Cholesterol Auxotrophy as a Targetable Vulnerability in Clear Cell Renal Cell Carcinoma

Renal cell carcinoma (RCC) ranks among the 10 most common cancers in men and women, with approximatively 74,000 new cases of RCC and 15,000 deaths in the United States in 2020 (American Cancer Society). Clear cell RCC (ccRCC) represents 70% to 80% of all RCC subtypes, with more than 90% displaying copy-number deletion, inactivating mutation, and/or epigenetic silencing of the von Hippel-Lindau (VHL) tumor suppressor gene (1). VHL encodes the recognition component of a protein complex that catalyzes the oxygen-dependent ubiquitination and subsequent proteasomal degradation of the hypoxia inducible factor (HIF) transcriptional regulators (2). Recently, ccRCC has been classified as a “metabolic disease” due to common alterations in cellular metabolic pathways associated with cancer initiation and progression (3). HIFs are critical drivers of ccRCC metabolic rewiring, partly by promoting the “Warburg effect” that provides cancer cells with ATP and glycolytic intermediates to fulfill the anabolic demands of proliferating cells (4, 5). Moreover, recent genome-wide, metabolic gene set analyses reveal that ccRCC tumors downregulate enzymes in the pathways opposing glycolysis, such as gluconeogenesis, as well as pathways that consume potential nutrients, such as the urea cycle pathway relative to normal tissues. For example, the rate-limiting gluconeogenic enzyme fructose-1,6-bisphosphatase 1 (FBP1) is universally depleted in ccRCC tumors, and its reexpression in ccRCC cells confirms an unexpected tumor suppressor activity (6). Similarly, expression of the urea cycle enzyme arginase 2 (ARG2) is also repressed in ccRCC, and ectopic ARG2 expression in ccRCC cells results in toxic polyamine accumulation and depletion of the essential biosynthetic cofactor pyridoxal phosphate (7).

Extensive accumulation of intracellular lipid droplets (LD) and glycogen, which confer the “clear cell” histologic phenotype, is an additional metabolic hallmark of ccRCC (8). LDs are dynamic cytoplasmic organelles protected by a phospholipid monolayer surrounding a neutral lipid core, composed primarily of cholesterol esters (CE) and triacylglycerols (TG), and have been linked with many cellular functions (9). We previously reported that reduced expression of the LD coat protein Perilipin-2 (PLIN2), or its transcriptional regulator HIF2α, impairs LD formation and reduces ccRCC cell viability (10). Cholesterol and esterified cholesterol are the most prominent lipids stored in ccRCC LDs, accumulating to 8-fold and 35-fold higher levels, respectively, compared with normal kidney tissue (11). Currently, the functional importance of cholesterol accumulation and turnover for ccRCC tumor growth and progression remains obscure.

In this study, we performed integrative studies of human ccRCC tumors and cell lines and demonstrate that ccRCC cells suppress de novo cholesterol biosynthesis, despite accumulating high levels of cholesterol and CEs. We demonstrate that ccRCC cells are, surprisingly, functional auxotrophs for exogenous cholesterol and induce apoptotic cell death in its absence. Furthermore, elevated dietary cholesterol promotes, whereas dietary cholesterol restriction inhibits, ccRCC xenograft tumor growth. The scavenger receptor B1 (SCARB1) transports cholesterol into ccRCC cells in the form of high-density lipoproteins (HDL), circulating 5–17 nm diameter particles composed of proteins (primarily apolipoproteins apo A-I and apo A-II), phospholipids, cholesterol, CEs, triglycerides, and other neutral lipids. Previous genome-wide association studies (GWAS) identified a common single-nucleotide polymorphism (SNP) within the SCARB1 locus linked to increased ccRCC risk (12, 13). We demonstrate that SCARB1 is functionally required for HDL transport in ccRCC; mutating the SCARB1 gene, reducing its expression, or inhibiting SCARB1 activity compromises ccRCC cell viability and tumor growth. Importantly, Mendelian randomization analyses of GWAS data indicate a causal relationship between genetic alleles associated with elevated circulating HDL (and HDL cholesterol) levels and ccRCC risk. Finally, we demonstrate that inhibiting SCARB1 or limiting cholesterol availability increases ROS levels in ccRCC cells, revealing an important role for cholesterol in redox homeostasis. The dependence of ccRCC on exogenous cholesterol import suggests a metabolic vulnerability that can be explored as a novel therapeutic strategy for ccRCC.

To investigate cholesterol metabolism in ccRCC, we conducted gene set enrichment analysis (GSEA) of RNA-sequencing data from The Cancer Genome Atlas (TCGA) and observed that gene sets associated with cholesterol metabolism and biosynthesis (Fig. 1A) are significantly downregulated in ccRCC tumors (Fig. 1B and C; Supplementary Fig. S1A). The majority of these genes encode rate-limiting enzymes in the de novo cholesterol biosynthetic pathway, including HMG-coA reductase (HMGCR), lanosterol synthase (LSS), and squalene monooxygenase (SQLE). Integrative analysis indicated that mRNA transcripts encoding other enzymes in the mevalonate pathway are coordinately repressed in ccRCC tumor samples relative to normal kidney (Fig. 1D). These genes were similarly underexpressed in 12 pairs of ccRCC human patient samples, relative to surrounding normal tissue (Fig. 1E). A large-scale proteogenomic study (14) also confirmed that mevalonate pathway gene and protein expression is downregulated in ccRCC (Supplementary Fig. S1B and S1C). In human A498 and 786-O ccRCC cell lines, HMGCR, LSS, SQLE, delta-24-sterol reductase (DHCR24), and farnesyl-diphosphate farnesyltransferase 1 (FDFT1) levels were reduced relative to the nontransformed HK-2 tubular epithelial cell line derived from normal kidney (Fig. 1F).

Structural alterations in mevalonate pathway genes analyzed were found only rarely in multiple TCGA ccRCC patient data sets (cBioPortal), in contrast to VHL mutations (Supplementary Fig. S1D) observed in a majority of cases (Fig. 1G; Supplementary Fig. S1E–S1H). Interestingly, anaplastic large-cell lymphoma (ALCL) cells with SQLE mutations accumulate squalene, which protects these cells from ferroptotic cell death (15). In contrast, pan-metabolomic analyses performed on 68 primary human ccRCC tumors and 114 normal kidney tissues from two independent cohorts (6, 16) revealed that squalene levels were underrepresented relative to corresponding normal kidney tissues (Fig. 1H; Supplementary Fig. S1I and S1J). We reasoned that because ccRCC cells are not actively synthesizing cholesterol, they should be relatively resistant to treatment with statins, which inhibit HMGCR, catalyzing the first committed step in cholesterol biosynthesis. Indeed, atorvastatin had no significant effect on ccRCC cell lines (Fig. 1I and J; Supplementary Fig. S1K and S1L), whereas it dramatically impaired HK-2 cell proliferation (Fig. 1K; Supplementary Fig. S1M). Taken together, these data suggest that intracellular cholesterol metabolism is deregulated in ccRCC tumor cells.

Given that ccRCC cells do not engage in cholesterol biosynthesis, we anticipated that they depend on exogenous cholesterol and CEs imported from the microenvironment potentially in the form of low-density lipoproteins (LDL) and high-density lipoproteins (HDL). To test this hypothesis, ccRCC cells (A498 and 786-O) were cultured in medium with 10% delipidated serum (DLPS), resulting in proliferation defects and increased cell death relative to cells grown in 10% FBS (Fig. 2A–C; Supplementary Fig. S2A–S2C). Supplementing DLPS media with exogenous cholesterol fully rescued ccRCC cell proliferation and viability (Fig. 2A–C; Supplementary Fig. S2A–S2C). Lipidomic profiles revealed that CEs (Fig. 2D) and total cellular cholesterol levels were reduced (Fig. 2E; Supplementary Fig. S2D) in ccRCC cells cultured in DLPS, and were restored by adding exogenous cholesterol. In contrast, viability of nontransformed human kidney cells (HK-2), which rely primarily on de novo cholesterol synthesis, was unaffected by DLPS media (Fig. 2F; Supplementary Fig. S2E and S2F). To investigate whether cholesterol dependence is relevant for other kidney cancers, we evaluated Caki-2 cells which present characteristics of papillary RCC and confirmed that these are more resistant to DLPS conditions than ccRCC cell lines, at least up to day 4 (Supplementary Fig. S2E and S2G). These results indicate that ccRCC cells, and not normal kidney epithelial cells, are strikingly dependent on exogenous cholesterol for proliferation and survival, identifying cholesterol homeostasis as a metabolic vulnerability in ccRCC.

To further investigate the role of circulating cholesterol in RCC risk, we employed two-sample Mendelian randomization (MR; ref. 17) using summary association statistics from a GWAS describing 10,784 RCC cases and 20,406 controls (18–20) to estimate the causal relationship between circulating metabolites and RCC. Out of 123 measured metabolites, MR analysis revealed that the odds of developing RCC were significantly higher when genetic alleles predict elevated serum levels of HDL or HDL cholesterol (e.g., OR = 1.17; 95% CI, 1.05–1.31; P = 0.005 for RCC per SD increment total cholesterol in HDL; Supplementary Table S1; Fig. 2G; Supplementary Fig. S3A and S3B). SNP heterogeneity was high for many of the metabolites (P value range: 0.005–0.96, Supplementary Table S1), although not unexpected due to the many underlying biological pathways that contribute to metabolite variation. Beyond the HDL metabolite subclass, only six metabolites were associated with RCC risk at a nominal level of significance (Supplementary Table S1). Evidence for causality in the reverse direction (i.e., the effect of RCC on serum HDL metabolites) was weak (e.g., −0.0003 SD total cholesterol in HDL; 95% CI = −0.06–0.06 per 1 log odds of RCC; Supplementary Table S2), further supporting evidence that perturbed HDL is a cause and not a consequence of RCC risk. These data highlight circulating HDL cholesterol as a key metabolite in RCC risk.

To test the hypothesis that elevated extracellular HDL and cholesterol levels can promote ccRCC tumor growth, we injected nude mice with A498 ccRCC cell lines and provided either a cholesterol-free (NC; 0% cholesterol) or high cholesterol (HC; 2% cholesterol) diet. Mice receiving an HC diet exhibited larger tumor volumes compared with mice receiving an NC diet (Fig. 2H–J), although neither diet affected overall mouse body weights (Supplementary Fig. S3C). As anticipated, circulating HDL and total cholesterol levels were significantly higher in mice fed the HC diet compared with mice fed the NC diet (Fig. 2K; Supplementary Fig. S3D). These data suggest that circulating cholesterol levels promote human ccRCC tumor growth and progression.

The apparent addiction of ccRCC cells to exogenous cholesterol, coupled with reduced expression of cholesterol biosynthetic genes, suggest that cholesterol transporters could represent therapeutic targets in ccRCC. Cells import cholesterol from the extracellular environment through a variety of receptors, including the very-low-density-lipoprotein receptor (VLDL-R), low-density-lipoprotein receptor (LDL-R), and the HDL transporter SCARB1. Metabolic gene set analysis revealed that SCARB1 is significantly overexpressed in human ccRCC clinical samples relative to normal human kidney (Fig. 3A and B), whereas LDLR expression is repressed (Fig. 3A and C), suggesting that ccRCC tumors rely primarily on SCARB1 for cholesterol import. Elevated SCARB1 expression across all ccRCC tumor stages implies that cholesterol metabolism is deregulated early in ccRCC initiation and progression (Fig. 3D).

Analysis of multiple TCGA data sets revealed no structural alterations in the SCARB1 gene in human ccRCCs (Fig. 3E); however, previous studies identified a common SNP within the SCARB1 locus associated with renal cancer susceptibility (12, 13). Cancer Cell Line Encyclopedia (CCLE) and Cancer Dependency Map (DepMap) datasets showed elevated SCARB1 mRNA and protein abundance in a variety of cell lines representing ccRCC and other cancer types (Fig. 3F and G). Interestingly, no significant dependencies on SCARB1 were observed among these cancer types, suggesting that the screen was performed under different culture conditions (Fig. 3H; see also Discussion). SCARB1 mRNA and protein levels were also upregulated in tumors, relative to matched normal kidney tissue, in samples obtained from a cohort of 14 patients with ccRCC (Fig. 3I and J). Of note, human tissue microarrays (TMA) exhibited SCARB1 accumulation in ccRCC, but not in other kidney cancer subtypes like oncocytoma, papillary, and chromophobe (Fig. 3K and L; Supplementary Fig. S4A–S4C), confirming that SCARB1 expression is a particular characteristic of ccRCC and correlates to the large LDs and CE accumulation in these cells.

Previous reports demonstrated that using siRNAs to inhibit SCARB1 expression in ccRCC cells reduced cholesterol import, cell growth, migration, and AKT phosphorylation in vitro (21). To assess the effect of SCARB1 inhibition on “normal” kidney cells, we targeted SCARB1 in HK-2 and RPTEC cells. Interestingly, SCARB1 is expressed at similar levels in HK-2 cells, ccRCC cell lines and tumor tissue (Supplementary Fig. S4D), suggesting that adaptation to tissue culture conditions explains the sensitivity of HK-2 to SCARB1 inhibition (Supplementary Fig. S4E–S4I). On the other hand, RPTEC cells, which express low SCARB1 levels similar to normal healthy tissue, were completely resistant to genetic and pharmacologic SCARB1 inhibition (Supplementary Fig. S4J–S4M). It should be noted that established cell lines do not always maintain expression levels of critical proteins consistently observed in our studies of a large number of primary human ccRCC (n = 831) or healthy tissue (n = 188). Therefore, the RPTEC cells are more representative of what would be expected for “normal” renal cells, at least in this regard. However, the functional importance of SCARB1 in human ccRCC tumor growth and progression has not been formally addressed. We generated A498 cells expressing doxycycline-inducible SCARB1 shRNAs and confirmed that SCARB1 knockdown inhibited HDL import (Fig. 4A), and blocked cell proliferation (Fig. 4B–D) by inducing apoptosis (Fig. 4E) and cell-cycle arrest (Fig. 4F) in vitro. Similar results were obtained using CRISPR/Cas9 technology (see Supplementary Fig. S5A–S5D).

To determine the uncharacterized role of SCARB1 in ccRCC tumor growth in vivo, A498 cells expressing doxycycline-inducible SCARB1 shRNAs were used to generate palpable (100 mm³) xenograft tumors, at which time mice were fed chow containing doxycycline. shSCARB1 expression substantially reduced xenograft tumor growth, relative to controls, without affecting mouse body weight (Fig. 4G–I; Supplementary Fig. S5E). As validated by Western blot and IHC, SCARB1 expression was decreased during doxycycline treatment (Supplementary Fig. S5F and S5G), and SCARB1-depleted tumors exhibited reduced tumor proliferation and elevated cell death as indicated by Ki-67 and cleaved caspase-3 staining, respectively (Supplementary Fig. S5H and S5I). These data demonstrated that SCARB1 depletion compromises ccRCC tumor growth in vivo.

Whereas free cholesterol or HDLs rescued cell proliferation defects in parental A498 cells exposed to DLPS media (Fig. 2A–C), HDL supplementation failed to rescue proliferation, cell death, and cell-cycle arrest phenotypes in shSCARB1-expressing cells cultured in DLPS media (Fig. 4J–L; Supplementary Fig. S5J). Collectively, these results indicate that ccRCC cells require SCARB1 for cholesterol import, proliferation, and survival.

Given the importance of SCARB1 for HDL import into ccRCC cells, we tested responses to SCARB1 antagonists (22, 23), including the small molecule Block Lipid Transporter-1 (BLT-1), a potent inhibitor of SCARB1 mediated lipid transport (22). BLT-1–treated ccRCC cells exhibited proliferation defects associated with apoptosis and cell-cycle arrest (Fig. 5A–C; Supplementary Fig. S6A–S6C), as well as significantly decreased intracellular cholesterol (Fig. 5D; Supplementary Fig. S6D). As expected, the proliferation and apoptotic phenotypes were not rescued by providing HDL to BLT-1–treated cells (Fig. 5E–H; Supplementary Fig. S6E–S6H). Treating ccRCC tumor-bearing mice with BLT-1 via oral gavage for 21 days significantly inhibited ccRCC tumor growth while increasing serum HDL levels (Fig. 5I–N; Supplementary Fig. S6I), indicating that ccRCC tumors rely on circulating HDL to sustain tumor growth. Surprisingly, BLT-1 treatment of ccRCC tumors only slightly lowered free cholesterol levels and actually increased CE levels (Supplementary Fig. S7A and S7B). Cholesterol efflux through SCARB1 is a process used by cells to avoid toxic accumulation of free cholesterol. Taken together, BLT-1 is blocking both import and efflux, and HDL cholesterol uptake through SCARB1 in and of itself appears to be critical to sustain AKT signaling and control cholesterol and CE homeostasis. Reducing HDL availability through SCARB1 inhibitors and/or dietary restriction could therefore offer a potential combination therapy for patients with ccRCC.

TCGA analysis revealed frequent alterations in the PI3K/AKT signaling pathway (Fig. 6A) in human ccRCC, consistent with its designation as a metabolic disease (24). We hypothesized that our observed cholesterol-dependent proliferation phenotype was at least partially due to an impaired PI3K/AKT signaling axis in ccRCC. AKT phosphorylation was decreased in ccRCC cells grown in DLPS and restored by supplementation with HDL (Fig. 6B and C). We further observed that SCARB1-deficient ccRCC cells also showed reduced AKT phosphorylation (Fig. 6D), consistent with a previous report (21). Moreover, providing HDL to BLT-1 treated ccRCC cells did not fully restore AKT phosphorylation (Fig. 6E). Treatment with PI3K and AKT inhibitors (LY-294002 and GSK-690693, respectively) also led to ccRCC proliferation defects that were not rescued by HDL (Fig. 6F–H), confirming the importance of the PI3K signaling downstream of SCARB1 to sustain ccRCC proliferation. In macrophages and endothelial cells, the PDZ domain containing 1 (PDZK1) protein has been shown to link SCARB1, HDL import, and AKT activation (25, 26); however, initial analysis indicates that AKT activation appears to be insensitive to PDZK1 inhibition in ccRCC cells (Supplementary Fig. S7C). Therefore, how SCARB1 is connected to AKT signaling will required more investigation in the future.

Cholesterol is a critical component of biological membranes and an important constituent of lipid rafts (LR). LRs serve as critical membrane platforms to amplify signaling cascades, including the PI3K/AKT pathway (27, 28). Moreover, SCARB1 clusters have been found in caveolin-enriched membrane domains that may be required for HDL binding and lipid uptake (29). To investigate whether cholesterol regulates LR-mediated PI3K/AKT signaling pathways, ccRCC cell lines were treated with edelfosine, an alkyl-lysophospholipid cholesterol analogue that accumulates in LRs and alters cellular signaling (30–32). Edelfosine treatment decreased AKT phosphorylation and cell growth in vitro accompanied by increased apoptosis (Fig. 6I–N). Elevated levels of PDK1, a kinase that activates AKT and localizes to LR-like domains, were detected in cytosolic versus membrane fractions from cells cultured in 10% DLPS compared with 10% FBS (Fig. 6O), and increased membrane fractionation was observed upon HDL supplementation. Collectively, our findings indicate that cholesterol import through SCARB1 might contribute to lipid raft homeostasis and particularly PI3K/AKT pathway activation.

Squalene and other cholesterol pathway intermediates can protect cells from excess reactive oxygen species (ROS) mediated oxidative stress (15) that also promotes development and progression of multiple cancer types (33, 34). Previous analyses indicated that ccRCC cells express elevated levels of antioxidant proteins (35), suggesting that regulating ROS levels is critical for maintaining cellular redox homeostasis. Cholesterol depletion may therefore produce toxic levels of intracellular ROS in ccRCC cells.

2′,7′-dichlorodihydrofluorescein diacetate (DCFDA) fluorescence analysis was used to evaluate ROS levels in ccRCC cells. As predicted, lipid-deprived cells exhibited significantly higher ROS levels compared with cells in complete media (Fig. 7A). Supplementing DLPS medium with HDL returned intracellular ROS levels to baseline (Fig. 7B), as did treatment with the well-established reducing agents α-tocopherol or N-acetyl-cysteine (NAC; Fig. 7C; Supplementary Fig. S7D). Of note, NAC was unable to rescue cell proliferation mediated by DLPS media, explained by the inability to rescue AKT phosphorylation (Supplementary Fig. S7E and S7F). Depleting SCARB1 in ccRCC cells grown in replete medium was sufficient to elevate ROS to similar levels as control ccRCC cells grown in DLPS media (Fig. 7D and E). Interestingly, whereas NAC and α-tocopherol rescued ROS levels in SCARB1-deficient cells (Supplementary Fig. S7G and S7H), HDL did not (Fig. 7F), confirming that ROS homeostasis in ccRCC cells depends on cholesterol availability and import mediated by SCARB1.

Collectively, our findings reveal that ccRCC cells require the import of exogenous cholesterol through SCARB1 to maintain viability, proliferation, and PI3K/AKT signaling, and to prevent ROS-mediated damage (Fig. 7G). Our data also demonstrate that elevated HDL levels are associated with increased risk of ccRCC, and limiting cholesterol availability, or inhibiting SCARB1 genetically or pharmacologically, reduces the growth of extant ccRCC tumors in murine models. These results suggest new therapeutic strategies that, either singly or in combination, could benefit patients with ccRCC.

The clear cell phenotype observed in ccRCC is associated with intracellular accumulation of neutral lipids, including triglycerides and esterified sterols. In particular, ccRCCs exhibit 8-fold and 35-fold higher levels of cholesterol and CEs, respectively, compared with normal kidney tissue (11). In mammals, cholesterol is considered a nonessential metabolite, the majority of which is synthesized de novo from acetyl-coenzyme A (Fig. 1A). Gene expression, genomic, and metabolomic data reveal that ccRCC cells exhibit a global suppression of the cholesterol biosynthetic pathway not associated with mutations in mevalonate pathway genes. Of note, the biosynthetic pathway is not suppressed via the VHL/HIF system, based on careful evaluation of RNA, proteomic, and lipidomic publicly available datasets for ccRCC cells and tumors with HIF2α inhibited pharmacologically or genetically (36–40). Using functional in vitro and in vivo studies, we demonstrate that ccRCC cells behave as cholesterol auxotrophs. Future studies will explore the mechanistic basis of the loss of cholesterol biosynthesis in ccRCC. Additional cancers are also dependent on exogenous cholesterol, including breast cancers and glioblastomas (41, 42), although a SCARB1 dependency was not revealed by DepMap in Fig. 3H. We are unsure why DepMap is not reflecting our findings other than their culture conditions may differ from ours. Interestingly, prostate tumors appear to rely more on intracellular cholesterol than on exogenous uptake (43), suggesting that extracellular cholesterol dependency varies among cancer types. More recently, Garcia-Bermudez and colleagues reported that ALK⁺ ALCLs are auxotrophs for cholesterol, due to a deficiency in squalene monooxygenase that results in accumulation of the antioxidant squalene (15). In contrast, squalene and other pathway intermediates are less abundant in ccRCC compared with healthy kidney tissue, consistent with general suppression of the mevalonate biosynthetic pathway (44).

We show here that ccRCC cells require extracellular cholesterol to proliferate and survive. Providing ccRCC tumor-bearing mice with a high-cholesterol diet, which increased serum HDL levels as expected, was sufficient to increase xenograft tumor growth. This was surprising, as low HDL levels have been associated previously with increased risk of various cancer types including myeloma, lymphoma, breast, lung, and brain (45). As HDL has reported antioxidant, anti-inflammatory, antithrombotic, antiapoptotic, and vascular-protective effects (46), we hypothesized that HDL fulfills one or more cytoprotective functions for ccRCC cells. Our preclinical results mesh with human data describing 31,190 individuals analyzed by two-sample Mendelian randomization, which demonstrated that genetic variants associated with elevated levels of circulating HDL are also positively associated with increased RCC risk. In contrast, genetic predisposition to RCC does not cause increased HDL levels, indicating that the causal pathway is clearly from HDL to RCC. High serum HDL levels may promote chronic kidney disease (CKD) by damaging renal tubular epithelial cells, although CKD has not been firmly established as a risk factor for renal cancer.

Our data reveal that the HDL receptor SCARB1 is essential for ccRCC cell proliferation and survival and represents a novel therapeutic target. SCARB1 is typically expressed in liver and adrenal tissues, where it contributes to the reciprocal loading and unloading of HDL particles by forming plasma membrane channels. Previous GWAS identified an association between a SCARB1 gene polymorphism (chr12q24.31-rs4765623) and increased ccRCC risk (12, 13), although our analysis of available TCGA gene expression data failed to uncover a significant correlation between SCARB1 transcript levels and patient prognosis or survival, in contrast to a previous report (21).

SCARB1 mRNA and protein expression are highly elevated in ccRCC tumors and kidney cancer cell lines compared with normal controls, although the molecular basis of increased SCARB1 expression remains unclear. The index RCC-associated SCARB1 polymorphism is adjacent to a functional HIF binding sequence motif, suggesting that SCARB1 may be transcriptionally regulated by the HIFs. However, reexpression of pVHL and subsequent HIF suppression, or treatment with a HIF2α inhibitor (37, 47), had no effect on SCARB1 expression in ccRCC cells or xenograft tumors, respectively (36–40). More recent work demonstrated that the chr12q24.31-rs4765623 SCARB1 polymorphism is contained within a superenhancer characterized by elevated H3K27 acetylation (48), and treatment of ccRCC cell lines with the BRD4 inhibitor JQ1 significantly decreased SCARB1 expression. In aggregate, these data support a model in which high HDL levels and elevated SCARB1 expression contribute directly to ccRCC tumorigenesis and growth, apparently independent of the pVHL tumor suppressor.

The underlying signaling events linking cholesterol metabolism to ccRCC cell proliferation and survival are likely to be complex. Previous TCGA analysis revealed that approximately 30% of ccRCC tumors harbor genetic alterations in the PI3K/AKT/mTOR pathway, consistent with an important role in the initiation and/or progression of the disease (24, 49). HDL is known to activate the PI3K/AKT, Src, and AMPK signaling pathways, which could reinforce oncogenic signaling (50) in ccRCC cells. We demonstrate that exposure to delipidated serum or SCARB1 inhibition dramatically reduces AKT activation in ccRCC cells, which is restored in a SCARB1-dependent manner by addition of HDL or cholesterol. Interestingly, SCARB1 has also been implicated in promoting HDL efflux (51), consistent with our data indicating that BLT-1 treatment of ccRCC tumors surprisingly increased CE levels. Therefore, BLT-1 appears to block both import and efflux. Overall, HDL cholesterol uptake through SCARB1 sustains AKT signaling and controls cholesterol and CE homeostasis. Cholesterol is also required for cell membrane fluidity and protein trafficking and is a central component of LRs, specialized membrane microdomains that serve as organizing centers to promote signal transduction. We demonstrate that the LR-disrupting agent edelfosine suppresses ccRCC growth and induces apoptosis while inhibiting AKT signaling, suggesting that SCARB1 may interact with LR-enriched plasma membrane domains and facilitate the incorporation of cholesterol to maintain PI3K/AKT pathway signaling. Further investigation is needed to fully elucidate the underlying signaling mechanisms linking SCARB1 and HDL to oncogenic signaling in ccRCC.

Our data also indicate that HDL import contributes to ROS homeostasis in ccRCC cells. Enhanced metabolic activity of cancer cells results in ROS overproduction leading to changes in many processes necessary for tumor initiation, growth, and progression (52). On the other hand, permanently elevated ROS levels have cytotoxic effects and maintaining redox homeostasis is a challenging feature for many cancer types. ccRCC tumors express high levels of antioxidant pathways, such as reduced glutathione (GSH) and glutathione peroxidases (GPX) to prevent multiple ROS accumulation and the induction of ferroptosis (35, 53). Therefore, HDL import contributes to redox balance, but other mechanisms are critical as well, underscoring the importance of redox balance in this disease. In addition, in noncancerous liver cells, excess cholesterol is detoxified in part by conversion to bile acid, the initial step of which is hydroxylation of the cholesterol 7α-position. This hydroxylation is rate-limiting and catalyzed by the enzyme CYP7A1; however, cholesterol oxidation also occurs nonenzymatically in a ROS-dependent manner to produce oxysterols that are ultimately converted to bile acids. We found that CYP7A1 gene expression was not significantly upregulated in ccRCC tumors compared with surrounding normal tissue in the TCGA dataset. In contrast, the enzyme HSD3B7, which is required for subsequent conversion and detoxification of oxysterols, is significantly overexpressed in ccRCC tumors. These observations suggest that catabolism of cholesterol by reactivating the bile acid pathway may contribute to ROS homeostasis in ccRCC cells.

In addition to its critical role in HDL uptake, SCARB1 has very recently been reported to contribute to SARS-CoV-2/COVID-19 and Hepatitis C Virus (HCV) entry in cells (54, 55). This feature makes SCARB1 an attractive antiviral target, leading to the development of a SCARB1 inhibitors (56) including ITX-5061, which has entered phase I clinical trials in HCV-infected humans (NCT01292824, NCT01560468, NCT01165359) and been well tolerated with no significant drug interactions. Future studies will be required to assess the efficacy and clinical utility of repurposing ITX-5061 to inhibit SCARB1 in ccRCC. It should also be noted that deregulated cholesterol metabolism has recently been implicated in regulating tumor-infiltrating T lymphocyte activity and tumor-associated macrophage (TAM) polarization (57, 58). Elucidating how cholesterol availability affects the tumor microenvironment, tumor-infiltrating immune cells, and immune checkpoint function could suggest new combination therapies. SCARB1 may also be a valuable therapeutic target for patients with ccRCC whose tumors exhibit either inherent or acquired resistance to the FDA-approved HIF2α inhibitor, who may also benefit from dietary HDL and cholesterol restriction.

Further information and request for resources and reagents should be directed to and will be fulfilled by the corresponding author, M. Celeste Simon (celeste2@pennmedicine.upenn.edu). This study did not generate new unique reagents.

Matched tumor/normal samples were obtained from the Cooperative Human Tissue Network (CHTN). Tumors were homogenized in TRIzol (see quantitative real-time PCR) or whole cell elution buffer (see Western blot analysis) and analyzed for SCARB1 mRNA and protein expression. RNA-seq data for 538 ccRCC and 72 normal kidney samples were downloaded from TCGA in April 2020. Differential gene expression analysis of tumor and normal samples was performed using DeSeq (Bioconductor Version 2.12). TCGA mutation and copy-number data were downloaded from cBioPortal for Cancer Genomics.

MR uses genetic variants to estimate causal associations, reducing the potential for confounding and reverse causation that limit observational studies (59). We conducted two sample MR to estimate the causal relationship between 123 circulating metabolites and RCC risk (60). Summary statistics from a genome wide association study of circulating metabolites (n = 24,925; ref. 61) and RCC (n = 31,190; ref. 18) were obtained for the purpose of this analysis. We identified SNPs that were independently associated (r² < 0.001 and P < 5 × 10⁻⁸) with 119 of 123 metabolites detected by targeted nuclear magnetic resonance (NMR) metabolomics to instrument circulating metabolites; these included lipoprotein subclass–specific lipids, amino acids, fatty acids, inflammatory glycoproteins, and others (61). For each metabolite, we estimated the unconfounded OR for RCC per SD increment in metabolite using the inverse variance weighted method under a random-effects model (Supplementary Table S2; ref. 62). As a genetic variant in the HDL receptor (SCARB1) had been previously identified in an RCC GWAS (12), we wanted to confirm this was not sorely responsible for driving the causal association. To test the “exchangeability” and “exclusion restriction” assumptions (that the instrument does not share a common cause with the outcome and that any effect of the instrument on the outcome is exclusively through its potential effect on the exposure, respectively; ref. 63), we conducted a sensitivity analysis excluding a SCARB1 variant contained in a number of the metabolite instruments (Supplementary Table S3). F-statistics for genetic instruments indicated sufficient strength to satisfy the “relevance” assumption (Supplementary Tables S1 and S2) (64). Cochrane Q-statistics was used to assess heterogeneity of SNP effects (smaller P values were interpreted to indicate higher heterogeneity, suggesting directional pleitropy, Supplementary Tables S1–S3; ref. 65). All analyses were performed in R version 3.6.1 (R Core Team (2019). R: A language and environment for statistical computing) using the TwoSampleMR package (66, 67).

Experiments were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania (Philadelphia, PA). NIH-III nude mice (Charles River Laboratories, 4–6 weeks old) were injected in each flank with 2 × 10⁶ A498, or A498 shcontrol (shSCR) or A498 shSCARB1 cells. Prior to injection, cells were grown in complete media (DMEM containing 10% FBS) in 15-cm dishes. Cells were then resuspended in ice-cold PBS and were mixed 1:1 with Matrigel (BD Biosciences, catalog no. 356234) in a final volume of 200 μL per injection. For the experiments using an NC diet (Teklad Global 2018C) or HC diet (Teklad Custom Diet TD.01383 2018 + 2% Cholesterol), mice were fed the appropriate diet for 2 weeks before cell injection. Tumor volumes were recorded at the indicated timepoints using caliper measurements, calculated by the formula V = (π/6)(L)(W²), where L was the longer measurement and W was the shorter measurement. For the mice bearing xenografts and receiving vehicle 2-hydroxypropyl-β-cyclodextrin (20%) or BLT-1 (50 mg/kg) by oral gavage, tumor growth was closely monitored and when tumor volume reached 100 mm³, mice were randomized and divided into groups showing similar tumor volumes. At the end of the experiment, mice were sacrificed by CO₂ inhalation, and tumors were dissected for further analyses. Total cholesterol and HDL levels were assessed using a Lipidocare Analyzer following manufacturer's instructions.

Human ccRCC cell lines (786-O, A498), Caki-2, hTERT immortalized RPTEC and 293T cells were obtained from the ATCC and were cultured in DMEM (Thermo Fisher Scientific, catalog no. 11965092) supplemented with 10% FBS (Gemini Bio-Products, catalog no. 900–108) or 10% delipidated FBS, referenced as DLPS, (Gemini Bio-Products, catalog no. 900–123) supplemented with either cholesterol (10 μg/mL) or HDL (100 μg/mL). ATCC carries out routine authentication via short tandem repeat (STR) analysis. Immortalized proximal tubular renal epithelial cells (HK2) were obtained from ATCC and cultured in Keratinocyte Serum Free Media with appropriate supplements (Thermo Fisher Scientific, catalog no. 17005042). Cells were routinely tested to for Mycoplasma (MycoAlert) and experiments performed with early passages.

Human single-guide RNAs (sgRNA) targeting SCARB1 #1 (ctccgtggatctctccgtaa) and control gRNA targeting mouse Rosa26 locus (aagatgggcgggagtcttct) were cloned into LentiCRISPRv2 plasmid. Mature antisense human SCARB1 shRNA #1 sequence (clone ID: TRCN0000001673) along with scrambled (SCR) control were cloned into a doxycycline-inducible pLKO lentiviral plasmid (AddGene, catalog no. 21915). Lentivirus was prepared by cotransfection of 293T cells with shRNA or CRISPR plasmid of interest along with packaging plasmids pMD2.G (AddGene, catalog no. 12259), psPAX2 (AddGene, catalog no. 12260) and Lipofectamine PLUS and LTX transfection reagent (Invitrogen, catalog no. 15338). Lentivirus-containing media was collected from plates at 48 hours posttransfection, filtered using a 0.45-μm filter, and stored at −80°C. For viral transduction, cells were incubated with lentivirus-containing medium and 8 μg/mL polybrene for 6 hours. Cells were allowed to recover for another 48 hours before selection with puromycin. All experiments were performed with cells that survived puromycin selection and displayed knockdown/knockout of SCARB1 as assayed by Western blot analysis.

Xenograft tumors were dehydrated, embedded in paraffin, and sectioned for staining. IHC was performed as described previously (68) using 1:200 SCARB1 (Abcam, catalog no. Ab52629), 1:100 Ki-67 (Abcam, catalog no. Ab15580), and 1:400 cleaved caspase-3 (Cell Signaling Technology, catalog no. 9661). Briefly, xenograft slides or TMA slides (US Biomax KD20812 or Penn TMA) were deparaffinized by baking slides at 60°C for 30 minutes. The slides were rehydrated in series of ethanol solutions and endogenous peroxidase activities were quenched by 1% H₂O₂ in distilled water for 20 minutes. After three washes in TT buffer (500 mmol/L NaCl, 10 mmol/L Trizma, and 0.05% Tween-20), antigen retrieval was performed by boiling slides for 20 minutes in a citrate-based antigen unmasking solution (Vector Labs, H3300). After cooling down to room temperature, slides were blocked in 2% normal goat serum and 4% BSA in TT buffer for 1 hour. Next, tissue slides were incubated with various primary antibodies at 4°C overnight. After three washes in TT buffer, biotin-ylated secondary antibody was added onto these slides for 1 hour, following by 1-hour treatment of the Vectastain Elite ABC reagents (Vector Labs, PK-6100). After three TT washes, the slides were processed with DAB peroxidase substrate kit (Vector Labs, SK-4100), and hematoxylin solutions for IHC staining, dehydrated in a standard ethanol/xylenes series, and mounted in 75%v/v Permount (Thermo Fisher Scientific, SP15–500) in xylenes.

Total RNA was processed and extracted with TRIzol reagent (Thermo Fisher Scientific, catalog no. 15596026) and RNeasy Mini Kit (Qiagen, catalog no. 74104). cDNAs were synthesized using a High-Capacity RNA-to-cDNA kit (Applied Biosystems, 4368814). qRT-PCR were then performed using TaqMan Master Mix (Life Technologies) and a ViiA7 Real-Time PCR Instrument (Applied Biosystems). Predesigned Taqman primers were obtained from Life Technologies for the following genes: B2M (Hs0187842_m1), SQLE (Hs01123768_m1), HMGCR (Hs00168352_m1), SCARB1 (Hs00969821), LSS (Hs01552331_m1), DHCR24 (Hs00207388_m1), FDFT1 (Hs00926054_m1), 18S (Hs03928985_g1)

Cells were washed with PBS and lysed using lysis buffer (150 mmol/L NaCl, 10 mmol/L Tris pH 7.6, 0.1% SDS, and 5 mmol/L EDTA) containing Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific, 78445). For Western blots of xenograft tissue, approximately 10–20 mg of tissue was suspended in 500 μL lysis buffer and homogenized on ice using a Tissue-Tearor (Biospec, 985370). Samples were centrifuged at 12,000 rpm for 20 minutes at 4°C. Protein lysates were resolved by SDS-PAGE and were transferred to nitrocellulose membranes (Bio-Rad, 162–0115, 0.45-μm pore size for all experiments). All membranes were incubated with the indicated primary antibodies overnight at 4°C and were diluted in TBS-T (20 mmol/L Tris pH 7.5, 150 mmol/L NaCl, 0.1% Tween-20) supplemented with 5% BSA (Sigma-Aldrich, A7906). The next day and after TBS-T washes, membranes were incubated with secondary antibody and Western Lightning Plus-ECL, Enhanced Chemiluminescence Substrate (PerkinElmer, catalog no. NEL103E001EA) was used to visualize proteins.

Cell death was determined using the FITC–Annexin V, PI Kit (catalog no. 556547) from BD Biosciences according to the manufacturer's instructions. Briefly, 20,000 cells of each cell line were plated in triplicate in 6-well plates under the indicated conditions. Four days later, cells were prepared using the FITC–Annexin V, PI Kit (BD Biosciences, catalog no. 556547) according to the manufacturer's instructions. For the drug treatment, 24 hours after plating, cells were treated or not for 48 hours with either BLT-1. Flow cytometry was performed using the BD FACSCalibur instrument, with dead cells represented as Annexin V–positive population.

Cells were plated in 6-cm plates in duplicates and harvested at 80% confluency. Cells were then resuspended in 1× PBS and fixed with 70% cold ethanol, and incubated at −20°C 1 hour. Then, cells were washed with ice-cold PBS and resuspended in 1 mL PBS. They were treated with RNAse (100 μg/mL), stained with PI (20 μg/mL), and then analyzed by flow cytometry using the BD FACSCalibur instrument.

Twenty thousand cells of each cell line were plated in triplicate on 6-well plates and supplemented with either DMEM containing 10% FBS, 10% delipidated FBS, 10% delipidated FBS and cholesterol, or 10% delipidated FBS and HDL. Every other day and as indicated on the figures, one set of cultures was collected and counted. Cells were trypsinized and counted using a cell counter (Countess II Life Technologies) with Trypan blue (Gibco, catalog no. 15250–061).

Twenty thousand cells of each cell line were plated in triplicate on 6-well plates. Three days later, live cells were washed twice in PBS and incubated in DCFDA (Abcam, catalog no. 113851) for 1 hour at 37°C. After staining and trypsinization, cells were washed twice in PBS and resuspended in PBS, passed through a cell strainer, and flow cytometry was performed on a BD FACSCalibur instrument under FL-1.

Cytosolic and membrane fractionation of indicated cells were performed using Mem-PER Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific, 89842) by following the manufacturer's protocol.

Twenty thousand cells of each cell line were plated in triplicate on 6-well plates. HDL-cholesterol uptake assay was performed using HDL Uptake Assay Kit (Fluorometric; Abcam, catalog no. ab204717). Assays were performed according to the manufacturer's protocols. Briefly, ccRCC cell lines were incubated in the assay media containing fluorescently labeled HDL for 3 hours, then were washed, and the fluorescence was measured in a microplate reader at Ex/Em = 540/575 nm. Unlabeled HDL included in the kit is used for assay validation. The cell uptake was normalized by total protein amounts of each well.

Cholesterol content in ccRCC cells was measured with a Cholesterol Quantitation kit (Sigma, catalog no. MAK043), following manufacturer's instructions. Briefly, samples (10⁶ cells) were extracted with 200 μL of chloroform:isopropanol:IGEPAL CA-630 (7:11:0.1). After centrifugation at 13,000 × g for 10 minutes to remove insoluble material, the organic phase of samples was transferred to a new tube and air dried in an incubator at 50°C. Then organic solvent residue was removed in a SpeedVac for 1 hour. Dried lipids were then dissolved with 200 μL of the Cholesterol Assay Buffer. Fifty microliters of samples and standards (1–5 ng) were added to 50 μL of reaction mixture and absorbance at 570 nm measured after 60-minute incubation at 37°C.

Optima grade methanol (MeOH), water, acetonitrile, methyl tert-butyl ether (MTBE), and 2-propanol were from Thermo Fisher Scientific. Gasses were supplied by Airgas. Glassware and HPLC vials were from Waters Corp. Lipid extraction was done according to Matyash and colleagues (69). For the HRMS analysis, a recently calibrated QE Exactive-HF mass spectrometer (Thermo Fisher Scientific) was used in positive ion mode with a HESI source. The operating conditions were: spray voltage at 3.5 kV; capillary temperature at 285°C; auxiliary temperature 370°C; tube lens 45. Nitrogen was used as the sheath gas at 45 units, the auxiliary gas at 10 units, and sweep gas was 2 units. Same MS conditions were used in negative ionization mode, but with a spray voltage at 3.2 kV. Control extraction blanks were made in the same way using just the solvents instead of the cell homogenate. The control blanks were used for the exclusion list with a threshold feature intensity set at 1e10⁵. Untargeted analysis and targeted peak integration were conducted using LipidsSearch 4.2 (Thermo Fisher Scientific). An external mass calibration was performed using the standard calibration mixture approximately every three days. All samples were analyzed in a randomized order in full scan MS that alternated with MS2 of top 20, with HCD scans at 30, 45, or 60 eV. Full scan resolution was set to 120,000 in the scan range between m/z 250–1800. The pool sample was run every 10 samples. The coefficient of variation of less than 10% of the pooled sample was used to assess a robust signal for cholesterols esters. Calibration curves were prepared in the range 2 ng to 800 ng for cholesterol, and were plotted against the area ratio of the dehydration product (m/z 369.3513) to the corresponding internal standard. The top 14 abundant CEs were identified based on their MS2 data from the LipidsSearch library and quantification was done from the full scan in positive mode. The areas were normalized on the basis of the 3,500 ng of CE internal standard (18:1-D7) added per sample (as found in the SPLASH LIPIDOMIX). All amounts were normalized to the original protein amount per plate.

During blood sample collection, 0.5 mL was collected through retro-orbital bleeding. After the blood was prepared in the heparin-coated tubes, plasma was collected by centrifugation at 2,000 rpm for 10 minutes. The plasma levels of triglycerides, total cholesterol (TC), and HDL were examined using a plasma lipid profile kit (Standard LipidoCare from SD Biosensor) according to manufacturer's instructions.

Data are expressed as mean ± SD. The results were analyzed by a Student t test, and P values less than 0.05 were considered to be statistically significant (*, P < 0.05; **, P < 0.01, and ***, P < 0.001).

J. Godfrey reports grants from NIH NCI during the conduct of the study. M. Simon reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

R. Riscal: Conceptualization, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C.J. Bull: Conceptualization, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Mesaros: Conceptualization, investigation, methodology. J.M. Finan: Conceptualization, investigation, methodology. M. Carens: Conceptualization, investigation, methodology. E.S. Ho: Conceptualization, investigation, methodology. J.P. Xu: Conceptualization, investigation, methodology. J. Godfrey: Conceptualization, investigation, methodology. P. Brennan: Conceptualization, investigation, methodology. M. Johansson: Conceptualization, investigation, methodology. M.P. Purdue: Conceptualization, investigation, methodology. S.J. Chanock: Conceptualization, investigation, methodology. D. Mariosa: Conceptualization, investigation, methodology. N.J. Timpson: Conceptualization, investigation, methodology. E.E. Vincent: Conceptualization, investigation, methodology. B. Keith: Conceptualization, supervision, validation, investigation, visualization, methodology, writing–original draft. I.A. Blair: Conceptualization, supervision, validation, investigation, visualization, methodology, writing–original draft. N. Skuli: Conceptualization, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.C. Simon: Conceptualization, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing.

We acknowledge the NIH grants P01CA104838 and R35CA197602 (to M.C. Simon, which support N. Skuli), and Damon Runyon postdoctoral fellowship DRG2326-18 (to R. Riscal). We also acknowledge the Department of Defense Kidney Cancer Research Program grant W81XWH-20-1-0856 to M.C. Simon. Mendelian randomization and analysis were performed by E.E. Vincent and C.J. Bull supported by Diabetes UK (17/0005587), the Wellcome Trust (202802/Z/16/Z), the CRUK Integrative Cancer Epidemiology Programme (C18281/A29019) and the World Cancer Research Fund (WCRF UK), as part of the World Cancer Research Fund International grant program (IIG_2019_2009). Mass spectrometry and analysis were performed by E.S. Ho, J.P. Xu, and C. Mesaros supported by the National Institute of Environmental Health Sciences grant P30ES013508. We are grateful to J. Tobias and the Penn Molecular Profiling Facility for help with processing the human TCGA data and P. Lal for providing the tissue microarrays. We thank members of the Simon laboratory for helpful comments on the manuscript. Illustrations were generated using Servier medical art.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.