Hep3B, SK-Hep1, and HeLa cells were grown in DMEM with penicillin (100 U/ml)-streptomycin (100 U/ml) (Invitrogen, Carlsbad, CA, USA) and 10% heat-inactivated fetal bovine serum (Hyclone, Logan, UT, USA). All cells were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). MG-132 and cycloheximide (CHX) were from Sigma (St. Louis, MO, USA). The hypoxic culture was kept in a gas-controlled chamber (Thermo Electron Corp., Marietta, OH, USA) maintained at 1% O₂, 94% N₂, and 5% CO₂ at 37°C. In some experiments, cobalt chloride was used to induce hypoxia mimicking conditions. Cobalt chloride was reported as a widely used mimetic of hypoxia in a large range of cells; the molecule is known to inhibit prolyl hydroxylases leading to HIF-1α stabilization (15). Celastrol was isolated from T. kirilowii and its structure is shown in Fig. 1A. The purity of celastrol was >98% in the HPLC analysis.

The ability of the compound to inhibit hypoxia inducible factor was determined by HRE-dependent reporter assay as previously described (16). In brief, at 50–80% confluence, Hep3B cells, which were co-transfected with the vectors for pGL3-HRE-Luciferase plasmid containing six copies of HREs derived from the human VEGF gene and with pRL-CMV (Promega, Madison, WI, USA) using Lipofectamine plus reagent (Invitrogen, Carlsbad, CA, USA). Following 24 h incubation, the cells were treated with various concentrations of celastrol and incubated for 16 h in hypoxia. Luciferase assay was performed using dual-luciferase reporter assay system according to the instructions of the manufacturer (Promega). Luciferase activity was determined in Microlumat plus luminometer (EG&G Berthold, Bad Wildbad, Germany) by injecting 100 μl of assay buffer containing luciferin and measuring light emission for 10 sec. The results were normalized to the activity of renilla expressed by cotransfected Rluc gene under the control of a constitutive promoter.

The ability of cells to invade through Matrigel-coated filters (invasion) was determined using a modified 24-well Boyden chamber (Corning Costar, Cambridge, MA, USA; 8 μm pore size) as previously described (17). Hep3B cells were seeded at a density of 5×10⁴ cells in 100 μl DMEM containing 10% FBS in the upper compartment of transwell. To determine the effect of celastrol, various concentrations of celastrol were added to the lower or upper compartment of transwell. After incubation for 24 h at 37°C in 5% CO₂, the cells that did not penetrate the filter were completely wiped out with a cotton swab, and the cells that had migrated to the lower surface of the filter were fixed, stained and counted in 5 randomly selected microscopic fields (×100) per filter. Cell viability was measured by an MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay (Sigma). Briefly, untreated cells or treated cells with celastrol in a 96-well plate were incubated for 24 h followed by the addition of MTT to the cells. Optical densities were determined on a microplate reader (Molecular Devices, Sunnyvale, CA, USA).

Whole-cell extracts were obtained by lysing cells in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 1% Nonidet P-40, 1 mM sulfonyl fluoride) supplemented with the protease inhibitor cocktail (BD Biosciences, San Diego, CA, USA). HIF-1α protein was analyzed in nuclear extracts prepared from cells using NE-PER reagent (Pierce, Rockford, IL, USA), according to the instructions of the manufacturer. An aliquot of protein extracts was used to determine protein concentration by the Bradford method. Fifty micrograms of whole-cell extracts or 30 μg of nuclear extract protein per lane was separated by SDS-polyacrylamide gels and followed by transferring to a polyvinylidene difluoride membrane (Millipore, Bedford, MA, USA). The membrane was blocked with 5% skim milk, and then incubated with the corresponding antibody. Antibody for HIF-1α was obtained from BD Biosciences (1:250). Antibodies for mTOR, phospho (Ser2448)-specific mTOR, p70S6K, phospho (Thr389)-specific p70S6K, ERK1/2, phospho (Thr202/Tyr204)-specific ERK1/2, eIF4E, phospho (Ser209)-specific eIF4E, p38, phospho-specific p38, SAPK/JNK and phospho-specific SAPK/JNK were purchased from Cell Signaling Technology (Beverly, MA, USA). Antibodies for HIF-1β and topo-I were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Antibody for α-tubulin was from Sigma. After binding of an appropriate secondary antibody coupled to horseradish peroxidase, proteins were visualized by enhanced chemiluminescence according to the instructions of the manufacturer (Amersham Pharmacia Biotech, Buckinghamshire, UK).

Hep3B cells were plated in 96-well plates at a density of 1×10⁵ cells/well and treated with various concentrations of celastrol for 16 h under normoxic or hypoxic conditions. The VEGF levels in the culture supernatant were determined by ELISA using the Duo-Set ELISA development kit (R&D Systems, Inc., Minneapolis, MN, USA), according to the manufacturer’s instructions.

Total RNA from Hep3B cells was obtained using RNA Mini kit (Qiagen, Valencia, CA, USA). Total RNA (2 μg) was used to perform reverse transcription-PCR (RT-PCR) using RT-PCR kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. The PCR primers were: VEGF, 5′-GCTCTACCTCCACCATGCCAA-3′ (sense) and 5′-TGGAAGATGTCCACCAGGGTC-3′ (antisense); EPO, 5′-CACTTTCCGCAAACTCTTCCG-3′ (sense) and 5′-GTCA CAGCTTGCCACCTAAG-3′ (antisense); HIF-1α, 5′-CTCA AAGTCCGACAGCCTCA-3′ (sense) and 5′-CCCTGCAGT AGGTTTCTGCT-3′ (antisense); GAPDH, 5′-ACCACAGTC CATGCCATCAC-3′ (sense) and 5′-TCCACCACCCTGTT GCTGTA-3′ (antisense). The oligonucleotide sequences of the reaction products were confirmed by sequencing.

All surgical procedures and careful handling of the animals were in accordance with IACUC guidelines. Six-week-old specific pathogen-free Crj:BALB/c nu/nu female athymic nude mice (Vital River, China) were randomly assigned to three groups, each of which consisted of five mice (n=5 per group), and were then subcutaneously inoculated with 0.2 ml of Hep3B cells (5×10⁷ cells/ml) in the right flank region. Celastrol, dissolved in DMSO, was administered orally three times a week for 35 days at a dose of 3 and 10 mg/kg body weight starting from the ten days post cell implantation with mice. Tumor volume was calculated every five days using the equation: (Length × (width)2)/2. Tumors were harvested 4 h after the last treatment, followed by homogenizing in RIPA for western blotting analysis.

All values are expressed as mean ± SD. A comparison of the results was performed with one-way ANOVA and Tukey’s multiple comparison tests (Graphpad Software, Inc., San Diego, CA, USA). Statistically significant differences between groups were defined as P-values <0.01.

To investigate whether celastrol inhibited HIF-1α transcriptional activation, we transfected Hep3B cells with a luciferase reporter gene driven by six specific HREs. A substantial increase of luciferase activity was observed in cells cultured in hypoxic conditions, whereas celastrol dose-dependently inhibited hypoxia-induced luciferase activity (Fig. 1B, left panel). Given that the inhibition of HIF-1α transcriptional activation might be correlated with celastrol-induced cytotoxicity, parallel studies of cell viability were performed (Fig. 1B, right panel). After the Hep3B cells were treated with celastrol (up to 10 μM) for 24 h, no significant alteration of cell viability was observed relative to the untreated control group.

To explore the mechanism underlying celastrol activity, we investigated its effect on HIF-1α protein levels. In Hep3B cells, HIF-1α protein is undetectable under normoxia, whereas it is stabilized under hypoxia or in the presence of CoCl₂ and becomes readily detectable by western blotting. Following 12-h treatment, celastrol exerted dose-dependent inhibition of HIF-1α protein levels induced by hypoxia or CoCl₂ in Hep3B cells. (Fig. 1C, top two panels). In contrast to the decrease of HIF-1α levels, celastrol had almost no effect on the levels of HIF-1β and topo-I proteins. Next, in order to address whether the inhibition of HIF-1α by celastrol was cell line-specific, we extended these studies to a diverse set of tumor cell lines with tissues of various origins, including the hepatic cancer cell lines SK-Hep1 and epithelial cervical cancer cell lines HeLa cells. Fig. 1C shows that, under hypoxic conditions, HIF-1α accumulation was strongly suppressed by celastrol in all cell lines.

Generally, the accumulation of HIF-1α is dependent on the balance between its protein synthesis and degradation. To specifically investigate whether celastrol modulates HIF-1α protein synthesis, we used the proteasome inhibitor MG-132 to prevent HIF-1α degradation. This allowed us to observe HIF-1α stabilization and accumulation following de novo protein synthesis. Due to rapid proteosomal destruction and undetectable HIF-1α levels in normoxic cells, the accumulation rate of HIF-1α via proteosomal inhibition reflects the synthesis rate of the protein (18). As shown in Fig. 2A, B and C, HIF-1α rapidly accumulated over a period of 4 h in the presence of MG-132 under both normoxia and hypoxia, as well as in the presence of CoCl₂. In contrast, co-treatment with celastrol and MG-132 resulted in a much slower rate of HIF-1α accumulation (compare lane 4 with lane 5 for CoCl₂ in Fig. 2A, lanes 2–4 with lanes 5–7 under normoxia in Fig. 2B, and lanes 2–4 with lanes 5–7 under hypoxia in Fig. 2C). No effects were observed on HIF-1β and topo-I. These results indicate that HIF-1α protein synthesis in Hep3B cells is markedly impaired in the presence of celastrol. To address the effect of celastrol on HIF-1α protein stability, the protein translation inhibitor CHX was used to prevent de novo HIF-1α protein synthesis. Accordingly, we determined whether celastrol affects HIF-1α protein stability by the addition of 200 μM CoCl₂ (Fig. 3). HIF-1α accumulation increased in a time-dependent manner, but the addition of celastrol resulted in the significant abrogation of HIF-1α accumulation at every time point (Fig. 3A). To determine whether HIF-1α protein half-life was affected by celastrol, we firstly induced HIF-1α accumulation in the presence of 200 μM CoCl₂ for 4 h, and then added CHX alone or in combination with celastrol. In the presence of CHX, HIF-1α levels rapidly declined in both the celastrol-untreated and celastrol-treated cells, showing a half-life was ~1.5 h in Hep3B cells. These results indicated that celastrol did not modify the degradation rate of HIF-1α (Fig. 3B, lanes 3–5 with lanes 6–8).

To determine whether HIF-1α synthesis inhibition by celastrol was a downstream effect from decreased HIF-1α gene transcription or HIF-1α mRNA stability, we analyzed HIF-1α mRNA levels by RT-PCR. Celastrol did not change HIF-1α mRNA levels under either normoxia or hypoxia during 16 h of treatment (Fig. 5A). This suggests that celastrol-mediated decrease of HIF-1α synthesis is due to downregulation of HIF-1α mRNA translation.

To reveal the underlying mechanism by which celastrol inhibits hypoxia-induced activation and translation of HIF-1α, we first examined the phosphorylation status of translation initiation factors such as mTOR, p70S6K, eIF4E, and ERK1/2 under hypoxia. As shown in Fig. 4A and B, in parallel with the alteration of HIF-1α protein, celastrol dose-dependently inhibited the expression of phospho-mTOR, phospho-p70S6K, phospho-eIF4E, and phospho-ERK1/2, induced by hypoxia; however, it had no inhibitory effect on phospho-p38 and phospho-JNK, and total protein levels for mTOR, p70S6K, eIF4E, ERK1/2, p38, and JNK. These findings were partially in line with a report showing that mTOR/p70S6K and ERK pathways are involved in HIF-1α protein synthesis via functional activation of the translational regulatory protein eIF4E in various cells (19). All of the results consistently support the idea that celastrol inhibits hypoxia-induced activation of HIF-1α through inhibition of mTOR/p70S6K/eIF4E and ERK signaling.

The expression of VEGF and EPO, which are involved in tumor cell proliferation, angiogenesis, invasion and metastasis, is known to be regulated by HIF-1α (11). We therefore examined whether celastrol can suppress the expression of these genes. VEGF and EPO mRNA levels were measured by RT-PCR analysis in Hep3B cells. Treatment of the cells with celastrol resulted in a dose-dependent inhibition of VEGF and EPO mRNA expression (Fig. 5A). The concentrations to inhibit the expression of HIF-1α target genes were comparable with those of HIF-1α protein accumulation. This result led us to measure the VEGF protein concentration in the culture supernatant by ELISA. Consistently, the hypoxic induction of secreted VEGF protein was dose-dependently inhibited by celastrol (Fig. 5B). Reduced expression of VEGF and EPO might be responsible for diminished invasion of tumor cells in celastrol treatment. Therefore, whether celastrol modulates invasion activity was examined in vitro with a Matrigel invasion assay. Hep3B cells were seeded in the top chamber of a Matrigel invasion chamber and were incubated with various concentrations of celastrol for 16 h. The result showed that celastrol significantly decreased invasiveness compared to the vehicle control, accounting for the anti-invasive activity of celastrol (Fig. 5C).

Since celastrol suppresses both the expression of HIF-1α target genes and the invasiveness of tumor cells, we next determined whether these results could be translated into an in vivo xenograft model. Hep3B cells were subcutaneously implanted in athymic nude mice, and the experimental mice were treated with celastrol (3 and 10 mg/kg) 3 days a week until the end of the study. As expected, the administration of celastrol (10 mg/kg) significantly inhibited Hep3B tumor growth up to 52.9%, compared to that of the vehicle-treated control group (Fig. 6A). Due to the key roles of HIF-1α in tumor angiogenesis, we studied its expression in the tumors by western blotting. Consistent with the finding in cultured cells, celastrol significantly decreased the protein levels of HIF-1α in the tumors, whereas no differences were observed in topo-I (Fig. 6B).