Modulation of Chaperone Function and Cochaperone Interaction by Novobiocin in the C-terminal Domain of Hsp90

The C-terminal domain of Hsp90 displays independent chaperone activity, mediates dimerization, and contains the MEEVD motif essential for interaction with tetratricopeptide repeat-containing immunophilin cochaperones assembled in mature steroid receptor complexes. An α-helical region, upstream of the MEEVD peptide, helps form the dimerization interface and includes a hydrophobic microdomain that contributes to the Hsp90 interaction with the immunophilin cochaperones and corresponds to the binding site for novobiocin, a coumarin-related Hsp90 inhibitor. Mutation of selected residues within the hydrophobic microdomain significantly impacted the chaperone function of a recombinant C-terminal Hsp90 fragment and novobiocin inhibited wild-type chaperone activity. Prior incubation of the Hsp90 fragment with novobiocin led to a direct blockade of immunophilin cochaperone binding. However, the drug had little influence on the pre-formed Hsp90-immunophilin complex, suggesting that bound cochaperones mask the novobiocin-binding site. We observed a differential effect of the drug on Hsp90-immunophilin interaction, suggesting that the immunophilins make distinct contacts within the C-terminal domain to specifically modulate Hsp90 function. Novobiocin also precluded the interaction of full-length Hsp90 with the p50cdc37 cochaperone, which targets the N-terminal nucleotide-binding domain, and is prevalent in Hsp90 complexes with protein kinase substrates. Novobiocin therefore acts locally and allosterically to induce conformational changes within multiple regions of the Hsp90 protein. We provide evidence that coumermycin A1, a coumarin structurally related to novobiocin, interferes with dimerization of the Hsp90 C-terminal domain. Coumarin-based inhibitors then may antagonize Hsp90 function by inducing a conformation favoring separation of the C-terminal domains and release of substrate.

Before binding to ligand, steroid receptors exist in complexes with heat shock protein 90 (Hsp90) 3 -based molecular chaperone machinery (1). Native receptor-Hsp90 complexes typically contain one of four Hsp90-bound immunophilins, two FK506-binding proteins, FKBP51 and FKBP52, cyclophilin 40 (CyP40), a binding protein for cyclosporin A, and PP5, a protein phosphatase with weak FK506-binding affinity (1)(2)(3). A common feature of these cochaperones is a tetratricopeptide repeat (TPR) domain that forms the Hsp90-binding site, linked to a peptidylprolyl isomerase/immunosuppressant drug-binding domain in CyP40 and the FKBPs or a protein phosphatase domain in PP5. By competing for Hsp90 binding, TPR immunophilins are rapidly exchanged during the dynamic assembly of steroid receptor complexes within the cell. There is now increasing evidence that receptor function is critically dependent on the selection of immunophilin within steroid receptor complexes (2)(3)(4). This may be governed, in part, by the selective preference of receptors for specific immunophilins.
PP5 has been reported to have a modulating role in glucocorticoid signaling (5), and there are strong indications that FKBP51 inhibits glucocorticoid receptor function. Elevated expression of FKBP51, resulting in greatly increased incorporation of FKBP51 into glucocorticoid receptor complexes, reduces hormone-binding affinity and promotes glucocorticoid resistance in primates (6,7). FKBP51 also sequesters glucocorticoid receptor within the cytoplasm (8,9), but hormone binding induces a functional exchange of FKBP52 for FKBP51 in receptor complexes allowing translocation of the complex to the nucleus (8). In a yeast model, FKBP52 was shown to dramatically potentiate glucocorticoid-dependent reporter gene activity through a mechanism that results in increased receptor hormone-binding affinity (10). Consistent with previous findings, coexpression of FKBP51 blocked the potentiating effects of FKBP52. These potentiating properties require FKBP52 catalytic activity as well as a functional interaction of the immunophilin with Hsp90. Receptor function, then, can be directly influenced by the prolyl isomerase activity of a TPR immunophilin. The Smith laboratory has now extended the study of FKBP52 function to a FKBP52 knock-out mouse model (11)(12)(13). Male mice exhibit many features in common with partial androgen insensitivity, reflecting loss of FKBP52-mediated potentiation of androgen receptor function (11). Female mice display a maternal defect linked to progesterone insensitivity in the uterus that impedes successful pregnancy (12). On the other hand, evidence that female FKBP52 mutant mice manifest normal ovulation suggests that signaling through estrogen receptor ␣ remains essentially intact (12,13). These results collectively argue against redundancy among the TPR immunophilins and provide support for distinct modes of action for these proteins in steroid hormone signaling. The results also identify the TPR immunophilins as potential drug targets, because specific inhibitors of their assembly into receptor complexes may be important therapeutically in hormone-dependent cancers of the breast and prostate.
By targeting a conserved ATP-binding pocket in the N-terminal region of Hsp90, antitumor drugs, such as the ansamycin antibiotic geldanamycin, interfere with Hsp90 function and promote the proteasomal degradation of several key regulatory proteins, including multiple tyrosine and serine/threonine kinases and steroid receptors, many of which are involved in promoting malignancy (14 -17). HER-2 receptor tyrosine kinase, overexpression of which is frequently linked with aggressive breast cancer and poor prognosis (18), and the androgen receptor, mutation of which may account, in part, for prostate cancer progression following anti-androgen therapy (19,20), appear to be among the most sensitive targets of the ansamycins (21,22). These novel anticancer agents form an important component of new strategies for the treatment of breast and prostate cancer. Novobiocin, a coumarin antibiotic, antagonizes Hsp90 function by binding within its C-terminal dimerization domain (23). Although their mode of action has yet to be properly defined, novobiocin and related coumarins have been shown to markedly reduce the cellular levels of several oncogenic protein kinases, including HER-2 and Raf-1 (24). Through its unique interaction with Hsp90, this drug family may provide a useful alternative in cancer therapies targeting Hsp90. The cochaperone p50 cdc37 interacts with the Hsp90 N-terminal domain and recruits protein kinases by blocking Hsp90 ATPase function (25). By dimerizing through its C-terminal domain, Hsp90 functions as a molecular clamp, utilizing ATP binding and hydrolysis to promote extensive interdomain conformational changes that result in transient contact between the N-terminal ends. Within the C-terminal dimerization domain, recent structural studies have defined a hydrophobic helix that is duplicated in a central, exposed position between the arms of the dimer, making it highly suited for interaction with potential chaperone substrates, including steroid receptors (26). The MEEVD motif located at the extreme C terminus of Hsp90 is the primary acceptor site for TPR immunophilins (27)(28)(29)(30)(31), but there is growing recognition that interaction of Hsp90 with these cochaperones may also involve additional elements within the Hsp90 dimerization domain (27). Therefore, TPR immunophilins are well positioned to exert modulating influences on steroid receptors within receptor-Hsp90 complexes.
We previously determined that, in addition to its TPR domain, CyP40 requires flanking N-terminal acidic and C-terminal basic regions for efficient binding to Hsp90 (32) and in subsequent studies identified key residues within the core TPR domain that contribute to this interaction (33). To better define the structural basis of CyP40-Hsp90 interaction, we have now dissected the C-terminal basic helix of CyP40 and confirmed the role of the so-called charge-Y sequence motif, previously identified in FKBP51 and FKBP52 to be essential for high affinity recognition of Hsp90 (34). Mutational analysis was used to further elucidate the role of a hydrophobic ␣-helical microdomain that is located upstream of the MEEVD motif in Hsp90. This C-terminal domain is targeted by novobiocin and contains elements critical for Hsp90 dimerization (23,26,35). In the present study we demonstrate that substitution of selected hydrophobic and basic residues within this region markedly altered the chaperone activity of a C-terminal Hsp90 fragment, but could not prevent Hsp90 dimer formation. However, coumermycin A1, a coumarin that is structurally related to novobiocin with more potent Hsp90 inhibitory properties (24), was shown to interfere with Hsp90 dimerization. Our studies have revealed that novobiocin has a differential ability to interfere with TPR immunophilin recognition by Hsp90. Furthermore, novobiocin and coumermycin A1 were shown to effectively reduce the level of glucocorticoid receptor protein in HeLa cells. Our results provide new insights into the unique mode of action of novobiocin-related coumarins in a region of Hsp90 that is exquisitely sensitive to modulation. These novel Hsp90 inhibitors may provide a ligand-independent approach to the treatment of breast and prostate cancers.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The expression plasmid pGEX4T.1-hCyP40-(185-370) wild type (33), containing the C-terminal half of human CyP40 encoding the TPR domain and flanking acidic and basic domains, was used as a template for the generation of hCyP40 C-terminal truncation mutants using PCR amplification. DNA fragments for the 185-367, 185-363, and 185-357 mutants were generated with flanking EcoRI/SalI sites then ligated into the pDrive cloning vector (Qiagen, Germany), and fidelity was confirmed by automated sequencing. Each DNA fragment was then released from pDrive by EcoRI/SalI digestion and ligated into the EcoRI/SalI sites of the glutathione S-transferase (GST) fusion expression vector pGEX4T.1 (Amersham Biosciences). Recombinants were transformed into the bacterial strain, Escherichia coli (E. coli) XLI Blue, and clones were confirmed for correct expression constructs by restriction enzyme analysis. Recombinant DNA was then transformed into the high protein expression bacterial strain, E. coli BL21 (codon ϩ). All expression plasmid constructs further described in this report, which were generated using PCR-amplified DNA, employed this pDrive protocol and were transformed into E. coli BL21 (codon ϩ) cells, along with expression constructs prepared by subcloning. The bovine CyP40 construct, pGEX2T-bCyP40-(185-352), has been described previously (32).
Site-directed Mutagenesis-The pVP16-hHsp90␤-(520 -724) plasmid (33) was used as a template for the generation of single amino acid substitution mutants using the QuikChange site-directed mutagenesis kit (Stratagene). Once the presence of mutations and the authenticity of the remaining sequence were confirmed, wild-type and mutant hHsp90␤-(520 -724) fragments were released from pVP16 by NotI digestion and cloned into the NotI site of pGEX4T.1, allowing expression of GST-hHsp90␤-(520 -724) fusion proteins for chaperoning studies. The pDrive-hHsp90␤-(527-724) plasmid was used as a template to generate tandem alanine mutations through residues 645-672 of hHsp90␤. DNA for the mutant hHsp90␤-(527-724) fragments was released from pDrive by NdeI digestion and subcloned into the NdeI site of pET28a(ϩ), facilitating expression of the C-terminal Hsp90 proteins for use in dimerization assays.
Protein Purification-The expression and purification of GST-tagged recombinant proteins has been described previously (33). Briefly, primary cultures of E. coli BL21 (codon ϩ) cells transformed with expression plasmids for the immunophilins, p50 cdc37 or GST alone were grown overnight and the next day, cells were pelleted by centrifugation, resus-pended in 100-ml culture medium, and grown for 1 h at 37°C. Protein expression was induced with 0.2 mM isopropyl ␤-thiogalactopyranoside (Promega Corp., Madison, WI) for a period of 1.5 h. Cultures were centrifuged and bacterial pellets resuspended in chilled MTPBS buffer (150 mM NaCl, 16 mM Na 2 HPO 4 , 4 mM NaH 2 PO 4 , pH 7.3) containing 1 mM phenylmethylsulfonyl fluoride (PMSF, Roche Molecular Biochemicals, Basel, Switzerland) and frozen overnight at Ϫ70°C. After thawing, the following reagents were added to the pellets in the concentrations indicated: 1 mM PMSF, 5 mM benzamidine, 5 mM dithiothreitol, 2 mM EDTA, 0.3 mg/ml lysozyme (Sigma-Aldrich) and 0.2% v/v Triton X-100 (Roche Applied Science). Cell suspensions were incubated for 5 min on ice, and then further lysed by sonication on ice, and the lysates were clarified by ultracentrifugation at 75,000 ϫ g for 50 min at 4°C. GSTtagged proteins present in the supernatant were then purified by affinity chromatography with glutathione-Sepharose 4B beads (Amersham Biosciences) for 45 min at 4°C in the presence of 1 mM PMSF and 5 mM benzamidine. Protein-bound beads were washed six times in chilled MTPBS buffer containing 1 mM PMSF, 5 mM benzamidine, and 1% v/v Triton X-100 (first three washes only). GST-tagged proteins were eluted twice with 600 l of 10 mM reduced glutathione (Sigma-Aldrich) in 50 mM Tris HCl, pH 8.0, at 4°C, followed by centrifugation to remove the beads.
The expression and purification of His-tagged hHsp90␤ full-length and C-terminal wild-type and mutant recombinant proteins were performed essentially as described previously (33). Purification conditions were modified to preserve the integrity of the Hsp90␤ C-terminal tandem alanine mutants. Primary cultures of E. coli BL21 (codon ϩ) cells transformed with Hsp90␤ expression plasmids were grown overnight, and on the next day they were pelleted by centrifugation, resuspended in either 100 ml or 1 liter of culture medium, and grown for a further 2-3 h at 37°C. Proteins were overexpressed by induction with 0.2 mM isopropyl ␤-thiogalactopyranoside at 37°C for 15 min, followed by centrifugation of cells and resuspension of the bacterial pellet in chilled Buffer 1 (50 mM sodium phosphate, pH 8.0, 1.8 M NaCl, 50 mM imidazole) with 1 mM PMSF before storage at Ϫ70°C. Cells were then thawed, and the following agents added to the specified concentrations: 5 mM benzamidine, 0.2% v/v Triton X-100, 15 mM ␤-mercaptoethanol (Sigma-Aldrich), and 1 mg/ml lysozyme. The mixtures were then incubated on ice for 5 min. After addition of PMSF to 1 mM, cells were lysed by sonication, followed by ultracentrifugation as described above. Supernatants containing the His-tagged fusion protein were purified by affinity chromatography through mixing with Ni-NTA-agarose beads (Qiagen) for 45 min at 4°C in the presence of 1 mM PMSF and 5 mM benzamidine. Protein-bound beads were washed six times with chilled Buffer 1 containing 1 mM PMSF, 5 mM benzamidine, and 1% v/v Triton X-100 (first three washes only). His-tagged fusion proteins were eluted twice with 600 l of Buffer 2 (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 250 mM imidazole), followed by centrifugation to remove the beads. Protein eluates were pooled, concentrated, and dialyzed against either 50 mM potassium phosphate, pH 7.4, for the dimerization studies or Buffer A (10 mM Tris-HCl, pH 7.3, 100 mM KCl) for other studies, in a Centricon YM-30 microconcentration tube (Millipore, Bedford, MA) to a final volume of ϳ400 l before storage at Ϫ70°C. All protein preparations were assessed for purity by SDS-PAGE, and protein concentration was determined by the Bradford assay (39).
ELISA Microtiter Plate Assays-This assay was modified from that previously described (33) to investigate the effect of novobiocin (Sigma-Aldrich) on the binding of GST-immunophilins to His-hHsp90␤-(527-724) and of GST-hp50 cdc37 to full-length His-hHsp90␤ protein. Using Immulon 4HBX 96-well microtiter plates (Dynex Laboratories Inc., Chantilly, VA), 80 nM of Hsp90 protein in 100 mM NaHCO 3 , pH 8.5, was applied in 50-l aliquots to triplicate wells for each sample and incubated at 4°C for 3 h. For each sample, an equivalent amount of bovine serum albumin (Promega) was added to adjacent wells in triplicate as a control for nonspecific binding. Wells were blocked with 100 l of 0.1% w/v bovine serum albumin in 100 mM NaHCO 3 , pH 8.5, for 16 h at 4°C, and then washed three times with chilled Tris-buffered saline (50 mM Tris, 150 mM NaCl, pH 7.5) supplemented with 0.075% v/v Tween 20 (TBST) to remove excess protein. Following washing, 50 l of novobiocin in TBST was added at the concentrations indicated and incubated for 1 h at 4°C. Then, 0.25 M GST-cochaperone was added in TBST, and the mixture was further incubated for 1 h at 4°C. The remainder of the protocol was identical to that described previously (33) except that a Titertek Multiskan PLUS plate reader (EFLAB, Finland) was used to measure absorbance. Levels of immunophilin and hp50 cdc37 binding to Hsp90 were determined after subtracting the mean absorbance for the nonspecific controls and the total binding for GST alone and were expressed as a percentage relative to binding observed in the absence of novobiocin. All assays were repeated at least three times.
Chaperone Assays-Chaperone function was measured using a rhodanese aggregation assay adapted from previously published methods (31,40). Bovine rhodanese (Sigma-Aldrich) was prepared in 50 mM HEPES, pH 7.0, to a concentration of 100 M and stocks were stored at Ϫ70°C. For the working concentration, one vial of rhodanese was denatured overnight at room temperature in denaturing buffer (6 M guanidine hydrochloride, 50 mM HEPES, pH 7.0, 100 mM NaCl), to a concentration of 25 M. GST-hHsp90␤-(520 -724) wild-type and point mutants were analyzed at concentrations of 0, 3, 5, or 10 M in assay buffer (50 mM HEPES, pH 7.0, 100 mM NaCl). As a negative control, 3-20 M of GST protein was monitored for chaperone activity. No decrease in rhodanese aggregation was observed for any GST concentration tested. Denatured rhodanese was added in the assay cuvette to a final concentration of 0.5 M, and the rate of its aggregation was monitored at 320 nm over a 20-min period at room temperature, using a GBC UV-Visible 916 spectrophotometer (GBC Scientific, Dandenong, Victoria, Australia). Absorbance was plotted as percent rhodanese aggregation over 20 min after normalizing against assays with rhodanese alone. To study the effect of novobiocin on Hsp90 chaperone function, 60 M novobiocin was incubated with 5 or 10 M GST-hHsp90␤-(520 -724) on ice for 1 h prior to the chaperone assay. Measurement of the rhodanese alone curve was monitored as described above but with the addition of 60 M novobiocin.
Dimerization Assays-C-terminal Hsp90 dimerization was assessed by chemical cross-linking using bis(sulfosuccinimidyl) suberate (BS) (Pierce), an amine-reactive cross-linker. His-hHsp90␤-(527-724) protein was added to a 20-l reaction volume containing 50 mM potassium phosphate, pH 7.4, to a final concentration of 2 M. As a control to validate the dimerization assay, 2 M His-hFKBP52 (41) was studied in a separate reaction. Optimization of effective cross-linker concentration was carried out for His-hHsp90␤-(527-724) (or His-hFKBP52 control) using 0 -60 M BS. Stock solutions of BS were freshly prepared immediately before addition to the reaction tubes. Hsp90 dimerization in the presence of coumermycin A1 (Sigma-Aldrich) was performed as follows: 2 M His-hHsp90␤-(527-724) was treated with 0 -1.0 mM coumermycin A1 dissolved in Me 2 SO and incubated on ice for 1 h, prior to chemical cross-linking with 15 M BS. For the dimerization study of Hsp90 tandem alanine mutants, 0 or 15 M BS was used for each mutant. Cross-linking, with or without BS, was carried out at room temperature for 1 h, and reactions were subsequently quenched by incubating with 50 mM Tris-HCl, pH 7.5, for a further 15 min. Each reaction mixture was then boiled in the presence of 2ϫ sample buffer, and 10 l was subjected to SDS-PAGE. Resolved protein bands were electroblotted overnight at 4°C onto polyvinylidene difluoride membranes (Amersham Biosciences), blocked with 5% w/v nonfat milk in TBST (0.02% v/v Tween 20), then probed with AC88 (42) or Hi52c antibody (43) to detect Hsp90 or FKBP52, respectively. Western blots were stained with enhanced chemiluminescence reagent (PerkinElmer) and exposed to Hyperfilm (Amersham Biosciences).
Steroid Receptor Protein Depletion-HeLa cells were plated in 6-well plates (Sarstedt, Numbrecht, Germany) in Dulbecco's modified Eagle's medium containing steroid-free fetal calf serum (Trace Biosciences, Noble Park, Victoria, Australia) at a density of 5 ϫ 10 5 cells/well. Cells were treated with increasing concentrations of novobiocin (0.25, 0.5, and 1.0 mM), coumermycin A1 (0.05, 0.1 mM), or control diluent for 16 h. Wells were aspirated, and cells washed with chilled phosphatebuffered saline, then scraped into 150 l of chilled 1ϫ sample buffer and boiled for 5 min. Lysates were quantitated for protein using the bicinchoninic acid assay (Pierce), and 100 g of total cellular protein was mixed with 5% v/v ␤-mercaptoethanol and 0.005% w/v bromphenol blue, boiled for 5 min, subjected to SDS-PAGE, and transferred to a Hybond-C Super nitrocellulose membrane (Amersham Biosciences). Glucocorticoid receptor was detected with the rabbit polyclonal antibody, H-300 (Santa Cruz Biotechnology Inc., Santa Cruz, CA). Protein loading was controlled by probing for ␣-tubulin with a mouse anti-␣tubulin monoclonal antibody (Sigma-Aldrich).
Statistical Analyses-For conventional pull-down assays, data were analyzed using the program GraphPad InStat (GraphPad Software, Inc., San Diego, CA). For the CyP40 C-terminal truncation mutants, analysis of variance performed among all groups demonstrated an overall statistically significant difference in the degree of binding. The Tukey-Kramer multiple comparison test was then performed to compare the GST-CyP40-(185-370) wild type with each truncation mutant for Hsp90 binding (p values Ͻ 0.05 were considered to be statistically significant). IC 50 data for novobiocin inhibition of immunophilin cochaperone binding to Hsp90 was analyzed using one-way analysis of variance followed by Duncan's post hoc test.

Hsp90 Binding by Truncation Mutants within the C-terminal Basic
Domain of CyP40-We have previously reported on the importance of a basic region, flanking the C-terminal end of the CyP40 TPR domain, for efficient interaction of this immunophilin with Hsp90 (32). The region is incorporated into the final helix (helix V) of CyP40, continuing the anti-parallel helical structure derived from the tandem arrangement of three units of the TPR motif (44). Although the C-terminal helix was predicted to form part of an extended Hsp90 binding surface, a subsequent mutational analysis of selected residues (Lys-349, Lys-353, Lys-360, Tyr-365, and Phe-369) within this region was unable to identify specific determinants of Hsp90 interaction, suggesting that multiple elements might be involved (33). To further define these binding determinants, we prepared C-terminal truncation mutants of GST-tagged CyP40-(185-370) for in vitro interaction studies with His-Hsp90␤-(527-724). We were guided in the design of these mutants by the recent results of Cheung-Flynn and colleagues (34), who have identified a conserved sequence, the charge-Y motif, that is essential for FKBP51 and FKBP52 binding to Hsp90 and is common to a number of Hsp90-binding cochaperones, including CyP40 (Fig. 1A). Thus, the GST fusion constructs for hCyP40-(185-367), -(185-363), and -(185-357) progressively deleted elements of the charge-Y motif conserved in the CyP40 protein (Fig. 1A). Pull-down assays, performed with the Hsp90␤-(527-724) C-terminal fragment, showed a gradual loss of Hsp90 binding with consecutive deletions (Fig. 1C). However, only the 185-357 construct, in which the charge-Y motif was completely removed, showed a significant decline in binding level from that achieved by the CyP40-(185-370) wild-type protein (Fig. 1, B and C). Interestingly, this truncation mutant still retained a 50% binding ability for Hsp90 relative to CyP40-(185-370), whereas the bCyP40-(185-352) deletion mutant displayed a much lower retention rate of 16%, consistent with our previous findings (32). The results suggest that, for CyP40, residues immediately upstream of the charge-Y motif also play a significant role in Hsp90 recognition.
Hsp90 Is Characterized by a Multifunctional C-terminal Domain-The preferential assembly of immunophilin cochaperones in steroid receptor complexes will likely be determined by specific contacts between receptors and immunophilins, and the receptor/cochaperone interface may be dependent on their own unique interactions with Hsp90 (3). The C-terminal domain of Hsp90 appears central to the function of this molecular chaperone. This region mediates Hsp90 dimerization, displays independent chaperone function, and contains structural elements important for steroid receptor interaction (2). The MEEVD motif at the extreme C terminus of Hsp90 serves as a common recognition site for the immunophilin cochaperones (27)(28)(29)(30)(31). The Hsp90 inhibitor, novobiocin, binds to a defined sequence within the dimerization domain (23). It is of interest that deletion of a hydrophobic microdomain, encompassing the putative novobiocin interaction site, disrupted Hsp90 dimerization and dramatically reduced binding to the immunophilin cochaperones (27). Therefore, this hydrophobic microdomain may have a stabilizing influence on Hsp90 binding to immunophilin cochaperones. Alternatively, the hydrophobic sequence may form part of a second interaction site for the immunophilin cochaperones, with the potential to impose additional specificities on Hsp90 interaction beyond those determined by the MEEVD peptide. Fig. 2 provides a summary of the essential elements associated with the multi-functional C-terminal domain of the Hsp90␤ isoform.

Influence of Novobiocin and Amino Acid Substitutions within the C-terminal Hydrophobic Microdomain of Hsp90 on Chaperone
Function-Translation of structural information recently determined for the E. coli Hsp90 homolog, htpG (26), shows the Hsp90␤ C-terminal domain to be comprised of five ␣-helices (Fig. 2). The C-terminal helices (H4 and H5) constitute the essential components of the dimerization motif. The hydrophobic microdomain, proposed as the binding site for novobiocin and which may mediate additional contacts with TPR-containing cochaperones (27), is located within the C-terminal half of helix 4 (Fig. 2). Separate studies previously investigated the influence on Hsp90 chaperone function of alanine substitutions for acidic residues within the C-terminal domain of Hsp90␣ (31), as well as serine substitutions of contiguous leucine residues (Leu-665/Leu-666 and Leu-671/ Leu-672) in the Hsp90␣ hydrophobic microdomain (35). In the present study, we generated alanine-substituted mutants for two basic residues (Lys-649 and Lys-652) and seven large hydrophobic amino acids (Leu-654, Leu-657, Phe-659, Leu-663, Leu-664, Phe-668, and Leu-670) within the helix 4 hydrophobic microdomain of Hsp90␤ and investigated the effects of these alterations on Hsp90 chaperone function. Wild-type GST-Hsp90␤-(520 -724) prevented aggregation of rhodanese in a concentration-dependent manner (Fig. 3A). Of the nine point mutants, only L670A displayed wild-type chaperone activity (Fig. 3A). Alanine substitution of residues Leu-657 and Phe-659 appeared to potently increase chaperone function compared with the wild-type protein, with the lowest concentration (3 M) of either mutant eliciting maximal chaperone activity (Fig. 3B). Compromised chaperone function was observed for the remaining six Hsp90 point mutants, which achieved almost 50% reduction in rhodanese aggregation at 10 M concentration compared with an 80% reduction in aggregation observed for the wildtype Hsp90␤ C-terminal fragment (Fig. 3C). Collectively, the results indicate that the C-terminal chaperone function of Hsp90 is sensitive to structural alterations within the hydrophobic microdomain located in helix 4.
Because novobiocin targets this region in Hsp90, we next determined whether the drug impacts on Hsp90 chaperone function. In preliminary  (40,49,50,69) and contains structural elements important for substrate interaction (26, 60, 70 -72). The location of ␣-helices (H2-H5) has been adapted from the structure of the C-terminal domain of htpG, the E. coli Hsp90 homolog (26). The hydrophobic helix 2 (H2) has been proposed to provide a surface for substrate binding, while helices 3-5 (H3-H5) interface with the 3Ј-5Ј helices of the opposite monomer allowing dimer formation (26). The minimal domain eliciting chaperone and folding activity has been delineated to helices 4 and 5 (H4 and H5) (50), but residues outside of this region within the C-terminal domain may also participate in chaperone function (31). Residues 645-673, located within helix 4 and part of helix 5, include the hydrophobic microdomain, which overlaps the putative interaction site for novobiocin (23). Deletion of the hydrophobic sequence (⌬653-669) has been shown to impact Hsp90 dimerization, substrate binding, and interaction with TPR-containing cochaperones, including Hop and the immunophilins (27,60,70,71). The MEEVD peptide at the extreme C terminus constitutes the primary binding site for the TPR-containing cochaperones (27)(28)(29)(30)(31). An unstructured acidic region of ϳ40 amino acids separates the MEEVD motif from helix 5 and has been identified as the binding site for cisplatin (53,54). MARCH 17, 2006 • VOLUME 281 • NUMBER 11 experiments we determined that 60 M novobiocin was the maximum concentration that could be used without adverse effects on the rhodanese aggregation curve (data not shown). We then treated GST-Hsp90␤-(520 -724) (5 M) with 60 M novobiocin prior to assessing its chaperone activity. A comparison of results with untreated GST-Hsp90␤-(520 -724) protein (Fig. 4A) showed that novobiocin blocked the chaperone activity of the Hsp90 C-terminal fragment (Fig. 4B). In a separate experiment, performed with 10 M of the GST-Hsp90␤ fusion protein, novobiocin caused a substantial inhibition of Hsp90 chaperone activity (Fig. 4, A and B). Novobiocin therefore has the ability to modulate the independent chaperone function of the Hsp90 C-terminal domain.

Functional Interactions of Hsp90 C-terminal Domain
Cochaperone Binding to Hsp90 Treated with Novobiocin-Novobiocin has been shown to interfere with the retention of Hsp70 and p23 chaperone components within Hsp90-Hop-Hsp70 and Hsp90-p50 cdc37 /immunophilin-p23 complexes, formed at early and late stages of the Hsp90-client assembly cycle, respectively (23). Because Hsp70 is not directly associated with Hsp90, loss of this chaperone may have resulted from a disruption of the interaction of Hop with the Hsp90 C-terminal domain (45). p23 binding requires Hsp90 in its ATP-bound state, and, although the precise binding site for p23 in Hsp90 has not been defined, it appears not to involve the putative interaction site identified for novobiocin (46). The negative impact of novobiocin on the association of p23 with Hsp90 multichaperone complexes may therefore be indirect. Using an ELISA microtiter plate assay previously described for the interaction of GST-CyP40 with Hsp90 (33), we next investigated the direct effect of novobiocin on the binding of His-Hsp90␤-(527-724) with the immunophilin cochaperones. Preliminary experiments with GST-CyP40 indicated that increasing concentrations of novobiocin impacted immunophilin binding only if the drug was incubated with Hsp90 prior to addition of GST-CyP40 fusion protein (data not shown), suggesting that the immunophilin may efficiently block novobiocin access to its interaction site. Fig. 5A shows that titration of the Hsp90␤-(527-724) fragment with increasing novobiocin concentrations had a differential effect on immunophilin binding, with Hsp90-FKBP52 interaction being most stable and binding of Hsp90 to PP5 being most sensitive to the drug. Intermediate inhibition profiles were seen for CyP40 and FKBP51. Almost complete inhibition of Hsp90 binding to all of the immunophilins was achieved with 10 mM novobiocin. IC 50 values of 4.00 Ϯ 0.03, 3.47 Ϯ 0.03, 3.27 Ϯ 0.03, and 3.08 Ϯ 0.07 mM novobiocin were calculated for FKBP52, CyP40, FKBP51, and PP5, respectively. These values were all determined to be significantly different (p Ͻ 0.05) from one another. A comparison of residual immunophilin binding levels at specific novobiocin concentrations was also informative, providing a further measure of stability of interaction with Hsp90. Thus at 2.5 mM novobiocin these were 97%, 90%, 78%, and 68% for FKBP52, CyP40, FKBP51, and PP5, respectively, while at 5 mM concentrations of the drug, the corresponding binding levels were 31%, 8%, 4%, and 8% (Fig. 5A). Our results have revealed that novobiocin has a differential ability to interfere with recognition of the immunophilin cochaperones by Hsp90, providing further evidence that upstream regions close to or overlapping the C-terminal hydrophobic microdomain, may contribute to the Hsp90-immunophilin interaction.
The cochaperone p50 cdc37 facilitates protein kinase loading by interacting with regions of the Hsp90 N-terminal domain implicated in nucleotide binding, thus arresting the Hsp90 ATPase cycle (25). Although novobiocin promotes the rapid depletion of cellular Hsp90 client protein kinases, the precise mechanism through which the drug antagonizes Hsp90 function has yet to be defined. To further address this question, we sought to determine the influence of novobiocin on Hsp90 interaction with the p50 cdc37 cochaperone. By applying the ELISA-based methodology we showed that pre-treatment of His-tagged full-length Hsp90␤ with increasing amounts of novobiocin caused a progressive decline in retention levels of GST-p50 cdc37 (Fig. 5B). The result suggests that novobiocin may act allosterically to prevent association of Hsp90 with p50 cdc37 and is consistent with evidence that mod- ulation of sequences within the Hsp90 C-terminal domain can significantly impact the functional activity of the Hsp90 amino terminus (47).
Role of Helix 4 Residues in Hsp90 C-terminal Domain Dimerization-Accumulating evidence from the work of others (23,27,31,35), together with our own data from the present study, has led to the understanding that manipulation of helix 4 in the Hsp90 C-terminal domain, either by deletion or substitution of amino acid residues or the targeting of a defined hydrophobic sequence by novobiocin, impacts important Hsp90 functions, including dimerization, chaperone activity, client protein assembly, and cochaperone interaction. To map elements within this region that might contribute to Hsp90 dimerization, we generated a series of tandem alanine mutants encompassing helix 4 and the adjoining loop to helix 5 (residues 645-673) (see Fig. 2). All mutants were derived from the His-Hsp90␤-(527-724) wild-type template.
After bacterial expression and purification, the mutants were assessed for their effect on Hsp90 dimerization using BS, a water-soluble cross-linker. This cross-linking reagent has previously been used by Chen et al. (27) in dimerization studies of wild-type and mutated forms of Hsp90. To determine an appropriate concentration of cross-linker to use in our experiments, 2 M of His-Hsp90␤-(527-724) protein was treated with 0 -60 M BS, and cross-linked dimers were observed by Western blotting using the Hsp90 antibody, AC88 (Fig. 6A). Increasing the amount of cross-linker led to a greater yield of Hsp90 dimer, although, under the conditions used, the highest concentration of BS (60 M) was unable to dimerize all of the available monomeric Hsp90. To verify that the cross-linking agent was specifically targeting dimers, His-FKBP52 (which functions as a monomer (48)) was also tested over the same concentration range of BS cross-linker. Dimers were not observed for FKBP52 protein at any concentration of BS tested (data not shown), thus validating the dimerization assay for Hsp90. A BS concentration of 15 M was chosen for the dimerization studies. The impact of the tandem alanine mutations on Hsp90 dimerization was determined under these conditions. All tandem alanine mutants displayed a dimer fragment when cross-linked with 15 M BS (Fig. 6B), indicating that mutation of these residues was not sufficient to preclude dimerization.
The Novobiocin-related Coumarin, Coumermycin A1, Disrupts Hsp90 Dimerization-As the dimerization domain overlaps the putative novobiocin interaction site, we were interested in testing whether this drug or related coumarin antibiotics could interfere with the formation of Hsp90 dimers. Because novobiocin reacts covalently with the primary amine-reactive BS cross-linker, the related coumarin, coumermycin A1, a potent Hsp90 inhibitor (24), was used in our experiments. The dimerization assay for the His-Hsp90␤-(527-724) fragment (2 M) was performed in the presence of 15 M BS, with increasing concentrations of coumermycin A1. Fig. 7A shows that dimer levels were progressively reduced with increasing coumarin concentration. The effect was maximal at a 300-fold excess of coumermycin A1 (0.6 mM), although higher concentrations of the drug to 1.0 mM did not completely abrogate Hsp90 dimer formation (Fig. 7B). Our results suggest that coumermycin A1 interferes with dimerization of the Hsp90 C-terminal domain.
Coumarin Antibiotics Deplete Glucocorticoid Receptor Levels in HeLa cells-Neckers and coworkers (23,24) previously reported that novobiocin interferes with Hsp70 and p23 association with Hsp90 during assembly of distinct multichaperone complexes involving Hsp90-Hop-Hsp70 and Hsp90-p50 cdc37 /immunophilin-p23, respectively, causing depletion of a number of oncogenic protein kinases. This, coupled with our own results that coumarin antibiotics may antagonize Hsp90 function by destabilizing the Hsp90 dimer and interfere with immunophilin cochaperone recognition by Hsp90, suggested that these novel Hsp90 inhibitors might also target steroid receptor-Hsp90 complexes. We therefore examined the influence of these drugs on the stability of the glucocorticoid receptor in HeLa cells. Fig. 8 shows that incubation of HeLa cells over 16 h with novobiocin or coumermycin A1 caused a dose-dependent depletion of cellular glucocorticoid receptor levels, with coumermycin A1 demonstrating a marked increase in potency relative to novobiocin.

DISCUSSION
Although the core TPR domain of CyP40 contains key residues that accommodate the MEEVD C-terminal peptide of Hsp90, it is not sufficient for high affinity Hsp90 interaction and requires a basic domain within a downstream helical region to maintain stable CyP40-Hsp90 association (32). It is understood that multiple elements within the C-terminal basic domain contribute to Hsp90 binding, because investigation of point mutations failed to identify specific residues that may be involved (33). In the present study, progressive C-terminal deletions through the basic domain of the CyP40 protein have enabled us to confirm the role of the charge-Y motif in Hsp90 binding. This sequence was previously found to play a critical role in FKBP51 and FKBP52 recognition of the Hsp90 chaperone (34). Sequences downstream of this motif in FKBP51 and FKBP52 were previously shown to respectively increase and attenuate the interaction of these immunophilins with Hsp90, but truncations that included the whole of the consensus motif caused almost a complete loss of binding (34). Our study also suggests that an adjacent sequence, 353 KIKAQ 357 , immediately upstream of the charge-Y motif (Fig. 1A), also contributes significantly to CyP40 interaction with Hsp90. Although this sequence is partially conserved in FKBP51 and FKBP52, it appears to have little influence on their binding capacity for Hsp90 (34). Thus, our results highlight similarities and differences between CyP40 and the FKBP cochaperones that may allow these immunophilins to recognize distinct binding surfaces in Hsp90 and differentially modify its function.
The initial observation that the C-terminal domain of Hsp90 can enhance the DNA-binding activity of MyoD1, a basic helix-loop-helix transcription factor (49), led to the identification of a 50-amino acid region in Hsp90 responsible for this specific folding activity (50). From the structural work of Harris et al. (26), we have located this chaperone function to helices 4 and 5 within the Hsp90 C-terminal domain (Fig. 2), a region that mediates dimerization (51,52), incorporates the binding site for novobiocin (23) and appears to be important for the association of Hsp90 with immunophilin cochaperones (27). Using a C-terminal Hsp90 fragment that also included helix 3 (Fig. 2) and all downstream residues to the C terminus, Young et al. (40) subsequently confirmed the chaperone activity of this domain. We have performed a mutational analysis of two basic and seven large hydrophobic residues within helix 4 of the C-terminal Hsp90 domain and shown that most of the mutations (K649A, K652A, L654A, L663A, L664A, and F668A) impacted negatively on Hsp90 chaperone function. Mutation of the Leu-670 residue, within the loop separating helices 4 and 5, had no influence on Hsp90 chaperone activity. Intriguingly, alanine substitution of the residues Leu-657 and Phe-659, both centrally located within helix 4, caused a dramatic increase in the chaperoning ability of Hsp90␤. Yamada et al. (35) have reported a reduced chaperone activity for a recombinant Hsp90␣ fragment containing a double serine mutation for the Hsp90␤ equivalent residues, Leu-657/Leu-658. Consistent with our own results, a more significant loss of chaperone function was noted on serine substitution of residues, Leu-663/Leu-664. The above results, specifically those involving alteration of large hydrophobic residues, provide an interesting comparison with the influence of modified acidic residues within helix 4. For example, the simultaneous mutation to alanine of Hsp90␤ equivalent residues Glu-643/Asp-645, which cap helix 4 (Fig.  2), caused a decrease in chaperone activity (31). On the other hand, alanine substitution of Asp-648 and Asp-653 maintained wild-type activity, whereas mutation of the Glu-660 residue to alanine enhanced chaperone function. Together the results support the notion that the structural integrity of helix 4 is critical for the normal chaperoning function of the Hsp90 C-terminal domain.
Although our experiments were limited by the low level of novobiocin (60 M) tolerated by the rhodanese-based chaperone assay, our study has demonstrated that the drug can inhibit the chaperone function of a recombinant C-terminal Hsp90␤ fragment. Cisplatin, which binds to an unstructured, acidic region of Hsp90, downstream of the novobiocin interaction site (Fig. 2), has also been shown to reduce Hsp90 chaperone activity (53). Some overlap between the interaction domains of these drugs is suggested by the ability of novobiocin to hinder cisplatin binding when preincubated with Hsp90 (54). It has been speculated that the binding of cisplatin to the Hsp90 C-terminal FIGURE 5. Inhibition of cochaperone-Hsp90 interaction by novobiocin. Using an ELISA microtiter plate method, purified His-Hsp90␤-(527-724) protein fragment (A) or full-length His-Hsp90␤ protein (B) was bound to the wells and incubated with 0 -10 mM novobiocin prior to the addition of GST-cochaperone fusion proteins. Bound GST-immunophilins (A) or GST-p50 cdc37 (B) were detected using a goat anti-GST antibody followed by rabbit anti-goat IgG antibody conjugated to horseradish peroxidase and 3,3Ј,5,5Јtetramethylbenzidine as substrate. Absorbance values were read at 450 nm for three independent assays, each assay being performed in triplicate. After subtracting the mean binding for the GST control (which displayed minimal Hsp90 binding), absorbance values were calculated as a percentage of that observed in the absence of novobiocin (100% binding).
domain might restrict access of Hsp90 chaperone substrates (53). The binding of novobiocin to both interaction sites in the Hsp90 dimer is highly co-operative, and there is evidence that this dual occupancy is required for novobiocin to induce distinct conformational changes in Hsp90 (55). The altered conformation then may have diminished chaperone activity. Indeed, it has been suggested that the unique structure imposed by novobiocin on Hsp90 might favor the release of substrate (55).
Using a purified recombinant Hsp90 C-terminal fragment incorporating the putative novobiocin interaction site, together with GSTtagged immunophilins, we have shown for the first time that, prior incubation of Hsp90 with increasing concentrations of novobiocin leads to a direct blockade of Hsp90 binding to immunophilin cochaperones. The variation between IC 50 values generated for specific Hsp90-immunophilin cochaperone interactions suggests that Hsp90-bound novobiocin has a differential effect on immunophilin binding and that regions upstream of the well characterized C-terminal MEEVD binding motif, including those overlapping or close to the helix 4 hydrophobic microdomain, contribute to the Hsp90/immunophilin binding interface. This is supported by additional observations that novobiocin had little influence once immunophilin binding to Hsp90 had taken place, suggesting that the immunophilin cochaperones might effectively mask the novobiocin interaction site. It is noteworthy that Yun et al. (55) also observed the loss of FKBP52 (as well as Hsp70 and p23) from Hsp90 complexes in rabbit reticulocyte lysate with similar high levels of novobiocin (5-10 mM). These workers showed that, in the presence of the drug, the Hsp90 C-terminal domain adopts a protease-resistant conformation unfavorable for the continued association of immunophilin cochaperones. Our results are consistent with earlier reports that the immunophilins and the TPR cochaperone, Hop, make distinct and extensive contacts within the Hsp90 C-terminal domain leading to specific modulation of Hsp90 function (27,31,34,56).
The ATPase cycle is critical for the chaperoning function of Hsp90 and is regulated through the combined interplay of the C-terminal dimerization domain and ATP-driven association of the N-terminal domain, forming a molecular clamp, the opening and closing of which is linked to Hsp90 ATPase activity (57). Although p50 cdc37 was initially thought to share or have an overlapping interaction site with immunophilin cochaperones in Hsp90 (58), the binding domain has recently been defined within the N-terminal domain of Hsp90 (25). p50 cdc37 arrests the ATPase cycle by blocking ATP hydrolysis and holds Hsp90 in an "open" conformation to facilitate kinase substrate loading. Adopting a similar approach to that already described for the immunophilin cochaperones, we showed that novobiocin also precluded the interaction of GST-p50 cdc37 with full-length Hsp90. On the other hand, Yun et al. (55) reported that the basal interaction between p50 cdc37 and Hsp90, in reticulocyte lysate, was unaffected by the drug. Presumably, the differences in results have arisen from differences in experimental protocols. Our results provide further evidence of cross-talk between the Cand N-terminal domains of Hsp90. There is now accumulating evidence that the conformational effects of novobiocin on the Hsp90 C terminus are communicated to the N-terminal domain (23,54). ATP binding in the Hsp90 N terminus leads to the exposure of a cryptic nucleotidebinding site within the C-terminal domain, increasing accessibility to novobiocin (54,59). Drug interaction with the C terminus disrupts nucleotide binding at both C-and N-terminal sites (23,54,59). Evidence suggests that novobiocin induces alterations in Hsp90 proteolytic fragmentation patterns caused by structural changes throughout multiple domains of the Hsp90 protein (55).
Studies have shown that internal deletion of helix 4 residues equivalent to 653-669 in Hsp90␤ compromises dimerization of the molecular chaperone, disrupting a structural domain critical for its normal function and interaction with TPR-containing cochaperones, including Hop and the immunophilins (27,60). This region also overlaps the putative binding site for novobiocin within the Hsp90 C-terminal domain (Fig.  2). With a series of tandem alanine mutations through helix 4 of Hsp90, and using a dimerization assay incorporating low concentrations of the BS cross-linker to allow ready evaluation by SDS-PAGE, we found that none of the modified segments structurally altered the dimerization domain sufficiently to prevent the cross-linking reaction. The dimerization interface in htpG involves the cradling of helix 5 within a hydrophobic groove elaborated by helices 3Ј, 4Ј, and 5Ј of the partner mono-mer (26). Conserved large hydrophobic residues (almost exclusively leucines) at 654, 657, 664, and 670 in helix 4 form hydrophobic contacts with key hydrophobic amino acids in helix 5Ј, but extensive hydrophobic links also exist between residues in helices 5 and 5Ј and helices 3 and 5Ј. It is possible, therefore, that at least part of the dimer interface remains viable in Hsp90 C-terminal constructs containing the tandem alanines 653 DLVV 656 , 657 LL 658 , 663 LL 664 , and 669 SLED 672 . Yamada et al. (35), however, have used a bacterial two-hybrid binding assay to show that serine substitution of the contiguous leucine residues Leu-657/ Leu-658 and Leu-663/Leu-664, within helix 4, disrupts dimer formation.
Our observation that coumermycin A1, a coumarin antibiotic that is structurally related to novobiocin (24), interferes with dimerization of the Hsp90 C-terminal domain signifies a novel mode of action for the coumarin-based inhibitors through which they might antagonize Hsp90 function by destabilizing the Hsp90 dimer. Nucleotide binding to the cryptic site, exposed within the C-terminal domain after the primary N-terminal nucleotide site has bound ATP (54), may trigger conformational changes leading to the transient dissociation of the C-terminal dimer, as intimated in the "back door" model for substrate release from the Hsp90 chaperone (26), the latter being based on molecular mechanisms proposed for the GHKL family member topoisomerases (61). Recent crystal structures of the human DNA topoisomerase II␣ ATPase domain in different nucleotide-bound states, suggest that the complex process to higher order DNA structure is achieved through the opening and closing of molecular "gates" in the protein, regulated by nucleotidedependent switching (62). There is evidence that novobiocin binds to mature Hsp90-substrate complexes (in which the C-terminal site might be accessible), giving rise to unique changes in Hsp90 conformation that favor substrate release (55). Harris et al. (26) have speculated that on transient separation of the C-terminal domains, the hydrophobic helix 2, which is proposed to have a putative role in substrate binding (Fig. 2), may have sufficient flexibility to function as an intramolecular mimic for helix 5 of the opposite monomer, thus contributing to the synchronous release of substrate. The possibility remains that, by destabilizing the dimer interface, the coumarin-related inhibitors promote the repositioning of helix 2. Structure determination of the Hsp90 C-terminal domain in association with high affinity novobiocin analogues, such as those recently described by Yu et al. (63), is likely to provide further insight into the mechanism of Hsp90 regulation by the coumarin antibiotics.
We have confirmed that the coumarin-based Hsp90 inhibitors, novobiocin and coumermycin A1, both induced a reduction in cellular glucocorticoid receptor protein levels in a dose-dependent manner. Coumermycin A1, the more potent inhibitor, was an order of magnitude more effective, as previously reported for the depletion of a number of oncogenic Hsp90-dependent signaling proteins, including HER-2 and Raf-1 (24). The observed effects on glucocorticoid receptor protein are similar to those seen with geldanamycin (64) and cisplatin (65), Hsp90 inhibiting agents that target the N-terminal nucleotide-binding site (66 -68), and a 40-residue acidic region at the C-terminal end of the chaperone (53, 54), respectively. As already noted, the interaction site for novobiocin lies within the dimerization domain of Hsp90 (24) and is distinct from that of cisplatin. Although novobiocin has the capacity to interfere with nucleotide binding within the N-and C-terminal domains (24,54), cisplatin selectively disrupts only the C-terminal site (54). Yu et al. (63) have recently reported a dramatic depletion of both wild-type and mutant androgen receptors in prostate cancer cell lines following treatment with A4, a novobiocin analogue that potently inhibits Hsp90 function. The development of related second generation com-  HeLa cells were treated with the indicated concentrations of novobiocin (A) or coumermycin A1 (B) for 16 h at 37°C followed by lysate preparation in sample buffer. Total cellular protein (100 g) was fractionated by SDS-PAGE, then probed for GR with H-300 antibody and for ␣-tubulin (loading control) by Western immunoblotting.
pounds with increased inhibitory activity may provide alternative ligand-independent therapies aimed to control signaling pathways linked to estrogen and androgen receptors in breast and prostate cancers.