RPMI 1640 medium, Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum (FBS), and streptomycin/penicillin for cell culture were obtained from PAA Laboratories (Linz, Austria). Wang resin was obtained from Synthesis Technologies Inc. (Tianjin, China). DMAP, DCC and DMF were purchased from Dikma Technologies Inc. (Beijing, China). Fmoc-L-Gly-OH and HBTU came from Tianmapharma Co., Ltd (Suzhou, China). Piperidine and NMM were obtained from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). Carboxyfluorescein diacetate, succinimidyle ester (CFSE) was from Molecular Probe (Eugene, OR) and 7-Amino-actinomycin D (7-AAD) was purchased from Anaspec (San Jose, CA). Novobiocin (NB) and dimethyl sulfoxide (DMSO) were purchased from Sigma (St. Louis, MO). 17-allylamino-17- demethoxygeldanamycin (17-AAG) was obtained from Invivogen (San Diego, CA). Anti-HSF1 antibody (Cat. 4356) was obtained from Cell Signaling Technology Inc. (Boston, MA). Anti-HSP70, anti-HSP40, anti-HSP90, and anti-p-HSF1 (Ser326) antibodies were purchased from ENZO Life Sciences Inc. (Farmingdale, NY, USA). Anti-p-HSF1 (Ser303) was purchased from Abcam (Cambridge, MA, US). Nuclear and Cytoplasmic Protein Extraction Kit, BCA protein assay reagent kit and Beyo ECL Plus for western blot were purchased from Beyotime Biotechnology (Jiangsu, China). Phosphatase inhibitor cocktail tablets were obtained from Roche (Mannheim, Germany). All reagents were stored according to manufacturer recommendations.

Celastrol was extracted as previously reported by us 12 . Celastrol and 17-AAG were dissolved in 50 mM and 1 mg/ml in DMSO, respectively. NB was dissolved in ddH₂O. All of these drugs were stored at -20°C and used within 3 months of preparation. The stored solution was further diluted with RPMI 1640 medium or DMEM to a proper lower concentration immediately before experiments.

The seven kinds of human cancer cell lines used in this study were obtained from the Shanghai Cell Bank of the Chinese Academy of Sciences, including breast cancer cell lines MCF-7 and MDA-MB-468, prostate cancer cell line PC3, hepatic cancer cell line HepG2, leukemic cell lines THP-1, U937, and NB4. Cells were maintained in RPMI 1640 or DMEM supplemented with 10% FBS, 100 IU/ml penicillin and 100 μg/ml streptomycin in a humidified 5% CO₂ incubator at 37°C. Exponentially growing cells were used for experiments. Cells were seeded into 96-well or 6-well culture plates followed by exposure to the indicated doses of celastrol, 17-AAG, or NB for the indicated times. The culture medium with DMSO (vehicle) served as control. The final concentration of DMSO never exceeded 0.1%. Each experiment was repeated at least three times.

Cells were incubated in lysis buffer and cleared by centrifugation at 13,000 × g for 10 min. For the phosphorylation protein assay, phosphatase inhibitor was added to suppress the activity of phosphatase. The extraction of cytoplasmic and nuclear protein was performed according to product manufacturer instructions. A BCA protein assay reagent kit determined protein concentrations. Aliquots of samples were subjected to 10% SDS-polyacrylamide gels and then transferred to polyvinylidenedifluoride (PVDF) membranes. Membranes were probed with the indicated antibodies. Detection was accomplished using corresponding horseradish peroxidase (HRP)-conjugated secondary antibodies followed by development with Beyo ECL Plus; pictures were captured by G: BOX iChemi XR (Syngene Inc., UK).

Cellular localization of HSF1 was assessed by immunofluorescence. PC3 monolayers grown on coverslips were exposed to 600 nM of celastrol for 10 min. At the end of the experimental period, PC3 monolayers were washed twice in cold PBS and fixed with 2% paraformaldehyde for 20 minutes. After being permeabilized with 0.1% Triton X-100 in PBS at room temperature for 20 minutes, monolayers were then incubated in blocking solution composed of bovine serum albumin and normal donkey serum in PBS for 1 hour. Cells were then labeled with primary antibodies in blocking solution overnight at 4°C. After being washed with PBS, the cells were incubated in Alxa Fluor 488 conjugated-secondary antibody for 1 hour at room temperature. Before being mounted on microscope slides, cells were incubated in PI at 37°C. Immunolocalization of HSF1 protein was visualized using an Olympus fluorescence microscope (Tokyo, Japan).

The celastrol analogue we used was celastrol attached to tri-petite of cylcine, a structure with similar length but less complicated than those used by Klaić et al to modify celastrol 13 . It was synthesized as follows: Wang resin was put into a tube and DMF (15 mg/ml) was added. After shaking for 30 min, the solution was filtered out. Then, Fmoc-L-Gly-OH, DMAP, DCC, and DMF were added sequentially, after reaction with shaking for 30 min, acetic anhydride was added as a blocker. 20% piperidine in DMF was used to remove the protective group Fmoc, followed by successive washing with DMF, methanol, and DMF, then, Fmoc-L-Gly-OH, HBTU and NMM were added, and a condensation reaction was carried out for 30 min, and the system was washed and 20% piperidine in DMF used to remove the Fmoc group. The system was washed again, and the condensation reaction repeated until the third glycine was coupled. Then, celastrol was added and connected to the tri-peptide of glycine through another 30 min round of condensation reaction in the presence of HBTU and NMM. The reaction system was washed as in previous steps and additionally washed with DCM three times. The resin was dried, and the synthesized product was cut and purified with semi-preparative HPLC (SHIMADZU, Japan). Celastrol analogues and tri-peptide alone were identified by API 150EX Mass Spectrometer System (PE Sciex Corp. USA).

At the end of the indicated time points, cells were collected and the cells were enumerated. Accurate enumeration was carried out by FCM based on a single-tube platform with self-made cell-Beads as internal controls, a method originally reported by Harrison et al. 14 and modified by us 15 . Briefly, samples were collected followed by the addition of a known number of self-made CFSE-containing Cell-Beads. Before analysis by FACScalibur flow cytometer (Becton-Dickinson, CA), 7-AAD was added with a final concentration of 1 μg/ml for separating dead cells. The FL1 detector was used for discrimination between Cell-Beads and tested cells. The FL3 detector was used to discriminate vital cells from dead. 10,000 events were detected. The number of vital (or dead) cells was calculated using the following equation:

Number of vital dead cells = [ number of vital dead cells detected / number of Cell − Beads detected ) ] × number of Cell − Beads input

U937 cells were incubated with different doses of celastrol or the inhibitors for 24 h, and the number of vital, dead, and total cells were counted by FCM. D_m value (the median-effect dose or concentration) of each drug was obtained using Calcusyn 2.0 software. According to the D_m value, various concentrations of the single agents and the combinations with a fixed constant ratio were tested. The drug interaction study was analyzed with Calcusyn 2.0 software.

Data in this study are presented as mean ± SD. Student’s t-test or One-way analysis of variance (ANOVA) was used for statistical evaluation of significant differences among the groups using SPSS statistics 17.0 software. A value of P < 0.05 was considered to be statistical significance. Experiments were repeated at least three times.

For the first strategy, we selected 7 cancer cell types and detected the expressions of HSPs in these cancer cell lines when treated with 600 nM celastrol for 24 h. (This dose was selected by pre-experimental screening, a process that was also applied to HSP90 inhibitors 17-AAG and NB). Western blot results showed that celastrol use resulted in significant up-regulation of HSP70 in all 7 of these cancer cell lines (Figure 1A). Celastrol treatment-caused HSP90 elevation was not found in 5 of the 7 tested cell lines, excepting PC3 and NB4. Another member of the HSP family, HSP40, was not significantly affected by celastrol (Figure 1A).

The HSP90 inhibitor 17-AAG has been reported to induce HSPs while another HSP90 inhibitor, NB, cannot 16 17 ; thus these two agents were chosen as positive and negative agents to confirm the reliability of our experimental system. The effects of 17-AAG and NB in inducing HSP70 were observed in these 7 cell lines. The results showed that 17-AAG could significantly induce HSP70 elevation, while NB could not (Figure 1B). These results for 17-AAG and NB were consistent with previous reports on these two agents, indicating that our system is reliable for evaluating celastrol’s ability to induce HSP70.

The above results suggested that the induction of HSPs, especially HSP70, is a common action of celastrol. That celastrol’s effects are not cancer cell type dependent might be explained by another suggested effect of celastrol, HSF1 activation. HSF1 is ubiquitous molecule in mammalian cells, and its activation could cause HSP70 expression. Therefore, we observed the effects of celastrol on HSF1 in PC3 cells. Western blot analysis showed that celastrol treatment at 600 nM for 10 min could significantly phosphorylate ser303 and ser326 on HSF1 when compared to the control (Figure 1C). 17-AAG was also found to cause HSF1 phosphorylation, while NB did not. Then, we observed the effects of celastrol on HSF1 distribution. The total amounts of HSF1 were similar in both the DMSO-treated control and celastrol-treated cells, however, a reduction in cytoplasm and elevation in the nuclear extract of this protein were revealed in cells treated with celastrol at 600 nM for 10 min (Figure 1D), implying that some of the HSF1 was rapidly relocated from cytoplasm to nuclei after celastrol was loaded. Consistent with this analysis, the antibody (Cat. 4356 from Cell Signaling Technology Inc., specific to the nuclear form of HSF1) presented more intensified nuclei staining in cells treated by celastrol for 10 min, indicating celastrol’s nuclear accumulation of this transcript factor (Figure 1E). Phosphorylation and cellular distribution detections showed that celastrol rapidly activated HSF1 in our system.

For the second strategy, we modified celastrol’s carboxyl group (suggested to be responsible for HSP70 induction) by attaching tri-peptide of glycine via peptide bond formation (Figure 2A), and then observed the effects of this modified celastrol on HSP70 induction and cellular survival. The purity of celastrol analogues or tri-peptide of glycine was over than 95% (Figure 2B). The results showed the modified celastrol did not induce HSP70 and phosphorylate HSF1 (Figure 2C and D), but also could not inhibit proliferation (Figure 2E). Since this attached structure, when used alone, showed no effects on cellular HSP70 levels, disability of the modified celastrol to inhibit proliferation should not be due to the attached tri-peptide of glycine (Figure 2C and D). We also tried modifying celastrol’s carboxyl group by attaching one glycine rather than tri-peptide of glycine, but the results were the same (data not shown).

In the third strategy, we tried to find signaling protein inhibitors that might specifically inhibit the HSP70 induction pathway but not interfere with celastrol’s proliferation inhibition. 11 inhibitors towards 10 signaling molecules were observed, including PI3K/AKT inhibitor wortmanin, PKC inhibitor staurosporine, mTOR1/2 inhibitor KU-0063794, JNK inhibitor SP600125, peptide deformylase (PDF) inhibitor actinonin, MEK1/2 or MEK1 inhibitor U1026 and PD98059, IKK inhibitor IKK-16, p38 MAPK inhibitor SB203580, and JAK inhibitor AG490, NF-κB inhibitor PDTC. 5 of the inhibitors significantly reduced celastrol-induced HSP70: PI3K/AKT inhibitor, PKC inhibitor, mTOR1/2 inhibitor, JNK inhibitor, and PDF inhibitor. Among these, the PDF inhibitor, actinonin, had the most obvious HSP70 reduction effect. NF-κB inhibitor PDTC exerted an enhancing action on HSP70 induction. The remaining 5 inhibitors had no significant effects on celastrol-induced HSP70 (Figure 3).

To determine if the inhibitors that reduced HSP70 expression could enhance the proliferation inhibition caused by celastrol, we observed the combinative effects of celastrol and these inhibitors in U937 cells. The results showed that only the combination of actinonin and celastrol had a synergetic action in proliferation inhibition (based on combination index (CI) values <1) (Figure 4A). The other four inhibitors (PI3K/AKT inhibitor, mTOR1/2 inhibitor, JNK inhibitor, and PKC inhibitor) reduced HSP70 levels, but also antagonized celastrol’s proliferation inhibition (CI value >1) (Figure 4A).

Next, we observed the relationship between the inhibitors for HSP70 induction and HSF1 activation. When treated with celastrol for 10 min, the phosphorylations of AKT, mTOR and HSF1 in U937 cells were dramatically increased. JNK phosphorylation was slightly elevated. The inhibitors of PI3K/AKT, mTOR, JNK, PKC could reduce AKT and mTOR phosphorylation (activations) caused by celastrol, but had no obvious or elevating effect on p-HSF1. Interestingly and strikingly, the PDF inhibitor, actinonin, significantly enhanced celastrol-induced HSF1 phosphorylation (Figure 4B). This demonstrates that these compatible HSP70 inhibitors worked downstream of HSF1 activation.