ARTICLE

Structural Basis for Depletion of Heat Shock 90 Client by Deguelin Seung Hyun Oh , Jong Kyu Woo , Yasemin Dakak Yazici , Jeffrey N . Myers , Woo-Young Kim , Quanri Jin , Soon Sun Hong , Hyun-Ju Park , Young-Ger Suh , Kyu-Won Kim , Waun Ki Hong , Ho-Young Lee

Background The molecular chaperone heat shock protein 90 (Hsp90) participates in preserving the expression and

activity of various oncoproteins, including hypoxia-inducible factor 1 ␣ (HIF-1 ␣ ) and Akt. Deguelin is a Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 rotenoid with antitumor activities. We investigated whether the antitumor activities of deguelin involve the functional inhibition of Hsp90.

Method Human xenograft tumors were generated in mice from H1299 (n = 6 per group) and A549 (n = 4 per group) non– small-cell lung cancer cells, UMSCC38 (n = 5 per group) head and neck cancer cells, MKN45 (n = 5 per group) stomach cancer cells, and PC-3 (n = 3 per group) prostate cancer cells. Tumor-bearing mice were treated with deguelin at 4 or 8 mg/kg or with vehicle (as a control) twice a day by oral gavage for 15– 28 days. Protein expression was assessed by western blot analysis. Akt and Hsp90 were assessed by use of adenoviral vectors expressing constitutively active Akt or Hsp90. Binding of deguelin to Hsp90 was exam- ined by docking analysis and by competition binding experiments with ATP – Sepharose beads. The protea- some inhibitor MG132 was used to investigate deguelin’s effect on the induction of -mediated proteasomal degradation of HIF-1␣ . All statistical tests were two-sided.

Results Deguelin bound to the ATP-binding pocket of Hsp90 and disrupted Hsp90 function, leading to ubiquitin- mediated degradation of HIF-1␣ . Administration of deguelin to xenograft-bearing mice statistically signifi- cantly decreased tumor growth by inducing apoptosis and decreasing the expression of Hsp90 client proteins, without detectable toxic effects. For example, at 15 days after the start of deguelin treatment, the volume of untreated control H1299 xenograft tumors was 798 mm3 and that of xenograft tumors treated with deguelin at 4 mg/kg was 115.9 mm3 (difference = 682.1 mm3 , 95% confidence interval = 480.4 to 883.9 mm 3 ; P <.001).

Conclusions The antitumor activities of deguelin appear to involve its binding to the ATP-binding pocket of Hsp90, which suppresses Hsp90 function.

J Natl Cancer Inst 2007;99: 949 – 61

Targeted molecular therapies have produced encouraging results cells may be dependent on Hsp90 for their growth and survival in the management of some cancers (1 , 2 ). However, when an (5 ). Small-molecule inhibitors —such as radicicol, 17-allyl- essential oncogenic pathway is blocked in cancer cells, those cells amino,17-demethoxygeldanamycin (17-AAG) (or NSC 330507), often survive by activating other overlapping signaling pathways 17-demethoxy,17-(2-dimethylamino)ethylaminogeldanamycin, with similar functions. One approach to counteract this problem CCT018159, PU3, and novobiocin— associate with Hsp90 is to block signaling from the overlapping pathways with small- and modulate its chaperone function, thereby accelerating molecule inhibitors that are specific for each pathway. A promising therapeutic target for such an approach is heat Affiliations of authors: Departments of Thoracic/Head and Neck Medical shock protein 90 (Hsp90), a molecular chaperone that is required Oncology, (SHO, JKW, WYK, QJ, WKH, HYL) and Head and Neck Surgery for the stability and function of mutated, chimeric, and over- (YDY, JNM), The University of Texas M. D. Anderson Cancer Center, Houston, TX; Research Institute of Pharmaceutical Sciences, College of expressed proteins. These proteins include mutated , Bcr– Abl, Pharmacy, Seoul National University, Seoul, South Korea (SSH, YGS, KWK); cyclin-dependent kinase 4 (CDK4), mitogen-activated protein College of Pharmacy, Sungkyunkwan University, Suwon, Korea (HJP). kinase kinase (MEK1/2), Raf1, Akt, HER2, and hypoxia-inducible Correspondence to: Ho-Young Lee, PhD, Unit 432, Department of Thoracic/ factor 1␣ (HIF-1␣ ) from various signaling pathways (3 ). The Head and Neck Medical Oncology, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030 (e-mail: hlee@ mechanism of action of Hsp90 includes the stepwise recruitment mdanderson.org ). cdc37 hop of cochaperones, including Hsp70, p50 , p60 , and p23 (4 ). See “Notes” following “References.” Hsp90 is involved in adaptation of cancer cells to many environ- DOI: 10.1093/jnci/djm007 mental stresses, such as hypoxia and nutrient deprivation (5 ), and © The Author 2007. Published by Oxford University Press. All rights reserved. is commonly overexpressed in cancer cells. Consequently, cancer For Permissions, please e-mail: [email protected]. jnci.oxfordjournals.org JNCI | Articles 949 RPMI-1640 medium supplemented with 10% fetal bovine serum CONTEXT AND CAVEATS and 1% penicillin– streptomycin (GIBCO-BRL Life Technologies, Prior knowledge Gaithersburg, MD). Heat shock protein 90 (Hsp90) is a molecular chaperone that is Adenoviral vectors expressing constitutively active, hemaggluti- required for the stability and function of various normal proteins nin (HA)-tagged Akt (Ad-HA-MyrAkt, in which Myr is the Src and oncoproteins (i.e., its client proteins), including hypoxia- myristoylation signal) (10 ), HA-tagged Hsp90 (Ad-HA-Hsp90) ␣ ␣ inducible factor 1 (HIF-1 ). Deguelin is a rotenoid with antitumor ( 16 ), and an empty adenoviral vector (Ad-EV) were amplifi ed as activity in a variety of tumor types. described previously (10 ). Briefl y, an adenoviral vector expressing Study design a full-length human Akt1 sequence with the Src myristoylation The molecular mechanism of action of deguelin was investigated signal fused in-frame to the c-Akt– coding sequence or expressing in human xenograft mouse model systems for five different tumors a full-length human Hsp90 sequence, each fused with HA, under and by computer modeling the binding of deguelin to the ATP- the control of the cytomegalovirus was constructed by binding pocket of Hsp90. use of the corresponding pAd-shuttle vector system. The viruses Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 Contribution were amplifi ed in human embryonic kidney 293 cells that were The antitumor activities of deguelin appear to involve its binding maintained in Dulbecco’s modifi ed Eagle medium containing 10% to the ATP-binding pocket of Hsp90, which suppresses Hsp90 fetal bovine serum and purifi ed by use of centrifugation through a function. CsCl gradient (17 ). Viral titers were determined by plaque assays. ␣ Implications The expression vector for HIF-1 mutated at two proline hydrox- Deguelin appears to be an effective antitumor agent that targets ylation sites (Pro-402 and Pro-564) was a kind gift from Dr Len Hsp90. Complete and extensive screening for toxic effects in clini- Neckers (National Institutes of Health, Rockville, MD) (18 ). cal trials is required before further development because of the Female nude mice, 6 weeks old, were purchased from Harlan- wide expression of Hsp90 in normal tissues. Sprague Dawley (Indianapolis, IN). Deguelin was purchased from Limitations Gaia Chemical Corp (Gaylordsville, CT). The Hsp90 inhibitor Deguelin’s mechanism of action on various Hsp90 client proteins 17-AAG was from Invivogen (San Diego, CA). Bovine serum albu- has not been fully elucidated. min, gelatin, the inhibitor MG132, CoCl2 , and 3-(4,5- dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide were from Sigma-Aldrich (St Louis, MO). Recombinant Hsp90 protein was degradation of numerous oncogenic client proteins that participate from Stressgen (San Diego, CA). in the regulation of cell proliferation, survival, and angiogenesis (6 ). The following antibodies were used for western blot or immu- Chemically distinct inhibitors of Hsp90 have been used as probes nohistochemical analyses: mouse monoclonal antibodies against to defi ne the biologic functions of Hsp90 at the molecular level HIF-1 ␣ (BD Transduction Laboratories, Lexington, KY), HIF-2␣ and/or to validate Hsp90 as a novel target for anticancer drugs in (Novus, Littleton, CO), p53 (BD Pharmingen, San Diego, CA), or clinical trials ( 7 , 8 ). Phase II clinical trials evaluating one of these HIF-1 ␤ (BD Transduction Laboratories); rabbit polyclonal anti- agents, 17-AAG, have begun, and several second-generation com- bodies against phosphorylated Akt (Ser-473) (Cell Signaling, pounds are now in late preclinical development (7 , 9 ). Beverly, MA), Akt (Cell Signaling), Hsp90␣ (Stressgen), and endo- We have previously shown (10 , 11 ) that the rotenoid deguelin thelial nitric oxide synthase (eNOS) (BD Pharmingen); goat poly- effectively prevents tobacco carcinogen– induced lung carcinogen- clonal antibodies against CDK4, HA, MEK1/2, and ␤ -actin (Santa esis by blocking Akt activation. Deguelin has also shown potential Cruz Biotechnology, Santa Cruz, CA); and a rabbit anti-mouse as a chemopreventive agent against breast, colon, and skin cancers immunoglobulin G (IgG)– horseradish peroxidase conjugate, a ( 12 – 14 ), and it has shown antiproliferative and apoptotic activities goat anti-rabbit IgG– horseradish peroxidase conjugate, and a rab- in some non – small-cell lung cancer (NSCLC) cell lines in vitro bit anti-goat IgG– horseradish peroxidase conjugate (Santa Cruz ( 12 ). However, anticancer activities of deguelin and its mechanism Biotechnology). of action have not been defi ned. In this study, we investigated the therapeutic effi cacy of deguelin in several human xenograft tumor Generation of Xenograft Tumors and models and potential mechanisms of its antitumor activity. Immunohistochemical Staining All animal procedures were performed in accordance with a proto- col approved by the M. D. Anderson Institutional Animal Care and Materials and Methods Usage Committee. Xenograft tumors were generated by subcuta- Cells, Animals, Adenoviral Vectors, Antibodies, and neous injection of H1299, A549 UMSCC38, MKN45, or PC-3 Materials cells, as described elsewhere (19 ). Briefly, nude mice were injected The human NSCLC cell lines H1299, A549, and H322; stomach at a single dorsal flank site with 5 × 107 H1299 (n = 12 mice), A549 cancer cell line MKN45; and prostate cancer cell line PC-3 were (n = 8 mice), UMSCC38 (n = 10 mice), MKN45 (n = 10 mice), or obtained from the American Type Culture Collection (Manassas, PC-3 (n = 6 mice) cells in 100 µL of phosphate-buffered saline VA). The human head and neck squamous cell carcinoma cell line (PBS). Injection of these cells into nude mice induced exponentially UMSCC38, which was established originally by Dr Thomas E. growing tumors. When tumors reached a volume of 40– 80 mm3 Carey (University of Michigan, Ann Arbor), was obtained from (termed day 0 for our experiments), mice were treated orally with Dr Reuben Lotan (15 ). These cancer cell lines were cultured in vehicle (50 µ L of cotton seed oil and 50 µ L of dimethyl sulfoxide)

950 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 or deguelin at 4 mg/kg— which is the maximum tolerable dose of insulin-like growth factor I (IGF-I; 100 ng/mL) or CoCl2 (100 µ M) intragastrically administered deguelin for rats (14 ) — twice a day for under normoxic conditions (20% O2 ) or exposed to hypoxia for 15– 28 days. We also included a deguelin dose of 8 mg/kg for 6 hours and then incubated with deguelin. We treated cells with

H1299 xenografts to obtain the maximum antitumor activity IGF-I or CoCl2 under normoxic conditions or exposed them to against this tumor. Tumor size was measured every 2 –4 days, and hypoxia for 6 hours, which showed the great induction of HIF-1␣ tumor growth was quantified by measuring the tumors in two in most of the cell lines used in our study. We did not test what dimensions. Volumes were calculated by the formula 0.5 × a × b2 , happened after 6 hours. To assess the role of ROS in deguelin- where a and b are the longer and shorter diameters, respectively. mediated decreased expression of HIF-1 ␣ , H1299 cells were Tumor volumes were expressed as the mean and 95% confidence preincubated under hypoxic conditions for 6 hours and then interval (CI). Mice with necrotic tumors or with tumors that had a incubated with 4 mM succinate in the presence or absence of diameter of more than 1.5 cm were killed by use of CO2 , and 100 nM deguelin for an additional 6 hours. tumors were removed. Histopathologic evidence of pulmonary To assess the effects of deguelin on the expression of HIF-1␣ toxicity (i.e., edema or inflammation of the bronchial epithelium and to investigate the roles of Akt, mitogen-activated protein Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 and alveoli), inflammation, and injury in other organs, such as liver, kinase, and Hsp90 in deguelin action on HIF-1␣ stability, H1299 kidney, spleen, stomach, and ovary, were evaluated by a veterinary cells were uninfected or infected with Ad-HA-MyrAkt, Ad-MEK, pathologist. Tumors were separated into three parts for analyses Ad-HA-Hsp90, or Ad-EV (each at 25– 50 plaque-forming units per

immediately after they were removed. One part was fixed in 10% cell) as described previously (15 ); treated with IGF-I or CoCl2

buffered formalin, embedded in paraffin, and sectioned for CDK4 under normoxic conditions or exposed to hypoxia (1% O2 ) for 6 staining. Another part was frozen and sectioned for terminal deoxy- hours; and then incubated with complete medium containing deg- nucleotidyltransferase-mediated UTP end-labeling (TUNEL) uelin for 1 – 6 hours. To test the effect of deguelin on protesome- staining, a measure of apoptosis. Immunohistochemical staining for mediated HIF-1␣ degradation, H1299 cells were infected with CDK4 and TUNEL was as previously described (15 ). For apopto- Ad-EV or Ad-HA-Hsp90 for 2 days and then treated with 100 nM sis analysis, we counted all nuclei in the field, as well as all TUNEL- deguelin in the absence or presence of 10 µ M MG132 (Z-Leu- positive cells in three to five randomly selected fields of view, at a Leu-Leu-al; a proteasome inhibitor) for 6 hours. magnification of ×100. The third part of the tumor was frozen for The effects of deguelin on the degradation of Hsp90 client western blot analysis. proteins were tested by western blot analysis in H1299 cells treated with 100 nM deguelin for 1 or 2 days in the absence or presence of Flow Cytometry Analysis of Reactive Oxygen 10 µ M MG132 in complete medium under hypoxic conditions. Species Production H1299 cells (3 × 105 cells) were preincubated under normoxic (20% Western Blot and Coimmunoprecipitation Analyses

O2 ) or hypoxic (1% O2 ) conditions for 6 hours and then incubated Western blot analysis was performed as described previously (15 ). with 5 mM N -acetylcysteine (a reactive oxygen species [ROS] scav- Briefly, whole-cell lysates were prepared in a lysis buffer containing enger; Sigma-Aldrich), 4 mM succinate (a substrate of mitochon- 20 mM Tris– HCl (pH 7.6), 1 mM EDTA, 140 mM NaCl, 1% drial complex II; Sigma-Aldrich), or 100 nM deguelin for an Nonidet P-40, 1% aprotinin, 1 mM phenylmethylsulfonyl fluoride, additional 6 hours. The relative cellular ROS concentration was and 1 mM sodium vanadate as described elsewhere (16 ). Equivalent obtained by measuring the hydrogen peroxide concentration. amounts of protein (20– 80 µg) were resolved by sodium dodecyl Briefly, cells were treated with 20 µM dichlorofluorescin diacetate sulfate– polyacrylamide gel electrophoresis (SDS– PAGE) in 7.5% (Molecular Probe; Invitrogen Corporation, Carlsbad, CA) for to 12% gels and transferred by electroblotting to a polyvinylidene 30 minutes at 37 °C, washed with PBS, trypsinized, and collected fluoride membrane. Nonspecific binding to the blot was blocked in 500 µ L of PBS. The hydrogen peroxide content in 2000 cells per by incubation in Tris-buffered saline containing 0.1% Tween-20 sample was analyzed by flow cytometry (FACS Calibur; Becton (TBST) supplemented with 5% nonfat powdered milk. The blot Dickinson, San Jose, CA). was incubated with primary antibody against HIF-1␣ , HIF-1␤ , phosphorylated Akt (Ser-473), Akt, CDK4, HA, MEK1/2, ␤ -actin, Cell Treatments or Hsp90␣ (each at an appropriate dilution) in TBST containing To assess the effects of deguelin on protein and mRNA expression, 5% bovine serum albumin at 4 °C for 16 hours. The membrane H1299 cells were seeded in six-well plates, at 4 × 10 5 to 5 × 10 5 was then washed three times with TBST and incubated with the cells per well, 1 day before the start of treatment. When the cul- appropriate horseradish peroxidase– conjugated secondary antibody ture was 80%– 90% confluent, cells were treated with deguelin for 2 hours at room temperature. The protein– antibody complexes (0.1– 100.0 nM) or 17-AAG (10 µM) in complete medium for 16 were detected by using enhanced chemiluminescence (Amersham, hours under hypoxic or normoxic conditions, and then prolyl Arlington Heights, IL), according to the manufacturer’s recom- hydroxylase (PHD)– or von Hippel– Lindau protein (pVHL) – mended protocol. independent HIF-1␣ regulation by deguelin was investigated. For coimmunoprecititation assays, H1299 cells were uninfected H1299 cells (2 × 105 cells per well in six-well plates) that were or infected with the indicated adenoviral vectors, treated with transfected with control or VHL small-interfering RNA (siRNA; IGF-I or CoCl2 (100 µ M) under normoxic conditions to induce 60 nmol) or with an expression vector for HIF-1␣ mutated at two HIF-1 ␣ expression or incubated under hypoxic conditions, and proline hydroxylation sites (Pro-402 and Pro-564; 1 µ g ), which then untreated or treated with deguelin (100 nM) as described exhibits diminished hydroxylation by PHD, were treated with above. Cells were lysed in a buffer containing 50 mM Tris – HCl jnci.oxfordjournals.org JNCI | Articles 951 (pH 7.5), 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, Sonics & Materials) twice for 15 seconds, followed by centrifugation 1 mM phenylmethylsulfonyl fl uoride, and a protease inhibitor (15 000g , 15 minutes, at 4 °C). To investigate the interaction mixture by pulse sonifi cation of 2 watts for 5 – 10 seconds (Vibra between Hsp90 and HIF-1 ␣ , 200 µg of H1299 cellular protein was Sonic sonicator, Sonics & Materials, Danbury, CT). The level of added to recombinant His-HIF-1␣ bound with Ni-NTA –agarose HIF-1 ␣ ubiquitination in deguelin-treated H1299 cells was deter- beads (Qiagen) (5 mg/mL of resin in a total volume of 0.5 mL) and mined by immunoprecipitation with an anti– HIF-1 ␣ antibody and deguelin (100 nM) in a binding buffer containing 25 mM Tris – HCl western blot analysis with an anti-ubiquitin antibody. To investi- (pH 7.4), 150 mM NaCl, and 0.2% Triton X-100; and the mixture gate the interaction between HIF-1 ␣ and Hsp90, 200 µ g of H1299 was incubated for 1 hour at 4 °C. Beads were then collected and cellular protein was immunoprecipitated with anti– HIF-1 ␣ washed three times with the same buffer, 30 µ L of 5× SDS – PAGE antibodies overnight at 4 °C in a binding buffer containing 25 mM sample buffer was added, and the mixture was boiled at 95 °C for Tris– HCl (pH 7.4), 150 mM NaCl, and 5 mM EDTA. Thereafter, 10 minutes to release bound proteins. These proteins were resolved protein-G agrose beads were added and incubated for 1 hour. The by SDS– PAGE and subjected to western blotting for Hsp90␣ as beads were collected and washed three times with the same buffer, described above. Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 30 µ L of 2× SDS– PAGE sample buffer was added, and the mixture Competition between deguelin and Hsp90 recombinant protein was boiled at 95 °C for 10 minutes to release bound proteins. or between deguelin and JNK1 recombinant protein (as a negative Bound proteins were resolved by SDS– PAGE gel and then sub- control) for binding to ATP– Sepharose was examined as described jected to western blot analysis with an anti-Hsp90 antibody. previously ( 20). Briefl y, 100 ng of each recombinant protein was incubated with vehicle (0.1% dimethyl sulfoxide), 10 mM ATP, Reverse Transcription – Polymerase Chain Reaction deguelin (at 1, 5, and 10 µ M), or 17-AAG (at 1, 5, and 10 µ M) in 200 Total RNA was prepared from H1299 cells, and cDNA was synthe- µL of TNE buffer (20 mM Tris– HCl at pH 7.4, 150 mM NaCl, 1 sized from the total RNA, as described previously (15 ). Primer mM EDTA, 10% glycerol, and 1 mM dithiothreitol) for 2 hours, sequences used for reverse transcription– polymerase chain reaction and then 50 µ L of ␥ -phosphate – linked ATP – agarose (Fluka, assays were as follows: for HIF-1␣ , 5 ′ -CTCAAAGTCGGACAG Milwaukee, WI) was added. Samples were incubated for another 2 CCTCA-3′ (sense) and 5 ′ -CCCTGCAGTAGGTTTCTGCT-3′ hours at 4 °C and washed three times with ice-cold TNE buffer. (antisense); for VHL, 5′ -′ GAGATGCAGGGACACACGAT-3′ Material bound to the agarose beads was eluted in 5× SDS– PAGE (sense) and 5 ′ -TCACAACTGTAGAATGTCAATCC-3 ′ (anti- sample buffer and the concentrations of JNK1 and Hsp90 proteins sense); and for ␤ -actin, 5′ -ACTACCTCATGAAGATC-3′ (sense) bound to ATP– Sepharose were determined by western blot analysis and 5′ -GATCCACATCTGCTGGAA-3′ (antisense). The thermo- with antibodies against JNK1 or Hsp90, respectively. Hsp90 protein cycler conditions used for amplification were 94 °C for 5 minutes bound to the beads was quantifi ed by densitometry analysis. and then 28 –32 cycles of 94 °C for 45 seconds, 54 –60 °C for 45 seconds, and 72 °C for 1 minute. Docking Modeling Flexible docking, a method to dock the conformational ensemble Expression and Purification of Epitope-Tagged Proteins (i.e., collection ) of a ligand into a target receptor, of deguelin onto Glutathoine S -transferase– tagged JNK1 proteins and hexahistidine Hsp90 was performed by use of the SYBYL version 7.0 software (His)-tagged HIF-1␣ proteins were expressed in Escherichia coli program (SYBYL molecular modeling software, Tripos, Inc, version strain BL21 (DE3) pLysS (Novagen, Madison, WI) and cultured in 7.0, St Louis, MO) and the Red Hat Linux operating system, version Luria-Bertani (LB) medium as previously described (19 ). Cells were 7.0. The chemical structures of deguelin, 17-AAG, and geldanamy- lysed with a lysis buffer containing 20 mM sodium phosphate (pH cin were prepared in MOL2 format by use of the sketcher module 7.8), 500 mM NaCl, lysozyme (1 mg/mL), 0.1 M phenylmethylsul- in SYBYL, and partial atomic charges were calculated by the fonyl fluoride, leupepstatin (0.5 mg/mL), pepstatin (1 mg/mL), and Gasteiger– Huckel method (21 ) and assigned to the ligand atoms. To apropronin (1 mg/mL) by sonification. After centrifugation (15 000 g optimize the ligand structures, the conjugate gradient energy mini- for 5 minutes at 4 °C), the lysate was loaded on columns of glutathi- mization (22 ) of the ligand was run until the value converged to a one– Sepharose beads (Pharmacia-LKB Biotechnology, Uppsala, maximum derivative of 0.001 kcal/mol·Å, and the final coordinates Sweden) or of nickel-nitrilotriacetic acid (Ni-NTA)– agarose beads were stored in a database. The x-ray crystal structure (23 ) of complex (Qiagen, Valencia, CA). Glutathoine S -transferase– tagged JNK1 between human Hsp90 and geldanamycin ( entry and His-tagged HIF-1␣ proteins were eluted from the matrix by the = 1YET) was used as a target for the docking of flexible ligand by addition of 20 mM glutathione or 500 µ M imidazole, respectively. FlexX algorithm in SYBYL. All crystallographic water molecules Eluted proteins were dialyzed against 1× PBS overnight at 4 °C. were removed from target structure except for one that was involved in the hydrogen bond network inside the binding pocket. The active Binding of Deguelin to Heat Shock Protein 90 site was defined as all the amino acid residues enclosed within 6.5-Å We investigated whether deguelin could disrupt the interaction radius sphere centered by the bound geldanamycin. The main set- between HIF-1␣ and Hsp90 by comparing the amounts of HIF-1␣ – tings for FlexX run were set in 1000 solutions to generate the maxi- mutant recombinant protein (1 = amino acid residues 1– 603; 2 = mum number of conformers for each compound during FlexX amino acid residues 401– 827; 3 = amino acid residues 604– 827) docking. For database ranking, the scores for all FlexX solutions bound to Hsp90 in the presence or absence of deguelin. Briefly, were calculated by use of a consensus scoring method (CScore) that H1299 cells were scraped off, lysed in 150 µ L of buffer A, and soni- was based on the following five scoring functions: FlexX score, cated with a standard probe-type sonicator (Vibra Sonic sonicator, G_score, D_score, ChemScore, and potential of mean force score

952 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 A H1299 B UMSCC38 1200 1200 con con

) 4mg/kg ) 3 1000 3 1000 4mg/kg 8mg/kg 800 800

600 600

400 400 P < .001 P Fig. 1 . Deguelin and tumor growth. Mice bear- Tumor volume (mm 200 Tumor volume (mm 200 = .037 ing a xenograft tumor (one tumor per mouse) P <.001 were treated with vehicle (con) or deguelin at 4 0 0 or 8 mg/kg, as indicated. Day 0 represents the 01357911 1315 03 7 10 14 17 21 24 28 fi rst day of deguelin treatment. Tumor volume days days was measured every 2– 4 days for 15 or 28 days. A ) H1299 non– small-cell lung cancer cells (n = 6 Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 mice per group). B ) UMSCC head and neck can- C MKN-45 D PC-3 1200 1200 cer cells (n = 5 mice per group). C ) MKN45 stom- con con ach cancer cells (n = 5 mice per group). D) PC-3 ) )

3 1000 4mg/kg 3 1000 4mg/kg prostate cells (n = 3 mice per group). Data are the mean and its 95% confi dence interval. 800 800

600 600

400 400

P = .010 Tumor volume (mm 200 Tumor volume (mm 200 P = .044 0 0 03 7 10 14 17 21 24 28 01 35 8 10 12 15 days days

( 24 ). After visual inspection, final docking models were created by MKN45 stomach (Fig. 1, C), and PC-3 prostate (Fig. 1, D ) xeno- selecting the conformers with the highest consensus scores (Cscore = graft tumors in athymic nude mice (six mice with H1299 tumors, 5) and then docking these conformers into the active site of Hsp90. five with UMSCC38 tumors, five with MKN45 tumors, and three with PC-3 tumors). Mice bearing xenograft tumors were treated Generation of the Molecular Surface Map with deguelin at 4 or 8 mg/kg or with vehicle (as a control) twice a To visualize the hydrophobic interactions between the ligand (17- day by oral gavage. Tumor growth was statistically significantly AAG and deguelin) and Hsp90, surface hydrophobicity (lipophilicity) suppressed by deguelin treatment throughout the 15- or 28-day potential physicochemical property maps of the ligand-binding sites treatment regimen, indicating the potent therapeutic efficacy of of Hsp90 were generated on the solvent-accessible surface of Hsp90 deguelin in various human cancers (Fig. 1 and Table 1). For exam- (25 ) by use of the MOLCAD module in SYBYL. Surface colors ple, at 15 days after the start of treatment, the volume of control ranged from dark blue (hydrophilic) to dark brown (lipophilic). H1299 xenografts was 798 mm 3 and that of xenografts treated with deguelin at 4 mg/kg was 115.9 mm3 (difference = 682.1 mm3 , 95% Statistical Analysis CI = 480.4 to 883.9 mm3 ; P <.001). Data are expressed as the mean and 95% confidence interval from at least triplicate samples. Data were calculated with the Microsoft Excel software program (version 5.0; Microsoft Corporation, Table 1. Xenograft tumor volume at 15 days after the start of treatment in mice treated with vehicle (control) or deguelin Redmond, WA). The statistical significance of differences between groups was analyzed with Student’s t test after we determined the Mean tumor volume, mm3 normal distribution of the data. Shapiro-Wilk’s test was applied for Cell line Treatment (95% confidence interval) P value * testing the normality of the data in each experimental group. The H1299 Control 798 (607.8 to 988.3) test of normality was not rejected at 5% level. All statistical tests Deguelin (4 mg/kg) 115.9 (48.6 to 183.1) <.001 Deguelin (8 mg/kg) 190.9 (139.7 to 242.2) <.001 were two-sided. A P value of less than .05 was considered to be PC-3 Control 854.3 (781.7 to 926.9) statistically significant. Deguelin (4 mg/kg) 533.0 (369.5 to 696.5) .044 UMSCC38 Control 746.9 (533.9 to 966.0) Deguelin (4 mg/kg) 379.0 (295.1 to 462.8) .037 Results MKN45 Control 993.3 (783.3 to 1203.2) Deguelin and Tumor Growth Deguelin (4 mg/kg) 492.4 (326.0 to 658.9) .010

We evaluated the effect of deguelin treatment on the growth of * The statistical significance of differences between groups was analyzed with H1299 NSCLC (Fig. 1, A), UMSCC38 head and neck (Fig. 1, B), Student’s t test (two-sided). jnci.oxfordjournals.org JNCI | Articles 953 Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021

Fig. 2 . Deguelin and reactive oxygen species (ROS) –, prolyl hydroxy - of HIF-1␣ expression in H1299 cells untreated or treated with deguelin, lase (PHD)– , von Hippel– Lindau protein (pVHL)– , and phosphatidylino- as indicated, and/or succinate for 6 hours. F ) PHD-independent HIF-1 ␣ sitol 3-kinase-Akt– independent regulation of hypoxia-inducible factor regulation by deguelin. Western blot analysis of HIF-1␣ expression was 1 ␣ (HIF-1␣ ) protein expression. H1299 cells were untreated or treated performed in H1299 cells transfected with mutant HIF-1 ␣ (Pro-402 and with 100 nM deguelin for 16 hours and then also treated, as indicated, Pro-564), which exhibits diminished hydroxylation by PHD, and then with insulin-like growth factor I (IGF-I) (50 ng/mL) or CoCl 2 under nor- incubated with deguelin under normoxic condition. G ) VHL-independent ␤ ␣ moxic (20% O2 ) or with IGF under hypoxic (1% O2 ) conditions. -Actin HIF-1 regulation by deguelin. H1299 cells were transfected with mRNA and protein were used as loading controls. A ) Western blot anal- control (Con) or VHL small-interfering RNA (siRNA) and treated with ysis of HIF-1 ␣ and HIF-1 ␤ protein. B) Reverse transcription – polymerase deguelin under normoxic or hypoxic conditions, as indicated. RT – PCR chain reaction (RT– PCR) analysis of vascular endothelial growth factor assay of VHL and ␤ -actin mRNA (as a control) and immunoblot analysis (VEGF) mRNA. C) RT – PCR analysis of aldolase A mRNA. D and E ) of HIF-1 ␣ or ␤ -actin protein (as a control) were performed. H ) PI3K-Akt Determination of ROS. Normoxic or hypoxic H1299 cells were untreated, pathway – independent HIF-1 ␣ regulation by deguelin. H1299 cells were treated with deguelin, treated with 4 mM succinate (a ROS activator), or uninfected ( upper ) or infected ( lower ) with an adenoviral vector treated with 5 mM N -acetylcysteine (a ROS scavanger). D ) Generation expressing hemagglutinin (HA)-tagged constitutively active Akt (Ad- of ROS as determined by fl ow cytometry by measuring cellular hydro- HA-MyrAkt) and then treated with 100 nM deguelin for 6 or 16 hours. gen peroxide level. Data are average values of relative fl uorescence of Western blot analysis of HIF-1 ␣ , phosphorylated Akt (pAkt), Akt, and ␤ -

the peak from one representative experiment of two experiments; actin expression in normoxic IGF- or CoCl2 -stimulated or hypoxic cells results from the other experiment were similar. E) Western blot analysis was conducted.

Deguelin and Expression of Proteins Downstream of We then investigated whether deguelin regulates HIF-1␣ Hypoxia-Inducible Factor 1  expression transcriptionally or posttranscriptionally in H1299 To explore the antitumor mechanisms of deguelin, we first investi- cells. After treatment with 100 nM deguelin for 6 hours, HIF-1␣ gated its effect on proteins whose transcription is stimulated by mRNA expression changed only marginally (data not shown), HIF-1␣ , such as vascular endothelial growth factor (VEGF) and whereas the expression of HIF-1␣ protein was effectively sup- aldolase, which are used by cancer cells to adapt to environmental pressed (Fig. 2, A). Thus, deguelin appears to regulate the expres- stress ( 26). Under normoxic conditions, HIF-1 ␣ protein was rap- sion of HIF-1␣ protein posttranscriptionally, and so we focused idly degraded, and so its expression was very weak (27 ). To stabilize our attention on whether deguelin treatment changed the stability ␣ ␣ HIF-1 protein in cells, we treated cells with IGF-I or CoCl2 under of HIF-1 protein.

normoxic conditions (20% O2 ) or incubated cells under hypoxic Hypoxia-induced ROS may be involved in the stability of HIF- ␣ conditions (1% O2 ). Treatment of H1299 cells with 0.1– 100 nM 1 , and because complex I inhibitors, such as rotenone and diphe- deguelin for 16 hours under all conditions reduced the expression nylene iodonium, block the electron fl ow of the mitochondrial of HIF-1␣ protein (Fig. 2, A ) and VEGF mRNA (Fig. 2, B ) in a respiratory chain and thus ablate the production of hypoxia- dose-dependent manner without affecting the expression of HIF- induced ROS (26 , 28 – 32 ). We, therefore, investigated whether 1␤ protein, which is expressed constitutively (Fig. 2, A). Deguelin ROS are involved in deguelin modulation of HIF-1␣ expression by also inhibited the expression of aldolase A, another HIF-1␣ – comparing the effect of deguelin on HIF-1␣ expression with its regulated (Fig. 2, C). effect on the generation of hydrogen peroxide. Hydrogen peroxide

954 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 transmits an oxygen-sensitive signal that regulates HIF-1␣ cultured with 100 nM deguelin and infected with Hsp90-expressing ( 29 , 32 , 33 ). H1299 cells that were grown under hypoxic conditions adenovirus (Ad-HA-Hsp90) (41 ). The total Hsp90 protein levels produced substantially more hydrogen peroxide than H1299 cells (endogenous intracellular and HA-tagged virus–induced), as deter- that were grown under normoxic conditions. Hydrogen peroxide mined by western blot analysis of Hsp90 and HA, showed no production was further increased by treatment with 4 mM succi- change after deguelin treatment. Moreover, overexpression of nate (a mitochondrial complex II substrate for an indirect route of Hsp90 resulted in a rapid decrease in HIF-1␣ ubiquitination after electron fl ow to complex III) and was decreased by treatment with a 1 hour of treatment with 100 nM deguelin and 10 µM MG132 5 mM N -acetylcysteine (an ROS scavenger) (Fig. 2, D). Treatment (a proteasome inhibitor) (Fig, 3, B), indicating that Hsp90 is with 100 nM deguelin for 6 hours resulted in a slight increase in involved in the deguelin-mediated HIF-1␣ ubiquitination and hydrogen peroxide production under normoxic conditions but not destabilization. under hypoxic conditions. Moreover, treatment with succinate We next investigate the involvement of Hsp90 in deguelin- failed to restore HIF-1␣ protein expression under hypoxic mediated regulation of HIF-1␣ expression. Because treatment conditions in the presence of 100 nM deguelin (Fig. 2, E). Thus, a with deguelin for up to 16 hours did not change the total level of Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 deguelin-mediated ROS-independent mechanism appears to regu- Hsp90 protein in H1299 cells, unlike the deguelin-mediated late HIF-1␣ expression. destabilization of HIF-1␣ (Fig. 3, C ), we investigated whether HIF-1 ␣ protein expression is regulated by many oxygen- HIF-1␣ interacted with Hsp90 in H1299 cells by use of coimmu- dependent and -independent mechanisms (27 , 34 ). Under nor- noprecipitation assays and whether deguelin treatment modifi ed moxic conditions, two proline sites (Pro-402 and Pro-564) within the interaction. We found that Hsp90 was co-precipitated with the oxygen-dependent degradation (ODD) domain of HIF-1␣ are anti-HIF-1␣ antibodies. The interaction between Hsp90 and subjected to PHD-mediated hydroxylation, which triggers binding HIF-1␣ was abolished in cells that had been treated with 100 nM of the pVHL and ubiquitin-mediated protein degradation (34 – 36 ). deguelin for 1 hour, but HIF-1␣ protein expression was not affected For these reasons, we investigated whether PHDs and pVHL are (Fig. 3, D). involved in the effect of deguelin on HIF-1 ␣ regulation. For these We further confi rmed the effects of deguelin on the interac- experiments, H1299 cells were transiently transfected with an tion between HIF-1␣ and Hsp90 by use of various HIF-1␣ expression vector containing HIF-1␣ mutated at two proline mutant proteins. Because the helix– loop – helix and Per– Arnt – Sim hydroxylation sites (Pro-402 and Pro-564). We found that the (HLH– PAS) domain of HIF-1 ␣ is essential for complex forma- mutated HIF-1␣ protein expression was also decreased by the tion with Hsp90 and sensitivity to Hsp90 inhibitors (38 , 40 , 42 ), we deguelin treatment (Fig. 2, F). Moreover, treatment with deguelin examined the interaction between Hsp90 and various recombi- decreased normoxic IGF-induced or hypoxia-induced HIF-1␣ nant proteins containing the HLH– PAS – ODD – N- transactivation protein expression in H1299 cells transfected with a pVHL- domain (N-TAD) domain, the ODD– N-TAD – C-terminal trans- specifi c siRNA to inhibit pVHL expression (Fig. 2, G). Thus, activation (C-TAD) domain, or C-TAD domains of HIF-1␣ , PHDs and pVHL are not required for the deguelin-mediated instead of intact HIF-1␣ . Consistent with previous fi ndings decrease in HIF-1␣ expression. We then explored whether the (40 , 43 ), the HLH– PAS – ODD – N-TAD domain of HIF-1␣ inter- phosphatidylinositol 3-kinase (PI3K) –Akt signaling pathway, acted strongly with Hsp90, and this interaction was inhibited by which induces cap-dependent mRNA translation and stabilization the addition of 100 nM deguelin to the H1299 cells (Fig. 3, E). In of HIF-1 ␣ protein ( 26 , 37 – 39 ), was involved in deguelin-mediated contrast, recombinant ODD –N-TAD – C-TAD and C-TAD HIF-1 ␣ decrease. Our fi ndings were consistent with previous fi nd- domains bound weakly to Hsp90, and this binding was not ings (10 ); i.e., treatment with 100 nM deguelin for 16 hours affected by deguelin. Thus, deguelin appears to interfere with the decreased the expression of phosphorylated Akt in normoxic IGF- function of Hsp90 by inhibiting the interaction between Hsp90 ␣ I – stimulated or normoxic CoCl2 -stimulated or hypoxic H1299 cells and HIF-1 . ( Fig. 2, H, left). However, after a 6-hour treatment with 100 nM deguelin, when HIF-1␣ protein was almost completely depleted, Molecular Modeling of Interaction Between Deguelin and the level of phosphorylated Akt was similar to that in untreated Heat Shock Protein 90 cells (Fig. 2, H, right). Moreover, when constitutively active Akt We investigated whether deguelin could bind to the ATP-binding (expressed from Ad-HA-MyrAkt) was overexpressed (10 ), deguelin pocket of Hsp90 and inhibit the association between HIF-1␣ and had no effect on the level of HIF-1 ␣ ( Fig. 2, H, lower). Thus, the Hsp90. By using the FlexX computer program, docking analysis of PHD, pVHL, and PI3K– Akt pathways do not appear to be deguelin was conducted to determine whether deguelin could bind involved in the effect of deguelin on HIF-1␣ expression. to the ATP-binding pocket of Hsp90. We used the structures of geldanamycin and 17-AAG, Hsp90 inhibitors that bind to the Heat Shock Protein 90 and Deguelin-Mediated Regulation ATP-binding pocket of Hsp90 (44 ), as model structures. As shown of Hypoxia-Inducible Factor 1 Stability in Fig. 4, A, the conformation of deguelin with the highest binding Hsp90 physically interacts with HIF-1␣ and protects it from score among the FlexX docking results is a good match for the oxygen-independent (i.e., pVHL independent) degradation by an benzoquinone moiety and polyketide backbone of the geldanamy- unidentified proteasome pathway (15 , 18 , 38 , 40 ). We next investi- cin and 17-AAG structures (23 ). The cis -fused ring system of deg- gated the role of Hsp90 in deguelin-mediated destabilization of uelin conferred a C-shaped conformation that is similar to the HIF-1␣ . As shown in Fig. 3, A, the levels of HIF-1␣ were restored overall shape of the Hsp90-bound geldanamycin or 17-AAG. The

in normoxic IGF- or CoCl2 -stimulated and hypoxic H1299 cells docked conformations of deguelin and 17-AAG with the highest jnci.oxfordjournals.org JNCI | Articles 955 Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021

Fig. 3 . Deguelin and the interaction between heat shock protein 90 D ) Deguelin and the interaction between Hsp90 and HIF-1 ␣ protein. (Hsp90) and hypoxia-inducible factor 1␣ (HIF-1 ␣ ) under expression in Whole-cell extracts from H1299 cells that were treated with 100 nM ␣ normoxic insulin-like growth factor I (IGF)- or CoCl 2 -stimulated or deguelin for 1 hour were immunoprecipitated with an anti – HIF-1 anti- hypoxic conditions. A ) Deguelin and Hsp90–mediated HIF-1␣ protein body, and then proteins in the immunoprecipitates were examined by ␣ stability. Normoxic IGF- or CoCl2 -treated or hypoxic H1299 cells were western blot analysis with antibodies against Hsp90 or HIF-1 . Whole- infected with adenoviruses Ad-EV (control) or Ad-HA-Hsp90 expressing cell extracts from IGF-treated cells under normoxic conditions were HA-tagged Hsp90 and then untreated or treated with 100 nM deguelin immunoprecipitated with preimmune rabbit serum (ps) and included as for 6 hours. Western blot analysis of HIF-1␣ , Hsp90, HA, and ␤ -actin was negative controls. Western blot analysis for Hsp90 was performed with conducted. B ) Deguelin and ubiquitination of HIF-1 ␣ . H1299 cells were antibodies against Hsp90. E ) Inhibition of the interaction between HIF- infected with an empty adenoviral vector (Ad-EV) or Ad-HA-Hsp90 and 1 ␣ and Hsp90. Upper ) Hsp90 binding to HIF-1 ␣ . Hsp90 from H1299 then treated with 100 nM deguelin and/or 10 µ M MG132 for 1 hour, as whole-cell extracts was incubated with HIF-1␣ – Sepharose beads in the indicated. The level of HIF-1␣ ubiquitination in deguelin-treated H1299 absence (– ) or presence (+) of 100 nM deguelin and then western blot cells was determined by immunoprecipitation (IP) with an anti-HIF-1␣ analysis was conducted with antibodies against Hsp90. Lower ) antibody and western blot analysis (WB) with antibodies against ubiq- Schematic representation of the various HIF-1␣ expression constructs. uitin or HIF-1␣ . Ub = ubiquitin. C ) Deguelin and Hsp90 protein expres- FL = full-length HIF-1␣ (amino acid residues 1– 827); 1 = amino acid resi-

sion. Normoxic IGF- and CoCl2 -stimulated or hypoxic H1299 cells dues 1– 603; 2 = amino acid residues 401– 827; 3 = amino acid residues were untreated or treated with deguelin, as indicated, for 16 hours. 604 – 827; bHLH = basic helix– loop – helix (bHLH) domain; N-TAD and The expression of Hsp90 was determined by western blot analysis. C-TAD = N- and C-terminal transactivation domains. consensus binding scores showed that 17-AAG bound to the ATP- deguelin shows that deguelin is able to bind to the Hsp90 pocket binding pocket of Hsp90 through various hydrophobic interactions (Fig. 4, F). (Fig. 4, B). In addition, we predicted that the carbonyl oxygen atom of deguelin formed hydrogen bonds both with the backbone amide Deguelin Binding to Heat Shock Protein 90 and Inhibition of Phe-138 and side-chain amide of Asn-51 (Fig. 4, C) and that the of the Chaperone Function of Heat Shock Protein 90 oxygen atom of the methoxy group also formed a hydrogen bond To further investigate the mechanism through which deguelin with the side-chain amide of Asn-51. In contrast, 17-AAG formed influences Hsp90 function, we examined the effects of deguelin on many hydrogen bonds involving the water-bridged hydrogen bond ATP binding to Hsp90 in vitro by use of ATP – Sepharose. The network between the carbamate group and surrounding polar binding of Hsp90 to ATP– Sepharose was more strongly inhibited amino acid residues, similar to the hydrogen bonds formed between by deguelin (difference = 95.65%, 95% CI = 93.58% to 97.71%; geldanamycin and the ATP-binding pocket of Hsp90 ( 23). Deguelin P <.001) than by 17-AAG (difference = 80.67%, 95% CI = 67.5% occupied the binding pocket of Hsp90 mainly through hydropho- to 93.8%; P <.001) (Fig. 5, A and B ). To confirm that the effects of bic interactions, with its dimethoxy benzopyran moiety contacting deguelin on Hsp90 were specific to ATP binding, we used the the interior hydrophobic bottom of Hsp90. The complementarity same assay to test the effects of deguelin on ATP binding to JNK1, of ligands with the ATP-binding pocket of Hsp90 demonstrated which has an ATP-binding pocket that is structurally different that the complex between Hsp90 and deguelin correlated highly from that of Hsp90, as a negative control. Deguelin did not alter with that between Hsp90 and 17-AAG (Fig. 4, D and E). The 17- the binding of JNK1 to ATP– Sepharose, whereas soluble ATP AAG fitted into the entire cavity of Hsp90 pocket, and the carba- completely inhibited the binding of JNK1 to ATP– Sepharose mate moiety contacted a trapped water molecule that was involved ( Fig. 5, B). In addition, consistent with the finding that ATP bind- in the hydrogen-bond network in the structure of the geldanamycin ing to Hsp90 was inhibited better by deguelin than by 17-AAG, complex. The geometry of the final complex between Hsp90 and treatment of H1299 cells with 100 nM deguelin for 16 hours

956 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 Fig. 4 . Structural analysis of deguelin and heat shock protein 90 (Hsp90). Geldanamycin and 17- allylamino,17-demethoxygeldanamycin (17-AAG), which bind to the ATP-binding pocket of Hsp90 were used as model structures. A ) Docked conformation of deguelin (yellow carbon) with the highest binding score in comparison with that of 17-AAG ( magenta carbon) and geldanamycin (white carbon). The col- ors of the other atoms are red for oxygen and blue for nitrogen. The hydrogen atoms are not displayed for the sake of clarity. B and C ) Structures of 17-AAG and deguelin bound to the ATP-binding pocket of Hsp90. Amino acid residues that form hydrogen bonds within the binding site are represented in line form, and the bound 17-AAG (B ) or deguelin (C ) are shown by capped stick models. The yellow dotted lines are hydrogen bond interactions (<2.5 Å). D and E ) Space- fi lling model of 17-AAG ( D ) and deguelin ( E ) in the Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 ATP-binding pocket of Hsp90. The binding pocket is a molecular surface (hydrophobicity map) of the sur- rounding amino acid residues. A trapped water mol- ecule in the x-ray structure of the geldanamycin complex is shown as a magenta sphere . F ) Orthogonal view of the complex between deguelin and Hsp90, as obtained by docking simulations. decreased HIF-1␣ expression more than treatment with 5 µ M 17- Discussion AAG (Fig. 5, C). We have shown in this study that deguelin binds directly to Hsp90 We then investigated whether blocking Hsp90 function would and inhibits its chaperone activity, resulting in ubiquitin-mediated lead to the proteasome-mediated degradation of Hsp90 client pro- degradation of Hsp90 client proteins. We have also shown that teins. As shown in Fig. 6, A, after treatment of A459, H1299, and deguelin has antitumor activity against xenografts of many human H322 cells with 0, 10, or 100 nM deguelin for 0 – 2 days, the expres- cancers, including lung, prostate, head and neck, and stomach sion of many Hsp90 client proteins was reduced— including that of cancers. These results are important because intense research CDK4, Akt, eNOS, MEK1/2, and mutant p53 proteins (in H322 efforts have been directed at developing more effective targeted cells)— in a time- and dose-dependent manner, with no detectable anticancer therapies. Hsp90 is a molecular chaperone that has many effect on the expression of Hsp90 and ␤ -actin. We found that pro- client proteins that play critical roles in cancer progression, tumor longed treatment with deguelin also decreased the expression of angiogenesis, and therapy resistance. An interaction between HIF-2 ␣ , another HIF- ␣ subunit and a Hsp90 client protein (45 ). Hsp90 and its client proteins is required for the stability and func- However, HIF-2␣ was less sensitive than HIF-1␣ to the deguelin tion of these client proteins. Consequently, Hsp90 is a potentially treatment because treatment with more than 100 nM deguelin for promising target for the development of new cancer treatments, longer than 48 hours was required to suppress HIF-2␣ expression and several small-molecule inhibitors of Hsp90 have shown prom- in H1299 and H322 cells (Fig. 6, A). Treatment with the protea- ising antitumor activity in preclinical and/or clinical systems (8 , 46 ). some inhibitor MG132 prevented the deguelin-mediated decreased In this article, we also present evidence that 1) deguelin reduces expression of CDK4 and Akt in H1299 cells (Fig. 6, B ). Thus, the ability of deguelin to disrupt Hsp90 function appears to have the expression of a subset of Hsp90 client proteins, including the ␣ altered the expression of many signaling components that have HIF-1 protein, under normoxic IGF-stimulated and hypoxic ␣ roles in cancer cell proliferation, survival, and angiogenic conditions; 2) deguelin-induced degradation of HIF-1 protein is activities. independent of ROS, PHD, pVHL, and PI3K– Akt pathways under normoxic and hypoxic conditions; and 3) deguelin binds directly to Deguelin Treatment, Expression of Client Proteins for the ATP-binding pocket of Hsp90 and thus interferes with the Heat Shock Protein 90, and Apoptosis In Vivo chaperone function of Hsp90. Thus, deguelin appears to be a novel We further investigated the in vivo effect of deguelin treatment on Hsp90 inhibitor with potential chemotherapeutic activity in many expression of Hsp90 client protein and apoptosis in H1299, A549, cancers. UMSCC38, and MKN45 xenograft tumors. In immunohisto- HIF-1 ␣ plays an important role in tumor growth and angio- chemical studies, we found that the expression of CDK4 in the genesis by regulating the transcription of several in response tumors from deguelin-treated mice was lower than that in control to hypoxic stress and changes in growth factor expression (26 ). tumors from untreated mice (Fig. 7, A). Conversely, tumors from Because we found that deguelin treatment affected the expression deguelin-treated mice showed higher TUNEL staining than of HIF-1␣ , we investigated whether HIF-1␣ is involved in antitu- tumors from untreated mice (for example, for control H1299 mor activities of deguelin. Our observations showed that deguelin tumors, mean = 1.69%, 95% CI = 0.65% to 2.72%; and for treatment decreases the half-life of HIF-1␣ . Because ROS produc- deguelin-treated H1299 tumors, mean = 8.84%, 95% CI = 4.73% to tion may be involved in the HIF-1 ␣ stability and because rotenone 12. 95%; P = .005) (Fig. 7, B). Thus, the cells with altered expression (a rotenoid) inhibits the complex I mitochondrial respiratory chain of Hsp90 client proteins were undergoing apoptosis (Fig. 7, B). and thus blocks hypoxia-induced ROS production (26 , 28 – 32 ), jnci.oxfordjournals.org JNCI | Articles 957 Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021

Fig. 6 . Inhibition of heat shock protein 90 (Hsp90) chaperone function by deguelin. A ) Hsp90 client proteins and deguelin. Western blot analysis of the indicated Hsp90 client protein expression in H1299, A549, and H322 cells untreated or treated with 10 or 100 nM deguelin for 1 or 2 days is presented. CDK4 = cyclin-dependent kinase 4; eNOS = endothe- lial nitric oxide synthase; MEK1/2 = mitogen-activated protein kinase kinase; HIF = hypoxia-inducible factor. B ) Deguelin and protein degra- dation. Western analysis of CDK4, Akt, Hsp90, and ␤ -actin expression in H1299 cells untreated or treated with 100 nM deguelin for 1 day in the absence or presence of the proteasome inhibitor MG132 (10 µ M). Fig. 5 . Deguelin and the binding of ATP to heat shock protein 90 (Hsp90). A and B ) Competition between deguelin and ATP for Hsp90 investigated whether PHD, pVHL, and Akt are involved in degue- binding. Recombinant Hsp90 and JNK1 (as a negative control) proteins ␣ were mixed with vehicle (as a control [Con], 0.1% dimethyl sulfoxide), lin-mediated regulation of HIF-1 expression by examining HIF- soluble ATP, and 17-allylamino,17-demethoxygeldanamycin (17-AAG) 1 ␣ proteins mutated at two proline hydroxylation sites (Pro-402 ␥ or deguelin, at the indicated concentrations and incubated with -phos- and Pro-564) or HIF-1␣ proteins expressed in pVHL siRNA– phate – linked ATP – Sepharose. Material bound to the beads was ana- ␣ lyzed by western blotting with antibodies against Hsp90 or JNK1. A) transfected H1299 cells. In both cases, the expression of HIF-1 Laser densitometric quantifi cation of the density of the Hsp90 bands protein was decreased by deguelin treatment through a PHD- and from the competition experiment. Results are expressed as percent pVHL-independent mechanism. Moreover, the effect of deguelin density of each condition relative to the control value. Data are ␣ expressed as the mean; error bars are 95% confi dence intervals. B ) on HIF-1 expression was not abrogated by overexpression of Representative competition binding between ATP and deguelin to constitutively active Akt, indicating that PI3K-Akt– independent Hsp90. Independent experiments were repeated three times. The statis- mechanisms may mediate the effect of deguelin on HIF-1␣ expres- tical signifi cance of differences between groups was analyzed with Student's t test. All statistical tests were two-sided. C ) HIF-1␣ and HIF-1␤ sion. One such mechanism could involve Hsp90, which increases expression in H1299 cells untreated or treated with 10 µ M 17-AAG or the stability of HIF-1␣ protein (15 , 40 , 42 , 43 ). Indeed, we found ␣ ␤ 100 nM deguelin in hypoxia for 16 hours. HIF-1 and HIF-1 expression that the deguelin-mediated decreased expression of HIF-1␣ pro- was assessed by western blot analysis with antibodies against HIF-1␣ or HIF-1 ␤ . tein was attenuated by overexpression of Hsp90. Furthermore, the results of our structure– function analyses (computer modeling and ATP-binding assay) support a mechanism through which deguelin ROS could have contributed to deguelin’s activity on HIF-1␣ inhibits Hsp90 function by 1) inhibiting the binding of Hsp90 protein expression. However, we found that the decreased HIF-1␣ to the PAS domain of HIF-1␣ , which mediates the interaction expression mediated by deguelin treatment was not associated with between HIF-1 ␣ and Hsp90 and thus the sensitivity of HIF-1 ␣ to ROS production and was not affected by the treatment with succi- Hsp90 inhibitors (40 , 43 ); 2) binding to the ATP-binding pocket nate, indicating that deguelin acts through a ROS-independent of Hsp90; 3) competing with ATP for binding with Hsp90; and mechanism to destabilize HIF-1␣ . 4) destabilizing HIF-1␣ and many Hsp90 client proteins. Because the PHD, pVHL, and PI3K– Akt pathways have roles A limitation of this study is that the mechanisms through which in the expression or stabilization of HIF-1␣ (27 , 35 , 36 ), we next deguelin mediates its effects on various Hsp90 client proteins have

958 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021

Fig. 7 . Deguelin and heat shock protein 90 function in vivo. A ) Immuno- deguelin-treated nude mice after 15 – 28 days are shown. B ) Apoptosis. histochemical analysis. Cyclin-dependent kinase 4 (CDK4) expression Percentage of apoptotic (TUNEL-positive) cells compared with DAPI- and terminal deoxynucleotidyltransferase-mediated UTP end-labeling positive cells in H1299, A549, and MKN45 xenograft tumors obtained (TUNEL), both detected immunohistochemically, and 4′ ,6-diamidino-2- from vehicle (Con) or deguelin (Deg)-treated nude mice after 15– 28 phenylindole (DAPI) staining for nucleus identifi cation in H1299, A549, days is shown. Data are expressed as the mean; error bars are 95% and MKN45 xenograft tumors obtained from vehicle control (Con) or confi dence intervals.

not been fully elucidated. When we tested deguelin as an inhibitor the half-life and expression of each client protein, the intracellular of Hsp90, we observed decreased expression of several Hsp90 level and activity of the various E3 ubiquitin ligases and/or deubiq- client proteins (including CDK4, Akt, eNOS, MEK1/2, mutant uitylases [which are involved in polyubiquitination, which marks p53, and HIF-2␣ ) in a subset of deguelin-treated cancer cell lines. proteins for degradation by the proteasome (49 )] required by the Each client protein showed different levels of sensitivity to the client proteins, and the metabolism of deguelin. The conformation deguelin treatment, depending on the cellular context. We previ- of Hsp90 in multichaperone complexes could be another factor ously showed that treating premalignant human bronchial epithe- that affects the sensitivity of various cells to deguelin, as has been lial cells with deguelin for less than 1 day reduced the level of found for other Hsp90 inhibitors (50 ). phosphorylated Akt (Ser-473) but did not affect that of total Akt Our fi ndings provide evidence that deguelin binds directly to protein (10 ). However, in this study prolonged treatment with the ATP-binding pocket of Hsp90, interferes with Hsp90 chaper- deguelin decreased Akt expression in NSCLC cell lines, consistent one function, decreases the expression of many Hsp90 client pro- with the observation that the Hsp90 inhibitor 17-AAG has a teins, and induces apoptosis in cancer cells, which reduces tumor greater effect on phosphorylated Akt than on Akt (47 , 48 ). Of the growth. A concomitant reduction in the expression of Hsp90 client client proteins that we tested for deguelin sensitivity, HIF-1␣ and proteins induced by deguelin could have resulted in the broad phosphorylated Akt were the most sensitive. The differential levels antitumor activities of deguelin in vitro and in vivo in various types of deguelin sensitivity may depend on various factors, including of cancer cells. jnci.oxfordjournals.org JNCI | Articles 959 Deguelin has several features that make it an attractive tar- and apoptosis in premalignant human bronchial epithelial cells . J Natl geted antineoplastic drug in clinical trials. The calculated molecu- Cancer Inst 2003 ; 95 : 291– 302 . (11) Lee HY , Oh SH , Woo JK , Kim WY , Van Pelt CS , Price RE , et al . lar volume of deguelin is approximately 70% of that of 17-AAG, Chemopreventive effects of deguelin, a novel Akt inhibitor, on tobacco- an Hsp90 antagonist that is currently being evaluated in clinical induced lung tumorigenesis . J Natl Cancer Inst 2005 ; 97 : 1695 – 9 . trials but is proving unsatisfactory because of problems related to (12) Lee HY . Molecular mechanisms of deguelin-induced apoptosis in trans- its metabolism and poor solubility (51 ). The binding of the ansa- formed human bronchial epithelial cells . Biochem Pharmacol 2004 ; mycin ring of 17-AAG to Hsp90 requires a large conformational 68 : 1119 –24 . (13) Udeani GO , Gerhauser C , Thomas CF , Moon RC , Kosmeder JW , change with a high entropic penalty ( 52). The molecular size and Kinghorn AD , et al . Cancer chemopreventive activity mediated by degue- structural rigidity of deguelin could make its binding to Hsp90 lin, a naturally occurring rotenoid . Cancer Res 1997 ; 57 : 3424 – 8 . more entropically favorable than the binding of 17-AAG. Indeed, (14) Udeani GO , Zhao GM , Shin YG , Kosmeder JW 2nd , Beecher CW , deguelin is more potent than 17-AAG in vivo. For example, at Kinghorn AD , et al . Pharmacokinetics of deguelin, a cancer chemopreven- 100 nM— a dose that is considerably below that achievable in rats tive agent in rats . Cancer Chemother Pharmacol 2001 ; 47 : 263 – 8 . (15) Han JY , Oh SH , Morgillo F , Myers JN , Kim E , Hong WK , et al . (13 , 14 , 53 ) — deguelin inhibited expression of multiple Hsp90 cli- Hypoxia-inducible factor 1alpha and antiangiogenic activity of farnesyl- Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 ent proteins and induced apoptosis in NSCLC cells, whereas transferase inhibitor SCH66336 in human aerodigestive tract cancer . 1 µM 17-AAG had antiproliferative and apoptotic effects in can- J Natl Cancer Inst 2005 ; 97 : 1272 – 86 . cer cells (10 , 12 ). Moreover, when toxic effects, including infl am- (16) Fontana J , Fulton D , Chen Y , Fairchild TA , McCabe TJ , Fujita N , et al . mation and injury, in organs (including lung, liver, kidney, spleen, Domain mapping studies reveal that the M domain of hsp90 serves as a molecular scaffold to regulate Akt-dependent phosphorylation of endo- stomach, and ovary) were evaluated by a veterinary pathologist in thelial nitric oxide synthase and NO release . Circ Res 2002 ; 90 : 866 – 73 . xenograft models, deguelin was well tolerated, with no detectable (17) Becker TC , Noel RJ , Coats WS , Gomez-Foix AM , Alam T , Gerard RD , organ or systemic toxic effects after a 4-week treatment to achieve et al . Use of recombinant adenovirus for metabolic engineering of mam- therapeutic effi cacy or after a 19-week treatment to achieve effec- malian cells . Methods Cell Biol 1994 ; 43 (Pt A) : 161 – 89 . tive chemopreventive activities (11 ). Several clinical trials of (18) Isaacs JS , Jung Y-J , Mimnaugh EG , Martinez A , Cuttitta F , Neckers LM . Hsp90 regulates a von Hippel Lindau-independent hypoxia-inducible Hsp90 inhibitors that have been completed or are in progress factor-1alpha-degradative pathway . J Biol Chem 2002 ; 277 : 29936 – 44 . have observed various side effects, including liver, gastrointesti- (19) Lee HY , Moon H , Chun KH , Chang YS , Hassan K , Ji L , et al . Effects of nal, and constitutional symptoms, although it is not clear whether insulin-like growth factor binding protein-3 and farnesyltransferase these side effects are related to the Hsp90 mechanism per se, to inhibitor SCH66336 on Akt expression and apoptosis in non-small-cell the specifi c chemical nature of the benzoquinone ansamycin lung cancer cells . J Natl Cancer Inst 2004 ; 96 : 1536 – 48 . (20) Marcu MG , Chadli A , Bouhouche I , Catelli M , Neckers LM . The heat structure, or to the formulation of the drug used (39 , 54 , 55 ). shock protein 90 antagonist novobiocin interacts with a previously unrec- Thus, deguelin appears to be an effective antineoplastic drug ognized ATP-binding domain in the carboxyl terminus of the chaperone . targeting Hsp90 client proteins. However, extensive and complete J Biol Chem 2000 ; 275 : 37181 – 6 . chronic toxicity testing in clinical trials is warranted before the fur- (21) Gasteiger J , Marsili M . Iterative partial equalization of orbital electro- ther development of deguelin as an anticancer therapeutic agent. negativity —a rapid access to atomic charges . Tetrahedron 1980 ; 36 : 3219 – 28 . (22) Powell M . Restart procedures for the conjugate gradient method . Math Program 1977 ; 12 : 241 – 54 . References (23) Stebbins CE , Russo AA , Schneider C , Rosen N , Hartl FU , Pavletich NP . (1) O’Dwyer ME , Druker BJ . STI571: an inhibitor of the BCR-ABL tyrosine Crystal structure of an Hsp90-geldanamycin complex: targeting of a pro- kinase for the treatment of chronic myelogenous leukaemia . Lancet Oncol tein chaperone by an antitumor agent . Cell 1997 ; 89 : 239– 50 . 2000 ; 1 : 207 – 11 . (24) Clark RD , Strizhev A , Leonard JM , Blake JF , Matthew JB . Consensus (2) Swanton C , Futreal A , Eisen T . Her2-targeted therapies in non-small cell scoring for ligand/protein interactions . J Mol Graph Model 2002 ; 20 : lung cancer . Clin Cancer Res 2006 ; 12 : 4377s – 83s . 281 – 95 . (3) Neckers L , Ivy SP . Heat shock protein 90 . Curr Opin Oncol 2003 ; (25) Kuo GH , Wang A , Emanuel S , Deangelis A , Zhang R , Connolly PJ , et al . 15 : 419 – 24 . Synthesis and discovery of pyrazine-pyridine biheteroaryl as a novel series (4) Pearl LH , Prodromou C . Structure and in vivo function of Hsp90 . Curr of potent vascular endothelial growth factor receptor-2 inhibitors . J Med Opin Struct Biol 2000 ; 10 : 46 – 51 . Chem 2005 ; 48 : 1886 – 900 . (5) Gabai VL , Kabakov AE . Induction of heat-shock protein synthesis and (26) Semenza GL . HIF-1: mediator of physiological and pathophysiological thermotolerance in EL-4 ascites tumor cells by transient ATP depletion responses to hypoxia . J Appl Physiol 2000 ; 88 : 1474 – 80 . after ischemic stress . Exp Mol Pathol 1994 ; 60 : 88 – 99 . (27) Semenza GL . Targeting HIF-1 for cancer therapy . Nat Rev Cancer (6) Sharp S , Workman P . Inhibitors of the HSP90 molecular chaperone: cur- 2003 ; 3 : 721– 32 . rent status . Adv Cancer Res 2006 ; 95 : 323 – 48 . (28) Brar SS , Kennedy TP , Whorton AR , Murphy TM , Chitano P , Hoidal JR . (7) Gyurkocza B , Plescia J , Raskett CM , Garlick DS , Lowry PA , Carter BZ , Requirement for reactive oxygen species in serum-induced and platelet- et al . Antileukemic activity of shepherdin and molecular diversity of hsp90 derived growth factor-induced growth of airway smooth muscle . J Biol inhibitors . J Natl Cancer Inst 2006 ; 98 : 1068 – 77 . Chem 1999 ; 274 : 20017 – 26 . (8) Neckers L , Neckers K . Heat-shock protein 90 inhibitors as novel can - (29) Chandel NS , McClintock DS , Feliciano CE , Wood TM , Melendez JA , cer chemotherapeutics— an update . Expert Opin Emerg Drugs 2005 ; 10 : Rodriguez AM , et al . Reactive oxygen species generated at mitochon- 137 – 49 . drial complex III stabilize hypoxia-inducible factor-1alpha during (9) Heath EI , Gaskins M , Pitot HC , Pili R , Tan W , Marschke R , et al . hypoxia: a mechanism of O2 sensing . J Biol Chem 2000 ; 275 : 25130 – 8 . A phase II trial of 17-allylamino-17-demethoxygeldanamycin in patients (30) Haddad JJ , Land SC. A non-hypoxic, ROS-sensitive pathway mediates with hormone-refractory metastatic prostate cancer . Clin Prostate Cancer TNF-alpha-dependent regulation of HIF-1alpha . FEBS Lett 2001 ; 2005 ; 4 : 138 – 41 . 505 : 269 –74 . (10) Chun KH , Kosmeder JW 2nd , Sun S , Pezzuto JM , Lotan R , Hong WK , (31) Sato H , Sato M , Kanai H , Uchiyama T , Iso T , Ohyama Y , et al . et al . Effects of deguelin on the phosphatidylinositol 3-kinase/Akt pathway Mitochondrial reactive oxygen species and c-Src play a critical role in

960 Articles | JNCI Vol. 99, Issue 12 | June 20, 2007 hypoxic response in vascular smooth muscle cells . Cardiovasc Res 2005 ; (46) Isaacs JS , Xu W , Neckers L . Heat shock protein 90 as a molecular target 67 : 714 – 22 . for cancer therapeutics . Cancer Cell 2003 ; 3 : 213– 7 . (32) Vaux EC , Metzen E , Yeates KM , Ratcliffe PJ . Regulation of hypoxia- (47) Basso AD , Solit DB , Munster PN , Rosen N . Ansamycin antibiotics inhibit inducible factor is preserved in the absence of a functioning mitochondrial Akt activation and cyclin D expression in breast cancer cells that overex- respiratory chain . Blood 2001 ; 98 : 296 – 302 . press HER2 . Oncogene 2002 ; 21 : 1159 – 66 . (33) Kietzmann T , Porwol T , Zierold K , Jungermann K , Acker H . Involvement (48) Solit DB , Basso AD , Olshen AB , Scher HI , Rosen N . Inhibition of heat of a local fenton reaction in the reciprocal modulation by O2 of the gluca- shock protein 90 function down-regulates Akt kinase and sensitizes tumors gon-dependent activation of the phosphoenolpyruvate carboxykinase gene to Taxol . Cancer Res 2003 ; 63 : 2139 – 44 . and the insulin-dependent activation of the glucokinase gene in rat hepa- (49) Thornton BR , Toczyski DP . Precise destruction: an emerging picture of tocytes . Biochem J 1998 ; 335 (Pt 2) : 425 – 32 . the APC . Genes Dev 2006 ; 20 : 3069 – 78 . (34) Harris AL . Hypoxia — a key regulatory factor in tumour growth . Nat Rev (50) Kamal A , Thao L , Sensintaffar J , Zhang L , Boehm MF , Fritz LC , et al . A Cancer 2002 ; 2 : 38 – 47 . high-affi nity conformation of Hsp90 confers tumour selectivity on Hsp90 (35) Semenza GL . HIF-1 and tumor progression: pathophysiology and thera- inhibitors . Nature 2003 ; 425 : 407 – 10 . peutics . Trends Mol Med 2002 ; 8 : S62 – 7 . (51) Beliakoff J , Whitesell L . Hsp90: an emerging target for breast cancer (36) Ivan M , Kondo K , Yang H , Kim W , Valiando J , Ohh M , et al . HIFalpha therapy . Anticancer Drugs 2004 ; 15 : 651– 62 .

targeted for VHL-mediated destruction by proline hydroxylation: impli- (52) Jez JM , Chen JC , Rastelli G , Stroud RM , Santi DV . Crystal structure and Downloaded from https://academic.oup.com/jnci/article/99/12/949/965680 by guest on 23 September 2021 cations for O2 sensing . Science 2001 ; 292 : 464 – 8 . molecular modeling of 17-DMAG in complex with human Hsp90 . Chem (37) Fukuda R , Hirota K , Fan F , Jung YD , Ellis LM , Semenza GL . Insulin-like Biol 2003 ; 10 : 361 – 8 . growth factor 1 induces hypoxia-inducible factor 1-mediated vascular (53) Gerhauser C , Lee SK , Kosmeder JW , Moriarty RM , Hamel E , Mehta RG , endothelial growth factor expression, which is dependent on MAP kinase et al . Regulation of ornithine decarboxylase induction by deguelin, a and phosphatidylinositol 3-kinase signaling in colon cancer cells . J Biol natural product cancer chemopreventive agent . Cancer Res 1997 ; 57 : Chem 2002 ; 277 : 38205 – 11 . 3429 – 35 . (38) Zhou J , Schmid T , Frank R , Brune B . PI3K/Akt is required for heat (54) Ramanathan RK , Trump DL , Eiseman JL , Belani CP , Agarwala SS , shock proteins to protect hypoxia-inducible factor 1alpha from pVHL- Zuhowski EG , et al . Phase I pharmacokinetic-pharmacodynamic study of independent degradation . J Biol Chem 2004 ; 279 : 13506 – 13 . 17-(allylamino)-17-demethoxygeldanamycin (17AAG, NSC 330507), a (39) Banerji U , O’Donnell A , Scurr M , Pacey S , Stapleton S , Asad Y , et al . novel inhibitor of heat shock protein 90, in patients with refractory Phase I pharmacokinetic and pharmacodynamic study of 17-allylamino, advanced cancers . Clin Cancer Res 2005 ; 11 : 3385 – 91 . 17-demethoxygeldanamycin in patients with advanced malignancies . (55) Grem JL , Morrison G , Guo XD , Agnew E , Takimoto CH , Thomas R , et al . J Clin Oncol 2005 ; 23 : 4152 – 61 . Phase I and pharmacologic study of 17-(allylamino)-17-demethoxygel- (40) Katschinski DM , Le L , Schindler SG , Thomas T , Voss AK , Wenger RH . danamycin in adult patients with solid tumors . J Clin Oncol 2005 ; 23 : Interaction of the PAS B domain with HSP90 accelerates hypoxia-inducible 1885 – 93 . factor-1alpha stabilization . Cell Physiol Biochem 2004 ; 14 : 351 – 60 . (41) Yamada S , Ono T , Mizuno A , Nemoto TK . A hydrophobic segment within the C-terminal domain is essential for both client-binding and Notes dimer formation of the HSP90-family molecular chaperone . Eur J S. H. Oh and J. K. Woo contributed equally to this work and should be consid- Biochem 2003 ; 270 : 146 – 54 . ered joint fi rst authors. (42) Minet E , Mottet D , Michel G , Roland I , Raes M , Remacle J , et al . This work was supported by National Institutes of Health grants RO1 Hypoxia-induced activation of HIF-1: role of HIF-1alpha-Hsp90 interac- CA100816-01 and R01 CA109520-01 (to H.-Y. Lee), American Cancer tion . FEBS Lett 1999 ; 460 : 251 – 6 . Society grant RSG-04-082-01-TBE 01 (to H.-Y. Lee), and partly by a (43) Isaacs JS , Jung Y-J , Neckers L . Aryl hydrocarbon nuclear translocator National Foundation for Cancer Research grant LF01-065-1 (to W. K. (ARNT) promotes oxygen-independent stabilization of hypoxia-inducible Hong) and by the Creative Research Initiatives (NeuroVascular Coordination factor-1{alpha} by modulating an Hsp90-dependent regulatory pathway . Research Center) of the Ministry of Science and Technology, Korea Science J Biol Chem 2004 ; 279 : 16128 – 35 . and Engineering Foundation , Korea (to K.-W. Kim). W. K. Hong is an (44) Neckers L . Development of small molecule Hsp90 inhibitors: utilizing American Cancer Society clinical research professor. We wish to thank Dr both forward and reverse chemical genomics for drug identifi cation . Curr Garth Powis, DPhil, for useful discussion. The authors had full responsibil- Med Chem 2003 ; 10 : 733 – 9 . ity for the design of the study concept, data analysis, data interpretation, and (45) Wenger RH . Cellular adaptation to hypoxia: O2-sensing protein hydroxy- preparation of the manuscript. lases, hypoxia-inducible transcription factors, and O2-regulated gene Manuscript received August 26 , 2006 ; revised March 29 , 2005 ; accepted expression . FASEB J 2002 ; 16 : 1151 – 62 . May 9 , 2007.

jnci.oxfordjournals.org JNCI | Articles 961