Development of a novel Hsp90 inhibitor NCT-50 as a potential anticancer agent for the treatment of non-small cell lung cancer

Despite the development of advanced therapeutic regimens such as molecular targeted therapy and immunotherapy, the 5-year survival of patients with lung cancer is still less than 20%, suggesting the need to develop additional treatment strategies. The molecular chaperone heat shock protein 90 (Hsp90) plays important roles in the maturation of oncogenic proteins and thus has been considered as an anticancer therapeutic target. Here we show the efficacy and biological mechanism of a Hsp90 inhibitor NCT-50, a novobiocin-deguelin analog hybridizing the pharmacophores of these known Hsp90 inhibitors. NCT-50 exhibited significant inhibitory effects on the viability and colony formation of non-small cell lung cancer (NSCLC) cells and those carrying resistance to chemotherapy. In contrast, NCT-50 showed minimal effects on the viability of normal cells. NCT-50 induced apoptosis in NSCLC cells, inhibited the expression and activity of several Hsp90 clients including hypoxia-inducible factor (HIF)-1α, and suppressed pro-angiogenic effects of NSCLC cells. Further biochemical and in silico studies revealed that NCT-50 downregulated Hsp90 function by interacting with the C-terminal ATP-binding pocket of Hsp90, leading to decrease in the interaction with Hsp90 client proteins. These results suggest the potential of NCT-50 as an anticancer Hsp90 inhibitor.

Despite the development of advanced therapeutic regimens such as molecular targeted therapy and immunotherapy, the 5-year survival of patients with lung cancer is still less than 20%, suggesting the need to develop additional treatment strategies. The molecular chaperone heat shock protein 90 (Hsp90) plays important roles in the maturation of oncogenic proteins and thus has been considered as an anticancer therapeutic target. Here we show the efficacy and biological mechanism of a Hsp90 inhibitor NCT-50, a novobiocin-deguelin analog hybridizing the pharmacophores of these known Hsp90 inhibitors. NCT-50 exhibited significant inhibitory effects on the viability and colony formation of non-small cell lung cancer (NSCLC) cells and those carrying resistance to chemotherapy. In contrast, NCT-50 showed minimal effects on the viability of normal cells. NCT-50 induced apoptosis in NSCLC cells, inhibited the expression and activity of several Hsp90 clients including hypoxia-inducible factor (HIF)-1α, and suppressed pro-angiogenic effects of NSCLC cells. Further biochemical and in silico studies revealed that NCT-50 downregulated Hsp90 function by interacting with the C-terminal ATPbinding pocket of Hsp90, leading to decrease in the interaction with Hsp90 client proteins. These results suggest the potential of NCT-50 as an anticancer Hsp90 inhibitor.
To maintain homeostasis during various extracellular and intracellular insults, cancer cells rely on heat shock protein 90 (Hsp90) to stabilize many proteins, constructing signaling networks responsible for cell survival, growth, and proliferation 1,2 . Indeed, Hsp90 client proteins are associated with the hallmarks of cancer 3,4 and thus targeting Hsp90 has been considered an efficient anticancer therapeutic strategy 4 . Several Hsp90 inhibitors with various structural backbones have shown potent anticancer activities in vitro and in vivo 5 . However, none of these inhibitors have been approved for use in the clinic, and several clinical trials using Hsp90 inhibitors have been discontinued due to minimal effects and toxicity. Hence, it is necessary to develop safer and more effective Hsp90 inhibitors.
Lung cancer is the leading cause of cancer-related human deaths worldwide. Despite extensive efforts to develop efficient therapeutic interventions, the 5-year survival rate for lung cancer is less than 20% 6,7 . Based on several problems with chemotherapy, such as unselective toxicity, drug resistance, and recurrence, agents selectively targeting overactivated signaling pathways in cancer cells have been extensively developed and now several drugs are used as the first-line therapy for patients with lung cancer carrying relevant genetic alterations 8 . Although some therapies have initially shown remarkable anticancer effects 9,10 , recurrence due to drug resistance is an inevitable consequence of these targeted therapies in most cases. Thus, there is an urgent unmet need to develop more efficacious anticancer agents that can be utilized as a monotherapy or an adjuvant therapy with other anticancer drugs.
Natural products are considered an important source for the development of anticancer drugs 11 . We have demonstrated the anticancer and cancer chemopreventive effects of a naturally occurring rotenoid deguelin [12][13][14][15][16][17] . Further studies revealed Hsp90 as the molecular target for deguelin's antitumor effect 18 . However, the potential toxic effects of deguelin, such as a Parkinson's disease-like syndrome 19 , hamper the further clinical utility of deguelin. Novobiocin, another natural product, is derived from Streptomyces and is a coumarin antibiotics with Hsp90 inhibiting activity 2,20,21 . However, it's half maximal inhibitory concentration (IC 50 ) has been reported to be up to 700 μM 22 , making it difficult to utilize as a clinical therapeutic.
To circumvent the limitation of deguelin, we have developed several deguelin analogs that harbor less or no potential toxicity to various normal cells 16,[23][24][25][26][27] . We and others previously demonstrate the involvement of the dimethoxy benzopyran moiety of deguelin in the interaction with Hsp90 18 and the crucial association of the acylamino moiety with the antiproliferative activity of novobiocin 28 . Importantly, the structural features of deguelin and novobiocin, which are highlighted in three different colors, indicate that they share principal pharmacophores (Fig. 1a). Thus, in addition to our previous efforts on the development of ring-truncated deguelin analogs [23][24][25][26][27] , here we synthesized a novel novobiocin-deguelin analog, designated 5-methoxy-N-(3-methoxy-4-(2-(pyridin-4-yl)ethoxy)phenyl)-2,2-dimethyl-2H-chromene-6-carboxamide (NCT-50), by hybridizing pharmacophores of these compounds and evaluated the effects of NCT-50 on the viability of non-small cell lung cancer (NSCLC) cells, which account for 80-85% of lung cancer 29 . NCT-50 displayed significant inhibitory effects on the viability and colony formation of NSCLC cells with minimal effects on the viability of normal cells. NCT-50 also induced apoptosis in NSCLC cells and inhibited pro-angiogenic activities of NSCLC cells. Further mechanistic investigation revealed that NCT-50 disrupted Hsp90 function by interacting with the ATP-binding pocket in the C-terminal domain of Hsp90 and concomitantly suppressed the expression and activity of multiple Hsp90 clients, including hypoxia-inducible factor (HIF)-1α. These results demonstrate the potential use of NCT-50 as an Hsp90-targeting anticancer agent.
NCT-50 significantly inhibits the viability and colony-forming ability of NSCLC cells with minimal effect on the viability of normal cells. We then examined the effect of NCT-50 on the viability and colony formation of human NSCLC cells, either naïve or resistant to anticancer drugs 23 , considering that resistance to chemotherapy is a large obstacle for efficient anticancer therapeutics. NCT-50 significantly inhibited the viability and anchorage-dependent colony formation of NSCLC cells in a concentration-dependent manner (Fig. 2a,b). The IC 50 values of NCT-50 on the inhibition of NSCLC cell viability were about 2 μM on average (Table 1), which was comparable or superior to previously developed deguelin analogs 23,24,26 , novobiocin, and its analogs 22,28,30 . The inhibitory effect of NCT-50 on the viability and anchorage-dependent colony formation of drug-resistant sublines (H1299/CsR, H1299/PmR, and H226B/PcR) was comparable with that on the corresponding naïve cells (H1299 and H226B), suggesting the effectiveness of NCT-50 for both chemo-naïve and chemo-resistant NSCLC cells (Fig. 2a,b). Consistent with these results, the colony formation of NSCLC cells (either chemo-naïve or chemoresistant), under anchorage-independent culture conditions, was also markedly suppressed by treatment with NCT-50, especially at the concentration of 5 μM (Fig. 2c), indicating the effectiveness of NCT-50 in suppressing the tumorigenicity of NSCLC cells, regardless of their drug resistance status. NCT-50 displayed weak but comparable cytotoxic effects in NSCLC cells compared with known Hsp90 inhibitors that have been evaluated in clinical trials such as ganetespib and PU-H71 31,32 (Fig. 2d). These results suggest that NCT-50 suppresses the survival and the colony-forming ability of both chemo-naïve and chemo-resistant NSCLC cells.

Induction of apoptosis is associated with NCT-50-medated suppression of NSCLC cell viability.
According to the efficient inhibitory effects of NCT-50 on the viability and colony-forming ability of NSCLC cells, we investigated the underlying mechanism. First, using flow cytometry, we examined the effect of NCT-50 on cell cycle progression. As shown in Fig. 3a, approximately 30% of the cell population was accumulated in the sub-G1 phase in H1299 and H460 cells after treatment with 5 μM NCT-50. Hoechst 33258 staining also showed increased chromatin condensation, a feature of apoptotic cells 33 , in NCT-50-treated cells (Fig. 3b). Western blot analysis also revealed obviously increased levels of poly-(ADP-ribose) polymerase (PARP) cleavage in the two cell lines after the drug treatment (Fig. 3c). Consistently, the Annexin V-positive population was markedly increased in the drug-treated cells (Fig. 3d). These results indicated apoptotic activities of NCT-50 in NSCLC cells, suggesting that the reduced NSCLC cell viability after treatment with NCT-50 is due, in part, to increased apoptosis.

NCT-50 displays improved safety profiles compared with known Hsp90 inhibitors and deguelin.
Considering the potential toxic effect of Hsp90 inhibitors 34 and deguelin 19 , we next examined whether NCT-50 has improved safety profiles compared to these known Hsp90 inhibitors. First, we assessed cytotoxicity of NCT-50 at the cellular levels by testing the effects of NCT-50 on the viability of several normal cells, including two human normal lung epithelial cells (HBE and BEAS-2B), human retinal pigment epithelial cells (RPE), and human vascular endothelial cells (HUVECs). Compared to various cancer cell lines treated with NCT-50 ( Fig. 2), these normal cell lines showed minimal changes in their viability after the treatment with NCT-50 (10 μM) (Fig. 4a). The IC 50 values of NCT-50 on the viability of these normal cells was over 10 μM (Table 1). Moreover, in contrast to minimal cytotoxic effect of NCT-50, known Hsp90 inhibitors such as ganetespib and PU-H71 displayed remarkable cytotoxic effects in BEAS-2B cells, suggesting reduced toxicity of NCT-50 compared with these Hsp90 inhibitors (Fig. 4b). We further performed several in vivo experiments to evaluate toxicity profiles of NCT-50. Mice in a FVB background were orally administered with 4 mg/kg NCT-50 twice a day for 7 consecutive days. Compared with vehicle-treated mice, NCT-50-treated mice displayed no significant changes in body weight (Fig. 4c). The serum levels GOT (glutamate oxaloacetate transaminase), GPT (glutamate pyruvate transaminase), and blood urea nitrogen (BUN), indicators of liver and renal function 35,36 , were not significantly different between vehicleand NCT-50-treated mice (Fig. 4d). Moreover, histological analyses of H&E-stained tissue samples obtained from several organs (lung, liver, brain, and kidney) of NCT-50-treated mice revealed no remarkable histopathological changes (Fig. 4e). These results collectively indicate minimal toxicities of NCT-50.
According to the concern that deguelin may induce Parkinson's-like syndrome 19 , which was manifested by decreased tyrosine hydroxylase immunoreactivity in the rat brain 19 , we evaluated the potential neurotoxicity of NCT-50. Because inhibition of NADH dehydrogenase activity was responsible for the potential neurotoxicity of deguelin 19 , we first determined the effects of NCT-50 and deguelin on NADH dehydrogenase activity using a mitochondria enriched fraction. In contrast to the significant inhibitory effects of deguelin, NCT-50 showed no detectable impact on the mitochondrial NADH dehydrogenase activity (Fig. 4f). We further compared the effect of NCT-50 and deguelin on the viability of mouse hippocampal neuronal cell line . Compared to deguelin, NCT-50 induced minimal changes in HT-22 cells viability (Fig. 4g). To verify these in vitro results, we determined in vivo neurotoxicity of NCT-50. To this end, mice were orally administered with NCT-50 or deguelin (4 mg/kg) twice a day for 7 consecutive days. We compared the effects of NCT-50 and deguelin on the immunoreactivity of tyrosine hydroxylase (TH), an enzyme in the late-limiting step of dopamine synthesis that has been used as a marker of dopaminergic neuron 37,38 , in the mouse midbrain. Consistent with the previous findings in the rat brain 19,25 , the TH immunoreactivity was significantly decreased by deguelin treatment in the mouse midbrain (Fig. 4h). In contrast, NCT-50 treatment minimally altered the level of the TH immunoreactivity. Taken together, these results indicate the markedly improved safety profile of NCT-50 compared with deguelin.

NCT-50 inhibits expression of client proteins of Hsp90 and shows anti-angiogenic activities.
Based on the previous studies demonstrating the inhibitory effect of novobiocin 20 and deguelin 18 , we assessed whether NCT-50 could suppress expression of Hsp90 client proteins. Treatment with NCT-50 in hypoxic conditions decreased HIF-1α expression in a dose-dependent manner (Fig. 5a). The NCT treatment also inhibited the expression of several Hsp90 client proteins, including epidermal growth factor receptor (EGFR), insulin-like    growth factor receptor-1 (IGF-1R), Akt, and MEK1/2 4,39 in normoxic conditions. Moreover, NCT-50 markedly suppressed the expression of HIF-1α target genes (VEGF, TGFB3, and PDGFB 40 ) (Fig. 5b) but not non-HIF target genes [DDIT3 (encoding GADD153/CHOP) 41 , PPPIR15A (a GADD153-target gene 42 ), and TRIB3 43 ] in H1299 and H460 cells (Fig. 5c). According to previous studies on the role of HIF-1α in angiogenesis 44 and the antiangiogenic effects of deguelin and its derivatives 23,25,45,46 , we further examined the antiangiogenic property of NCT-50. Because vascular endothelial cell tube formation ability in Matrigel is a general feature related to angiogenesis 47 and angiogenic factors are produced by cancer cells and endothelial or stromal cells 48 , we assessed the effects of NCT-50 on the tube formation and proliferation of vascular endothelial cells, using the conditioned medium (CM) collected from NCT-50-treated NSCLC cells. In contrast to CM from vehicle-treated H1299 and H460 cells, exposure of HUVECs to CM from NCT-50-treated cells resulted in decreased proliferation (Fig. 5d) and tube formation (Fig. 5e) of HUVECs, indicating that NCT-50 inhibits the production of angiogenic factors by NSCLC cells. Moreover, NCT-50 exhibited improved suppressive effects on HIF-1α expression compared with the known Hsp90 inhibitors, including ganetespib and PU-H71 (Fig. 5f). These findings suggested that Hsp90 function was effectively suppressed by treatment with NCT-50.

NCT-50 inhibits Hsp90 function by binding to the C-terminal ATP binding pocket of Hsp90.
We investigated the inhibitory effect of NCT-50 on Hsp90 function. Because the interaction of Hsp90 with client proteins is essential for its chaperone function 49 , we examined the effect of NCT-50 on the interaction between Hsp90 and HIF-1α. Under hypoxic conditions, the physical interaction between Hsp90 and HIF-1α was clearly reduced in H1299 and H460 cells after treatment with NCT-50 for 1 hour, when HIF-1α and Hsp90 expression was minimally affected (Fig. 6a). These findings indicated that NCT-50 biologically induce the dissociation of HIF-1α with Hsp90. Hsp90 has been proposed to possesses the ATP-binding pockets in the N-terminal and C-terminal domains 2,25,50 . To obtain evidence for NCT-50 binding to the ATP-binding pockets of Hsp90, we analyzed the effects of NCT-50 on the binding capacity of recombinant Hsp90 protein to ATP-agarose. We found that the ATP interaction with full length (FL) Hsp90 protein was suppressed by treatment with NCT-50 in a dose dependent manner (Fig. 6b). Subsequent analysis with truncated N-terminal (N) and C-terminal (C) domains revealed that ATP binding to the C domain, but not to the N domain, was markedly affected by NCT-50 treatment (Fig. 6c). To ensure the binding of NCT-50 to the C-terminal ATP-binding site of Hsp90, we performed a competition assay using biotinylated NCT-50 and a known C-terminal Hsp90 inhibitor cisplatin. Preincubation with cisplatin gradually suppressed the binding of biotinylated NCT-50 to the C-terminal domain of Hsp90 in a concentration-dependent manner (Fig. 6d). These results collectively suggest the potential of NCT-50 as a C-terminal Hsp90 inhibitor We further conducted molecular docking analysis using the Surflex-Dox program to evaluate the binding capability of NCT-50. Previously published Hsp90 homology model structure 23,25 was used as a receptor for docking in this study. As shown in Fig. 6d, NCT-50 fits well into the C-terminal ATP-binding cavity, locating at central region of dimerization interface of Hsp90 homodimer 51 . The oxygen atom of acylamino group forms a hydrogen bond with the side-chain of Lys615 in chain A, and the oxygen in the methoxy group forms additional hydrogen bond with the side-chain of Ser677 (A). In addition, benzopyran ring of NCT-50 engages in a cation-π interaction with the side-chain amine of Lys615. These two residues (Lys615 and Ser677) were confirmed as key residues for ATP binding in Hsp90 C-terminal domain by ATP-agarose binding assay in our previous study 25 . The pyridine and benzene fragment locate in a polar region and oxygen in the methoxy group in the benzene ring forms a hydrogen bond with side chain of Asn622. Overall, our docking model suggests that NCT-50 can efficiently bind to the C-terminal ATP binding site, stabilizing the open conformation of Hsp90 homodimer, which leads to the inhibition of Hsp90.

Discussion
In this study, we demonstrate the potential of NCT-50 as an Hsp90-targeting anticancer agent with reduced toxicity. NCT-50 significantly inhibited the viability, colony-forming ability, and proangiogenic effect of NSCLC cells. NCT-50 induced the apoptotic cell death of NSCLC cells, but the viability of normal cells was minimally affected by NCT-50 treatment. Moreover, NCT-50 displayed comparable inhibitory effects on the survival and colony formation of NSCLC cells with acquired chemoresistance. Mechanistically, NCT-50 effectively disrupted Hsp90 function by directly binding to the C-terminal ATP-binding pocket of Hsp90, leading to downregulation of the expression and activity of several Hsp90 client proteins including HIF-1α. These data suggest that NCT-50 may have potential as an anticancer drug targeting Hsp90.
An ATPase-containing molecular chaperone, Hsp90 plays an important role in the maintenance and stabilization of numerous client proteins in response to heat shock or other stresses 1,2 . Cancer cells easily undergo proteotoxic pressure due, in part, to the accumulation of mutated molecules that could lead to cell lethality 52 . Numerous Hsp90 client proteins are often mutated or overexpressed in several types of human cancer, including NSCLC 2,53 , and these proteins mediate cancer cell proliferation, survival, angiogenesis, invasion, metastasis, and resistance to conventional or targeted anticancer drugs, such as erlotinib, trastuzumab, paclitaxel, and cisplatin 1,2,54-57 . Hence, targeting Hsp90 would be an effective way to treat cancer and overcome anticancer drug resistance. Although several clinical trials evaluating the effectiveness of Hsp90 inhibitors in lung cancer treatment are ongoing 58 , the main issues with Hsp90 inhibitors are undesirable side effects, poor water solubility, and toxicity 1 . Therefore, it is necessary to develop a novel anticancer Hsp90 inhibitor with additional chemical entities or mechanisms of action.
Previously, we demonstrated the potential of a naturally occurring rotenoid deguelin as a potent anticancer agent targeting Hsp90 through direct interaction with the ATP-binding pocket of Hsp90 18 . However, the potential neurotoxicity of deguelin due to its inhibitory effect on NADH dehydrogenase 19 raises concern for its use in cancer patients. Hence, we have attempted to synthesize deguelin analogs to develop novel Hsp90 inhibitors 16,23-27 . According to the similar structural and mechanistic features between deguelin and novobiocin, here we synthesized a novobiocin-deguelin analog NCT-50 and demonstrated its potential to suppress NSCLC cell viability and proangiogenic ability by inhibiting Hsp90 function. We observed some features of NCT-50 that identify it as a potential hit or chemical lead for the development of anticancer Hsp90 inhibitor. NCT-50 displayed significant cytotoxic and proapoptotic effects on NSCLC cells without significantly affecting the growth of normal cells derived from lung epithelium, retinal pigment epithelium, vascular endothelium and hippocampus. The potency of the inhibitory effect of NCT-50 on the viability of NSCLC cells and Hsp90 function were comparable or superior to that of previously developed deguelin analogs, novobiocin and its analogs, and Hsp90 inhibitors in clinical trials [22][23][24]26,28,30 , emphasizing the potential of NCT-50 as a novel and efficacious Hsp90 inhibitor. In addition, considering liver and ocular toxicities are drawbacks of the currently available Hsp90 inhibitors 58 , the reduced toxicity of NCT-50 in vitro and in vivo compared with known Hsp90 inhibitors or deguelin appears to be a clinically favorable feature. In addition, NCT-50 significantly suppressed proangiogenic ability of NSCLC cells. Because angiogenesis is crucial for tumor growth and metastasis 59 , the antiangiogenic effect of NCT-50 may disrupt primary tumor growth and lower metastatic burden. Moreover, consistent with previous reports suggesting an association of Hsp90 with anticancer drug resistance and overcoming the resistance to chemo-or targeted anticancer therapies by using Hsp90 inhibitors 54-57 , NCT-50 was effective in both chemo-naïve and chemoresistant NSCLC cells.. Although additional studies such as animal experiments should be performed to evaluate the effectiveness and toxicity of NCT-50, either alone or in combination with chemotherapy, our results suggest the potential utility of NCT-50 to overcome resistance to paclitaxel, cisplatin, or pemetrexed, which are generally used to treat patients with NSCLC 60,61 .
Our study reveals that Hsp90 is a molecular target of NCT-50 for its anticancer and antiangiogenic effects. Consistent with the ability of deguelin analogs and novobiocin to interact with the C-terminal ATP binding pocket of Hsp90 23,25,62 , NCT-50 can interact with the C-terminal ATP binding pocket of Hsp90. Importantly, the acylamino moiety and the benzopyran ring of NCT-50 are crucial for the interaction with Lys615 and Ser677 residues of Hsp90, key residues for ATP binding in the C-terminal ATP binding pocket of Hsp90 25 . These moieties were previously defined as an important pharmacophore for the interaction of deguelin with Hsp90 18 and the antiproliferative effect of novobiocin 28 . Thus, together with previous findings, our study provides key structural backbone for the development of C-terminal Hsp90 inhibitors.
In summary, the present study demonstrates that NCT-50 exhibits significant cytotoxic and antiangiogenic activities in NSCLC cells by suppressing Hsp90 function through directly binding to the C-terminal ATP-binding pocket of Hsp90, suggesting that NCT-50 can be considered a novel hit compound to further develop as an anticancer Hsp90 inhibitors. Further studies are warranted to investigate the effectiveness of NCT-50 in additional preclinical and clinical settings.

Methods
Reagents. Antibodies against EGFR, Akt, MEK1/2 and tubulin were purchased from Cell Signaling Technology (Danvers, MA, USA). Antibodies against cleaved PARP and HIF-1α were purchased from BD Biosciences (San Jose, CA, USA). Antibodies against IGF-1R and actin were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). HRP-conjugated anti-rabbit and anti-goat secondary antibodies were purchased from GeneTex (Irvine, CA, USA). An anti-Hsp90 antibody was purchased from Enzo Life Science (Farmingdale, NY, USA). Ni-NTA agarose was purchased from Invitrogen (Carlsbad, CA, USA). ATP-agarose was acquired from Innova Biosciences (Cambridge, UK). The first-strand cDNA synthesis kit was purchased from Dakara Korea Biomedical, Inc. in VascuLife basal medium supplemented with VascuLife VEGF life factors (Lifeline Cell Technology, Frederick, MD, USA). Cells were incubated at 37 °C with 5% CO 2 in a humidified atmosphere.

MTT assay. Cells [2 × 10 3 cells/well (for NSCLC and HT-22 cells) or 3 × 10 3 cells/well (for HBE, BEAS-2B,
RPE, and HUVEC) in 96-well plates] were treated with increasing concentrations of NCT-50, deguelin, ganetespib, or PU-H71 (0.1, 1, and 10 μM) for 2 or 3 days. After incubation, the cells were incubated with MTT solution (final concentration of 200 μg/ml) and incubated for 2 h at 37 °C. The formazan products were dissolved in DMSO, and the absorbance was measured at 570 nm. The data are presented as a percentage of the control group.
Anchorage-dependent and anchorage-independent colony formation. For the anchoragedependent colony formation assay, cells were plated at a density of 300 cells/well in six-well plates and were treated with increasing concentrations of NCT-50 for 2 weeks. Colonies were fixed with 100% methanol, stained with 0.02% crystal violet solution at room temperature, photographed, and counted.
For the anchorage-independent colony formation assay, cells were mixed with sterile 1% agar solution (final concentration of 0.4%) and then poured onto the 1% base agar in 12-or 24-well plates. After solidification, NCT-50 diluted in complete medium was added to the agar and incubated for 2 weeks. Colonies were stained with the MTT solution, photographed, and counted. Western blot analysis. Cells were treated with increasing concentrations of NCT-50 (0, 1, and 5 μM) for 24 h. When necessary, the cells were further incubated under hypoxic conditions (1% O 2 for 4 h). The cells were harvested with RIPA lysis buffer as described previously 18 . Equal amounts of cell lysates were resolved by 8-15% SDS-PAGE and transferred onto a PVDF membrane. Membranes were blocked with blocking buffer [3% skim milk in TBS containing 0.1% Tween-20 (TBST)] for 1 h at room temperature and incubated with primary antibodies diluted in TBST containing 3% bovine serum albumin (BSA) for overnight at 4 °C. Membranes were washed three times with TBST for 1 h at room temperature and incubated with the corresponding secondary antibodies diluted in 3% skim milk in TBST (1:5000) for 1 h at room temperature. Membranes were washed three times with TBST and visualized using an enhanced chemiluminescence (ECL) detection kit (Thermo Fisher Scientific, Waltham, MA, USA).
SCIenTIfIC REPORTS | (2018) 8:13924 | DOI:10.1038/s41598-018-32196-6 NADH dehydrogenase activity assay. NADH dehydrogenase activity assay was performed according to the previously reports with some modifications 64,65 . Confluent H460 cells in 100 mm culture dishes were washed twice with ice-cold PBS and harvested by scraping in PBS. After centrifugation, cell pellets were suspended in 1 ml ice-cold 10 mM Tris (pH 7.4) and mechanistically disrupted by sonication using Vibra-cell ultrasonic processor (Sonics & Materials Inc., Newtown, CT, USA). After adding 0.2 ml ice-cold 1.5 M sucrose, the suspension was centrifuged at 600 × g for 10 min at 4 °C. Supernatants were further centrifuged at 14,000 × g for 10 min at 4 °C. Pellets were suspended in 100 μl 10 mM Tris (pH 7.4). Protein concentration was determined using BCA assay kit (Thermo Fisher Scientific). A mitochondria enriched fraction (20 μl, containing 20 μg of protein) was incubated with 960 μl of the incubation mixture containing 25 mM potassium phosphate (pH 7.8), 3.5 g/L fatty acid-free bovine serum albumin (BSA), 60 μM 2,6-dichloroindophenol, 70 μM decylubiquinone, and 1 μM antimycin A for 1 min at room temperature. The mixture was treated with vehicle or test compounds (5 μM NCT-50 or deguelin). After measuring the absorbance at 600 nm, 20 μl of 10 mM NADH solution was added, and the absorbance was monitored at 1 min intervals for 10 min. Enzyme activity of the test group was calculated according to the following formula and expressed as a percentage of the vehicle-treated control group; enzyme activity (nmol/min/ mg) = (Δ Absorbance/min × 1,000)/[(extinction coefficient of DCIP × volume of sample used in 1 ml) × (sample protein concentration in mg/ml)] 65 . Extinction coefficient of DCIP was 19.1 mM −1 cm −1 64,65 .
Immunoprecipitation and pull-down assay. Immunoprecipitation, the purification of Hsp90 proteins, and a pull-down assay to determine the competitive binding to the ATP-binding pocket using ATP-agarose were performed according to our previous report 18 . In vivo toxicity test. Animal experiments were performed according to protocols approved by the Seoul National University Institutional Animal Care and Use Committee. Mice were fed standard mouse chow and water ad libitum and housed in temperature-and humidity-controlled facilities with a 12-h light/12-h dark cycle. 6-weeks-old FVB mice were treated with vehicle (10% DMSO in corn oil) or test compounds (deguelin or NCT-50, 4 mg/kg) twice a day for 7 consecutive days. Body weight changes were monitored during the treatment. In vivo toxicity test was perform as described previously 66 . Briefly, blood was collected from euthanized mice by cardiac puncture. Serum was collected by centrifugation at 3,000 rpm for 10 min at 4 °C. The level of GOT, GPT, and BUN in the serum was analyzed using a veterinary hematology analyzer (Fuji DRI-Chem 3500 s, Fujifilm, Tokyo, Japan) according to the manufacturer's recommended procedure. The histopathological changes in liver, lung, brain, and kidney were evaluated by using H&E-stained section of the tissues.
Immunofluorescence. Frozen sections of the mouse midbrain were fixed with 4% paraformaldehyde for 30 min at room temperature. After washing twice with PBS for 5 min each, sections were permeabilized with 0.2% Triton X-100 solution for 15 min at room temperature. Sections were washed three times with PBS for 5 min each. Sections were blocked with blocking buffer (3% BSA solution containing 10% normal donkey serum) for 1 h at room temperature, and then incubated with anti-tyrosine hydroxylase primary antibody (EMD Millipore, Burlington, MA, USA; 1:200 dilution) for overnight at 4 °C. Slides were washed three times with PBS for 10 min each at room temperature. Slides were further incubated with corresponding Alexa 488-conjugated secondary antibody (Thermo Fisher Scientific; 1:500 dilution) for 1 h at room temperature. Slides were washed six times with PBS for 10 min each and then mounted with mounting medium with DAPI. Images were acquired with a confocal microscope (LSM 700; Carl Zeiss Microscopy, Jena, Germany). Docking modeling. All performances of computational works were carried out on the Tripos Sybyl-X 2.1 (Tripos Inc, St Louis, MO, USA) molecular modeling package.
Preparation of ligands. The ligands were prepared as Mol2 format using sketch modules embedded in Sybyl. Gasteiger-Hückel charges were assigned to all ligands atoms. Each ligand was energy-minimized using a standard Tripos force field with convergence to maximum derivatives of 0.001 kcal mol-1.Å-1.
Homology modeling of open conformation of hHsp90 homodimer. We conducted the homology modeling of full-length of human Hsp90α by using ORCHESTRAR module. It was reported that the function of Hsp90 C-terminal inhibitors was related with binding to the open form of homodimer Hsp90 structure 67 . Based on this information, we built up an open conformation of human Hsp90 dimer based on the extended SAXS model of E. coli Hsp90 homodimer including N-, middle and C-terminal domain. The template structure was retrieved from Agard' lab website (http://www.msg.ucsf.edu/agard, PDB id: hsp90). The sequence identity of Hsp90 between E-coli and human is 42.9%. First, alignment file was defined using complete partial model. Then, based on this alignment file, structurally conserved regions (SCRs) were built automatically, and structurally variable region was identified. SCRs including loops were optimized by loop-search option. Side chains of amino acids were fixed with set side chain option. To remove the bad contacts in the protein structure, the brief energy-minimization of model was performed with assigning Kollman all charges to all the atoms of protein and convergence to maximum derivatives of 0.05 kcal mol-1.Å-1. of ligand was calculated by Surflex-Dock score (-log Kd). Based on the docking score and visual inspection, top-ranked docking model was selected.

Data Availability
All data generated or analyzed during this study are included in this published article