We retrospectively studied the prospective kidney cancer database of the second Xiangya Hospital of Central South University, including follow-up data of 362 consecutive patients from 2011 to 2017. A total of 4 ccRCC patient samples were obtained from the second Xiangya Hospital of Central South University between 2017 and 2019. The samples included both cancer tissue and adjacent normal tissue (1 cm away from the margin of the tumor tissues). All patients included in the study had no history of adjuvant therapy (chemotherapy or radiotherapy).

Total RNA was extracted from human ccRCC and patient-matched normal tissues with TRIzol (Invitrogen, Thermo Fisher Scientific, Inc. United States) according to the manufacturer’s protocol. RNA quality was examined by gel electrophoresis, and only paired RNA of high quality was used for RNA sequencing. RNA sequencing libraries were prepared according to the manufacturer's instructions and then sequenced with the Illumina HiSeq 2000 at Riobo company (Guangzhou).

We obtained the standardized mRNA expression of CAV-1 in ccRCC and disease-free survival time (DFS) from the TCGA database and GTEx database.

Cultured cells, clinical sample or xenograft tumor tissues were extracted from radio-immunoprecipitation assay (RIPA, Cwbiotech, China) buffer and protease inhibitor cocktail (Cwbiotech, China). Protein concentrations were determined by the BCA™ Protein Assay Kit according to the protocol provided by the manufacturer (Cwbiotech, China). Protein (50 μg) loaded in each lane was subjected to 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and then transferred to a polyvinylidene difluoride (PVDF, Merck Millipore, Billerica, MA, United States) membrane. After blocking with 5% skimmed milk for 2 h, PVDF membranes were incubated with a primary antibody overnight at 4°C, and then incubated with a secondary antibody for 1.5 h. The protein bands were detected using visualizer (Tanon, China).

Human ccRCC cell lines (786-O, A498, SN12C and OS-RC-2) and renal epithelial cells (HK-2 cell) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). ccRCC cells were cultured in RPMI1640 (Gibco, United States) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA) and 1% streptomycin-penicillin (Gibco, Grand Island, NY, United States). Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Grand Island, NY, United States) containing 10% fetal bovine serum (FBS; Gibco, Grand Island, NY, United States) and 1% streptomycin-penicillin were used for HK-2 culture. All cells were maintained at 37°C in a humidified atmosphere with 5% CO2.

The total RNA was isolated using the TRIzol isolation methods following the manufacturers’ recommendations. The RNA concentrations were quantified using Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). The mRNA expression was calculated based on the 2^−ΔΔCq method. qPCR was conducted using an Applied BioRad CFX96 Real-Time PCR. The primers used for qPCR are available on request.

Male BALB/c nude mice (Slac, SCXK (Xiang)2014-0002, 4 weeks old) were kept in a specific pathogen-free animal laboratory. The animal study was reviewed and approved by Hunan University of Chinese Medicine. BALB/c nude mice were originated from the mus musculus. The two main defects of BALB/c nude mice are failure of hair growth and hypoplasia of thymic epithelium due to developmental failure of the thymic anlage. BALB/c nude mice also have poor response to thymus-dependent antigens because of the defect in helper T-cell activity. For subcutaneous xenograft models, approximately 1.0 × 10⁷ 786-O cells were subcutaneously into the left flank of mice, which were randomly assigned to 6 groups (n = 3) by using blinded allocation. BALB/c nude mice without tumorigenesis were excluded. BALB/c nude mice fed with a normal diet (control group) and high-fat diet (HFD group). The groups treated with CeT were fed with HFD. Body weight of mice and tumor volumes were measured every two days. Tumor volumes were recorded at the indicated time points by measuring the tumors with calipers and calculating the volumes with the formula L×W²/2 (L is the length of the longer axis and W is the length of the shorter axis). Four weeks after the injection, all mice were euthanized under anesthesia with isoflurane inhalation (2%), and the tumors were excised, weighed, and photographed. Some tumors were stored at −80°C for western blot analysis, others were fixed in 4% paraformaldehyde for staining experiment.

Tissues were fixed in 4% paraformaldehyde and processed for histologic examination, including embedding in paraffin, sectioning, and staining with hematoxylin and eosin (H&E). The formaldehyde-fixed epithelial tissues were dehydrated using gradient ethanol (70%, 80%, 90%, 95%, 100%, 5 min each), cleared twice with xylene (10 min each), embedded with paraffin, and subsequently cut into 4 μm serial sections. The sections were baked at 60 °C for 1 h and dewaxed with xylene. Following the removal of xylene using gradient ethanol, the sections were washed, stained with hematoxylin for 10 min, and washed with distilled water for a 1 min period. After being differentiated with 1% hydrochloric ethanol for 20 s, the sections were washed with distilled water for 1 min and turned blue using 1% ammonia for 30 s. The sections were then additionally stained with eosin for 3 min, dehydrated with gradient ethanol (2 min for each), cleared twice with xylene (5 min for each), and mounted using neutral balsam. Then the samples were observed under an optical microscope (×40) (Olympus, Tokyo, Japan).

Blood was collected after 4 weeks of treatment with the corresponding medicine into heparinized tubes. Biochemical parameters, including plasma high density lipoprotein (HDL), triglyceride (TG), total cholesterol (TC), low density lipoprotein (LDL), and very low density lipoprotein (VLDL) levels, were measured using an autoanalyzer (Beckman Coulter, Miami, FL, USA).

Freeze tumors were cryo-sectioned using Leica CM3050 cryotome (Leica Biosystems Inc., Wetzlar, Germany) and 7 μm sections were obtained. The slides were placed to 25°С and washed in running water to remove the optimal cutting temperature (OCT) compound (Tissue-Tek, Torrance, CA, United States). Slides were placed in 50% isopropanol for 3 min and in 100% isopropanol for 3 min and stained with 0.5% Oil red O (Solarbio, China) in 100% isopropanol for 2 h. Then, slides were differentiated in 85% isopropanol for 3 min three times, washed with running water, and stained with Mayer’s hematoxylin for 15 s followed by bluing in running water for 10 min. Slides were mounted with Glycerol Jelly Mounting Medium (Beyotime, China) before analyzed.

Tissue samples collected from ccRCC patients and tumor xenograft mice were made into 4% paraformaldehyde blocks using the following steps: formalin fixation, dehydration and paraffin embedding. The sections were incubated overnight at 4°С with each primary antibody and incubated for 1 h at 25°С with secondary antibodies. Proteins were visualized using EnVision Detection Systems (Dako Japan, Tokyo, Japan). Immunochemical results were evaluated by an experienced pathologist with a semiquantitative approach assigning an H-score (or “Histoscore”) to the tumor sample. The percentage of tumor cells was determined for each different nuclear staining intensity (0/+/++/+++), and the sum of the individual H-scores for each intensity level was then calculated using the following equation: H-score = 1 × (% of cells with an intensity of 1+) + 2 × (% of cells with an intensity of 2+) + 3 × (% of cells with an intensity of 3+).

Cell viability was evaluated using the CCK-8 kit (Beyotime, China) according to the manufacturer's instructions. Cells were seeded in 96-well flat bottom microtiter plates at a density of 5 × 10³ cells per well. The absorbance was measured on a microplate reader (Synergy HT, Bio-Tek, Biotek Winooski, Vermont, United States) at 450 nm.

We examined cholesterol efflux by using fluorescent sterol, boron dipyrromethene difluoride linked to sterol carbon-24 (BODIPY-cholesterol; No. GC42964, Glpbio, Montclair, CA, United States). 786-O cells were cultured in 6-well plates, and then cultured in serum-free medium containing 0.1 ml labeling media for 1 h. The cells were washed three times with MEM-HEPES (Gibco, Grand Island, NY, United States), and then cultured in serum-free medium containing treatment factors for 24 h. Fluorescent intensity was determined by a fluorescence microscope unit with ×40 magnification and photographed with OLYMPUS Stream system (Olympus DP73, TH4-200, Tokyo, Japan). Fluorescent intensity was quantified from at least 3 random fields per slide, from three slides per experimental condition and graphed.

Cholesterol content was determined by enzymatic assay kit (Applygene Technologies Inc., China). Cholesterol standard solution was appropriately diluted using cholesterol assay buffer. Detection reagent was added to 50 μl of cholesterol standard or collected cell lysate supernatant for 5 min. Optical density values at 535/590 nm were measured. Standard curve was prepared using cholesterol standard solution, and the concentrations of cholesterol were calculated according to the standard curve.

Cultured cells were fixed by 4% paraformaldehyde, treated with 0.1% Triton X-100 (Sigma-Aldrich, St. Louis, MO, United States) for 5 min and blocked in 3% bovine serum albumin (BSA; Solarbio, China) for 1 h. The cells were then incubated with primary antibodies with a dilution of 1:100 overnight at 4°C. After incubation with fluorescent-dye-conjugated secondary antibodies and DAPI. Immunofluorescent microscopic images of the cells were obtained and viewed by using an inverted fluorescence microscope (Olympus DP73, TH4-200, Tokyo, Japan). Quantitative analysis of fluorescence intensity was performed with Image Pro Plus 6.0 software (Media Cybernetics, United States).

The 786-O cells were plated at a density of 5 × 10³ cells per well in ultra-low attachment six well plates (Corning Incorporated, Corning, NY, United States) and incubated in DMEF/F12 (1:1) supplemented with B-27 (Gibco, Grand Island, NY, United States), 10 ng/ml fibroblast growth factor-basic (bFGF; Gibco, Grand Island, NY, United States) and 10 ng/ml epidermal growth factor (EGF; Gibco, Grand Island, NY, United States). 786-O cells were treated with corresponding agent after plating for 24 h. Every three days, half of the medium was replaced. Cells were cultured for 10 days. The generated spheroids were counted and photographed using inverted microscope (Olympus DP73, TH4-200, Tokyo, Japan). The number of spheres with a diameter of >50 μm was quantified by ImageJ software.

Three double-stranded siRNAs targeting CAV-1 or LOX-1 and a negative control siRNA were designed and synthesized by Ribobio Company (China) and transfected into 786-O cells using riboFECT™ CP Reagent (Ribobio, China) for 24 h following the manufacturer’s guidelines. CAV-1-01, CAV-1-02 and CAV-1-03 correspond to 3 different siRNA against CAV-1. LOX-1-01, LOX-1-02 and LOX-1-03 correspond to 3 different siRNA against LOX-1.

The 786-O cells were fixed in 1.6% glutaraldehyde prior to post-fixation in osmium tetroxide and uranyl acetate en bloc staining. Samples were processed and embedded in Spurr epoxy resin, thin sectioned, and counterstained with lead citrate. Digital images were obtained with a Hitachi HT7700 TEM (Hitachi, Tokyo, Japan).

The 786-O cells were lysed on ice in 20 mM Tris-HCl (pH 7.4), 1% Triton X-100, 0.025% SDS, 100 mM NaCl, 1 mM Na₃VO₄, 10 mM NaF, and 1% protease inhibitor cocktail (Sigma, United States). Soluble extracts were incubated for 2 h at 4°C with relevant antibodies: anti-CAV-1 (Abcam, United States), anti-LOX-1 (Proteintech, United States) and a negative isotype control mouse immunoglobulin (IgG) (Santa Cruz Biotechnology, United States). Immune complexes were precipitated with protein A/G agarose (Santa Cruz Biotechnology, United States). Western blotting was performed as described before.

All values are expressed as mean ± SDs. All experiments were independently repeated at least three times. A paired t-test was used to compare patient tissues. Student’s t test was used in the two groups of experiments. One-way analysis of variance (ANOVA) ANOVA with Tukey’s post-test (One-way ANOVA for comparisons between groups) were used to compare values among different experimental groups. All statistical analyses were carried out using GraphPad Prism software (La Jolla, CA, United States). The statistical significance was set at p < 0.05.

A total of 362 ccRCC patients were included in this retrospective investigation. The detailed clinicopathological data of the patients are shown in Table 1. The patients included 245 men and 117 women, with a median age of 53 years (range: 2–88 years). Among the surgical patients, 29 cases had lymph node metastasis and 27 cases had distant metastasis (Table 1). Of these patients, 168 cases had HDL levels <1.04 mmol/L, 180 cases had TG levels >1.21 mmol/L, and 60 cases had high levels of TC (>5.89 mmol/L). There were 108 cases with LDL level >3.4 mmol/L (Table 1). These data indicated a positive correlation between lipoproteins and ccRCC risk.

Microarray gene expression analysis revealed differentially expressed genes in HK-2 cells, 786-O cells and celastrol-treated 786-O cells. We first filtered genes by applying a change of ≥ 1-fold as a cut-off value to identify genes that are significantly up-and down-regulated (p < 0.05). Comparing with HK-2 cells, 493 genes were upregulated, and 472 genes were downregulated in 786-O cells (Figure 1A). The gene expression of 786-O cells in response to celastrol treatment were evaluated. There were 143 differentially expressed genes identified in 786-O cells with celastrol treatment, including 57 up-regulated and 86 down-regulated genes (Figure 1B). Then, lipid uptake-associated genes were examined, and the results were plotted in the heat map (Figure 1C). Among lipid uptake genes, CAV-1 was specifically up-regulated in 786-O cells and down-regulated by celastrol. Then, the mRNA expression of CAV-1 in ccRCC tissues and adjacent tissues from publicly available TCGA-KIRC database and GTEx-KIRC database were assessed. Bioinformatics analysis showed that the expression of CAV-1 in ccRCC tissues was increased compared with that in adjacent tissues (Figure 1D). Then, the prognostic level of CAV-1 in ccRCC was determined. The disease-free survival time of the high CAV-1 expression group was lower than that of the low CAV-1 expression group (Figure 1E). Consistently, the protein expression of CAV-1 in ccRCC tissues was remarkably increased relative to adjacent tissues (Figures 1F,G). The mRNA and protein levels of CAV-1 in ccRCC cell were also significantly higher than that in HK-2 cells, especially in 786-O cells (Figures 1H–J). These findings indicated that CAV-1 might be a potential biomarker for ccRCC diagnosis.

To assess the effect of celastrol in ccRCC in vivo, the tumor xenograft model by subcutaneous injection of 786-O cells was employed. As shown in Figure 2A, nude mice fed with HFD promoted tumor growth, while celastrol attenuated tumor growth rate in a dose-dependent manner. Tumor weight were higher in the HFD group, whereas significantly lower in the celastrol group and atorvastatin group (Figure 2A). Atorvastatin is the most commonly used anti-cholesterol drug and has been reported to inhibit the progression of various cancers, including RCC. The body weight of mice was monitored throughout the whole study, and the results showed that the HFD made the mice get heavier, but their body weight was dose-dependently decreased after each exposure to celastrol (Figure 2A). Atorvastatin at 1.5 mg/kg kept the body weight steady of nude mice even fed with HFD (Supplementary Figure S1A). Subsequently, the results of H&E staining showed that tumor xenograft model fed with normal diet exhibited cylindrical shape, large nuclei, increased nuclear/cytoplasmic ratio and cellular cleavage, as well as the glands with abnormal sizes and shapes (Figure 2B). The tumor xenograft model fed with HFD showed that the cancer was poorly differentiated and invaded into the mucous membrane, accompanied by degeneration and necrosis of crypt cells and an amount of infiltrative inflammation (Figure 2B). Celastrol significantly relieved these symptoms, only showing moderate edema in lamina propria interstitial or mild inflammatory cell infiltration (Figure 2B), and similar results were observed in atorvastatin group (Figure 2B). The lipid accumulation was increased in the high-fat diet-fed animal model and celastrol significantly reduced lipid deposition in tumor tissues (Figures 2C,D). Biochemical analysis showed that celastrol increased HDL level, while reduced plasma levels of TC, TG, LDL and VLDL in nude mice fed HFD (Figures 2E–I). In addition, the tumor resected from the HFD group expressed higher levels of CAV-1, LOX-1, CD133, Bmi-1, SOX-2, p-GSK-3β (S9)/GSK-3β and β-catenin than the tumor from the control group, while celastrol down-regulated the levels of these proteins (Figure 2J). The protein levels of these molecules were also confirmed by IHC (Supplementary Figures S1B–I). These results indicated that celastrol inhibited tumor growth of ccRCC in mice fed with high-fat diet through reducing lipid deposition and stemness proteins expression.

Celastrol significantly inhibited the viability of 786-O cells in a dose- and time-dependent manner (Figure 3A). In addition, celastrol reduced lipid accumulation in 786-O cells in a dose-dependent manner (Figure 3B). TC, free cholesterol (FC), and cholesterol ester (CE) were also decreased in celastrol-treated 786-O cells (Figure 3C). The inhibitory effect of celastrol on lipid accumulation was consistent with that of atorvastatin. Celastrol significantly decreased the protein expression of CAV-1 and LOX-1, while did not change flotillin-1 (FLOT-1) expression, a marker of lipid rafts (Figure 3D). Furthermore, celastrol inhibited the localization of CAV-1 and LOX-1 (Figure 3E). In addition, celastrol reduced the number of tumor spheres (Figure 3F). Celastrol also downregulated the levels of stemness markers, such as clusters of differentiation 133 (CD133), B-cell-specific Moloney murine leukemia virus insertion site 1 (Bmi-1), SRY-box 2 (SOX-2) in 786-O cells (Figure 3G). Collectively, these findings suggested that celastrol reduced intracellular lipid accumulation and impaired the stem cell properties, especially at 1 μΜ. Thus, 1 μΜ celastrol was chosen for subsequent experiments.

LDL is the major carrier of cholesterol to tissues. Early findings indicated that native LDL increased cholesterol content during in vitro incubations by receptor-independent, fluid-phase pinocytosis (Kruth 2011). Instead, oxidized LDL (ox-LDL), a pathologically modified lipoprotein, induces pronounced accumulation of cholesterol via its receptors LOX-1. Elevated ox-LDL levels are usually positively correlated with an increased risk of carcinogenesis (González-Chavarría, Fernandez, Gutierrez, González-Horta, Sandoval, Cifuentes, Castillo, Cerro, Sanchez and Toledo 2018). Herein, celastrol inhibited cell proliferation induced by ox-LDL (Supplementary Figure S2A). In addition, celastrol attenuated lipid accumulation and reduced the levels of TC, CE, FC induced by ox-LDL (Figure 4A, Supplementary Figures S2B–D). To confirm the presence of caveolae in 786-O cells, we performed ultrastructural studies by using TEM. Ox-LDL enrichment resulted in higher caveolae number and more caveolin-enriched vesicles (Figure 4B). In contrast, celastrol reduced the number of caveolae and increased the size of caveolar structure (Figure 4B). Representative microscopic images showed that celastrol inhibited the co-localization of CAV-1 with LOX-1 promoted by ox-LDL (Figure 4C). Celastrol also significantly decreased the expression of CAV-1 and LOX-1 in 786-O cells exposed to ox-LDL (Figure 4D). In addition, the ability of 786-O cells to generate tumor spheres was enhanced and the protein expression of CD133, Bmi-1, SOX-2 were upregulated under the culture with ox-LDL (Figures 4E,F). After treatment with celastrol, the number of spheres and stemness markers were significantly decreased (Figures 4E,F). Overall, these data demonstrated that celastrol reversed stemness induced by ox-LDL through reducing lipid accumulation.

We then examined whether celastrol inhibited lipid accumulation and prevents stemness properties via CAV-1. Firstly, 786-O cells were transfected with CAV-1-targeting siRNA (siCAV-1). As determined by western blotting, the expression of CAV-1 in three groups was more strongly decreased (Supplementary Figure S3A). Celastrol and siCAV-1 independently, significantly reduced lipid levels in general (Figure 5A). Treatment with celastrol on siCAV-1 cells significantly reduced FC and CE (Supplementary Figure S3B). Celastrol caused the loss of morphologically defined caveolae in siCAV-1 cells (Figure 5B). Furthermore, celastrol reduced the co-localization of CAV-1 and LOX-1 through inhibiting CAV-1 (Figure 5C). We also confirmed that celastrol reduced the protein expression of CAV-1 and LOX-1 in siNC cells or siCAV-1 cells (Figure 5D). Besides, silencing CAV-1 in 786-O cells impaired the potential for tumor sphere formation and the levels of stemness markers (Supplementary Figure S3C, Figure 5E). However, celastrol reduced sphere number and exhibited lower expression of stemness markers (Supplementary Figure S3C, Figure 5E). The celastrol treatment was more effective when combined with siCAV-1 as CAV-1 expression was reduced. These results demonstrated the synergistic effect of celastrol and siCAV-1 on lipid accumulation and stemness features.

We performed co-IP analysis for CAV-1 and LOX-1 in 786-O cells treated with or without celastrol. Indeed, celastrol prevented the binding of CAV-1 and LOX-1 (Figure 6A). Then, we tested whether double knockdown of CAV-1 and LOX-1 enhanced the anti-cancer activity of celastrol in 786-O cells. 786-O cells were transfected with three different siRNAs against LOX-1. The siLOX-1-03 showed the best interference efficiency (Supplementary Figure S4A). Furthermore, compared with knockdown CAV-1 or LOX-1 alone, double knockdown of CAV-1 and LOX-1 showed better interference efficiency (Supplementary Figure S4B). After double knockdown of CAV-1 and LOX-1 in 786-O cells, celastrol further remarkably inhibited lipid accumulation and TC, FC, CE levels (Figure 6B, Supplementary Figure S4C). Immunofluorescence staining further confirmed that celastrol inhibited co-localization of CAV-1 and LOX-1 in double knockdown of CAV-1 and LOX-1 cells (Figure 6C). In addition, celastrol attenuated the stem-like properties and stemness markers expression after knockdown of CAV-1 and LOX-1 (Figures 6D,E). These results suggested that celastrol exerted its inhibitory activity on lipid accumulation and stemness by suppressing the binding of CAV-1 and LOX-1.

Under this premise, we next explored whether celastrol exerted anticancer effect by inhibiting CAV-1/LOX-1 and then blocking Wnt/β-catenin pathway. Lithium chloride (LiCl) was used as an activator of the Wnt/β-catenin pathway. The siCAV-1-LOX-1 cells were treated with Wnt/β-catenin pathway activator LiCl or celastrol, respectively. LiCl increased the protein level of the phosphorylation of glycogen synthase kinase-3β (p-GSK-3β) at serine 9 and promoted nuclear translocation of β-catenin (Figure 7A). In contrast, celastrol caused a decrease of p-GSK-3β (S9) and an increase of β-catenin level in the cytoplasm. We next determined the role of celastrol on stemness mediated by LiCl. As expected, celastrol inhibited sphere formation ability induced by LiCl (Figure 7B). When treated with celastrol, the expression of stemness markers induced by LiCl were reduced (Figure 7C). Interestingly, we observed that LiCl contributed to lipid accumulation and resulted in an increment of TC, FC, and CE levels, and these effects were reversed by celastrol (Supplementary Figures S5A,B). Based on the above findings, we concluded that Wnt/β-catenin signaling pathway is involved in the inhibitory effect of celastrol on the stemness-like properties of the 786-O cells.