Isolation of a Novel Thioflavin S–Derived Compound That Inhibits BAG-1–Mediated Protein Interactions and Targets BRAF Inhibitor–Resistant Cell Lines

RAF kinases are part of an evolutionarily conserved core-signaling cascade downstream of activated receptor tyrosine kinases and the small G-protein RAS (1). The three RAF isoforms, ARAF, BRAF, and CRAF (also RAF-1), are involved in mitogen, survival, and differentiation signaling. Their activation is a complex process involving interactions with proteins and lipids as well as phosphorylation and requires the assembly of signalosomes, which, along with the core components of the signaling cascade, contain accessory molecules required for modulation, compartmentalization, and specificity (1–3). RAF kinases are upstream of a three-tiered mitogen–activated protein kinase (MAPK) cascade comprising the dual-specificity Ser/Thr MAPK kinases (MAP2K) MEK1/2 and their substrates ERK1/2, which are required for most reported RAF effects (4, 5).

Several proteins have been described, which may be important for RAF activation and signaling under specific conditions at distinct cellular sites. These include the Bcl-2–associated athanogene 1 (BAG-1; refs. 6–10). BAG-1 was originally identified as a Bcl-2–interacting protein with antiapoptotic activity (11). BAG-1 is a member of a family of proteins characterized by at least one copy of an approximately 100 amino acid–long evolutionarily conserved α-helical BAG domain that allows them to interact with and regulate the Hsp70 family of molecular chaperones (12, 13). BAG-1 can bind to the kinase domain of CRAF (14) or BRAF (15), and in vitro experiments showed that BAG-1/CRAF interaction leads to the activation of CRAF independently of RAS (14). RAS-independent RAF activation, which results in the activation of ERK1/2, is also observed in cells overexpressing BAG-1 (16). ERK1/2 activation can be terminated by stress-induced upregulation of Hsp70 (16). Mechanistically, this may be explained by the competition between CRAF and Hsp70 for a common binding site in helix 2 of the BAG domain. Genetic disruption of BAG-1 in mice has no effect on the activation of ERK1/2 by mitogens (15). Nevertheless, subcellular localization studies have suggested that BAG-1 forms a mitochondrial survival signaling complex with Hsp70/AKT and BRAF, the absence of which may account for increased apoptosis in the liver and developing nervous system of BAG-1–deficient mouse embryos (15). Evidence for BAG-1 involvement in RAF-driven transformation has been provided by the demonstration that BAG-1 heterozygosity in mice expressing a constitutively active form of CRAF in type II pneumocytes significantly reduces oncogene-induced lung adenoma growth (17). In this model, RAF-downstream signaling was unaffected, whereas reduced BAG-1 expression specifically targeted tumor cells to apoptosis (17).

Activating mutations in RAF kinases are almost exclusively restricted to BRAF and are present in approximately 10% of all human tumors with the highest incidence of approximately 42% (Catalogue of Somatic Mutations in Cancer Database, Wellcome Sanger Trust) detected in melanoma (1, 5, 18). Most commonly BRAF mutant melanoma are V600E (74%–90%; ref. 19) and 16% to 29% are V600K mutations (20, 21). Because of the role of RAF kinases in promoting cancer cell growth, mainly through enforcing cell-cycle progression and enhancing cell survival (1, 5, 18, 22, 23), the RAS–RAF–MEK module has become a promising target for therapeutic intervention. This led to the development of small molecular weight inhibitors of RAF and MAP–ERK kinase (MEK; refs. 5, 24), with recent clinical studies reporting that highly specific BRAF inhibitors are effective in the treatment of metastatic melanoma (25–27). However, initial promise has been hampered by the development of resistance (28–30), which is characterized by the reactivation of ERK1/2 (31–33) and has been attributed to various mechanisms including activating NRAS mutations (29), CRAF overexpression (34), compensatory upregulation of MAP2K kinase COT (28), activating MEK1 mutations (35), and amplification of mutant BRAF (36).

Disruption of specific protein–protein interactions (PPI) within signalosomes may provide an alternative approach, in addition to existing RAF protein conformation–specific inhibitors, to interfere with aberrant signaling in cancer cells. Indeed, it has been shown over the past few years that PPI surfaces are druggable by small molecules (37–39), and the first PPI inhibitors are now entering the clinic (39).

Thioflavin S (NSC71948) is a complex mixture of different components resulting from the methylation and sulfonation of primulin base, a derivative of dehydrothio-p-toluidine (40, 41). Thioflavin S interferes with the ability of BAG-1 to interact with Hsc70 in vitro and in human breast cancer cell lines it reduces the interaction of BAG-1 with Hsc70, Hsp70, and CRAF, without affecting interaction between BIM and MCL-1, and results in decreased ERK1/2 phosphorylation, cell proliferation, and viability (42). Thioflavin S also reduces viability of wild-type (wt) but not BAG-1–deficient mouse embryonic fibroblasts (42). Until now, hit-to-lead development of drug-like inhibitors has been precluded by complexity of the Thioflavin S mixture. Here, we report the isolation and purification of the compounds Thio-2, Thio-3, Thio-5, and Thio-6 from Thioflavin S and their structural characterization. We also demonstrate the ability of Thio-2 to block signal propagation via the MAPK pathway, to impede the proliferation of RAF-transformed cells, and inhibit the interaction of BAG-1 with Hsc70 and to a lesser degree RAF in HEK293 cells.

General chemicals were of molecular biology grade and were purchased either from Sigma-Aldrich or Fisher Scientific, unless otherwise stated. Thioflavin S practical grade (CAS 1326-12-1) was purchased from Sigma-Aldrich. Calculation of molarity was based on information provided by the National Cancer Institute Developmental Therapeutics Program (NCI-DTP; Rockville, MD; http://dtp.cancer.gov). Analytical grade reagents (n-hexane, diethyl ether, ethyl acetate, n-butanol, petroleum ether, acetone, dichloromethane, and methanol) were supplied by VWR. High-performance liquid chromatography (HPLC)-grade acetonitrile and methanol were obtained from Merck. Deionized water (18.2 MΩ × cm) was obtained from an Arium 611 UV system (Sartorius Stedim Biotech GmbH). The MEK-inhibitor U0126 was purchased from Promega Corporation and prepared as a 10 mmol/L stock solution in dimethyl sulfoxide (DMSO; Sigma-Aldrich). AZD6244 (selumetinib) was obtained from Eubio and prepared as a 10 mmol/L stock in DMSO. Staurosporine (100 mmol/L stock) was obtained from Sigma-Aldrich. RAF inhibitors sorafenib (BAY43-9006) and PLX4032 were obtained from Axon Medchem BV and prepared as 50 mmol/L and 10 mmol/L stocks in DMSO, respectively. Thioflavin S (Sigma-Aldrich) and purified compounds (Thio-2, Thio-3, and Thio-5; for isolation and purification see below) were prepared as a 10 mmol/L stock solution in DMSO. All inhibitors were finally diluted in culture medium to reach working concentrations. Cell culture media and supplements were from PAA Laboratories, unless otherwise stated. The ECL reagents (SuperSignal West Pico and Femto Chemiluminescent Substrates) were from Pierce Biotechnology. Glutathione Sepharose 4B beads were from GE Healthcare Bio-Sciences AB. Protein concentrations were routinely assessed using Bio-Rad DC Protein Assay (Bio-Rad Laboratories). Etoposide (Sigma-Aldrich) was prepared as a 20 mg/mL stock in DMSO. Formaldehyde and glutaraldehyde (50% aqueous solution) were both obtained from Sigma-Aldrich. X-Gal was purchased from GIBCO, Life Technologies. Carboxyfluorescein diacetate succinimidyl ester (CFSE) was obtained from Molecular Probes, Life Technologies and stored at a 10 mmol/L stock in DMSO.

Thioflavin S practical grade (CAS 1326-12-1; ca. 450 g) was first soxhlet extracted with 1 l diethyl ether and the remaining powder extracted with water and diethyl ether to yield 8.5 g crude extract, which was fractionated over silica gel (for further details see Supplementary Materials and Methods). The crystalline fractions were purified over an SPE-C18 cartridge based on the color of the eluting bands; the eluate was divided into 10 and 12 subfractions, respectively. The subfractions eluting at CH₃OH/H₂O (90:10; 26.0 and 16.6 mg, respectively) were combined and finally purified over a Sephadex LH-20 column to yield 32.2 mg of compound Thio-3. Fraction F4 (219.1 mg) was purified over an SPE-C18 cartridge. The subfraction eluting at dichloromethane/acetone 85:15 (48.7 mg), was finally purified by Sephadex LH-20 CC to yield 39.0 mg of compound Thio-6. Fraction F5 (429.5 mg) was fractionated over an SPE-C18 cartridge. The subfraction, which eluted at CH₃OH/H₂O (85:15; 12.4 mg) was identified as Thio-2. Fraction F-6 (534.5 mg) was fractionated over an SPE-C18 cartridge. The subfraction, which eluted at CH₃OH/H₂O (80:20 and 85:15; 70.8 mg), was also recognized as Thio-2. The powder-like crystals (318.1 mg), which precipitated from fraction F-8 (367.9 mg), were combined with fraction F-10 (119.1 mg). The combined fraction was purified over an SPE-C18 cartridge. The subfraction eluted with dichloromethane/acetone 85:15 (v/v; 105.31 mg) was finally purified by Sephadex LH-20 CC to yield 35 mg of compound Thio-5.

Electrospray–mass spectrometry (ESI–MS) spectra were measured in ESI-positive mode on a Bruker (BrukerDaltonics) Esquire 3000 plus ion-trap mass spectrometer in positive mode (spray voltage +4.5 kV; endplate offset −500 V; nebulizer gas 40 psi; drying gas flow rate 9.00 L/min; dry temperature 350°C; and m/z range 100–1,500) coupled to an HPLC Hewlett Packard HP 1100 instrument with autosampler, diode-array detector, and column thermostat at 40°C using a Gemini C18 110A column (5 μm; 150 × 3.00 mm; Phenomenex) with guard column at a flow rate of 400 μL/min using an acetonitrile gradient in water both containing 0.1% formic acid (same as for HPLC analysis).

High-resolution ESI–MS spectra were recorded on a micrOTOF-Q II mass spectrometer Bruker (BrukerDaltonics) in positive mode (spray voltage +4.5 kV; endplate offset −500 V; nebulizer gas 10 psi; drying gas flow 5 L/min; dry temperature 180°C; and m/z range 100–1,000).

Analytical HPLC separations were performed on a HPLC Hewlett Packard HP-1050 system equipped with autosampler, diode-array detector, and column thermostat at 40°C using a Gemini C18 110A column (5 μm; 150 × 3.00 mm; Phenomenex) with guard column at a flow rate of 400 μL/min using an acetonitrile gradient in water containing 0.1% formic acid [ACN/H₂O (20:80) to (98:2) in 25 minutes, isocratic for 10 minutes, in 5 minutes back to ACN/H₂O (20:80) and isocratic for 5 minutes]. The detection was performed at 200, 254, 280, 350, and 400 nm.

Infrared spectra were recorded on a Bruker (Bruker Optics) IFS 25 FTIR spectrometer in transmission mode (4,000–600 cm⁻¹) using ZnSe disks of 2-mm thickness. Melting points were obtained on a Kofler hot-stage instrument (uncorrected).

The pLNC-HA-MEK2-(S220D, S226D; constitutively active) and pMCEF-myc-BRAF wt constructs were kindly provided by Prof. Piero Crespo (Universidad de Cantabria, Cantabria, Santander, Spain) and Prof. Richard Marais (The Patterson Institute for Cancer Research, The University of Manchester, Manchester, UK), respectively. The HA-mBAG-1S construct was a kind gift of Prof. John Reed (Sanford-Burnham Medical Research Institute, La Jolla, CA). To generate pGEX-2TK-BAG-1S H3AB, BAG-1S H3AB was excised from pcDNA3-BAG-1S H3AB (43) by HindIII and XhoI double digestion and ligated into pGEX-2TK in between the same flanking restriction sites.

MCF-7, MCF-10A, and HEK293 cells were obtained from LGC Standards. MCF-10A cells are derived from adherent cells from a mammary epithelial cell line produced from long-term culture in serum-free medium with low Ca²⁺ concentration (44). In contrast with MCF-7 estrogen receptor–positive breast cancer cells (45), these cells are nontumorigenic in immunosuppressed mice (46). Mouse fibroblast cell lines NIH 3T3 (wt) and NIH 3T3 expressing constitutively active forms of BRAF^V600E (kindly provided by Prof. R. Marais; ref. 47) were grown in Dulbecco's Modified Eagle Medium (DMEM). MCF-7 cells were maintained in minimum essential medium (Sigma-Aldrich), HEK293 cells were kept in DMEM. UACC257 cells (provided by Dr. Roland Houben, University of Würzburg, Würzburg Germany) were kept in RPMI, MDA-MB-231 and MDA-MB453 (LGC standards) cells were maintained in DMEM. All media contained 10% fetal calf serum (FCS), 2 mmol/L l-glutamine and 100 × penicillin/streptomycin. For MCF-10A, cell culture DMEM/F12 (1:1 v/v) medium (GIBCO, Life Technologies) was supplemented with 5% FCS, 20 ng/mL EGF, 0.5 μg/μL hydrocortisone, and 10 μg/mL ITS (insulin–transferrin–sodium selenite). Parental (M229 and M238) and PLX4032-resistant (M229 R5 and M238 R1) melanoma cell lines (29) were provided by Dr. Antoni Ribas (University of California, CA) and were maintained in complete cell culture media or media supplemented with 1 μmol/L PLX4032, respectively. All cell lines were cultured inhibitor free for 96 hours before use. Cells were maintained at 37°C in a 5% CO₂ humidified atmosphere. The cells were passaged for less than 6 months in our laboratory after receipt or resuscitation. No authentication was done by the authors.

Cells were transfected using Lipofectamine 2000 (Life Technologies-Invitrogen) according to the manufacturer's protocol.

MTT assay was carried out as previously described (48). Absorbance was measured at 550 nm with a 620 nm reference filter using an Anthos 2010 Microplate Reader (Biochrom, Holliston, MA).

Cell proliferation was determined using a Neubauer counting chamber or by performing crystal violet staining (49). Cell viability was assessed by trypan-blue exclusion assay of adherent and nonadherent cells.

MCF-7 and MCF-10A cells were seeded (1 × 10³ cells/well) in triplicate and treated for 24 or 48 hours as described. Cells were kept in complete growth medium until colonies were visible to the naked eye. Colonies were fixed and stained by adding 1.5 mL of a 0.5% crystal violet (Sigma-Aldrich) solution (in 50% methanol) for 45 minutes and excess stain removed by immersing plates in water. Plates were air-dried and colonies counted with a manual counter.

Protein separation and immunoblot analysis was performed as described previously (50). Antibodies raised against the following antigens were used: pERK (sc-16982R) and ERK (sc-94; Santa Cruz Biotechnology); PARP (9542), caspase-7 (9492), pAKT (4058), AKT (4685), and 9E10 myc-tag (2276; Cell Signaling Technology); BRAF (sc-5284, Santa Cruz Biotechnology) and GAPDH (glyceraldehyde-3-phosphate dehydrogenase; AM4300; Ambion). For coimmunoprecipitation, HEK293 cells stably expressing wt HA-mBAG-1S were seeded in triplicate in a 6-well plate at a density of 6.5 × 10⁵ cells per well. Samples were transiently transfected with equal amounts of pMCEF-myc-BRAF. Thirty hours posttransfection cells were treated with Thio-2, Thioflavin S (50 and 100 μmol/L), or DMSO for 16 hours and cells were lysed in HKEM buffer [50 mmol/L HEPES, pH 7.2, 5 mmol/L MgCl₂, 142 mmol/L KCl, 2 mmol/L EGTA, and 0.2% (v/v) Nonidet P40] supplemented with 1:100 protease inhibitor. To pull down HA-mBAG-1S, 350–500 μg total protein was incubated with 30 μL of anti-HA–affinity matrix (Roche Diagnostic) overnight on an orbital shaker at 4°C. Immune complexes were washed five times with HKEM buffer, beads were resuspended in 50 μL 2 × Laemmli buffer and heated for 5 minutes at 95°C. Immunoblot analysis was performed as described previously (50). Antibodies raised against the following antigens were used: BAG-1 (sc-8348) and Hsc70 (sc-7298; Santa Cruz Biotechnology); 9E10 myc-tag (2276; Cell Signaling Technology); and HA-tag (12013819001; Roche Diagnostic). Densitometric analyses were performed using Image J software (http://rsbweb.nih.gov/ij/). PPI was additionally tested using recombinant bacterially expressed and purified Hsc70, which had been labeleled with Fluorescein isothiocyanate (FITC) and recombinant GST-BAG-1. Binding was determined spectrophotometrically using a Varioskan Flash reader with SkanIt RE 2.4.3 Software (Thermo Scientific). For further details see Supplementary Materials and Methods.

Protein preparation was using the software program MOE (51). Docking studies were performed with GOLD version 5.1 (GOLDScore, 100% search efficiency; ref. 52) and without specific constraints. Protein Data Bank (PDB) (53) entry 1HX1 (54) was selected as protein template. Docking results were postprocessed using the software package LigandScout (55–57). For protein and binding site visualization VMD (58) and LigandScout were used.

All data are presented as mean ± SD unless otherwise stated. Statistical analysis was performed using GraphPad Prism 5 Software (GraphPad Software). One-way ANOVA and Tukey Multiple Comparison Test were used for comparison of significant effects of one independent variable in multiple groups. To calculate statistical differences of multiple independent variables in two groups, two-way ANOVA with Bonferroni posttest was applied. Significance values were designated as follows: *, P < 0.05; **, P < 0.01; ***, P< 0.001.

Because Thioflavin S is a mixture of two major constituents [2-(4-(diethylamino)phenyl)-6-methylbenzo[d]thiazole-7-sulfonic acid and 2′-(4-(diethylamino)phenyl)-6-methyl-2,6′-bibenzo[d]thiazole-7-sulfonic acid; Reaxys Registry Number 22509892; Fig. 1A], plus many more compounds at lower concentrations, we used a simplified bioguided isolation protocol to identify the compounds responsible for the biologic activity of Thioflavin S (37). Thioflavin S was fractionated by means of liquid–liquid extraction with different solvents yielding five subfractions. The highest growth inhibitory activity was found in the most lipophilic subfractions (MTT assay, data not shown). Furthermore, extraction yielded four pure compounds: Thio-2, -3, -5, and -6 (Fig. 1A). Additional substantive information is provided in Supplementary Tables S1–S4 and Supplementary Figs. S1–S4.

The stability of Thio-2, -3, -5, and -6 under tissue culture conditions was determined by LC-ESI–MS analysis and detected no decline of compounds Thio-2, -5, and -3 (purity >98%) after 72 hours. However, as degradation of Thio-6 was seen (data not shown), this compound was not investigated further.

To test the biologic activity of Thioflavin S and the pure compounds Thio-2, Thio-3, and Thio-5, MCF-7 cells were treated with 0 to 100 μmol/L of these compounds for 72 hours. Treatment with Thioflavin S caused a 14% growth inhibition at 100 μmol/L as monitored by the MTT assay (Fig. 1B). Of note, 100 μmol/L Thio-3 and -5 caused 26% and 38% cell growth inhibition, respectively. MCF-7 cells maintained in the presence of Thio-2 proliferated significantly slower compared with control cells (P < 0.001) or Thioflavin S–treated cells (P < 0.001). Cell proliferation was inhibited in a dose-dependent manner with a mean IC₅₀ of 29.6 ± 7 μmol/L for Thio-2. The growth inhibitory effect of Thio-2 was not limited to MCF-7 cells but was observed in two other breast cancer cell lines (MDA-MB-231 and MDA-MB-453) and the melanoma cell line UACC257 (Supplementary Fig. S5).

The number of viable cells was also counted 72 hours following treatment with 25 μmol/L of each compound using U0126 (an inhibitor of the RAF effector kinase MEK) as a positive control (Fig. 1C). Significantly reduced cell numbers were observed after 72 hours in the presence of Thioflavin S, Thio-2, -3, and -5, whereas the effect of U0126 became apparent at 48 hours (Fig. 1C). Compared with other Thio fractions, Thio-2 exhibited the most pronounced inhibition by 72 hours (Fig. 1C).

Cell death induced by Thio-2 treatment was analyzed by monitoring proteolytic processing of caspase-7 and PARP, which was highest with Thio-2 concentrations ≥25 μmol/L (Fig. 1D). In contrast, no PARP cleavage was observed with Thioflavin S (Fig. 1D), Thio-3, and Thio-5 at concentrations up to 100 μmol/L (data not shown). A clonogenic survival assay was also performed using 100 μmol/L of Thio-2 or Thioflavin S (Fig. 1E). Staurosporine was included as a positive control. Thio-2 potently reduced colony yield compared with Thioflavin S or DMSO control. On the basis of this initial characterization, Thio-2 was selected for all subsequent analyses.

To gain insights into possible regulation of RAF by Thio-2, we studied its effects on the activation of ERK1/2 in MCF-7 cells compared with untransformed MCF-10A cells (Fig. 2 and Supplementary Fig. S6). Serum stimulation resulted in a pronounced phosphorylation of ERK1/2, which was effectively blocked by MEK inhibitors U0126 and AZD6244 used as a positive control in both cell lines (Fig. 2A). Thio-2 in contrast only blocked ERK1/2 activation in MCF-7, whereas in MCF-10A cells even increased phosphorylation is observed. However, Thio-2 had no significant effect on the activation of AKT (a prosurvival kinase downstream of RAS but outside the RAF–MEK–ERK pathway) in MCF-7 cells, whereas AKT phosphorylation was decreased in MCF-10A cells (Fig. 2A). Specificity for RAF-dependent signaling in transformed cells was also supported by experiments using a combination of Thio-2 and an inhibitor of RAF. As shown in Fig. 2B, application of 5 μmol/L sorafenib, an inhibitor of CRAF and BRAF proteins, caused a modest reduction (32%) in ERK1/2 phosphorylation, whereas the combination of 5 μmol/L sorafenib with 25 μmol/L Thio-2 significantly reduced ERK1/2 phosphorylation (P < 0.01) compared with sorafenib alone (Fig. 2C). These findings suggest that Thio-2 may act at the level of RAF or MEK. To further test this hypothesis, HEK293 cells overexpressing a constitutively active mutant form of MEK (MEK^2S220D/S226D; ref. 59) were treated with the MEK-specific inhibitor U0126 or Thio-2. ERK phosphorylation was inhibited by U0126 but was unaffected by Thio-2 (Fig. 2D). Finally, we also directly tested for the effect of sorafenib or Thio-2 on the phosphorylation of MEK following serum stimulation of MCF-7 cells. As shown in Fig. 2E, Thio-2 efficiently prevented MEK phosphorylation. All these findings are consistent with an inhibitory function of Thio-2 in the RAF–MEK–ERK pathway, most likely at the level of RAF.

Thio-2 ≥ 12 μmol/L significantly inhibited growth of transformed MCF-7 cells in a dose-dependent manner compared with untransformed MCF-10A cells (Fig. 3A). This difference in the response to Thio-2 was also obvious when activation of apoptotic signaling was tested by immunoblotting for the processing of caspase-7 and PARP (Fig. 3B or when a clonogenic survival assay was performed (Fig. 3C and D). Quantification of trypan-blue stained cells (nonviable) confirmed the pronounced sensitivity of MCF-7 to Thio-2 treatment compared with MCF-10A cells (Supplementary Fig. S7). As tumor cells may carry multiple genetic alterations, NIH 3T3 murine fibroblast cell line–expressing wt or the oncogenic V600E mutant human BRAF proteins were examined. Cell growth was inhibited by Thio-2 in a dose-dependent manner in BRAF^V600E but not wt cells as shown by trypan-blue exclusion and crystal violet cell–staining assays (Fig. 3E and F, respectively). Immunoblotting showed no increase in the processing of PARP or caspase-3 cells in BRAF^V600E-expressing fibroblasts following treatment with 25 to 100 μmol/L Thio-2 (Supplementary Fig. S8A), indicating the absence of apoptotic cell death under these conditions. Also, a β-gal activity assay has been carried out but yielded no evidence that Thio-2–treated cells show increased senescence (data not shown). Finally, CFSE staining hinted decreased proliferation in U0126- and Thio-2–treated cells as obvious from the comparison of histogram plots of DMSO- (green line) and Thio-2/U0126–treated (red lines) cells (Supplementary Fig. S8B).

NIH 3T3 cells are characterized by low basal phosphorylation of ERK1/2, which is, however, pronounced in BRAF-transformed cells due to constitutive RAF signaling (Fig. 3G). Phosphorylation of ERK1/2 was efficiently reduced in BRAF^V600E-expressing cells by Thio-2 but not in wt fibroblasts (Fig. 3G and H). When combined together, the MEK-inhibitor U0126 and Thio-2 had an additive effect in suppressing ERK1/2 phosphorylation in BRAF^V600E cells (Fig. 3I). To exclude the possibility that the effect of Thio-2 treatment is caused through BRAF destabilization, BRAF protein expression was analyzed by immunoblotting. No decrease in endogenous or mutant BRAF levels was observed in MCF-7 or NIH 3T3 cells after treatment with Thio-2 for up to 48 hours (Supplementary Fig. S9C) or 72 hours (Supplementary Fig. S9A and S9B), respectively.

We investigated whether purified Thio-2 could inhibit the interaction of the cochaperone BAG-1 with Hsc70 and BRAF. Thioflavin S (1 mmol/L) but not its structurally related analogs Thioflavin T (1 mmol/L) or BTA-1 (1 mmol/L), inhibited interaction between BAG-1S and Hsc70; none of the compounds affected protein integrity (Fig. 4A). Recombinant BAG-1S H3AB mutant protein, which is defective in binding Hsc70, was used to confirm specificity of interaction between BAG-1S and Hsc70. Thio-2–inhibited GST-BAG-1S/Hsc70 interaction relative to Thioflavin T control, but its effect was not as pronounced as that of Thioflavin S (Fig. 4B).

As the concentration of compounds required to inhibit BAG-1S/Hsc70 interaction was quite high, an in vitro ELISA assay was developed, with a Z score range 0.5 to 0.8, to accurately measure the effect of Thio-2. Glutathione S-transferase (GST)-tagged BAG-1S, but neither the negative control GST-BAG-1S H3AB nor GST alone interacted with FITC-labeled Hsc70 (typical degree of labeling was around 0.5 mol of FITC/mole of Hsc70). GST-tagged proteins did not interact with the unrelated protein Annexin V–FITC (Fig. 4C). Thioflavin S dose-dependently inhibited interaction between GST-BAG-1S and Hsc70-FITC with DMSO (vehicle) alone exhibiting no effect on binding; as expected there was negligible binding between GST-BAG-1S H3AB and Hsc70-FITC at concentrations of Thioflavin S up to 100 μmol/L (Fig. 4D). Thio-2 dose-dependently targeted GST-BAG-1S/Hsc70-FITC interaction, causing a mean ± SEM inhibition of binding of 28.45 ± 7.15 at a concentration of 75 μmol/L (Fig. 4E).

To corroborate our in vitro data, we further analyzed PPIs in cells. In HEK293 cells stably expressing HA-mBAG-1S, Thioflavin S, and Thio-2 significantly decreased BAG-1 binding to Hsc70 over the dose range applied (Fig. 4F and Supplementary Fig. S10). This effect extended to the BAG-1/myc-BRAF interaction, which was also decreased upon incubation with Thioflavin S or Thio-2.

Taken together our results confirm the ability of Thio-2 to disrupt BAG-1–mediated protein interactions and illustrate a possible mode of action of this Thioflavin S derivative.

The potential of Thio-2 treatment to overcome melanoma resistance to PLX4032, a drug that targets oncogenic mutant BRAF with high specificity (29), was examined using drug-responsive (M229 and M238) or -resistant (M229 R5 and M238 R1) cell lines. Thio-2–inhibited cell growth dose-dependent and this was statistically significant compared with DMSO control at concentrations ≥50 μmol/L in all melanoma cell lines tested (Fig. 5A). In PLX4032-resistant M229 R5 cells, growth was reduced more effectively compared with controls. Moreover, the degree of inhibition achieved with Thio-2 in PLX4032-resistant M229 R5 cells was comparable with that of PLX4032 in M229 cells (Fig. 5B). We also analyzed the effects of Thio-2 on ERK1/2 activation (Fig. 5C). To eliminate stimulatory effects of serum growth factors cells were maintained at 0.05% serum. All cell lines exhibited robust ERK1/2 phosphorylation, which was efficiently decreased by the treatment with PLX4032 in both, resistant and responsive cells, as reported previously (29) and also to a lesser degree by Thio-2.

To further strengthen our hypothesis that primarily the binding of Thio-2 to BAG-1 is the reason for the effects described, we performed molecular modeling experiments starting from the crystal structure of the complex of BAG-1 and Hsc70. On the basis of structural and mutational studies published by Sondermann and colleagues (54), a putative binding site for small-molecule inhibitors was predicted in the interface of the BAG-1/Hsc70 complex (Fig. 6A). Docking revealed plausible binding poses for Thio-2, showing supramolecular interactions with the C-terminal part of BAG-1. In particular, π-stacking of the aniline and benzothiazol moieties with Arg234 and Arg237 (helix 3 of the BAG domain), hydrogen bonding of the aniline nitrogen with Asp222 (helix 2 of the BAG domain), and a hydrophobic contact of the ethyl group with Ile225 could be observed (Fig. 6B). Notably, Asp222 and Arg237, which are evolutionarily highly conserved, are essential for forming the interaction surface with the Hsc70 ATPase domain (54).

We report the isolation and characterization of the Thioflavin S–derived compound Thio-2 (Fig. 1 and Supplementary Data), which is mainly responsible for the biologic activities and biochemical properties previously reported for Thioflavin S (42). In an ELISA-based assay using recombinant BAG-1 and Hsc70 proteins, Thio-2 was able to decrease binding of Hsc70 to immobilized BAG-1 in a dose-dependent manner, although less efficiently than the parental compound Thioflavin S (Fig. 4 and Supplementary Data). In contrast, comparable efficacies for Thioflavin S and Thio-2 were observed when protein binding was studied by coimmunoprecipitation in HEK293 cells. These cell-based experiments demonstrated that Thio-2 or Thioflavin S also targeted BAG-1/RAF interaction (Fig. 4 and Supplementary Data). The reasons for the higher potency of the PPI inhibitors in cells are unclear. While in the in vitro experiments with recombinant protein test substances were added to immobilized BAG-1 prior to the incubation with Hsc70 (in the absence of RAF), in cells, complexes of BAG-1 with Hsc70 or RAF (and presumably other partners) preexist and will also be formed de novo during incubation with the compounds. Competition between RAF and Hsp70 for binding to the BAG domain has been demonstrated (16) and other proteins may also bind to this region of BAG-1. Upregulation of Hsp70 during cellular stress displaces RAF from BAG-1 and can result in growth arrest (16). It is therefore conceivable that protein competition for binding to BAG-1 may occur alongside the inhibitory effect of Thio-2 or Thioflavin S thereby reducing binding of Hsc70 or RAF to BAG-1 in cells.

Our docking experiments show that Thio-2 potentially binds to the BAG domain of BAG-1 at the binding interface of Hsc70, thereby targeting PPIs (Fig. 6). Amino acids Asp222 (helix 2) and Arg237 (helix 3) within the BAG domain are critical for binding to the ATPase domain of Hsc70 (54) and may serve as contact sites for Thio-2 according to our model. The requirement of helices 1 and 2 of the BAG domain for interaction with RAF has also been shown (16), although the amino acids that are critical for binding at the protein interface have not been identified as yet.

Thio-2 showed a pronounced inhibition of RAF signaling, as is evident from the decreased phosphorylation of ERK1/2 in cells stimulated with serum or expressing an oncogenic RAF mutant (Figs. 2, 3 and 5). We failed to detect any such effect of Thio-2 on AKT, which is also activated by mitogen stimulation and lies downstream of RAS. Thio-2 cooperated with another inhibitors of the MAPK pathway, such as the RAF-inhibitor sorafenib in suppressing growth factor–induced ERK1/2 phosphorylation. The inhibitory effect of Thio-2 on ERK1/2 activation was overcome by the expression of a constitutively active form of MEK (59), further supporting that Thio-2 acts upstream of MEK, possibly at the level of RAF.

With regard to RAF signaling, two possible functions have been proposed for BAG-1: activation of RAF in an RAS-independent fashion (14, 16) and mitochondrial relocalization of RAF (15). The mechanistic details underlying these effects remain elusive. In the general scheme of RAF activation, no essential role has been demonstrated for BAG-1 by biochemical or genetic studies. Our data show that Thio-2 inhibits signaling downstream of RAF activated either by serum growth factors or as a result of activating mutation, suggesting that RAF/BAG-1 interaction is not primarily required for RAF activation. How Thio-2 prevents signaling from active RAF remains to be established. Analysis of BAG-1–deficient mice has demonstrated the requirement of a BAG-1–induced mitochondrial RAF translocation for cell survival (15), suggesting that localization of RAF is critical for determining signaling outcome. The analysis of signaling through RAF/MEK/ERK led to the identification of several proteins, which are not members of the core-signaling machine, but are required for compartmentalized signaling. In the case of RAS/RAF transformation signaling from sites other than the cell membrane has been suggested (60). It remains to be shown whether BAG-1 is also involved in localizing RAF to any of these sites. Our data also demonstrate that Thio-2 has a more pronounced effect on proliferation, survival, and signaling of transformed than untransformed cells. The frequently increased BAG-1 levels observed in tumors (61) may therefore be required for other signaling events in transformed cells in addition to BAG-1–mediated antiapoptotic functions.

Thio-2 was also tested in a panel of melanoma cells harboring the common BRAF^V600E mutation, which are responsive or resistant to PLX4032 a mutant BRAF-specific inhibitor (Fig. 5). Thio-2 proved effective at inhibiting growth of both PLX4032-responsive and -resistant cells. In particular, M229 PLX4032–resistant cells (M229 R5) were even more susceptible than PLX4032-responsive cells to Thio-2–induced growth inhibition. M229 R5 cells acquire resistance to PLX4032 by upregulating the platelet–derived growth factor receptor β (PDGFRβ; ref. 29) allowing for the activation of alternative prosurvival pathways. M229 R5 and M238 R1 cells also remained highly resistant to MEK inhibition, and only stable knockdown of PDGFRβ caused cell-cycle arrest and apoptosis (29). The effect of Thio-2 in these cells, which are unresponsive to further targeting of the RAF-MEK-ERK cascade (29), is surprising. One possible explanation is based on the observation that BAG-1 enhances PDGFR-mediated protection from apoptosis through direct binding to the receptor (62). The region involved on the side of BAG-1 again involves helices 2 and 3 of the BAG-domain, which are also necessary for Hsc70 binding. Thio-2–mediated inhibition of M229 R5 cell growth could be the result of the disruption of BAG-1/PDGFR binding, pointing to a novel mechanism for overcoming BRAF-inhibitor resistance.

We have identified and extensively characterized Thio-2, which impedes the growth of cell lines that had become resistant to treatment with PLX4032, an inhibitor of mutant BRAF. We demonstrate that Thio-2 inhibits signaling at the level of RAF. The mode of action of Thio-2 may depend on its ability to interfere with the binding of RAF and/or Hsc70 to BAG-1 and further work will be required to delineate this.

No potential conflicts of interest were disclosed.

Conception and design: G. Wolber, G. Packham, R.I. Cutress, H. Stuppner, J. Troppmair

Development of methodology: M. Enthammer, E.S. Papadakis, M. Deutsch, M.I. Ashraf, G. Wolber, G. Packham, R.I. Cutress

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Enthammer, E.S. Papadakis, K. Koziel, S. Khalid, R.I. Cutress

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Enthammer, E.S. Papadakis, M.I. Ashraf, G. Wolber, G. Packham, R.I. Cutress, J. Troppmair

Writing, review, and/or revision of the manuscript: M. Enthammer, E.S. Papadakis, M.S. Gachet, S. Schwaiger, K. Koziel, M.I. Ashraf, G. Wolber, G. Packham, R.I. Cutress, H. Stuppner, J. Troppmair

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Enthammer, E.S. Papadakis, M.S. Gachet, M. Deutsch, H. Stuppner, J. Troppmair

Study supervision: G. Wolber, R.I. Cutress, H. Stuppner, J. Troppmair

The authors thank Dr A. Ribas for the kind gift of PLX4032-responsive and -resistant cell lines and Profs. R. Marais, P. Crespo, and J. Reed for providing expression constructs. The authors also thank Dr. P. Duriez in the Cancer Research UK Protein Core Facility, Cancer Sciences Unit, University of Southampton (Southampton, UK), for protein expression, purification, and labeling. The authors thank Dr. L. Forse for critical reading of the article and all members of the J. Troppmair Laboratory for helpful discussions. The authors also thank the valuable assistance of Ruth Baldauf in the preparation of the article.

The work was supported by a Cancer Research UK grant (Southampton Cancer Research Centre Core Award grant, ref: C34999/A11344; R.I Cutress). The work in the laboratory of R.I. Cutress was also supported by a Research Support grant from the Royal College of Surgeons of Edinburgh and by a project grant from Breast Cancer Campaign (ref 2011NovPR39). J. Troppmair obtained financial support from Oncotyrol, Project 1.5. The competence center Oncotyrol is funded within the scope of COMET—Competence Centers for Excellent Technologies by the Federal Ministry for Transport, Innovation, and Technology (BMVIT) and the Federal Ministry of Economy, Family, and Youth (BMWFJ) as well as the federal states Tyrol and Styria. The Austrian Research Promotion Agency (FFG) manages the competence center program COMET. J. Troppmair was also supported by the FWF, Austrian Science Foundation, project MCBO ZFW011010-08. M. Enthammer is supported by a project from the Austrian Cancer Society/Tyrol.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.