The Quinone Methide Aurin Is a Heat Shock Response Inducer That Causes Proteotoxic Stress and Noxa-dependent Apoptosis in Malignant Melanoma Cells*

Background: Pharmacological induction of proteotoxic stress is a promising strategy for anti-melanoma intervention. Results: The quinone methide aurin displays Hsp90α-directed inhibitory activity causing proteotoxic stress and Noxa-dependent apoptosis in malignant melanoma cells. Conclusions: Aurin causes lethal proteotoxic stress in melanoma cells. Significance: Cancer cell proteostasis can be undermined using aurin as a proteotoxic experimental therapeutic. Pharmacological induction of proteotoxic stress is rapidly emerging as a promising strategy for cancer cell-directed chemotherapeutic intervention. Here, we describe the identification of a novel drug-like heat shock response inducer for the therapeutic induction of proteotoxic stress targeting malignant human melanoma cells. Screening a focused library of compounds containing redox-directed electrophilic pharmacophores employing the Stress & Toxicity PathwayFinderTM PCR Array technology as a discovery tool, a drug-like triphenylmethane-derivative (aurin; 4-[bis(p-hydroxyphenyl)methylene]-2,5-cyclohexadien-1-one) was identified as an experimental cell stress modulator that causes (i) heat shock factor transcriptional activation, (ii) up-regulation of heat shock response gene expression (HSPA6, HSPA1A, DNAJB4, HMOX1), (iii) early unfolded protein response signaling (phospho-PERK, phospho-eIF2α, CHOP (CCAAT/enhancer-binding protein homologous protein)), (iv) proteasome impairment with increased protein-ubiquitination, and (v) oxidative stress with glutathione depletion. Fluorescence polarization-based experiments revealed that aurin displays activity as a geldanamycin-competitive Hsp90α-antagonist, a finding further substantiated by molecular docking and ATPase inhibition analysis. Aurin exposure caused caspase-dependent cell death in a panel of human malignant melanoma cells (A375, G361, LOX-IMVI) but not in non-malignant human skin cells (Hs27 fibroblasts, HaCaT keratinocytes, primary melanocytes) undergoing the aurin-induced heat shock response without impairment of viability. Aurin-induced melanoma cell apoptosis depends on Noxa up-regulation as confirmed by siRNA rescue experiments demonstrating that siPMAIP1-based target down-regulation suppresses aurin-induced cell death. Taken together, our data suggest feasibility of apoptotic elimination of malignant melanoma cells using the quinone methide-derived heat shock response inducer aurin.


Pharmacological induction of proteotoxic stress is rapidly emerging as a promising strategy for cancer cell-directed chemotherapeutic intervention. Here, we describe the identification of a novel drug-like heat shock response inducer for the therapeutic induction of proteotoxic stress targeting malignant human melanoma cells. Screening a focused library of compounds containing redox-directed electrophilic pharmacophores employing the Stress & Toxicity PathwayFinder TM PCR Array technology as a discovery tool, a drug-like triphenylmethane-derivative (aurin; 4-[bis(p-hydroxyphenyl)methylene]-2,5-
cyclohexadien-1-one) was identified as an experimental cell stress modulator that causes (i) heat shock factor transcriptional activation, (ii) up-regulation of heat shock response gene expression (HSPA6, HSPA1A, DNAJB4, HMOX1), (iii) early unfolded protein response signaling (phospho-PERK, phospho-eIF2␣, CHOP (CCAAT/enhancer-binding protein homologous protein)), (iv) proteasome impairment with increased proteinubiquitination, and (v) oxidative stress with glutathione depletion. Fluorescence polarization-based experiments revealed that aurin displays activity as a geldanamycin-competitive Hsp90␣-antagonist, a finding further substantiated by molecular docking and ATPase inhibition analysis. Aurin exposure caused caspase-dependent cell death in a panel of human malignant melanoma cells (A375, G361, LOX-IMVI) but not in nonmalignant human skin cells (Hs27 fibroblasts, HaCaT keratinocytes, primary melanocytes) undergoing the aurin-induced heat shock response without impairment of viability. Aurin-induced melanoma cell apoptosis depends on Noxa up-regulation as confirmed by siRNA rescue experiments demonstrating that siPMAIP1-based target down-regulation suppresses aurin-induced cell death. Taken together, our data suggest feasibility of apoptotic elimination of malignant melanoma cells using the quinone methide-derived heat shock response inducer aurin.
Proteotoxic stress (i.e. deviations from proteostasis) occurs in response to cytotoxic stimuli that cause accumulation of unfolded and/or misfolded proteins including heat shock, oxidative stress, calcium dysregulation, and inhibition of proteasomal or autophagic-lysosomal function (1)(2)(3)(4)(5). It is now widely accepted that tumor cells are exposed to high levels of endogenous proteotoxic stress originating from mutation-driven expression of misfolded proteins, aneuploidy, and adverse conditions associated with the tumor microenvironment including hypoxia, energy crisis, and redox dysregulation (6). In contrast, it is thought that normal cells display lower constitutive levels of endogenous proteotoxic stress together with a diminished dependence on proteostasis-ensuring mechanisms including proteasomal and heat shock protein chaperone functions. Therefore, constitutively elevated levels of proteotoxic stress may represent a specific molecular vulnerability of malignant cells amenable to chemotherapeutic intervention (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13).
In an effort to identify novel drug-like molecules that might target malignant melanoma cells through the induction of proteotoxic and/or oxidative stress, we recently screened a focused library of compounds containing redox-directed electrophilic pharmacophores using the Stress & Toxicity PathwayFinder TM PCR Array technology (12, 39 -44). Here, we report for the first time that in a series of drug-like phenolic triphenylmethane derivatives, the quinone methide aurin (CAS #143-74-8; 4-[bis(p-hydroxyphenyl)methylene]-2,5-cyclohexadien-1-one; rosolic acid; see Fig. 1A) was identified as a potent inducer of lethal proteotoxic stress in human A375 and other malignant melanoma cells. Further analysis revealed that aurin displays inhibitory activity targeting Hsp90␣, causing early heat shock factor (HSF) transcriptional activation, up-regulation of heat shock response gene expression, and unfolded protein response (UPR) signaling followed by induction of Noxa-dependent apoptotic cell death.
Human Stress and Toxicity Pathfinder TM RT 2 Profiler PCR Expression Array-After pharmacological exposure, total cellular RNA (3 ϫ 10 6 cells) was prepared according to a standard procedure using the RNeasy kit (Qiagen, Valencia, CA).
Reverse transcription was performed using the RT 2 First Strand kit (Superarray, Frederick, MD) and 5 g of total RNA as described previously (40 -42). The RT 2 Human Stress and Toxicity Pathfinder TM PCR Expression Array (SuperArray) profiling the expression of 84 stress-related genes was run using the following PCR conditions: 95°C for 10 min followed by 40 cycles of 95°C for 15 s alternating with 60°C for 1 min (Applied Biosystems 7000 SDS). Gene-specific product was normalized to ACTB and quantified using the comparative (⌬⌬Ct) Ct method as described in the ABI Prism 7000 sequence detection system user guide as published earlier (42,45). Expression values were averaged across three independent array experiments, and standard deviation was calculated for graphing.
DDIT3, HSPA1A, HSPA6, HMOX1, and PMAIP1 Expression Analysis by Real Time RT-PCR-For expression analysis by real time RT-PCR, total cellular RNA (3 ϫ 10 6 cells) was prepared using the RNeasy kit from Qiagen. Reverse transcription was performed using TaqMan Reverse Transcription Reagents (Roche Molecular Biochemicals) and 200 ng of total RNA in a 50-l reaction. Reverse transcription was primed with random hexamers and incubated at 25°C for 10 min followed by 48°C for 30 min, 95°C for 5 min, and a chill at 4°C. Each PCR reaction consisted of 3.75 l of cDNA added to 12.5 l of TaqMan Universal PCR Master Mix (Roche Molecular Biochemicals), 1.25 l of gene-specific primer/probe mix (Assays-by-Design; Applied Biosystems, Foster City, CA), and 7.5 l of PCR water. PCR conditions were 95°C for 10 min followed by 40 cycles of 95°C for 15 s, alternating with 60°C for 1 min using an Applied Biosystems 7000 SDS and Applied Biosystems Assays On Demand primers specific to DDIT3 (assay ID Hs00358796_g1), HSPA6 (assay ID Hs00275682_s1), HSPA1A (assay ID Hs00359163_s1), HMOX1 (assay ID Hs00157965_m1), PMAIP1 (assay ID Hs00560402_m1), and ACTB (␤-actin, assay ID Hs99999903_m1). Gene-specific product was normalized to ACTB and quantified using the comparative (⌬⌬C t ) Ct method as described before (42,45).
Heat Shock Element (HSE)-luciferase Reporter Assay-Heat shock factor-responsive TransLucent TM reporter vector, pHSE-Luc, and empty TransLucent control reporter vector were purchased from Panomics (Fremont, CA). pHSE-Luc contains multiple copies of the heat shock factor binding cis-acting promotor element (HSE; 5Ј-CTGGAATTTTCTAGACTGGA-ATTTTCTAGACTGGAATTTTCTAGA-3Ј) upstream of the herpes simplex virus thymidine kinase TATA box promoter driving luciferase gene expression. A375 human melanoma cells were grown to 60% confluency in a 24-well plate. For transfection, 0.4 g of HSF-responsive TransLucent reporter vector, 0.8 ng of promoter-less Renilla reporter vector (phRG-B; Promega, Madison, WI), 1.32 l of FuGENE 6 (Roche Molecular Biochemicals), and 20.68 l of Opti-MEM I (Invitrogen) were added per well. Twenty-four hours after transfection cells were exposed to test the compounds for 8 h. Cells were then lysed in 100 l of passive lysis buffer using the Dual-Luciferase reporter assay system (Promega) according to the manufacturer's instructions. Luciferase activity was measured with a Turner Designs TD20/20 luminometer. Data are reported as normalized averages of the firefly/renilla luciferase ratio.
Cell Death Analysis-Viability and induction of cell death (early and late apoptosis/necrosis) were examined by AV-FITC/PI dual staining of cells followed by flow cytometric analysis as published previously (40). Cells (100,000) were seeded on 35-mm dishes and received drug treatment 24 h later. Cells were harvested at various time points after treatment, and cell staining was performed using an apoptosis detection kit according to the manufacturer's specifications (APO-AF, Sigma).
Proteasome Activity Assay-The Proteasome-Glo TM Cell-Based Assay kit (Promega) was used according to manufacturer's specifications. Cells were seeded at 4000 cells/well of an opaque 96-well plate. After 24 h cells were treated with test compound. At the end of the treatment period, cells were washed with PBS and an equal volume of reagent containing the appropriate luminogenic substrate (Suc-LLVY-aminoluciferin for chymotrypsin-like activity) was added to each well. Plates were incubated at room temperature (10 min) before measuring luminescence signal using a Synergy 2 microplate reader. Data are expressed as % activity relative to untreated control cells (means Ϯ S.D. (n ϭ 3)). Likewise, inhibition of 20 S chymotrypsin-like proteasomal activity was examined using the isolated human erythrocyte 20 S proteasome obtained from a commercial kit (EMD Millipore).
Detection of Intracellular Oxidative Stress by Flow Cytometric Analysis-Induction of intracellular oxidative stress by test compound was analyzed by flow cytometry using 2Ј,7Ј-dichlorodihydrofluorescein diacetate (DCFH-DA) as a sensitive nonfluorescent precursor dye according to a published standard procedure (46).
Determination of Reduced Cellular Glutathione Content-Intracellular reduced glutathione was measured using the GSH-Glo glutathione assay kit (Promega, San Luis Obispo, CA) as described recently (41). Cells were seeded at 100,000 cells/ dish on 35-mm dishes. After 24 h, cells were treated with test compound. At selected time points after the addition of test compound, cells were harvested by trypsinization and then counted using a Coulter counter. Cells were washed in PBS, and 10,000 cells/well (50 l) were transferred onto a 96-well plate. GSH-Glo reagent (50 l) containing luciferin-NT and glutathione S-transferase was then added followed by 30 min of incubation. After the addition of luciferin detection reagent to each well (100 l) and 15 min of incubation, luminescence reading was performed using a BioTek Synergy 2 Reader (BioTek, Winooski, VT). Data are normalized to GSH content in untreated cells and expressed as the means Ϯ S.D. (n ϭ 3).
Fluorescence intensity was then converted to fluorescence polarization using the Tecan Magellan6 software, and the fluorescence polarization data were analyzed using the computer software Graphpad Prism. The fluorescence polarization (FP t ) in the absence of the compound in each data set was defined as 100% activity. In the absence of protein and test compound, the value of fluorescence polarization (FP b ) in each data set was defined as 0% activity. The % activity in the presence of the test compound was calculated according to the equation, where FP ϭ fluorescence polarization in the presence of the compound.
For IC 50 measurements, the values of % activity versus a series of compound concentrations were then plotted using non-linear regression analysis of a sigmoidal dose-response curve generated with the equation, where Y ϭ % activity, B ϭ minimum % activity, T ϭ maximum % activity, X ϭ logarithm of compound, and Hill Slope ϭ slope factor or Hill coefficient. The IC 50 value was defined as the compound concentration causing half-maximal % activity.
Hsp90 ATPase Activity Assay-Inhibition of Hsp90 ATPase enzymatic activity by test compounds was assessed in 96-well plate format as described previously (58). This assay determines the amount of free phosphate generated due to enzymatic ATP hydrolysis and is based on the reaction of a phosphomolybdate complex with malachite green. Per reaction, human Hsp90␣ (10 g of pure full-length recombinant protein from Escherichia coli (ab48801; Abcam, Cambridge, MA)) was incubated with test compound in a buffer containing 50 mM Hepes (pH 7.5), 6 mM MgCl 2 , 20 mM KCl, 1 mM ATP (50 l total volume; 6 h, 25°C). Using the malachite green phosphate assay kit from Cayman Chemical (Ann Arbor, MI), the reaction was terminated by the addition of 5 l of malachite green acidic solution and 15 l of MG blue solution to each well. After 25 min at room temperature, the absorbance at 620 nm was determined. Data are expressed as % ATPase activity relative to Hsp90␣ incubated with solvent control (means Ϯ S.D. (n ϭ 3)).
Molecular Modeling-Docking analysis was performed using Glide software ("grid-based ligand docking with energetics"; available from Schrodinger Inc.). X-ray crystal structure coordinates of geldanamycin with HSP90␣ (Protein Data Bank code 1YET) were used for docking (49). Glide-generated ligand poses pass through a series of hierarchical filters that evaluate the ligand interaction with the target. Initial filters test the spatial fit between ligand and target examining complementarity of ligand-target interactions using a grid-based method patterned after the empirical ChemScore function. Final scoring is carried out on energy-minimized poses; Glide scores (G scores) approximating the ligand binding free energy were calculated for aurin and geldanamycin with the Hsp90␣ target protein (50).
Statistical Analysis-Unless indicated differently, the results are presented as the mean Ϯ S.D. of at least three independent experiments, and data were analyzed employing one-way analysis of variance with Tukey's post hoc test using the Prism 4.0 software. Differences were considered significant at p Ͻ 0.05. Means without a common letter differ (p Ͻ 0.05).
For comparison, an analogous expression array profile was generated using the established heat shock response inducer celastrol (10 M, 6 h exposure; Fig. 2C), It was observed that even though up-regulation of Hsp70 subtype encoding genes (HSPA1A, HSPA6) was a common feature of aurin-and celastrol-induced expression changes, the transcriptional profile induced by celastrol treatment differed significantly from that elicited by aurin and was characterized by pronounced up-regulation of proinflammatory genes including PTGS1 (ϳ170fold), CCL3 (ϳ68-fold), CXCL10 (ϳ7-fold), and TNFSF10 (ϳ64-fold).
Next, the dose-response relationship and time course of aurin-induced effects on the expression of HSPA1A and other heat shock response genes (mRNA and protein level) were examined using independent real time RT-PCR and immunoblot analysis (Fig. 3, A and C). Pronounced up-regulation of HSPA1A (ϳ28-fold), HSPA6 (ϳ150-fold), and HMOX1 (ϳ55fold) was observed at the mRNA level (Fig. 3A), and Hsp70, Hsp27, and heme oxygenase 1 (HO-1) protein levels were increased within 6 -12 h of continuous exposure to aurin at low micromolar concentrations (Fig. 3, B and C). In contrast, expression of Hsp60 and Hsp90 (encoded by HSPD1 and HSP90AA1, respectively) remained unaffected by aurin exposure both at the mRNA (Fig. 2, A and B) and protein (Figs. 3B and 6A) levels. Importantly, these changes affecting expression of HSPA1A, HSPA6, and HMOX1 were not observed using ATA (Fig. 3, F and G) confirming the results of our initial screen (Fig. 1, B and C).
To further elucidate the molecular mechanisms underlying aurin-induced heat shock response up-regulation, aurin-modulation of HSF transcriptional activity was assessed in A375 cells transfected with an HSE-luciferase reporter construct employing a dual luciferase reporter assay (Fig. 3D). Indeed, aurin exposure caused a pronounced (ϳ9-fold) activation of firefly luciferase expression indicative of HSF1 transcriptional activation. The established heat shock response inducer celastrol was employed as a positive control (54 -57). Immunoblot analysis (comparing cytosolic and nuclear fractions) indicated that aurin-induced HSF1 nuclear translocation occurs within 10 min exposure time (Fig. 3E).
Aurin Induces UPR Signaling, Proteasomal Impairment, and Oxidative Stress in Human A375 Malignant Melanoma Cells-Led by HSF1 nuclear translocation observed within minutes of aurin exposure followed by early up-regulation of proteotoxic stress and heat shock response gene expression (HSPA6, HSPA1A, HSPA1L, DNAJB4, CRYAB, HSPB1, HSPH1, DNAJA1, and DDIT3) as observed by array analysis (Fig. 2), we further examined the occurrence of UPR stress signaling in aurin-treated A375 melanoma cells (Fig. 5, A and B). Time course analysis by quantitative RT-PCR revealed early up-regulation of DDIT3 expression encoding CHOP (CCAAT/enhancer-binding protein homologous protein), the ER stressinducible transcription factor (Fig. 5A). DDIT3 mRNA levels exceeded control by almost 7-fold within 6 h of exposure. Immunoblot detection indicated that up-regulation of CHOP protein levels occurred within 1-3 h of continuous exposure, reaching a maximum within 12 h (Fig. 5B). As a positive control, CHOP protein up-regulation was also monitored in response to thapsigargin (T), an established inducer of ER stress and UPR through SERCA inhibition (sarco/endoplasmic reticulum)-FIGURE 2. Gene expression array analysis performed in human A375 malignant melanoma cells exposed to aurin or celastrol. A, the scatter blot depicts differential gene expression as detected by the RT 2 Human Stress and Toxicity Profiler TM PCR Expression Array technology (aurin, 10 M, 6 h exposure). Upper and lower lines, cut-off indicating 4-fold up-or down-regulated expression, respectively. Arrays were performed in three independent repeats and analyzed using the two-sided Student's t test. B, numerical expression changes (p Ͻ 0.05) elicited by aurin. C, for comparison, an analogous experiment was performed using the established heat shock response inducer celastrol (10 M, 6-h exposure), and celastrol-induced numerical expression changes are listed (p Ͻ 0.05). Ca 2ϩ ATPase (Fig. 5B) (39). Inhibitory phosphorylation of eIF2␣ (eukaryotic translation initiation factor) through activation of PERK (double-stranded RNA-activated protein kinase (PKR)-like endoplasmic reticulum kinase) is an established hallmark of cytotoxic ER stress known to occur upstream of DDIT3/CHOP up-regulation. We, therefore, examined aurininduced changes affecting phosphorylation status of eIF2␣ and PERK (Fig. 5B). Consistent with aurin induction of cytotoxic ER stress, pronounced PERK activation (autophosphorylation of its cytoplasmic kinase domain (Thr-980)) together with inhibitory phosphorylation of eIF2␣ (Ser-51) was detectable within 6 h of exposure and was sustained over the 24-h observation period. Likewise, inhibitory phosphorylation of eIF2␣ was detectable over a 24-h time course (Fig. 5B).
After confirming aurin induction of ER stress signaling, we examined aurin-induced changes in protein ubiquitination sta-tus and inhibition of proteasomal enzymatic activity (Fig. 5, C  and D). Consistent with aurin induction of a pronounced proteotoxic insult in A375 cells, elevated levels of ubiquitinated proteins were detectable within 1-3 h of exposure, further increasing over a 12-h observation time (Fig. 5C). Moreover, using a luminescent protease assay that measures proteasomedependent release of luciferin from specific precursor substrate peptides, we observed that aurin displays activity as an inhibitor of chymotrypsin-like proteasomal function. Specifically, chymotrypsin-like activity in cytosolic extracts prepared from aurin-treated A375 cells was impaired by Ͼ20% of control levels within 6 h of exposure. However, no impairment of proteasomal activity was detected at earlier time points (3 h), suggesting that HSF1 nuclear translocation, heat shock response gene expression, and UPR-related gene expression changes including DDIT3 up-regulation detect- able within minutes to hours of aurin exposure precede proteasomal impairment.
Next, we examined the occurrence of oxidative stress in aurin-exposed A375 malignant melanoma cells using flow cytometric determination of 2Ј,7Ј-dichlorodihydrofluorescein (DCF) fluorescence and luminometric detection of reduced glutathione levels (Fig. 5, E and F). Aurin treatment was associated with the induction of pronounced oxidative stress as obvious from significant 2Ј,7Ј-dichlorodihydrofluorescein fluorescence increase (almost 10-fold more than control levels) and depletion of reduced glutathione (by ϳ35%) detectable within 12-18 h of exposure time. The fact that aurin-induced redox dysregulation was detectable only after HSF1 nuclear translocation (Fig. 3E), transcriptional activation (Fig. 3D), and up-regulation of early heat shock response gene expression (Fig. 2, A and B) have already occurred suggests that oxidative stress does not represent the upstream event that triggers aurin-induced stress responses. However, consistent with a mechanistic involvement of redox dysregulation in the apoptotic mode and cytotoxic potency of aurin, it was observed that cellular glutathione depletion employing L-buthionine-S,R-sulfoximine (BSO; 1 mM, 24 h of pretreatment) strongly sensitized A375 cells to the apoptogenic activity of aurin (Fig. 5G).

Fluorescence Polarization and Molecular Modeling Identify Aurin as a Geldanamycin-competitive Inhibitor of Hsp90 Causing Down-regulation of Hsp90 Client Proteins in A375
Malignant Melanoma Cells-It has been shown earlier that heat shock response gene expression and UPR may be up-regulated in response to pharmacological inhibition of Hsp90, an established pathway leading to heat shock response gene expression as observed with the Hsp90 inhibitor geldanamycin (Fig. 1, B and C) (59 -63). Specifically, up-regulation of HSPA1A and other heat shock response genes downstream of Hsp90 antagonism may represent a counter-regulatory mechanism involving the known repression of heat shock transcription factor (HSF1) activation by Hsp90 that forms a stress-sensitive complex with HSF1 (59 -62). We, therefore, tested the hypothesis that aurin induction of heat shock gene expression and UPR may occur in part as a result of Hsp90-directed inhibitory activity.
First, we examined the possibility that aurin-induced impairment of Hsp90 function may cause destabilization and subsequent down-regulation of established Hsp90 client proteins in melanoma cells. It has recently been shown that V600E B-Raf (as expressed in A375 malignant melanoma cells) requires Hsp90 chaperone function for stability and that V600E B-Raf is degraded in response to Hsp90 inhibitor treatment (34,37). Likewise, Hsp90 inhibitors suppress aryl hydrocarbon receptor (AhR)-mediated activation of the AhR target gene CYP1A1 (63). Indeed, immunoblot analysis revealed a pronounced down-regulation of these Hsp90 client proteins in A375 melanoma cells exposed to aurin (10 M; Յ24 h; Fig. 6A). Consistent with pronounced reduction of cellular AhR protein levels, we also observed that aurin treatment suppressed expression of CYP1A1 at the mRNA level observable within 3 h of exposure time (Fig. 6B). This finding is in agreement with our initial expression array-based identification of CYP1A1 as the gene displaying the most pronounced down-regulation in response to aurin treatment in A375 melanoma cells (Fig. 2, A and B).
Next, using a validated commercial fluorescence polarization assay platform, we directly examined the possibility that aurin displays activity as a competitive Hsp90␣ N-terminal inhibitor. Based on the competition between a fluorescently labeled high affinity probe derived from the standard Hsp90 inhibitor geldanamycin (FITC-geldanamycin) and the test compound for binding to human recombinant Hsp90␣, reduction in FITCgeldanamycin fluorescence polarization was assessed to quantify competitive target binding (47,48). Measuring displacement of the geldanamycin-FITC/Hsp90␣ interaction over a broad concentration range of aurin versus geldanamycin (positive control) (Fig. 6, C and D), aurin displayed a significant geldanamycin-competitive Hsp90␣ target affinity (geldanamycin, 22.0 Ϯ 2.3 nM; aurin, 574.0 Ϯ 15.2 nM; IC 50 values (inhibitory activity on Hsp90␣; data are the mean Ϯ S.E.; n ϭ 2)). However, Hsp90␣-directed ligand activity of aurin was ϳ25fold smaller than that displayed by geldanamycin.
Further experimental evidence in support of an Hsp90-directed antagonistic activity of aurin was obtained by assessment of aurin-induced inhibition of ATPase enzymatic activity of FIGURE 6. Aurin as a structure-based, geldanamycin-competitive inhibitor of Hsp90; down-regulation of Hsp90 client proteins in aurin-exposed A375 malignant melanoma cells. A, immunoblot detection of Hsp90 client proteins and Hsp90 in A375 cells exposed to aurin (10 M; Յ24 h). Immunoblot analysis representative of three independent repeats is depicted. B, expression of the AhR target gene CYP1A1 induced by aurin treatment (10 M, Յ24 h) assessed at the mRNA level (quantitative RT-PCR; mean Ϯ S.D.; n ϭ 3). C and D, Hsp90␣ inhibition as assessed by fluorescence polarization. A competitive binding assay employing FITC-labeled geldanamycin and purified recombinant Hsp90␣ was performed comparing inhibitory activity of geldanamycin (panel C) and aurin (panel D). E (inset in D), aurin (Ͻ40 M)and geldanamycin (40 M)-induced inhibition of ATPase activity of recombinant human Hsp90␣ was assessed using the malachite green assay. Data are expressed as % ATPase activity relative to Hsp90␣ incubated with solvent control (means Ϯ S.D. (n ϭ 3)). F, molecular docking of aurin with HSP90␣ x-ray structure (ribbon view). Aurin interacts with the HSP90␣ active site formed by ␤-sheets at the bottom and ␣-helices H2, H4, and H7. G, close view of aurin docked with the HSP90␣ active site. Yellow lines indicate hydrogen bond interactions. Phenolic -OH groups of aurin engage in hydrogen bonding interactions with Asp-93 carboxylate and Asn-106 backbone carbonyl groups. The quinone carbonyl group forms a hydrogen bond with Lys-58. H, close view of geldanamycin docked with the HSP90␣ active site. Interactions with Asp-93, Lys-58, and Asn-106 are depicted. recombinant human Hsp90␣ (Fig. 6E) (58). Using the malachite green assay for the photometric detection of phosphate generated from enzymatic ATP hydrolysis, it was observed that low micromolar concentrations of aurin (10 -40 M) were sufficient to cause an almost 40% reduction in Hsp90␣ ATPase activity. Under our assay conditions, aurin surpassed the inhibitory activity of geldanamycin (ϳ20% inhibition of ATPase activity at 40 M), an observation that is consistent with the known dependence of geldanamycin and its derivatives on NQO1/NADH-mediated bioreductive activation for full inhibitory activity (58).
A computational strategy was then used to explore the potential molecular basis of aurin-induced target modulation by modeling ligand binding of aurin to Hsp90␣. It is well established that Hsp90 chaperone function depends on the conformational changes driven by its ATPase activity. Binding of the ansamycin antibiotic geldanamycin to the N-terminal ATP pocket restrains Hsp90 in its ADP-bound conformation, preventing the subsequent "clamping" of Hsp90 around a client protein (49,64). The N-terminal geldanamycin binding domain of Hsp90 (amino acid residues 9 -232) reveals a pronounced pocket that is highly conserved across species representing an ATP/ADP switch domain regulating Hsp90 conformation, and it has been demonstrated that geldanamycin adopts a compact structure inside this pocket, thereby competing with substrate binding and blocking the conformational maturation and refolding reaction of Hsp90 client proteins (49). To assess geldanamycin-competitive structure-based target inhibition, molecular modeling was performed based on the x-ray crystal structure coordinates of geldanamycin with HSP90␣ (Protein Data Bank code 1YET) employing the Glide docking methodology, an empirical scoring function that approximates the ligand binding free energy (Fig. 6, F-H) (49,50). Glide score (G score) analysis indicated that aurin displays significant binding affinity toward Hsp90␣, a finding consistent with the fluorescence polarization-based observation that aurin may operate as a structure-based Hsp90␣ inhibitor. A common hallmark of known Hsp90 N-terminal domain inhibitors is their ability to engage in molecular interaction with Asp-93, conserved throughout all known Hsp90 homologs from at least 35 species, positioned at the bottom of a mostly hydrophobic pocket. Indeed, comparative Glide docking revealed that aurin may interact favorably with the Hsp90␣ active site (formed by antiparallel ␤-sheets S3, S4, and S7 at the bottom surrounded by a-helices H2, H4, and H7) through hydrogen bonding interactions involving Asp-93, Asn-106, and Lys-58, residues also engaged in geldanamycin binding (G score ϭ Ϫ3.8 versus G score ϭ Ϫ5.3, respectively; Fig. 6, F-H).
Aurin-induced Melanoma Cell Apoptosis Depends on Upregulated PMAIP1 Gene Expression-Next, we aimed at elucidating details of the molecular pathway underlying aurin-induced A375 melanoma cell death. To this end, viability was monitored using flow cytometric analysis of annexin V-PIstained cells defining the dose-response relationship and time course of aurin-induced cell death (Fig. 7, A and B). Remarkably, viability of A375 cells exposed to aurin (10 M, 24 h) was maintained upon cotreatment with zVADfmk but not with the caspase 8-specific inhibitor IETD-CHO pointing toward a causative involvement of the intrinsic (mitochondrial) pathway of apoptosis (Figs. 3H and 7C). We also observed that pretreatment using the thiol antioxidant and glutathione precursor N-acetyl-L-cysteine (10 mM, 24 h) rescued cells from aurin-induced apoptosis, an observation consistent with the prior observation that glutathione-depletion caused sensitization to aurin-induced cell death (Fig. 5, F and G). Aurin-induced proteolytic procaspase-3 activation was observable within 12 h of continuous exposure (Fig. 7D), a time point at which a significant disruption of the mitochondrial transmembrane potential (⌬m) became apparent (Fig. 7E). In the context of a possible mitochondrial involvement in aurin-induced apoptosis, we then focused on the up-regulated expression of the proapoptotic gene PMAIP1 encoding the BH3-only Bcl2 family member Noxa as detected earlier at the mRNA level by expression array profiling (Fig. 2, A and B), confirmed subsequently by a more detailed time course analysis (Fig. 7F). Due to the established causative involvement of PMAIP1 up-regulation in cancer cell mitochondrial apoptosis that occurs in response to various genotoxic, proteotoxic, and oxidative stress stimuli (39,41,44,65,66,68), a genetic target modulation approach employing siRNA-based antagonism of PMAIP1 expression was used to further substantiate the causative involvement of Noxa in aurin-induced melanoma cell apoptosis. First, immunoblot analysis confirmed that aurin-induced up-regulation of Noxa protein levels (detectable within 12 h of exposure time) was fully suppressed in siPMAIP1-cotreated A375 melanoma cells (Fig. 7G). Strikingly, siPMAIP1-based suppression of Noxa expression protected A375 melanoma cells against aurin-induced (10 M; 24 h) apoptosis, a specific rescue effect that was not observed in siControl-treated cells exposed to aurin (Fig.  7H).
Here, we report the identification of aurin, a phenolic triphenylmethane-derived quinone-methide, as an apoptogenic inducer of proteotoxic and oxidative stress in human cultured malignant melanoma cells. Our data indicate that in A375 malignant melanoma cells, aurin treatment causes (i) early nuclear translocation and transcriptional activation of HSF1 (Fig. 3, D and E), (ii) up-regulation of heat shock response gene expression (Figs. 1, B and C, and 2, A and B), (iii) UPR induction (phosphorylation of PERK and eIF2␣, up-regulation of DDIT3/ CHOP; Fig. 5, A and B), (iv) proteasome inhibition with accumulation of ubiquitinylated proteins (Fig. 5, C and D), (v) downregulation of Hsp90 client proteins including BRAF and AhR (Fig. 6A), (vi) oxidative stress with glutathione depletion (Fig. 5, E-G), and (vii) induction of caspase-dependent melanoma cell death downstream of PMAIP1 (Noxa) up-regulation and impairment of mitochondrial transmembrane potential. Furthermore, fluorescence polarization-based experimentation revealed that aurin displays activity as a geldanamycin-competitive Hsp90␣ antagonist (Fig. 6, C and D), a finding further substantiated by aurin inhibition of ATPase enzymatic activity of recombinant human Hsp90␣ (Fig. 6E) and molecular docking analysis (Fig. 6, F-H). Consistent with Hsp90-directed antagonism, we observed that aurin treatment, although not affecting Hsp90 protein levels (Fig. 6A), caused pronounced down-regulation of Hsp90 client proteins including V600E B-RAF and AhR, changes observed in conjunction with early transcriptional down-regulation of the AhR target gene CYP1A1 (Fig. 6, A and B) as already indicated by array analysis (Fig. 2, A  and B) (34,37,63). Consistent with an aurin-induced up-regulation of heat shock response gene expression downstream of  , n ϭ 3)). F, time course of PMAIP1 expression changes induced by aurin (10 M) at the mRNA level (quantitative RT-PCR; n ϭ 3). G, immunoblot analysis of Noxa protein levels in aurin (10 M; Յ24 h) exposed cells with siControl or siPMAIP1 cotreatment. H, viability of aurin-exposed (10 M; 24 h) cells was monitored as a function of siControl or siPMAIP1 cotreatment using flow cytometric analysis (annexin V-PI staining; mean Ϯ S.D., n ϭ 3).
Hsp90 inhibition (Fig. 1, C and D), earlier experiments have demonstrated that Hsp90-directed pharmacological inhibitors up-regulate mRNA and protein levels of heat shock response genes by inhibiting the association of Hsp90 with HSF1 causing nuclear translocation and increased transcriptional activity of the latter (59 -62).
Earlier research has demonstrated that Hsp90 antagonism, proteasome inhibition, and UPR induction can all be associated with redox dysregulation and oxidative stress occurring downstream of the initial proteotoxic event (8,66). Indeed, our data indicate that redox dysregulation represents an additional factor involved in aurin-induced apoptosis as evident from antioxidant rescue (using N-acetyl-L-cysteine (NAC), Fig. 7C) and cell sensitization caused by pharmacological glutathione depletion (using L-buthionine-sulfoximine, Fig. 5G). Importantly, a detailed time course analysis indicated that induction of proteotoxic stress (as evident from HSF1 nuclear translocation observable within 10 min of aurin exposure (Fig. 3E), heat shock response gene expression (Figs. 1, B and C, and 2, A and B), DDIT3/CHOP up-regulation (Fig. 5, A and B), and protein ubiquitination (Fig. 3C)) precedes loss of proteasomal enzymatic activity (Fig. 3D), redox dysregulation (Fig. 5, E and F), and mitochondrial impairment (Fig. 7E) detectable only at later time points. This suggests that it is the Hsp90-directed inhibitory activity of aurin that is the upstream factor causing the induction of a cytotoxic heat shock response in malignant melanoma cells. It should also be mentioned that prior research has documented the specific mitochondriotoxic activity of cationic aniline-triphenylmethane dyes including crystal (gentian) violet that target the mitochondrion based on cationic lipophilic properties (71,72). However, at physiological pH aurin (rosolic acid) exists as an anionic phenolate devoid of mitochondrial tropism (70), and our data indicate that early proteotoxic effects of aurin precede loss of mitochondrial transmembrane potential.
However, more detailed follow up experimentation must aim at elucidating the mechanistic interplay between Hsp90-, proteasome-, and redox-directed effects elicited by aurin in A375 malignant melanoma cells. Importantly, our data identify aurin-induced PMAIP1 up-regulation (Figs. 2, A and B and 7, F-H) as a key factor underlying Noxa-dependent apoptosis in A375 melanoma cells. A tentative mechanistic scheme presenting the suggested molecular mechanism underlying induction of lethal proteotoxic stress in aurin-exposed malignant melanoma cells is presented in Fig. 8.
Remarkably, in non-malignant human skin cells (Hs27 fibroblasts, HaCaT keratinocytes, primary melanocytes) an aurininduced heat shock response (as indicated by HSPA1A up-regulation) occurred without impairment of viability (Fig. 4, D-L), an observation consistent with the emerging concept that normal cells display lower constitutive levels of endogenous proteotoxic stress together with a diminished dependence on proteostasis-ensuring mechanisms including proteasomal and heat shock protein chaperone functions pointing toward a molecular vulnerability of malignant cells amenable to therapeutic intervention as proposed before (4 -6, 39, 40). It is interesting to note that one prior publication has reported the cytoprotective activity of aurin observed in cultured human endothelial cells where aurin caused up-regulation of heme oxygenase 1 (HO-1) without loss of cell viability (70), consistent with our observation that in non-malignant human cultured cells aurin treatment up-regulates the heat shock response without inducing cell death (Fig. 4, D-L).
Due to its essential role in the stabilization of oncogenic client proteins and its emerging role in the suppression of mitochondrial apoptosis, the molecular chaperone Hsp90 represents a valid molecular target for chemotherapeutic intervention (64). Small molecule inhibitors of Hsp90 are, therefore, an important class of experimental and investigational chemotherapeutics that have reached the stage of clinical testing. Hsp90-directed inhibitory activity of aurin seems to be of particular relevance in the context of melanoma chemotherapy, as Hsp90 inhibition has emerged as a particularly promising therapeutic strategy based on the finding that V600E B-Raf is an obligatory Hsp90 client protein essential to the clonal evolution of melanoma tumors and that Hsp90 inhibitors cause the degradation of V600E B-Raf (34,37,63). Moreover, it has now been observed that V600E BRAF melanoma cells that have acquired resistance to vemurafenib due to a variety of molecular changes retain sensitivity to Hsp90 inhibition suggesting feasibility of FIGURE 8. Suggested molecular mechanism underlying induction of lethal proteotoxic stress in aurin-exposed malignant melanoma cells. Aurin induces rapid Hsp90 inhibition causing HSF1 nuclear translocation (detectable within minutes of exposure) and pronounced activation of heat shock response gene expression. Together with Hsp90 antagonism, destabilization of Hsp90 client proteins (e.g. oncogenic BRAF and AhR), UPR induction characterized by phosphorylation of PERK and eIF2␣ and up-regulation of DDIT3 (CHOP), and proteasomal impairment with accumulation of ubiquitinated proteins contribute to aurin-induced proteotoxic stress. Downstream of proteotoxic stress induction (with unknown involvement of transcription factors upstream of PMAIP1 such as ATF4 and p53) Noxa up-regulation facilitates the mitochondrial pathway of apoptosis associated with breakdown of mitochondrial transmembrane potential, induction of massive oxidative stress and glutathione depletion, and ultimately initiation of caspase-dependent cell death. Cell rescue as demonstrated in this study can occur by siRNA-based suppression of PMAIP1 up-regulation, antioxidant intervention (NAC), and pancaspase inhibition (zVADfmk). The molecular mechanism underlying the crucial transition from aurin-induced proteotoxic stress to apoptogenic lethal execution remains to be elucidated.
Based on the results of our prototype fluorescence polarization and ATPase enzymatic activity assays combined with Glide molecular docking analysis, aurin was identified as a novel structure-based geldanamycin-competitive Hsp90 antagonist (Fig. 6). Future research efforts must explore the structure activity relationship underlying aurin-induced Hsp90-directed proteotoxic stress. Interestingly, aurin contains a quinone methide functional group reminiscent of the triterpenoid celastrol, a reference heat shock inducer and Hsp90 inhibitory agent containing an essential quinone methide pharmacophore. However, our comparative transcriptional profiling analysis revealed that apart from the common induction of Hsp70-directed gene expression changes aurin (Fig. 2, A and  B)and celastrol-induced (Fig. 2C) responses differed significantly in A375 melanoma cells. Interestingly, in contrast to aurin targeting the N-terminal ATP pocket, it has been demonstrated that celastrol displays Hsp90 C-terminal domain-directed antagonistic activity inhibiting Hsp90 interaction with its cochaperone, cdc37 (67). Remarkably, in addition to Hsp90directed antagonistic activity, proteasome-directed inhibitory activity and depletion of cellular thiols are thought to be involved in celastrol-induced cancer cell apoptosis, suggesting a mechanistic overlap between celastrol-induced proteotoxic and oxidative stress pathways reminiscent of aurin-induced changes observed by us in malignant melanoma cells (54 -57). Interestingly, Noxa-induced apoptosis downstream of ER stress causing PERK/phospho-eIF2␣/ATF4-dependent PMAIP1 upregulation has been substantiated earlier in celastrol-exposed hepatocellular carcinoma cells (65). In this context, it will be interesting to examine how Hsp90-directed antagonistic activity synergizes with proteasomal impairment, UPR signaling (resulting in up-regulated expression of the pro-apoptotic transcription factor CHOP), and oxidative stress, all of which are potentially critical factors contributing to aurin-induced Noxadependent melanoma cell apoptosis (11,39,65,66,68).
In the context of our preclinical study examining feasibility of experimental intervention targeting malignant melanoma cells using the small molecule drug-like quinone-methide aurin, it should be mentioned that prior preclinical and clinical developmental studies have focused on the pharmacological activity of the tricarboxylated aurin derivative ATA (Fig. 1A), a promiscuous experimental therapeutic acting as an inhibitor of protein-protein and DNA-protein interactions (51)(52)(53)69). However, ATA was not active in our HSE-luciferase reporter (Fig.  3D), mRNA-(HSPA1A, HSPA6, HMOX1; Figs. 1B and 3F), and protein-directed (Figs. 1C and 3G) assays examining heat shock response induction. Based on the prototype evidence presented above, our ongoing preclinical experimentation aims at testing efficacy of aurin-induced proteotoxic stress and apoptosis for therapeutic intervention in relevant murine xenograft models of human malignant melanoma.