Betulinic acid acetate, an antiproliferative natural product, suppresses client proteins of heat shock protein pathways through a CDC37-binding mechanism

Abstract

CDC37 has emerged as a promising target in antitumor chemotherapy because of its significant role in oncogenic signaling networks. In an effort to discover novel CDC37 inhibitors, a shape-based model derived from celastrol was built. A high-throughput virtual screening of ChemDiv based on the ROCS model has led to the identification of several pentacyclic triterpenes with high shape similarity with celastrol, among which betulinic acid acetate was the most attractive. Bio-layer interferometry assay demonstrated that betulinic acid acetate binds to CDC37 with a KD value (25.7 μM) on the same order of magnitude as that of celastrol. This compound not only shows comparable antiproliferative activity against a series of cancer cell lines, but also causes the degradation of HSP90 client proteins in PC-3 cell lines without inducing the expression of HSP70. In addition, betulinic acid acetate inhibits the function of HSP90 in a manner different from celastrol. The HTRF assay indicated that it cannot inhibit the association between HSP90 and CDC37. Given that it is not an α,β-unsaturated carbonyl compound, betulinic acid acetate can be recognized as the first inhibitor that binds to CDC37 through non-covalent binding.

---

1. Introduction

Heat shock protein 90 (HSP90) is a conserved molecular chaperone that facilitates the correct conformation, maturation and biological activities of its client proteins by modulating processes such as trafficking, folding and degradation.¹ The majority of HSP90 clients, including kinases, transcription factors, hormone receptors and many other proteins, are involved in oncogenic signaling pathways that contribute to the proliferation and anti-apoptosis ability of cancer cells.² The inhibition of HSP90 simultaneously incapacitates these clients and perturbs multiple signaling pathways, resulting in a combined tumor-killing effect.³ Therefore, HSP90 has long been recognized as a therapeutic target for cancer, and various HSP90 inhibitors have been developed with different mechanisms of action.^4,5 The most reported inhibitors originate from natural products and act by targeting the binding sites of the N-terminal domain of HSP90. For example, geldanamycin and its derivatives have entered multiple clinical trials for cancer treatment.^6,7 However, these products suffer from various clinical complications; for example, they show toxicity in the liver and in ocular and cardiac systems.⁶

One possible reason for such toxicity is the non-selective degradation of client proteins, which may be important for normal biological activities. In order to alleviate these side effects observed in current clinical trials, continuing efforts have been made to facilitate selective client degradation by perturbing the interaction between HSP90 and co-chaperone that is specific for certain classes of clients.⁸ It is well established that HSP90, a homodimer, functions through the formation of oligomeric chaperone complexes with a cohort of co-chaperones to assist in the correct folding and release of specific clients.⁹ For example, peptidyl-prolyl isomerase co-chaperones (PPIases) play an important role in maintaining the three-dimensional structures of its clients. Co-chaperone p23 facilitates ATP binding of HSP90 by helping maintain the optimal conformation.¹⁰ Moreover, the maturation of many kinase clients, especially those tightly correlated to carcinogenesis, calls for the assistance of co-chaperone cell division cycle 37 (CDC37).¹¹ Additionally, CDC37 is reported to be up-regulated in various cancer cells, further supporting its important function in the development of cancers. Therefore, disrupting the activity of CDC37 using small molecules has emerged as a novel therapeutic strategy for the treatment of cancer.¹² Recently, celastrol, a quinonemethide triterpene extracted from Celastraceae plants, has been reported to bind to CDC37, disrupt the interaction between HSP90 and CDC37,¹³ and prevent the maturation of oncogenic kinases.¹⁴ An in vivo study indicated the potent antitumor activity of celastrol in nude mice bearing prostate or pancreatic tumors.^14,15

Despite the encouraging effects of celastrol in tumor growth inhibition, it can induce heat shock response (HSR), a similar drawback to N-terminal HSP90 inhibitors. HSR formation is caused by the dissociation of heat shock factor 1 (HSF-1) and leads to elevated cellular levels of heat shock proteins, complicating drug scheduling and dosing.⁸ As a result, increasing attention has been focused on searching for novel CDC37 inhibitors in order to obtain both superior therapeutic efficacy and HSR exemption.

Herein, we report the discovery of a series of pentacyclic triterpenes as new CDC37 inhibitors through shape-based virtual screenings. Among all the compounds, betulinic acid acetate (4) can interact with CDC37 complexes and subsequently halt the maturation of the downstream client proteins. Bio-layer interferometry (BLI) and western blot assays revealed that the effect of 4 on HSP90 mechanism resulted from binding to CDC37 but with a different pattern compared to celastrol. This hypothesis was further supported by homogeneous time-resolved fluorescence (HTRF); the results showed that celastrol could dissociate the HSP90–CDC37 complex, while 4 could not. Importantly, 4 exerted antiproliferative activities against a panel of cancer cell lines. More importantly, compound 4 did not induce HSR when inhibiting CDC37, which is a great advantage over celastrol. Considering the fact that 4 does not have an α,β-unsaturated carbonyl group and might bind to CDC37 non-covalently, it can serve as a promising leading compound for the discovery of anti-tumor agents through the regulation of the chaperone system.

2. Experimental section

2.1 Virtual screen for inhibitors to interrupt the HSP90–CDC37 interaction

Rapid overlay of chemical structures (ROCS) is a shape-based method that uses a Gaussian algorithm to perceive similarity between molecules based on their three-dimensional shapes.¹⁶ In this study, celastrol was used as a reference to generate the ROCS model. For virtual screening, the compound collection of the ChemDiv database was downloaded from the official web site (http://www.chemdiv.com/). Multiple conformations of the compounds were generated by Omega module of the Openeye platform with the following parameters. The root-mean-square distance (RMS) was set to 0.5 Å. Ewindow, the value used to discard high-energy conformations, was set to 10 kcal mol⁻¹. To ensure complete conformational coverage, the maximum allowed conformations per compound was 400, and the entire conformation generation process was performed under the Merck Molecular Force Field 94 (MMFF94).¹⁷ The ROCS program was carried out with the following parameters: rank by = combo and best hits = 1. ROCS assesses the candidate compound and celastrol by measuring the shape overlap between the two molecules. The similarity between the molecules was ranked by the combo score, which ranged from 0 to 2. The closer the combo score is to 2, the more similar the shapes and chemical features of the molecules are. In contrast, the closer the combo score is to 0, the less similar the shape and chemical features are. Finally, the 11 compounds with the maximum combo scores were retained and purchased from the Topscience database (http://www.tsbiochem.com). The purities of all compounds were all over 95%, as determined by HPLC.

2.2 Protein expression and purification

The plasmid-encoding His-tagged full-length HSP90α and GST-tagged CDC37 were purchased from Novobio (http://www.novobiosci.com). To purify these proteins, Escherichia coli BL21(DE3) cells were transformed with HSP90 or CDC37 plasmid and coated on LB-broth agar plates containing 50 μg mL⁻¹ kanamycin or ampicillin. After culturing overnight at 37 °C, a monoclonal colony was selected to grow in LB medium containing 50 μg mL⁻¹ antibiotic at 37 °C in an incubator shaker set to 220 rpm. When the OD₆₀₀ value reached 0.5, 0.5 mM isopropyl-β-D-thiogalactopy-ranoside (IPTG) was added to induce protein expression. For HSP90, cells were induced at 11 °C for 8 h, while for CDC37, cells were induced at 16 °C for 8 h. The bacteria were then collected by centrifugation at 4000 rpm and 4 °C for 15 min and washed with cold PBS. After resuspension in lysis buffer (20 mM Tris, 300 mM NaCl, 1 mM PMSF, 10% glycerol, 0.1% Tween, 20 mM β-mercaptoethanol), the bacterial pellet was lysed by sonication. The crude lysate was clarified by centrifugation (12 000 rpm, 30 min, 4 °C) and filtration (pore size = 0.22 μM) and then purified by AKTA (GE Healthcare)-His affinity chromatography or GST affinity chromatography.

For HSP90, the cell lysate was loaded onto a 5 mL HisTrap Heparin column (GE Healthcare) pre-equilibrated with balance buffer (20 mM Tris, 300 mM NaCl, 10% glycerol, 0.1% Tween). After washed with washing buffer containing 20 mM imidazole, the column was eluted with an elution buffer containing 100 mM imidazole over 10 column volumes with a flow rate of 4 mL min⁻¹. For CDC37, the lysate supernatant was loaded onto a 5 mL GST Trap Heparin column (GE Healthcare) pre-equilibrated with balance buffer (15 mM Tris, 300 mM NaCl). The wash and elution buffer contained 2.5 mM and 15 mM glutathione, respectively. Proteins were identified by western blotting and concentrated using the Bradford protein assay kit (Thermo Scientific # 23225) with purified bovine serum albumin (BSA, Thermo Scientific) as the standard. All the purification steps were performed at 4 °C, and the purified proteins were stored in 10% glycerol at −80 °C for further use.

2.3 Detecting the affinity of inhibitors against CDC37 by BLI measurement

The direct interactions between all molecules and CDC37 were quantified using BLI (Octet RED96; FortéBio Inc., Menlo Park, CA). Aminopropylsilane (APS) biosensors (FortéBio, Inc., Menlo Park, CA) were pre-incubated with 200 μL 1× PBS buffer/well to establish a baseline before protein immobilization. All of the binding data were collected at room temperature. The experiments included five steps: (1) the baseline for the biosensors to reach steady state in the buffer; (2) the loading of CDC37 onto the sensor; (3) the baseline for the CDC37-loaded biosensor to reach steady state in the buffer; (4) the association of CDC37 with natural products for the measurement of K_on; and (5) the dissociation of the natural products with CDC37 for the measurement of K_off. During the binding detection, the natural products were diluted at four concentrations (2.5, 5, 10 and 20 μM). Celastrol was defined as the positive control. The association and dissociation plot was fitted using a 2 : 1 (heterogeneous ligand) binding model and was calculated by double reference subtraction using the FortéBio Octet RED96 data analysis software 6.0.

2.4 Effect of inhibitors on cell proliferation

A549, MCF-7, PANC1, HCT116 and PC-3 cell lines were seeded in the flat-bottom 96-well plates (about 5000 cells per well) and cultured in 100 μL Gibco medium (supplied with 2 g L⁻¹ Na₂CO₃, 0.1 g L⁻¹ streptomycin, 0.1 g L⁻¹ penicillin and 10% fetal bovine serum) overnight. Various concentrations of natural products (diluted using 1% DMSO) were then added and incubated with the cells for 72 h at 37 °C and 5% CO₂. Celastrol and DMSO were added as the positive and negative controls, respectively. After 72 h, 20 μL of 5 mg mL⁻¹ MTT was added in each well followed by another 4 h of incubation at 37 °C. Subsequently, 150 μL DMSO was added into each well after the removal of the supernatant. The plate was then shaken for 10 minutes at room temperature. Finally, the absorbance at 490 nm was recorded using an Elx800 absorbance microplate reader (BioTek, Vermont, USA), and the IC₅₀ value was calculated in triplicate using GraphPad Prism 5.0.

2.5 Effect of compound 4 on HSP90 client proteins by western blot assay

PC-3 cell lines were seeded in seven 25 cm² cell culture flasks (Corning) with 5 × 10⁶ cells per flask. After culturing in 4 mL Gibco medium supplied with 2 g L⁻¹ Na₂CO₃, 0.1 g L⁻¹ streptomycin, 0.1 g L⁻¹ penicillin and 10% fetal bovine serum overnight, the cells in different flasks were treated with compound 4 at various concentrations (1, 2, 5, 10 and 20 μM). Celastrol and DMSO (2 μM) were defined as the positive and negative controls, respectively. After 24 h, the cells were harvested and lysed in cold RIPA lysis buffer supplemented with 50 mM Tris–HCl (pH 7.4), 1% NP-40, 0.25% Na–deoxycholate, 150 mM NaCl, 1 mM EDTA and the protease and phosphatase inhibitors (Roche). The supernatants were then obtained by centrifugation at 13 000g for 10 min at 4 °C. The concentration of protein was measured using a BCA assay kit (Thermo Prod # 23225). After equally loading the proteins for SDS-PAGE, the proteins were separated and transferred to the PVDF membrane (Millipore immobile transfer membranes CAT no. IPVH00010). After blocking with skim milk (BD, LOT4049465) for 1.5 h at room temperature, the membranes were incubated with the primary antibody [HER2 (Cell Signaling Technology, # 2165S), HSP70 (Cell Signaling Technology, # 4872), AKT (Cell Signaling Technology, # 9272S), CDK6 (Cell Signaling Technology, # 3136) and actin (Proteintech, 23660-1-AP)] overnight at 4 °C followed by secondary antibody (DyLight™ 800 labeled, Abbkine, A23920) at room temperature for 1 h. Finally, the membranes were screened through the Odyssey infrared imaging system (LI-COR, Lincoln, Nebraska, USA).

2.6 Effect of inhibitors on the HSP90–CDC37 interaction detected by HTRF measurement

XL665-labeled monoclonal anti-histidine antibody (product no. 61HISKLB) and europium cryptate-conjugated monoclonal antibody anti-glutathione S-transferase (product no. 61GSTKLB) were purchased from Cisbio Bioassays (Codolet, France). Celastrol (product no. HY-13067), which was identified as an HSP90–CDC37 interaction inhibitor in 2006, was obtained from Medchem Express (http://www.medchemexpress.cn/). The HTRF interrupting assay was carried out with a total volume of 20 μL using the 384-well low-volume plate (Greiner # 784076). The HTRF assay was performed as follows. First, 4 μL HSP90 (final concentration = 375 nM), 4 μL CDC37 (final concentration = 125 nM) and 4 μL natural products (diluted in buffer containing 50 mM PBS, 0.1% BSA and 0.05% Tween 20 at pH 8.0) were incubated at 37 °C for 1 h. Next, 4 μL anti-GST–cryptate (final concentration = 20 ng μL⁻¹) and 4 μL anti-6His–XL665 (final concentration = 20 ng μL⁻¹) were added to the wells. After another incubation at 37 °C for 2 h, the time-resolved fluorescence was measured using a Molecular Device (SpectraMax paradigm; λ_ex = 320 nm, λ_em = 665 nm and 620 nm). The HTRF ratio was calculated as a two-wavelength signal ratio as (signal 665 nm/signal 620 nm) × 10 000. Wells containing His–HSP90, GST–CDC37, XL66-labeled monoclonal anti-histidine antibody and europium cryptate-conjugated monoclonal antibody anti-glutathione S-transferase were used as negative controls, while wells containing only the anti-6His–XL665 and anti-GST–cryptate were used as blank controls. Data for the HTRF assay were analyzed using Prism software (GraphPad Software, Inc. San Diego, CA). The IC₅₀ was determined by fitting the data using the log(inhibitor) vs. normalized response equation. To ensure reproducibility, all the experiments were carried out in duplicate or triplicate.

2.7 Molecular docking of the compounds for interaction with the HSP90–CDC37 complex

Molecular docking was performed using LibDock (Accelrys Discovery Studio 4.0). LibDock is a fast and site-directed docking method that uses geometric and shape complementarity-based algorithms.¹⁸ In contrast to other algorithms such as Gold and CDOCKER, which generate random conformations using CHARMM-based molecular dynamics, LibDock is able to find various conformations of the ligands and then further optimize the receptor–ligand interaction by molecular dynamics.¹⁹ The crystallographic protein structure of CDC37 was retrieved from the Brookhaven Protein Data Bank (PDB ID: 1US7). Prior to docking, the proteins were prepared using the “clean water” and “add hydrogens” modules. After the ionization and tautomeric states of amino acid residues such as His, Asp, Ser and Glu were satisfied, the binding sites were defined from the protein cavity that contained the key interaction amino acid residues between HSP90 and CDC37. The molecules that were refined using energy minimization were then used as input structures for processing LibDock.

3. Results and discussion

3.1 Virtual screening for potential HSP90–CDC37 inhibitors using the ROCS model

ROCS is a ligand-based method based on the principle that molecules will form similar shapes if their volumes overlay well.²⁰ Compared to other structure pharmacophores such as the Hiphop model and Hypogen model, which highlight the common structural information, the ROCS model emphasizes the unique features of a reference molecule.²¹ In this study, celastrol was selected as the template to generate the ROCS model. As the bioactive conformation is not necessary for the shape-based screening,²² a 3D structure of celastrol, which was built using Discovery Studio and further energy minimized with a CHARMM force field, was used directly to generate the ROCS model. The molecular shape depicted in gray shadow (Fig. 1) contained hydrophones derived from the core of pentacyclic triterpenes. Both the carbonyl and carboxyl groups supplied the hydrogen-acceptor features. For the virtual screening, the multiple conformations of the compounds in the Topscience database were generated by OMEGA in Openeye. The 3D shape similarity of a given compound to celastrol was evaluated based on two distinct aspects: the shape Tanimoto coefficient and the score retrieved from the ROCS color force field. As each score varies between 0 and 1, the sum of both scores, called the combo score, varies between 0 and 2.²³ Finally, 11 compounds (Fig. 2) were retained and purchased from Topscience for further biological evaluation.

Fig. 1 The ROCS model generated from celastrol.

Fig. 2 2D structures of the 11 screened hits.

3.2 Binding affinity between natural products and CDC37 using BLI assay

BLI was carried out to confirm whether direct interactions occurred between the natural products and CDC37. In the BLI assay, the natural products were defined as ligands, while CDC37 protein was the receptor. In BLI assay, the fitting procedure is of key importance in determining the binding affinity. Many artifacts such as mass transport and immobilization can cause the data points to deviate from the ideal binding curve. In addition, the chosen binding model also affects the shape of the binding curve. In order to obtain an accurate result, we used the aminopropylsilane (APS) sensor for BLI in this study. Several binding models were also tested to eliminate the poor fitting caused by complex binding between the ligand and receptor. Finally, 2 : 1 (heterogeneous ligand) binding was demonstrated to be the best fit. We first detected the binding affinity of celastrol, which was found to directly bind to CDC37 with a KD value of 6.85 μM. This result was consistent with the results of HTRF assay (Section 3.5). We then further determined the binding affinity of all 11 natural product hits using the established method (Fig. 3, Table 1). The KD value of betulinic acid acetate (4) was determined to be 25.7 μM. Both its association rate constant (K_on) and disassociation rate constant (K_off) were smaller than those of celastrol. The significant difference in binding kinetics indicated that these two compounds have different binding patterns. As postulated by Sridhar et al.,¹³ the cysteine thiol groups in CDC37 might serve as nucleophilic groups, attracting electrophilic sites and forming covalent adducts with celastrol. As betulinic acid acetate lacks this type of α,β-unsaturated carbonyl group, its binding to CDC37 occurs mostly through non-covalent interactions, which explains its slower binding compared to celastrol. Under physiological conditions, the Michael reaction of celastrol and CDC37 is both fast and reversible, which could explain the larger dissociation constant of celastrol. In summary, compared to the covalent binding of celastrol, the interaction of 4 with CDC37 is more direct and more specific.

Fig. 3 Dose–response binding curves using the BLI assay. (A) Dose–response curves of celastrol. (B) Dose–response curves of betulinic acid acetate. (C) Binding constants determined by BLI.

Table 1 CDC37 binding ability and the inhibitory activity of the screened-out compounds. Four of the 11 hits exhibited the ability to bind CDC37 in a dose-dependent manner; the KD values of these hits are shown in the table

Cpd	CDC37 binding affinity (KD, μM)	Inhibitory activity (HTRF, % at 100 μM)
a The IC50 of celastrol was determined to be 11.6 ± 1.1 μM using HTRF assay.b NA means compound cannot dose-dependently bind to CDC37.
1	90.5	0.0
2	NA^b	0.0
3	125.0	1.1
4	25.7	6.0
5	NA	3.3
6	NA	0.0
7	NA	1.5
8	20.4	0.0
9	NA	17.3
10	NA	0.7
11	NA	3.9
Celastrol	6.85	81.8^a

---

We then explored the pharmacophore features of these compounds based on the activities (Fig. 4). As is known, the main hydrophobic force that drives the interaction between celastrol and CDC37 is provided by the triterpenoid skeleton. Thus, the scaffold of celastrol should not be simplified. The removal of this ring E (3) caused the decrease in activity, while the alternation of cyclohexane to cyclopentane (1, 4, 8) was tolerated. The polar acid group in ring E is significant for the activity. The modification of this group into an ether group (1) or a hydroxyl group (2) or its removal (5, 8, 10, 11) reduces or even eliminates the activity, indicating that this group might form strong polar interactions with CDC37 through ionic bonding. This ionic interaction might be the major polar force contributing to the interaction between celastrol and CDC37. Next, we focused on the substitution of ring A. It seems that the replacement of ring A with a saturated ring (4) has no significant effect on the activity. Ring A also shows great tolerance to multiple substitutions (carbonyl group, methyl group, dimethyl group), indicating that this ring could provide promising sites for structural modifications. Notably, the hydroxyl group in ring A contributed significantly to the activity. When it was substituted by a ester group (2, 5, 11), carbonyl group (6, 7) or oximido group (9, 10), the compounds showed abrogated activity, suggesting that the incorporation of a hydrogen-bond donor at this position would be of great importance to improve the activity. In addition, the hydrogen bond could be another source of polar interaction. The structures and activities of these compounds suggest that to be an active CDC37 inhibitor, the pentacyclic triterpene compound should have at least one polar group, either the carboxylic group in ring E or the hydroxyl group in ring A.

Fig. 4 Structure–activity relationship of the CDC37 inhibitors with pentacyclic triterpene scaffold.

3.3 Effect of pentacyclic triterpenes on cancer cell viability

3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2-H-tetrazolium bromide (MTT) assay was performed to investigate the anti-proliferative activity of these pentacyclic triterpenes. HCT116, MCF-7, A549, PC-3 and PANC-1 cancer cell lines were treated with various concentrations of the tested compounds. As indicated in Table 2, celastrol blocks cell proliferation in a dose-dependent manner. The growth of the HCT116, MCF-7, A549, PC-3 and PANC-1 cells was inhibited by 50% by treatment with 1.9, 9.8, 3.3, 8.7, and 2.7 μM celastrol, respectively. Compounds 1 and 4 showed higher cytotoxicities than other pentacyclic triterpenes (5, 6, 9, 10, and 11). Their IC₅₀ values for the inhibition of cell proliferation ranged from 5.5 to 39.6 μM. The IC₅₀ values of compounds 3 and 8 against MCF-7, PC-3 and PANC-1 cell lines were over 100 μM. Their cytotoxicity profiles against HCT116 and A549 cell lines were similar, suggesting modest activity. Among all the pentacyclic triterpenes, compounds 5 and 11 showed poor cytotoxicities, which might be attributed to their poor lipophilicity and stability.

Table 2 Effect of pentacyclic triterpenes on the proliferation of a panel of tumor cells. Data are presented as mean ± SD (n = 3)

Compound	Cytotoxicity IC50 (μmol L^−1)
	HCT116	MCF-7	A549	PC-3	PANC-1
1	36.2 ± 1.7	5.5 ± 0.7	11.8 ± 1.0	35.5 ± 3.9	39.6 ± 1.7
2	27.6 ± 1.4	15.0 ± 1.2	15.7 ± 1.2	>100	30.1 ± 1.9
3	51.5 ± 2.2	>100	47.8 ± 2.2	>100	>100
4	17.1 ± 0.8	21.8 ± 1.5	15.3 ± 0.8	27.5 ± 1.5	12.2 ± 1.6
5	>100	>100	>100	>100	>100
6	25.8 ± 1.2	>100	>100	>100	59.8 ± 2.2
7	11.7 ± 1.1	16.8 ± 1.1	13.9 ± 1.2	49.3 ± 7.7	45.8 ± 1.6
8	54.7 ± 1.6	>100	63.8 ± 2.4	>100	>100
9	30.6 ± 0.5	9.8 ± 0.9	28.0 ± 3.1	39.6 ± 1.2	54.1 ± 1.8
10	57.4 ± 3.5	>100	>100	26.5 ± 2.2	45.8 ± 1.5
11	>100	>100	>100	>100	>100
Celastrol	1.9 ± 0.3	9.8 ± 0.1	3.3 ± 0.4	8.7 ± 0.5	2.7 ± 0.2

---

3.4 Betulinic acid acetate decreases HSP90 client protein levels

Western blot assay was carried out to evaluate the effects of betulinic acid acetate (4), the most potent compound in BLI and MTT assays, on the degradation of some typical HSP90 client proteins (HER-2, AKT and CDK6) and on the up-regulation of HSP70 in the PC-3 tumor cell line, which is CDC37 over-expressed and HSP90 highly relied on. As shown in Fig. 5, when PC-3 cells were treated with 4, decreased protein levels of HER-2, AKT and CDK6 were observed in a dose-dependent manner. Treating PC-3 cells for 24 h with 5 μM 4 decreased HER-2, AKT and CDK6 levels by 58%, 43% and 42%, respectively. Taken together, these data demonstrate that 4 can inhibit the HSP90 signaling pathway and induce the degradation of client proteins. However, this degradation of client proteins was not accompanied by elevated levels of the co-chaperone HSP70, a signatory feature of HSP90 inhibition. In contrast, celastrol was able to increase HSP70 protein expression by more than 1.8-fold in PC-3 cells after 24 h of incubation. These data demonstrate that 4 did not induce HSR compared to celastrol. Given their different binding kinetics in the BLI assay, we might postulate that their difference in HSR induction might be attributed to their different binding patterns.

Fig. 5 Betulinic acid acetate down-regulates the expression of HSP90 client proteins. PC-3 cells were grown for 24 h in the presence of various concentrations of betulinic acid acetate, and the client proteins were detected with western blot analysis. β-Actin is shown as a loading control.

3.5 Effect of natural products on HSP90–CDC37 interaction detected by HTRF measurement

To further study the binding pattern of the hits, HTRF assay was performed to investigate whether these pentacyclic triterpenes can inhibit HSP90–CDC37 PPI through a similar mechanism as celastrol. After the incubation of HSP90 labeled with the 6His-tag, CDC37 labeled with GST-tag and 100 μM compound for 2 h at 37 °C, the detection reagent was added to detect the intensity of fluorescence. For fluorescence, 625 nm excitation and 665 nm emission filters were used in the competition binding assay. The results demonstrated that all these compounds could not compete with HSP90 to bind to CDC37 (Table 1). Even when their concentrations were up to 100 μM, they only inhibited HSP90–CDC37 complex formation by 10%, while celastrol exhibited an IC₅₀ of 11.6 μM. These data indicate the distinctive features of 4.

Interestingly, despite the competitive interaction of celastrol with CDC37–HSP90, the related betulinic acid acetate (4), which binds with higher affinity, cannot disrupt the complex between HSP90 and CDC37. Structural analysis revealed that although 4 and celastrol share the same scaffold, they have some subtle differences. Thus, we postulated that their structural differences contribute to their different HTRF assay results. First, celastrol has polar groups on both ring A and ring E. As reported,¹⁴ the hydroxyl and carbonyl groups on ring A were involved in a network of H-bonds with the side chains of the HSP90–CDC37 complex, weakening the HSP90–CDC37 interaction. However, unlike celastrol, betulinic acid acetate only has an acetoxyl group on ring E. From the aspect of polar interaction with CDC37, betulinic acid acetate might induce a different conformational change in the binding pocket. Second, celastrol has a conjugated system in the left two rings, which means that its planarity is much better than that of saturated betulinic acid acetate. Although these two compounds have similar scaffolds, their difference in planarity might result in the discrepancy in space conformation and binding sites, thus causing different biological activities. Third, celastrol is an α,β-unsaturated carbonyl compound. As mentioned above, its binding to CDC37 is covalent, whereas the binding of 4 to CDC37 occurs primarily through non-covalent interactions. Thus, celastrol and 4 might have completely different mechanisms from each other.

3.6 Molecular docking of the compounds for interaction with the HSP90–CDC37 complex

To further investigate the different mechanism of 4, we docked this compound into the HSP90–CDC37 complex. As mentioned by Tao Zhang et al., the ring A of celastrol binds to the polar and charged groove of HSP90–CDC37. The carbonyl group of celastrol can form H-bonds with the side chain of Arg32 and His197, which could block the essential H-bond interaction between Arg167 (CDC37) and Glu33 (HSP90).¹⁴ However, in our binding model of 4 (Fig. 6), compound 4 occupied a different binding conformation; the carboxylic group (ring E) protruded into the polar pocket of the HSP90–CDC37 complex without disrupting the H-bond between Arg167 and Glu33. The docking data further verified our assumption that the different polarities of these two compounds can make them adopt different binding conformations. Thus, the mechanistic difference between celastrol and 4 might be attributed to their structural differences.

Fig. 6 Molecular docking of compound 4 into the HSP90–CDC37 complex. (A) Compound 4 binds to the polar groove of the HSP90–CDC37 complex. (B) Ring E of compound 4 points to the residue side of Glu 33 and Arg 167 without blocking their interaction.

4. Conclusions

As a novel target, CDC37 modulates the maturation of multiple kinase proteins. Thus, targeting CDC37 could selectively decrease kinase levels in tumor cells. However, recently, few inhibitors have been reported to bind to CDC37 with a clear mechanism and high specificity. The key problem in developing CDC37 inhibitors for urgent problems is the discovery of hits with high specificity. In this study, we discovered a novel compound, betulinic acid acetate (4), that can bind to CDC37 with a binding affinity of 25.7 μM. Since 4 is not an α,β-unsaturated carbonyl compound, its binding to CDC37 is more direct and more specific than that of celastrol. In the anti-proliferative assay, compound 4 displayed a significant inhibition effect with micromolar IC₅₀ values. In the western blot assay, 4 resulted in the dose-dependent decrease of HSP90 client proteins. More importantly, it did not induce HSR in PC-3 cells. With further study of its mechanism and optimization of its activity, compound 4 will be useful not only as a molecular probe to help understand the biological function of CDC37, but also for the potential development of novel antitumor drugs.

Acknowledgements

This work was supported by projects 81573346, 81573281, 81502983 and 81502990 of the National Natural Science Foundation of China; BK20150691 of the Natural Science Foundation of Jiangsu Province; 2015ZD009 of the key program of China Pharmaceutical University; and 2013ZX09402102-001-005 of the National Major Science and Technology Project of China (Innovation and Development of New Drugs).

Notes and references

1. S. Paul and S. Mahanta, Mol. Cell. Biochem., 2014, 386, 45–61.
2. L. Whitesell and S. L. Lindquist, Nat. Rev. Cancer, 2005, 5, 761–772.
3. J. Trepel, M. Mollapour, G. Giaccone and L. Neckers, Nat. Rev. Cancer, 2010, 10, 537–549.
4. M. A. Biamonte, R. Van de Water, J. W. Arndt, R. H. Scannevin, D. Perret and W. C. Lee, J. Med. Chem., 2010, 53, 3–17.
5. R. Garcia-Carbonero, A. Carnero and L. Paz-Ares, Lancet Oncol., 2013, 14, 358–369.
6. J. M. Holzbeierlein, A. Windsperger and G. Vielhauer, Curr. Oncol. Rep., 2010, 12, 95–101.
7. J. R. Porter, C. C. Fritz and K. M. Depew, Curr. Opin. Chem. Biol., 2010, 14, 412–420.
8. J. A. Hall, S. Seedarala, N. Rice, L. Kopel, F. Halaweish and B. S. Blagg, J. Nat. Prod., 2015, 78, 873–879.
9. W. B. Pratt, J. Biol. Chem., 1993, 268, 21455–21458.
10. C. Ratzke, B. Hellenkamp and T. Hugel, Nat. Commun., 2014, 5, 4192.
11. M. MacLean and D. Picard, Cell Stress Chaperones, 2003, 8, 114–119.
12. J. R. Smith, P. A. Clarke, E. de Billy and P. Workman, Oncogene, 2009, 28, 157–169.
13. S. Sreeramulu, S. L. Gande, M. Gobel and H. Schwalbe, Angew. Chem., Int. Ed. Engl., 2009, 48, 5853–5855.
14. T. Zhang, A. Hamza, X. Cao, B. Wang, S. Yu, C. G. Zhan and D. Sun, Mol. Cancer Ther., 2008, 7, 162–170.
15. H. Yang, D. Chen, Q. C. Cui, X. Yuan and Q. P. Dou, Cancer Res., 2006, 66, 4758–4765.
16. T. S. Rush 3rd, J. A. Grant, L. Mosyak and A. Nicholls, J. Med. Chem., 2005, 48, 1489–1495.
17. M. D. AbdulHameed, S. Chaudhury, N. Singh, H. Sun, A. Wallqvist and G. J. Tawa, J. Chem. Inf. Model., 2012, 52, 492–505.
18. S. N. Rao, M. S. Head, A. Kulkarni and J. M. LaLonde, J. Chem. Inf. Model., 2007, 47, 2159–2171.
19. T. Usha, S. K. Middha, A. K. Goyal, M. Karthik, D. Manoj, S. Faizan, P. Goyal, H. Prashanth and V. Pande, J. Biomed. Res., 2014, 28, 406–415.
20. Y. Chen, Z. L. Liu, T. M. Fu, W. Li, X. L. Xu and H. P. Sun, Bioorg. Med. Chem. Lett., 2015, 25, 3442–3446.
21. H. P. Sun, J. M. Jia, F. Jiang, X. L. Xu, F. Liu, X. K. Guo, B. Cherfaoui, H. Z. Huang, Y. Pan and Q. D. You, Eur. J. Med. Chem., 2014, 79, 399–412.
22. J. Kirchmair, S. Distinto, P. Markt, D. Schuster, G. M. Spitzer, K. R. Liedl and G. Wolber, J. Chem. Inf. Model., 2009, 49, 678–692.
23. H. Sun, X. Xu, X. Wu, X. Zhang, F. Liu, J. Jia, X. Guo, J. Huang, Z. Jiang, T. Feng, H. Chu, Y. Zhou, S. Zhang, Z. Liu and Q. You, J. Chem. Inf. Model., 2013, 53, 2093–2210.

---

Footnote

† Electronic supplementary information (ESI) available: Celastrol inhibition of HSP90–CDC37 formation by HTRF. The IC₅₀ curves that produced the values in Table 2. See DOI: 10.1039/c6ra04776a

---