Oxidized Low-Density Lipoprotein Links Hypercholesterolemia and Bladder Cancer Aggressiveness by Promoting Cancer Stemness

An obesity-induced metabolic disorder is one of the most preventable causes of cancers (1). Hypercholesterolemia is a metabolic disorder commonly found in people with obesity. Therefore, hypercholesterolemia has been accepted as a risk factor in the development of the cancers derived from steroid-targeted tissues, such as breast and prostate (2–4). Because cholesterol is the precursor of steroid hormone syntheses, it is conceivable that elevated levels of cholesterol favor the synthesis of these hormones as well as the intermediates in the synthetic pathways (5, 6). It has been reported that the cholesterol metabolite 27-HC can not only interact with the estrogen receptor (ER) but also promote ER-positive breast cancer cell proliferation (4). However, little is known about the effects of hypercholesterolemia on cancers in the non–steroid-targeted tissues.

Although urinary bladder cancer (UBC) is the 10th most common malignancy worldwide, UBC is considered the costliest cancer in the United States partially due to a lengthy follow-up surveillance and related treatments (7, 8). In addition, UBC is mostly diagnosed in late life, with a median age of UBC diagnosis at 73 years (9). Therefore, reliable risk factors could be helpful in predicting UBC occurrence and subsequently leading to effective prevention of this disease. Reagents specifically targeting these risk factors could be developed as efficacious UBC therapeutics. Research in different epidemiologic studies tried but failed to establish a convincing relationship between hypercholesterolemia and UBC partially due to the relatively smaller sample sizes and incomplete patient information (10–12). However, a recent comprehensive study known as the Metabolic Syndrome and Cancer Project (Me-Can) 2.0 revealed an unambiguous positive association between high cholesterol and UBC risk (13). Nevertheless, this research was unable to establish a cause-effect relationship between hypercholesterolemia and UBC.

Cancer stem-like cells (CSC), also known as tumor-initiating cells, are the most tumorigenic subset of cancer cells involved in different processes of tumor development including initiation, recurrence, and metastasis (14). It has been reported that bladder CSCs not only exist but also express specific bladder basal cell markers CK5, CK14, and CD44. Such CSC signature can predict poor clinical outcomes (15, 16). Because transgenic mice expressing constitutively active STAT3 under the control of CK5 promoter are more prone to carcinogen BBN [N-Butyl-N-(4-hydroxybutyl) nitrosamine]-induced bladder cancer, it is suggested that STAT3 plays an important role in UBC development (17). In addition, maintenance of UBC stemness requires not only the intrinsic molecular alterations but also the extrinsic cues from the tumor microenvironment, such as cancer-associated fibroblasts, lipids, and cholesterol (18–20). In addition, it has been proposed that the elevated levels of lipids and cholesterol in the circulation resulted from metabolic syndrome could affect cancer stemness and cancer development (21).

In this study, we investigated the cause-effect relationship between hypercholesterolemia and UBC development in the diet-induced and Ldlr knockout–induced hypercholesterolemia mouse models. We first demonstrated that high level of cholesterol plays essential roles in UBC tumor growth and cancer stemness, and then addressed the following important points: (i) identifying the key factor related to hypercholesterolemia in promoting UBC cell proliferation and stemness; (ii) revealing the downstream signaling pathway induced by the key factor; and (iii) confirming the clinical relevance with intervention of hypercholesterolemia in UBCs and human UBC samples.

Human UBC cell lines T24 (RRID: CVCL_0554), 5637 (RRID: CVCL_0126), and UM-UC-3 (RRID: CVCL_1783) were purchased from the Cell Bank of Type Culture Collection, Chinese Academy of Sciences (Shanghai, P.R. China) and verified by short tandem repeat profiling (Genesky). The murine UBC cell line MB49 (RRID: CVCL_7076) was kindly provided by Jimin Gao. Cell lines were regularly tested for Mycoplasma contamination using Mycoplasma Detection Kit (40601ES20, Yeasen). Cells were maintained in RPMI1640 medium (Life Technologies), supplemented with 10% FBS (Hyclone) and 100 U/mL penicillin and 100 μg/mL streptomycin at 37°C with 5% CO₂.

For the oxidized low-density lipoprotein (ox-LDL)-related in vitro assays, cells were seeded in 3.5-cm dishes, maintained with conditioned medium with 2% FBS overnight. The medium was removed the next day while medium with different concentrations of ox-LDL (YB-002, purchased from Yiyuan Biotech, which was derived from human LDL-c oxidized with CuSO₄) was added. A total of 48 hours later, cells were harvested for other experiments referred below.

For the mouse serum–related in vitro assays, mouse sera were collected from wild-type (WT) mice and Ldlr^−/− mice (n = 3). Samples from mice with the same genotype were mixed and filtrated by 0.22-μm-diameter filter. Cells were cultured with medium with 2% FBS overnight, then changed to the conditioned medium supplemented with 10% mouse serum. For neutralizing assay, 20 μg/mL ox-LDL neutralizing antibody (TAB-775, Creative Biolabs) was added to conditioned medium supplemented with mouse serum and cultured at 37°C for 2 hours prior to treat cells. A total of 48 hours later, cells were harvested for further experiments.

Mice were maintained under specific pathogen-free conditions in the animal core facilities of the Model Animal Research Center of Nanjing University (Nanjing, Jiangsu, P.R. China) or Shanghai Institute of Materia Medica, Chinese Academy of Sciences (Shanghai, P.R. China). All animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals and approved by the Animal Care and Use Committees at Model Animal Research Center at Nanjing University (Nanjing, Jiangsu, P.R. China) and Shanghai Institute of Matera Medica, Chinese Academy of Sciences (Shanghai, P.R. China), respectively.

Five-week-old athymic nu/nu male mice were fed a low-fat/no-cholesterol diet (LFNC; 4 gm% fat, 0 gm% cholesterol, D12102C, Research Diets) for the first 2 weeks. Then the mice were divided into high-fat/high-cholesterol diet (HFHC; 20 gm% fat, 1.25 gm% cholesterol, D12108C, Research Diets) and LFNC diet groups with or without Ezetimibe (30 mg/kg/day, dissolved in H₂O with 0.5% CMC-Na). The mice were fed with these diets for additional 2 weeks. Blood was drawn via retro-orbital blood collection, and the serum lipid levels were determined by the biochemistry analyzer. Once the diet-induced hypercholesterolemia model was established, xenografts were initiated by subcutaneously injecting T24 cells (5 × 10⁶, 5 × 10⁵, and 5 × 10⁴ cells/site) with 1:1 volume of Matrigel (BD BioSciences). Xenografts were measured every 3 days. The tumor volume was calculated as π × width² × length/6. About 30 days after T24 cell inoculation, tumor-bearing mice were euthanized and T24 xenografts were harvested, weighed, and processed for further histologic and stem cell analysis. The extreme limiting dilution analysis (ELDA) was conducted using WEHI web interface (https://bioinf.wehi.edu.au/software/elda/) to evaluate whether T24 cells have different tumorigenicity in the animals with different diets.

Ldlr^−/− mice were characterized by spontaneous hypercholesterolemia due to loss of the LDL receptor. Thus, 8-week-old Ldlr^−/− male mice (Jackson Laboratory, RRID: IMSR_JAX:002207) and the C57BL/6 background WT male mice (RRID: IMSR_JAX:000664) were applied with 0.05% BBN in drinking water for 10 weeks. BBN was a carcinogen for bladder in rodents. To investigate the localized tumor progression difference between Ldlr^−/− mice and WT mice, animals were euthanized at 10, 11–15, 16–25, and >25 weeks after BBN feeding (n ≥ 8 for each group at each timepoint). Bladders were harvested, weighed, and processed for further histologic studies. Tumor grade was decided based on hematoxylin and eosin staining as following criteria: (i) normal urothelium: 3–6 cell layers with normal maturation from basal to luminal levels; (ii) inflammation with urothelial atypia and dysplasia (IAD): nuclear (pleomorphism) or architectural distortion without increased number of cells; (iii) cancer in situ (CIS): flat lesion within the urothelial layer displaying loss of polarity/differentiation with marked nuclear pleomorphism, high nuclear/cytoplasmic ratio, and mitotic figures; (d) invasive carcinoma: tumor cells were seen infiltrating the muscle layers of the bladder. To investigate the tumor-associated overall survival (OS) difference between Ldlr^−/− and WT mice, another animal cohort (n ≥ 20 for each group) was analyzed, with the survival endpoints defined as follows: weight loss >20% of initial weight, illness, impaired mobility, behavior, activity, blood on urethra, and any sign of pain or distress not relieved by standard treatment, unexpected death.

Eight-week-old Ldlr^−/− or WT male mice were subcutaneously injected with mouse UBC cell line MB49 cells (5 × 10⁴ cells/site, seven sites for either group). Allografts were measured every 3 days. The tumor volume was calculated as π × width² × length/6. A total of 24 days after MB49 cell inoculation, tumor-bearing mice were euthanized and allografts were harvested, weighed, and processed for further histologic and stem cell analysis.

For ELDA, 8-week-old Ldlr^−/− or WT male mice were subcutaneously injected with MB49 cells (5 × 10⁴, 5 × 10³, and 5 × 10² cells/site, 10–20 sites for each concentration group). About 30 days later, mice were euthanized and allografts were harvested and analysis was conducted using WEHI web interface.

Fresh T24 xenografts or MB49 allografts were washed three times in ice-cold PBS and mechanically minced with sterile surgical scissors. The fragments were digested with 0.5% trypsin-0.01% EDTA at 37°C for 5 to 10 minutes. The dissociated cells were then filtered with 70 μmol/L cell strainers (BD), centrifuged at 1,000 rpm for 4 minutes and resuspended in RPMI1640 supplemented with 10% FBS, in which antibiotics/antimycotics were added. The isolated primary cells were maintained in a 95% humidified atmosphere of 5% CO₂ at 37°C.

The study was conducted in accordance with Declaration of Helsinki. The study protocol using human UBC tissues was approved by the Ethics Committee of Nanjing Drum Tower Hospital, Nanjing University (Nanjing, Jiangsu, P.R. China) and the written informed consent was obtained from each patient (n = 98). All patients were Asians (from Jiangsu or Anhui province, China). The inclusion criteria were as follows: patients who were diagnosed as UBC for the first time, of 45 to 80 years old, with no smoking history, with no medication taken that could influence serum lipid levels, with no surgical operation before. Clinicopathologic information from the patients with UBC, such as, age, gender, serum lipid levels, tumor size, tumor number, tumor grade, tumor stage, and recurrence were analyzed. Serum specimens were collected right after their hospitalization. Paraffin sections from patients with UBC were also obtained.

About 1.0 × 10⁴ UBC cells were seeded in the 6-well ultralow attachment plates (Corning) in 2 mL DMEM/F12, supplemented with 10 ng/mL human recombinant EGF (AF-100-15; PeproTech) and 10 ng/mL human recombinant bFGF (100-18B; PeproTech). Then 0.5 mL fresh cell culture medium as above was added to each well every 3 days. After culturing for 5–7 days, spheres were photographed and counted for spheres with diameter ≥ 50 μm.

ALDH^high cell population from T24, 5637, and UM-UC-3 cells were examined by an ALDEFluor kit (Stem Cell Technologies). A total of 5 μL of activated ALDEFluor reagent was added to 1 mL single-cell suspension (2.5–5 × 10⁵ cells/mL) and mixed well. A total of 5 μL of an ALDH-specific inhibitor, diethylaminobenzaldehyde, was added to 500 μL of the cell suspension separated prior to staining as a negative control. These two samples were incubated away from light at 37°C for 40 minutes and were then analyzed on a BD FACScan flow cytometer.

Total RNA was extracted with TRIzol Reagent (Invitrogen) and then analyzed with GeneChip (Affymetrix) by Shanghai Baygene Biotechnologies, affiliated with Hong Kong Gene Group Holdings. The original and normalized microarray data from this study can be accessed at Gene Expression Omnibus (GEO): GSE134530.

Differential expression genes upregulated in ox-LDL group compared with control group (fold change≥2) were defined as the ox-LDL^high signature. Prognosis significance of ox-LDL^high signature was validated in The Cancer Genome Atlas (TCGA)-Bladder Urothelial Carcinoma (BLCA) dataset (n = 319) and GSE13507 (n = 165). For each sample, a signature score was computed using gene set variation analysis (GSVA), then all of the samples were sorted by signature score, and the top 1 of 3 samples with high signature score were regarded as Group T1, the rest were regarded as Group T2. Kaplan–Meier survival curves with log-rank (Mantel–Cox) test were applied to analyze disease-free survival (DFS) in TCGA-BLCA, and OS in GSE13507.

Data were analyzed by SPSS 19.0 software (SPSS, RRID: SCR_002865). Each experiment was repeated three times. An unpaired Student t test was used to analyze differences between two groups. The χ² test (Fisher exact test if necessary) was used to compare the associations among serum LDL-c, ox-LDL, CD36, p-STAT3 expression and clinicopathologic variables of patients with UBC. The log-rank test was used for survival analysis. The Pearson correlation method was used for correlations among serum LDL-c, ox-LDL, CD36, and p-STAT3 expression. Data were presented as mean ± SEM. The statistical significance was defined as P < 0.05. All graphs were plotted using GraphPad Prism 7.0 (GraphPad Software, RRID: SCR_002798).

To investigate the effects of hypercholesterolemia on UBC progression, we established the hypercholesterolemia mouse model by feeding nude mice with a HFHC diet for 2 weeks. Compared with those in mice fed with a LFNC, the serum levels of total cholesterol (TC), LDL-cholesterol (LDL-c) and high-density lipoprotein-cholesterol in mice fed with HFHC diet increased significantly (Supplementary Fig. S1A–S1C). It has been reported that ezetimibe is capable for blocking intestinal cholesterol absorption by inhibiting NPC1L1 (22, 23). We therefore pretreated a subgroup of mice with ezetimibe before the introduction of the HFHC diet and found that ezetimibe indeed could counteract HFHC-induced TC and LDL-c increase (Supplementary Fig. S1A–S1C). Next, we subcutaneously injected serial diluted human UBC T24 cells into the flanks of nude mice, and continuously treated with the same diets as before for another 4 weeks (Fig. 1A). More xenografts formed in mice fed with HFHC diet than those with the LFNC diet in low group (5 × 10⁴ cells/site) and intermediate group (5 × 10⁵ cells/site; Fig. 1B). Although tumors formed on all mice in high group (5 × 10⁶ cells/site), the slopes of tumor growth curves starting from approximately day 20 in all four groups were very similar, implying the reduced volume in the groups other than HFHC/Veh was possibly due to tumor latency (Fig. 1C). Xenografts were heavier (Fig. 1D) on mice fed with HFHC diet than those with LFNC diet. Notably, ezetimibe decreased HFHC diet–induced tumor incidence (Fig. 1B), growth (Fig. 1C), and weight (Fig. 1D). However, we did not observe obvious changes in body weight and serum levels of glucose, insulin, IL6, and TNFα in different groups (Supplementary Fig. S1D–S1H). The percentage of CK5⁺, CK14⁺, and p-STAT3⁺ cancer cells, the cell subpopulation enriching UBC CSCs, was also increased in the xenografts from mice fed with HFHC diet, and ezetimibe could reduce the percentage of HFHC diet–induced CK5⁺, CK14⁺, and p-STAT3⁺ cancer cells (Fig. 1E). We thus estimated the levels of additional cancer stemness markers, including ALDH1A1, CD44, KLF4, and Nanog, in xenografts. Increased protein expression of all these makers was observed in tumor tissues from mice fed with HFHC diet (Fig. 1F; Supplementary Fig. S1I). To further determine whether hypercholesterolemia enhances UBC cancer stemness, we analyzed the sphere formation capacity and ALDH^high population in primary cultured cells from UBC xenografts. It was shown that primary cultured cells from xenograft-bearing mice fed with HFHC diet not only formed bigger (Fig. 1G; Supplementary Fig. S1J) and more spheres (Fig. 1H), but also contained greater percentage of CSCs than those in LFNC group (Fig. 1I; Supplementary Fig. S1K–S1L). As expected, ezetimibe was capable of counteracting HFHC diet–induced cancer stemness. These data altogether indicated that hypercholesterolemia might accelerate UBC cell growth by increasing cancer cell stemness; and that blocking intestinal cholesterol absorption may abrogate hypercholesterolemia-enhanced UBC cell growth.

To understand the mechanism of hypercholesterolemia-enhanced UBC development, we also compared UBC development in Ldlr^−/− and C57BL/6 WT mice. We treated 8-week-old Ldlr^−/− and WT mice with BBN for 10 weeks, a carcinogen specifically insulting DNA of the urothelial layer, to induce UBCs (17, 24). Pathologically, BBN-induced UBCs in our cohort mimicked human malignant UBC development with the stages of IAD, CIS, and muscle-invasive bladder cancer (Supplementary Fig. S2A). Consistently, comparing with the age-matched WT animals, the Ldlr^−/− mice had an earlier tumor initiation and developed more aggressive UBCs (Fig. 2A) when they were treated with BBN, accompanied by shorter survival time (Fig. 2B). Considering the similar change trends in survival curves in BBN-fed WT and Ldlr^−/− group, the correlation between Ldlr deficiency and early tumor progression was implied. In addition, urothelial cells with CK14⁺ and Ki67⁺, which were regarded as the stem cells in urothelial basal cells (25), were increased in Ldlr^−/− mice after a 10-week BBN treatment (Fig. 2C and D; Supplementary Fig. S2B and S2C), suggesting the accelerated expansion of urothelial basal cells in the BBN-treated Ldlr^−/− mice with increased cancer stemness.

To investigate whether the hypercholesterolemia condition alone in the Ldlr deficiency mice provides a tumor-promoting macroenvironment for UBC aggressiveness, the murine UBC MB49 cells derived from C57BL/6 mice genetic background (26) were injected subcutaneously to both Ldlr^−/− and WT mice. The volume of MB49 allografts became significantly larger in Ldlr^−/− mice since day 12 from injection (Fig. 2E), and tumor weight was heavier of MB49 allografts in Ldlr^−/− mice than that in WT mice (Fig. 2F). Both IHC staining and Western blot assays showed the greater cancer stemness in allografts from Ldlr^−/− mice, comparing with those from WT mice (Fig. 2G and H). Furthermore, primary cultured cells of UBC allografts from Ldlr^−/− mice also formed bigger and more spheres than those from the WT mice (Fig. 2I and J). An ELDA in vivo interpreted more active stem cell proportion existing in allografts from the Ldlr^−/− mice than those from WT mice (Fig. 2K). The data above demonstrated that hypercholesterolemia could enhance UBC aggressiveness via increasing cancer stemness.

Because both diet- and Ldlr deficiency–induced hypercholesterolemia enhanced UBC stemness and aggressiveness, we hypothesized that some factors in the blood of hypercholesterolemic patients/animals are responsible for such promoting effect. Thus, we treated two human UBC cell lines T24 and 5637 with sera from either Ldlr^−/− or WT mice and found that the sera from Ldlr^−/−, but not the WT mice, could increase cell proliferation, sphere formation capacity, the ALDH^high subpopulation, as well as the expression levels of a wide spectrum of stemness markers, including ALDH1A1, CD44, KLF4, and Nanog (Supplementary Fig. S3A–S3F). We noticed that LDL-c was consistently and remarkably increased in both diet-induced and Ldlr knockout–induced hypercholesterolemic mouse models, along with that LDL-c was reduced in ezetimibe-treated HFHC mice (Supplementary Figs. S1B and S4A). Surprisingly, when T24 and 5637 cells were treated with LDL-c, it cannot significantly increase the cell proliferation (Supplementary Fig. S4B) and cell stemness evidenced by sphere formation, ALDEFluor and Western blot assays (Supplementary Fig. S4C–S4E). Given that ox-LDL is a derivative of LDL-c elevated in hypercholesterolemia and atherosclerosis, we then examined its level in the sera of hypercholesterolemic mice. The significantly increased ox-LDL level was confirmed in both HFHC diet– and Ldlr deficiency–induced hypercholesterolemic mice respectively, compared with their corresponding controls (Fig. 3A and B). Neutralization of ox-LDL by its specific antibody remarkably reduced Ldlr^−/− serum-enhanced sphere formation and ALDH^high population in T24 cells (Fig. 3C,–E); whereas the addition of ox-LDL was capable of increasing the cell proliferation rates, sphere formation capacities, percentages of ALDH^high subpopulation, and expression levels of stem markers in a dose-dependent manner in both T24 cells (Fig. 3F,–J) and 5637 cells (Supplementary Fig. S4F–S4I). Because previous literature has shown that human ox-LDL could be accumulated not only in the liver but also in tissues throughout the whole body of mice (27), ox-LDL were administered into the T24 xenograft-bearing mice to test its effect on tumor growth and cancer stemness in vivo (Fig. 3K,–L). Of note, during the 36 days of ox-LDL treatment, once we withdrew the ox-LDL at day 24 (ox-LDL+→−), decreased tumor growth and cell stemness of xenografts were observed, compared with the tumor-bearing animals treated with ox-LDL for the entire period (ox-LDL+→+). However, 27-HC, one metabolite of cholesterol in hypercholesterolemic sera, which was reported to enhance breast cancer progression (4), had no significant effect on T24 xenografts. Altogether, these results indicate that ox-LDL in the hypercholesterolemic mice sera plays an indispensable role in UBC growth by, at least partially, sustaining cancer stemness.

To understand the mechanism of ox-LDL–enhanced UBC stemness, we examined the role of two major ox-LDL receptors, LOX-1 and CD36. First, we knocked down LOX-1 by siRNAs in UBC cells. It was shown that knockdown LOX-1 alone increased the sphere formation capacity and the ALDH^high subpopulation; in addition, the sphere formation capacity and the ALDH^high subpopulation elevated more in presence of ox-LDL (Supplementary Fig. S5A–S5C). These data indicated that ox-LDL–mediated cancer stemness may not depend on LOX-1. Then, we knocked down CD36 with two short hairpin RNAs (shRNA) in T24 and 5637 cells (Supplementary Fig. S5D) and results from both sphere formation and ALDEFluor assays indicated that knockdown CD36 not only lowered the levels of stemness makers but also reduced CSC population (Supplementary Fig. S5D–S5F). In addition, ox-LDL–mediated cancer stemness was diminished when CD36 was knocked down in both T24 (Fig. 4A–C) and 5637 cells (Supplementary Fig. S5G–S5I). Next, we determined the in vivo role of CD36 using a diet-induced hypercholesterolemia mouse model. T24 cells transfected with either control shRNA (T24-shCTL) or shRNA against CD36 (T24-shCD36) were injected to nude mice fed with either HFHC or LFNC diet. Compared with those from T24-shCTL, the xenografts from T24-shCD36 grew much slower (Fig. 4D) and lighter (Fig. 4E) in mice fed with HFHC diet. Of note, results from both IHC and Western blot assays indicated that HFHC diet–enhanced stemness markers, including CK5, p-STAT3, ALDH1A1, CD44, KLF4, and Nanog, were decreased when CD36 expression was inhibited (Fig. 4F and G). These findings were further verified in primary cultured cells from the xenografts of T24-shCTL and T24-shCD36 with HFHC or LFNC diets, by sphere formation (Fig. 4H and I), ALDEFluor (Fig. 4J), and in vivo extreme limiting dilution assays (Fig. 4K). Similar to the effect of CD36 knockdown, neutralization of CD36 in vivo by an antibody against this receptor (JC63.1; refs. 28, 29) also suppressed tumor progression in Ldlr^−/− mice (Supplementary Fig. S5J and S5K) with concurrent downregulation of stemness markers CK5 and p-STAT3 (Supplementary Fig. S5L).

It has been reported that a mutant CD36 (K164A) disables its interaction with ox-LDL (30). We ectopically expressed either the WT (CD36-WT) or this mutant CD36 (CD36-Mut) in a human UBC cell line UM-UC-3 (Supplementary Fig. S5M) and found that only CD36-WT, but not CD36-Mut, could enhance the sphere formation, ALDH^high population, and stemness markers in presence of ox-LDL (Fig. 4L,–N). These data demonstrated the essentiality of CD36 in ox-LDL–enhanced UBC stemness.

To further dissect the downstream signaling of CD36 in regulating cell stemness, we treated T24 cells with ox-LDL and subsequently performed the gene expression profiling. The results from gene set enrichment analysis (GSEA) suggested that ox-LDL might activate STAT3 signaling (Fig. 5A; Supplementary Fig. S6A). Indeed, Western blot assays showed the levels of phosphorylated STAT3 increased significantly when the T24 and 5637 cells were treated with ox-LDL in a dose-dependent manner (Fig. 5B). Although STAT3 can be phosphorylated by multiple kinases including JAK1, JAK2, and Src, we found that only JAK2 and Src, but not JAK1, were upregulated by ox-LDL here (Fig. 5B; Supplementary Fig. S6B). However, when T24 cells were treated with either JAK1/2 inhibitor INCB018424, JAK2 inhibitor AZD1480, or Src family inhibitor dasatinib, we found that both INCB018424 and AZD1480, but not dasatinib, inhibited ox-LDL–mediated STAT3 activation (Fig. 5C; Supplementary Fig. S6C and S6D). Consistently, JAK2 knockdown by siRNA abolished ox-LDL–mediated STAT3 phosphorylation (Fig. 5D). Then, we want to determine the role of CD36 in the ox-LDL–activated JAK2-STAT3 signaling pathway. When CD36 was either inhibited by sulfo-N-succinimidyl oleate (SSO), which binds lysine 164 of CD36 and inhibits uptake of fatty acids and ox-LDL (31), or knocked down by shRNA in T24 cells, ox-LDL–mediated JAK2-STAT3 activation was suppressed (Fig. 5E and F). In addition, CD36 and JAK2 protein could be coimmunoprecipitated in T24 cells (Fig. 5G), and such interaction was enhanced in the presence of ox-LDL (Fig. 5H). The data above suggested that binding of ox-LDL to CD36 could increase the interaction between CD36 and JAK2, and subsequently activate the JAK2-pSTAT3 pathway. Consistently, ox-LDL–enhanced cell viability and stemness were abolished by either JAK2 inhibitor AZD1480 or STAT3 inhibitor Stattic (Fig. 5I,–K). On the basis of these findings, we concluded that ox-LDL enhances UBC stemness via the CD36/JAK2/STAT3 axis.

To determine the clinical relevance of ox-LDL/CD36/STAT3 axis, we recruited a cohort of 98 patients with UBC. High serum ox-LDL level was significantly associated with poor recurrence-free survival in patients with UBC, even after adjustment for most clinicopathologic risk factors (Fig. 6A; Table 1; Supplementary Table S1). Moreover, overexpression of CD36 protein level also predicted poor clinical outcome in patients with UBC (Fig. 6B and C). In addition, the levels of serum ox-LDL, CD36 protein, and p-STAT3^Y705 protein were positively correlated to high-grade and advanced tumor stage UBC, respectively (Supplementary Table S2). Positive correlations between ox-LDL and CD36 (r = 0.526), between CD36 and p-STAT3^Y705 (r = 0.631) and between ox-LDL and p-STAT3^Y705 (r = 0.530) were also shown in the patients with UBC (Fig. 6D,–F). Interestingly, serum LDL-c level did not exhibit significant correlation with recurrence-free survival in patients with UBC, although serum LDL-c level was also positively associated with tumor malignancy, serum ox-LDL level, CD36, and p-STAT3^Y705 in the same UBC cohort (Supplementary Fig. S7A–S7D; Supplementary Table S2).

Finally, we analyzed the risk of ox-LDL in UBC in the public UBC datasets with survival information. We defined the upregulated gene sets with fold change ≥2 in the microarray of ox-LDL–treated T24 cells (Fig. 5A) as the signature of ox-LDL^high level targets. As a result, we found that patients with UBC with ox-LDL^high signature expression (Group T1) exhibited worse DFS probability or OS probability than those with ox-LDL^low signature expression (Group T2) in TCGA or GEO database, respectively (Fig. 6G and H). These results indicated that ox-LDL signature predicts UBC progression and ox-LDL as a risk factor of UBC.

Hypercholesterolemia, which is defined as high plasma cholesterol levels, due to the rise of cholesterol and Apolipoprotein B–rich lipoproteins, called LDL-c, has been implicated in some cancer types, such as prostate cancer. To address whether it is also a risk factor for patients with UBC, herein we proved the cause-effect relationship that UBC progression, was significantly accelerated by hypercholesterolemia induced by either HFHC diet or conventional Ldlr knockout. The significant increase of ox-LDL in serum, but not LDL-c or cholesterol metabolite, 27-HC, under hypercholesterolemia condition promotes UBC cancer stemness and aggressiveness. Our data are consistent with and explain the recent finding of a comprehensive epidemiologic (Me-Can) 2.0 study on an association between TC and the risk for UBC (13).

As a complex particle consisting of lipids at the core and proteins outside, LDL-c functions as a cholesterol carrier into peripheral tissues (32). However, LDL-c may undergo various types of modifications, such as oxidative modifications including its esterification and lipid peroxidation (33). The elevated level of ox-LDL has been reported to be an increased risk of certain cancers, such as breast cancer (34, 35). In addition, only ox-LDL, but not native LDL-c, contributes heavily to atherosclerosis associated inflammation (36). Consistent with these findings, we also demonstrated that only ox-LDL, but not LDL-c in general, facilitated UBC progression. The mechanisms linking hypercholesterolemia to UBC progression remain largely elusive. Considering dyslipidemic and oxidative states in cancer, the roles of peroxidation metabolites, including oxysterols and ox-LDL, should be investigated in UBC. Here, we assessed the effect of 27-HC, the functional oxysterol, which has been extensively studied (37). We found that 27-HC had little effects on UBC stemness and this finding is consistent with a recent independent report that 27-HC does not increase and cannot even inhibit UBC cell proliferation in vitro (38). However, in current experimental settings, human ox-LDL were injected into the nude mice bearing human UBC xenografts, which is a limitation considering that the differences exist between mouse and human LDL-c. Overall, the role of ox-LDL as a key product for hypercholesterolemia and oxidation states in tumor macroenvironment is indicated in UBC.

The elevated levels of serum ox-LDL may target multiple cell types, besides cancer cells. Through interactions with its receptors CD36 or LOX-1 to activate downstream signaling, such as NFκB pathway, ox-LDL could increase cancer cell proliferation and induce the epithelial-to-mesenchymal transition in vitro (35). As manifested in atherosclerosis, ox-LDL can be taken up by macrophage through CD36 and transform it to foam cells with the secretion of pro-inflammatory cytokines and chemokines (39). In addition, ox-LDL can also target endothelial cells to stimulate angiogenesis (40). In this study, we demonstrated that hypercholesterolemia remarkably enhances cancer stemness, which is mostly mediated by the elevated levels of serum ox-LDL and its receptor CD36 in UBC cells. Overexpression of the WT CD36, but not the mutant CD36, enhanced cancer stemness in the presence of ox-LDL. As expected, antibody (JC63.1) neutralization of CD36 to block CD36-mediated ox-LDL uptake significantly inhibited ox-LDL–induced cancer stemness.

The STAT3 pathway was reported to account for the enhanced cancer stemness in UBC (16). We identified that the presence of ox-LDL promotes the interaction of CD36 and JAK2, eventually leading to the activation of STAT3. Meanwhile, ox-LDL was also reported to upregulate the phosphorylation of STAT3 in activated platelets and induce endothelial cells injury (41, 42). Different from cancer cells, ox-LDL mainly induces the expression of Toll-like receptor 4 and then affect NFκB/STAT3 signaling pathway. Thus, it is important to clarify the relevance between ox-LDL and STAT3 in different diseases. In this study, we established a novel link between ox-LDL/CD36/JAK2/STAT3 signaling axis and cancer stemness in the hypercholesterolemia-induced UBC.

In summary, we demonstrate that (i) ox-LDL links hypercholesterolemia with UBC progression by enhancing cancer stemness; (ii) the serum ox-LDL level elevates UBC stemness via CD36/JAK2/STAT3 axis; (iii) the activation of ox-LDL/CD36/STAT3 axis is positively associated with UBC malignancy and poor prognosis; (iv) ox-LDL is a high-risk factor of UBC. Blocking the ox-LDL receptor CD36, targeting the JAK2-STAT3 signaling pathway, even intervention on the systemic cholesterol level, might be therapeutic options for patients with UBC with hypercholesterolemia.

No disclosures were reported.

L. Yang: Data curation, software, formal analysis, validation, investigation, methodology, writing–original draft. J. Sun: Data curation, funding acquisition, investigation, writing–original draft. M. Li: Data curation, validation. Y. Long: Data curation, formal analysis. D. Zhang: Methodology. H. Guo: Resources, supervision, funding acquisition. R. Huang: Conceptualization, resources, supervision, funding acquisition, writing–original draft. J. Yan: Conceptualization, resources, supervision, writing–original draft.

The authors thank Wei Zhao, Junzun Li, Xiaohong Yu from Jun Yan's lab, and the Institutional Center for Shared Technologies and Facilities of Shanghai Institute of Materia Medica, Chinese Academy of Sciences for helpful technical supports.

The work was supported by National Natural Science Foundation of China (91859106 and 81771890 to R. Huang; 81902968 to J. Sun; 81972388 and 81772710 to H. Guo), the National Science & Technology Major Project “Key New Drug Creation and Manufacturing Program,” China (2018ZX09711002 to R. Huang), the research fund from Shanghai Science and Technology Committee (20S11901400 to R. Huang), One Hundred Talent Program of Chinese Academy of Sciences (to R. Huang), and Shanghai Sailing Program (19YF1456900 to J. Sun).

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.