsp90. To test binding of Aha1 to Hsp90 directly by an orthogonal method, SPR measurements were performed (Fig. 2D). Aha1 was titrated to Hsp90 in the presence of HAM-1. The affinity of Aha1 for Hsp90 is only marginally affected by the presence of HAM-1, both in the absence (from a K D of 3.12 Ϯ 0.65 to 4.70 Ϯ 0.74 M) or presence of the non-hydrolyzable ATP analog AMP-PNP (from a K D of 2.55 Ϯ 0.30 M to a K D ϭ 1.80 Ϯ 0.30 M) (Fig. 2D). These findings show that the inhibitor HAM-1 does not prevent binding of Aha1 to Hsp90, whereas at the same time, Aha1 is not able to stimulate Hsp90 efficiently anymore.

The modulator HAM-1 binds to the Hsp90 N-domain
To determine the binding site of the inhibitor HAM-1, we recorded 1 H, 15 (Fig. 3B). Especially the NMR signals assigned to residues in helix ␣2 are strongly affected. Interestingly, this region overlaps partly with the previously determined second Aha1-binding site in the NTD (20) (Fig. 3B). To test whether HAM-1 affects ATP binding, we recorded 1 H, 15 N correlation spectra of the Hsp90 NTD in the presence of ATP. Upon adding HAM-1 to the ATP-bound NTD, NMR chemical shifts move to the same position as observed for the NTD bound to HAM-1 in the absence of ATP. In addition, a number of signals with weaker intensity remains with chemical shifts of the ATP-bound NTD (Fig. 3A, blue spectrum). This indicates that binding of HAM-1 can release ATP from the nucleotidebinding pocket. Because NMR signals for both the HAM-1 and the ATP-bound state are visible, the exchange between these two states is slow on the NMR time scale. As control experiments, we tested binding of HAM-1 to the individual Aha1 domains. Here, the addition of HAM-1 to the Aha1 ND and CD domains showed no or very minor differences in chemical shifts, respectively, indicating no significant interaction (supplemental Fig. S4, B and C).

HAM-1 effects Hsp90 -Aha1 client processing in vivo
To test whether HAM-1 has an influence on Hsp90 client proteins in vivo, we used well established reporter assays in yeast (38 -41). We first analyzed the modulator's effects on steroid hormone receptor (SHR) activity, specifically on the glucocorticoid receptor (GR) and the mineralocorticoid receptor (MR). Both are highly dependent on Hsp90 and its co-chaperones (42)(43)(44)(45). Inhibition of Hsp90 with radicicol and GA showed reduced GR and MR activity in vivo (46,47). In contrast to GR, MR seems to be also dependent on Aha1. When we added HAM-1 to the yeast cultures, we observed that HAM-1 exhibits a strong inhibitory effect on MR activity by decreasing its activity to 40 Ϯ 6% in a concentration-dependent manner (Fig. 4A, black line), although the activity of GR was not affected (Fig. 4A, red line). Thus, we are able to specifically affect an Table 1 Apparent affinity of the modulators for Hsp90/Aha1 and their effects on ATP turnover in the presence and absence of Aha1 ATPase measurements were performed as described under "Experimental procedures." The K D app for the modulators was determined by assaying ATP turnover at different modulator concentrations and nonlinear curve fitting. The relative Hsp90 ATPase activity was determined by measuring the ATPase activity in the presence of saturating concentrations of the modulator relative to the turnover in the absence of modulators in the presence of 1% DMSO.

Chemical compound inhibiting Aha1-Hsp90 chaperone complex
HAM-1 resulted in a pronounced inhibition. This finding underlines the selectivity of HAM-1 on the Hsp90 -Aha1 interaction. Taken together, we could show that the identified compound is a potent modulator of Hsp90 function in vitro and in vivo. The selectivity of HAM-1 for MR, with a dependence on Aha1, demonstrates the Aha1 specificity of the effect.

HAM-1 alters CFTR degradation
Degradation of ⌬F508 CFTR, which causes cystic fibrosis, was shown to be triggered by the Hsp90 machinery (21,26). Because Aha1 has been described to be an important co-chaperone in this context (26), we tested whether the modulator identified in this study can also affect CFTR stability. To this end, we used a WT yeast strain (BY4741) expressing WT CFTR or ⌬F508 CFTR, respectively (48). We followed CFTR degradation after addition of the translation inhibitor cycloheximide over time in the absence and presence of HAM-1 (Fig. 5). CFTR levels were corrected for variations of a loading control. After 5 min, a decrease of the overall CFTR levels can be detected. 10% of WT CFTR and 40% of the ⌬F508 variant were degraded. The ⌬F508 CFTR variant was degraded faster than WT CFTR, which is in agreement with the literature and manifests its instability (49,50). In the presence of HAM-1, higher amounts of ⌬F508 CFTR and thus a prolonged half-life could be observed (40% degradation compared with 70% without the modulator; Fig. 5B). This stabilization of the ⌬F508 variant implies a specific effect of HAM-1 on the Hsp90 -Aha1 interplay. Consistent with a weaker dependence of WT CFTR on Hsp90 -Aha1 compared with the ⌬F508 variant (26), HAM-1 did not have a strong effect on the stabilization on WT CFTR (Fig. 5A).

Discussion
Our findings demonstrate that it is in principle possible to selectively target the influence of Aha1 on Hsp90 by chemical compounds. Screening a 15,000-compound-spanning library resulted in the identification of several modulators for the

Chemical compound inhibiting Aha1-Hsp90 chaperone complex
Hsp90 -Aha1 interaction. Six chemical compounds showed promising effects and characteristic profiles in vitro. HAM-1 was the most effective inhibitor in terms of binding and the potential to alter the Aha1-mediated stimulation of Hsp90. The presence of HAM-1 resulted in a roughly 90% inhibition of the stimulatory effect of Aha1 on Hsp90, whereas the effect on the Hsp90 ATPase in the absence of Aha1 was moderate. We would have expected that this inhibitor abolishes complex formation of Hsp90 with Aha1, which was surprisingly not the case. Analytical ultracentrifugation and SPR measurements showed that the Hsp90 -Aha1 complex is formed in the presence of HAM-1. However, this complex seems to be inefficient in progressing through the Hsp90 ATPase cycle because Aha1 cannot potently stimulate the N-terminal association of Hsp90 anymore. Modulation of the Hsp90 chaperone machinery by HAM-1 therefore seems to be specific for Aha1. These results may be counterintuitive at first glance because only the function of Aha1 is affected but not its interaction with Hsp90. However, one has to bear in mind that the interaction of Aha1 with Hsp90 is complex (Fig. 6), with the two domains of Aha1 interacting with two different domains of Hsp90: the Aha1 N-domain stably interacts with the Hsp90 M-domain (6,19,51). The interaction of this domain alone does not strongly affect the Hsp90 ATPase cycle. Only together with the Aha1 C-domain, which interacts at a specific point of the conformational transitions with the Hsp90 N-terminal domain, has an acceleration been observed (17,20,21). This transient interaction occurs only for a short period of time in the reaction cycle of Hsp90 to stabilize a specific closed state the formation of which is rate-limiting (12,20). Thus, we assumed that HAM-1 may interfere with this interaction. Our NMR results are consistent with this notion: HAM-1 interacts with the Hsp90 N-domain. This binding site has a significant overlap with the Aha1 binding site in the Hsp90 N-domain. Because Aha1 can still bind to Hsp90, our results suggest that the Aha1 N-domain is not affected in its interaction with the Hsp90 M-domain. In contrast, the Aha1 C-domain exhibits an impaired interaction with the Hsp90 N-domain in the presence of HAM-1 and is no longer able to stimulate the Hsp90 ATPase to its full potential. Thus, HAM-1 seems to exert its function by sterically blocking the interaction of Aha1 with the Hsp90 N-domains required to accelerate the formation of an N-terminally closed state, which is rate-limiting for ATP hydrolysis (Fig. 6). Furthermore, HAM-1 does not change Aha1 affinity for Hsp90 significantly as determined by SPR measurements (Fig. 4D). This further supports a mechanism where HAM-1 affects more the transient interaction between Hsp90 and Aha1, which is crucial for the accelerating effect of Aha1 on the Hsp90 cycle. In other words, in the presence of HAM1, Aha1 is still bound to Hsp90, but it is no longer functional as an accelerator of the ATPase, because the second, "catalytic" interaction can no longer occur. This impeded interaction also influences the processing of different Hsp90 clients in vivo. SHR such as GR are highly dependent on Hsp90 (52), and the inhibition of Hsp90 by RA or GA results in reduced GR activity in vivo (46,47). In all cases investigated so far, the activity of GR is not completely abolished, which may be due to spontaneous folding or assistance by the Hsp70 chaperone system. Interestingly, HAM-1 did not inhibit the processing and activation of GR in vivo. Because Aha1 is not required for GR processing and because HAM-1 only slightly affects the Hsp90 ATPase, these results point toward a selectivity of HAM-1 on the Hsp90 -Aha1 interplay. In contrast, the activation of MR, a client that requires Aha1, was impaired by 40%. The fact that the MR activity was not altered by HAM-1 in the ⌬Aha1 strain further supports a mech-

Chemical compound inhibiting Aha1-Hsp90 chaperone complex
anism where HAM-1 requires the presence of Aha1 to unfold its inhibitory potential. Additionally, MR processing in vivo was inhibited by HAM-1 in a dose-dependent manner, resulting in an IC 50 of 125.5 M, which is lower than the affinity of HAM-1 for Hsp90 -Aha1 measured in vitro (23.5 Ϯ 1.7 M). In this context, it should be noted that we do not know the effective inhibitor concentration in the yeast cell and have to assume that the uptake of the compound is counteracted by efflux pumps (53). Similarly, when a well established general inhibitor of the Hsp90 ATPase (RA) was used to inhibit Hsp90 in yeast, much higher concentrations were required to abrogate the effect of Hsp90 on MR and GR than for the inhibition of Hsp90 in vitro (54). In this case, the Hsp90-dependent activation of MR and GR was affected by ϳ80%, which may reflect the higher affinity of RA for Hsp90, differences in uptake/efflux or differences in mechanism because RA is inhibiting the entire Hsp90 system, and HAM-1 affects a specific aspect. Given its specificity for the effect of Aha1, HAM-1 could have a beneficial effect in the context of cystic fibrosis where the misfolded ⌬F508 CFTR protein, which most commonly causes the childhood disease cystic fibrosis, is an Hsp90 client (25,26). Together with the quality control system of the endoplasmic reticulum, the misfolded ⌬F508 CFTR is targeted by the Hsp90 machinery for degradation (55)(56)(57). Interestingly, the interplay between Hsp90 and Aha1 seems to be crucial for the ERADassociated degradation of the ⌬F508 CFTR because the downregulation of Aha1 was shown to promote the stability of the ⌬F508 channel activity (26 -28, 59). Our results further support this scenario because in the presence of HAM-1, a prolonged life time of ⌬F508 CFTR was observed. The inhibition of the Hsp90 -Aha1 interplay in this context might thus provide a novel approach to target the folding problem of the CFTR in cystic fibrosis. In a previous study by the Obermann lab (36), compounds were identified that, in combination with the VX-809 corrector, could support the activity of ⌬F508 CFTR. In this case, the mechanism seems to be different here because the dissociation of the Aha1-Hsp90 complex may be the underlying principle (36). Because Hsp90 and Aha1 are involved in the processing of several other disease-related clients like the kinases Raf and c-Abl (23, 60), modulating the Hsp90 -Aha1 interaction might also have beneficial effects in the context of cancer and other diseases. This concept may also be applicable to other Hsp90 co-chaperone interactions.

Protein purification and labeling
Yeast Hsp90 and Aha1 were expressed and purified as described previously (61). For NMR experiments, the Hsp90 NTD and MD, as well as the Aha1 ND and CD, were 15 Nlabeled. Fluorescent labeling of the yeast Hsp90 cysteine variant Q385C with Atto 550 (Atto-Tec, Siegen, Germany) was performed as described earlier (12). Yeast Aha1 was labeled randomly via cysteine residues with Alexa Fluor 488 C 5 maleimide dye (Life Technologies) according to the manufacturer's protocol. Free label was separated as previously described (22).

FRET-based screening
For the identification of specific modulators of the Hsp90 -Aha1 interaction, a compound library composed of ϳ15,000 small molecules (New Chemistry and Discovery Chemistry Collection, ChemDiv, San Diego, CA) was screened. We made use of an Hsp90 -Aha1 FRET system (22). To this end, 500 nM of labeled Hsp90 (yHsp90 Q385C* Atto 550) and 500 nM of labeled Aha1 (yAha1*Alexa 488) were mixed in low salt buffer (40 mM HEPES, 20 mM KCl, 5 mM MgCl 2 , pH 7.5) to allow formation of the FRET complex. This complex was incubated in a 96-well plate with the compounds of the library (dissolved in DMSO; final concentration, 50 g/ml; 1% DMSO). The conformational changes leading to FRET were induced by the addition of 2 mM AMP-PNP (Roche). DMSO and 100 M of the Hsp90 inhibitor GA (Sigma-Aldrich) were used as negative and positive controls, respectively. Changes in acceptor fluorescence were recorded for 20 min at 25°C in a Wallac EnVision Xcite multilabel plate reader (Perkin Elmer). A photometric 570/8 emission filter (center wavelength, 570 nm; bandwidth, 8 nm) and excitation at 490 nm (FITC, 485/14; center wavelength, 485 nm; bandwidth, 14 nm) were used. The Wallac Envision Manager Software 1.12 was used to overlay and evaluate the recorded kinetic traces. The Z-factor of the assay was calculated based on the following equation.
The Z-factor is at least 0.4 for the selection procedure to determine hits based on the kinetics of the closing reaction, which was used as primary screen and 0.7 for the ATPase reaction, which was used as secondary screen.

ATPase activity and affinity measurements
ATPase assays, using an ATP-regenerating enzyme system, were performed as previously described (62). For testing the effects of the modulators on the Hsp90 -Aha1 complex, an Hsp90 concentration of 1 M and an Aha1 concentration of 20 M was used. Compounds were titrated to the Hsp90 -Aha1 complex in varying concentrations from 0 to 0.1 mg/ml (1% DMSO, final concentration). To determine the ATPase activity of Hsp90 in the absence of Aha1, assays were performed with 3 M Hsp90. Measurements were carried out in low salt buffer (40 mM HEPES, 20 mM KCl, 5 mM MgCl 2 , pH 7.5) at 30°C in a Cary 50 UV-visible spectrophotometer (Varian Inc.; Agilent Technologies, Santa Clara, CA). Assays were performed with 2 mM ATP (Roche). The recorded kinetics were analyzed using the OriginPro software version 8.6 (OriginLab Corporation, Northhampton, MA) to determine the catalytic activity of Hsp90. Apparent binding affinities (K D app ) were determined by applying a Michaelis-Menten fit.
The apparent binding affinities of Aha1 to Hsp90 in the presence of the modulators (final concentration, 0.01 mg/ml) were measured under varying concentrations of Aha1 (0 -30 M). To determine the K D app of ATP to Hsp90 and Hsp90 -Aha1, 0 -3 mM ATP was titrated to 3 M Hsp90 and 1 M Hsp90/20 M Aha1, respectively. All ATPase assays were measured at least Chemical compound inhibiting Aha1-Hsp90 chaperone complex three times. The results are expressed as the mean catalytical activity and K D app , respectively, Ϯ standard deviation.

Surface plasmon resonance
SPR analysis was carried out in a Biacore X100 system (GE Healthcare Life Sciences). Hsp90 was coupled to a CM5 sensor chip using amine coupling (500 RU) following the manufacturer's instructions. Direct binding of Aha1 (0 -5 M) was measured in low salt buffer. Binding of Aha1 to Hsp90 was addressed under different conditions, adding nucleotide (2 mM AMP-PNP) and/or the inhibitor HAM-1 (60 M) to the measurement buffer. The recorded sensorgrams were analyzed with the Biacore X100 software (GE Healthcare Life Sciences). For the calculation of binding affinities, steady-state affinity was determined. The steady-state response against concentration was derived from the recorded sensorgrams and then fitted with the Biacore software to obtain the K D .

Analytical ultracentrifugation sedimentation velocity experiments
Analytical ultracentrifugation was carried out with a Pro-teomLab XL-I (Beckman, Krefeld, Germany) supplied with absorbance optics. 450 l of the samples were loaded into assembled cells with sapphire windows and 12-mm-path length charcoalfilled Epon double-sector centerpieces and centrifuged at 42,000 rpm in an eight-hole Beckman-Coulter AN50-Ti rotor. Sedimentation was monitored with an UV-visible spectrophotometer, equipped with a monochromator, at 280 nm. Data analysis was carried out with the program Sedfit (63), using a non-model-based continuous Svedberg distribution method (c(S)), with time and radial invariant noise on. 1 H, 15 N-HSQC NMR spectra were acquired on a Bruker AVIII 600 MHz spectrometer with a cryogenic triple resonance gradient probe. NMR samples of the yeast Hsp90 NTD (residues 1-210) or MD (residues 217-529) were dissolved in 20 mM sodium phosphate buffer (pH 6.5), 100 mM sodium chloride, 2 mM EDTA, 1 mM DTT, and 5% D 2 O. NMR samples of the yeast Aha1 ND (residues 1-156) and CD (residues 156 -356) were prepared accordingly. For the measurements with the Aha1 ND and CD and the Hsp90 MD, protein concentrations of 200, 167, and 200 M, respectively, and a HAM-1 concentration of 200 M were used. For the NTD of Hsp90 95 M protein and 285 M of HAM-1 were used. Binding was also observed at lower concentrations of HAM-1 (data not shown) but not to saturation. For the measurements of the NTD in complex with ATP, 6 mM magnesium sulfate, and 4 mM of ATP were added to a 100 M protein sample. Later on, 600 M of HAM-1 were added. All spectra were processed with NMRPipe/Draw (64) and analyzed with CcpNMR (65).

␤-Galactosidase assay
The influence of the modulators on client processing in vivo was tested with a ␤-galactosidase reporter assay system (39 -41). The wild-type yeast strain BY4741 was used to test for SHR activity (MR, GR) in the presence of the modulators. The BY4741 ⌬Aha1 strain was also included to test the effects of the modulators in the absence of Aha1. Compounds were titrated with concentrations ranging from 0 to 1000 M to a 10-l overnight culture of yeast cells and supplemented with 190 l of selection medium (Ura-His). The GR-and MR-dependent transcription was induced with 10 mM 11-deoxycorticosterone (Sigma-Aldrich) and aldosterone (Sigma-Aldrich), respectively. Cultures were incubated overnight at 30°C. 50 l of the yeast culture was centrifuged, and the pellet was lysed with 150 l of SDS buffer (82 mM Na 2 HPO 4, 12 mM NaH 2 PO 4 , 0.1% SDS (w/v), pH 7.5) for 15 min on a shaker at room temperature. The ␤-galactosidase activity was determined in triplicates after the addition of 50 l of o-nitrophenyl-␤-D-galactopyranosid (ONPG) (Sigma-Aldrich) (4 mg/ml stock in SDS buffer) by measuring the absorption in a Tecan Sunrise Absorbance Reader (Tecan, Mainz, Germany) at 405 nm. Differences in cell amounts were corrected by the respective A 600 . At least three independent experiments were performed in triplicates, and their mean Ϯ S.D. was determined.

CFTR degradation assay
To assess the effect of the identified modulators on CFTR stability in vivo, we used a yeast strain (BY4741) expressing HA-tagged CFTR WT and ⌬F508, respectively (the plasmids were a kind gift from Jeff Brodsky). CFTR degradation assays (48) were performed with yeast cells grown to log phase in the presence and absence of the modulator HAM-1 (200 M). The assays were performed at least five times in triplicate. Protein synthesis was arrested by the addition of cycloheximide (50 g/ml). For the analysis of the CFTR stability over time, samples were taken after 0, 5, 15, 25, and 40 min and prepared for Western blot with the TCA extraction protocol (48,58). Samples were run on a gradient gel and transferred to a PVDF membrane. For the detection of CFTR, an anti-HA antibody (monoclonal anti-HA antibody, clone HA-7 (mouse); Sigma-Aldrich) was used. As controls, antibodies against PGK, yHsp90, yAha1, and yHsp70 (Dr. Pineda, Berlin, Germany) were employed. Primary antibodies were detected with sheep anti-mouse IgG horseradish peroxidase-conjugated or goat anti-rabbit horseradish peroxidase-conjugated antibody (Sigma-Aldrich) and an ECL Western blotting system (Western-Bright ECL Spray, Advansta, Menlo Park, CA) according to the manufacturer's specifications. Quantification of the Western blots was performed with the ImageJ software. Means of at least five experiments and the standard error of the means were calculated.
Author contributions-S. C. S. and J. B. designed the study; S. C. S., M. R., and V. S. K. planned and performed the experiments; and S. C. S., M. R., C. J., V. S. K., M. S., and J. B. analyzed data and contributed to writing the paper. All authors reviewed the results and approved the final version of the manuscript.