Clear cell renal cell carcinoma (ccRCC) is characterized by accumulation of neutral lipids and adipogenic transdifferentiation. We assessed adipokine expression in ccRCC and found that tumor tissues and patient plasma exhibit obesity-dependent elevations of the adipokine chemerin. Attenuation of chemerin by several approaches led to significant reduction in lipid deposition and impairment of tumor cell growth in vitro and in vivo. A multi-omics approach revealed that chemerin suppresses fatty acid oxidation, preventing ferroptosis, and maintains fatty acid levels that activate hypoxia-inducible factor 2α expression. The lipid coenzyme Q and mitochondrial complex IV, whose biogeneses are lipid-dependent, were found to be decreased after chemerin inhibition, contributing to lipid reactive oxygen species production. Monoclonal antibody targeting chemerin led to reduced lipid storage and diminished tumor growth, demonstrating translational potential of chemerin inhibition. Collectively, the results suggest that obesity and tumor cells contribute to ccRCC through the expression of chemerin, which is indispensable in ccRCC biology.
Identification of a hypoxia-inducible factor–dependent adipokine that prevents fatty acid oxidation and causes escape from ferroptosis highlights a critical metabolic dependency unique in the clear cell subtype of kidney cancer. Targeting lipid metabolism via inhibition of a soluble factor is a promising pharmacologic approach to expand therapeutic strategies for patients with ccRCC.
See related commentary by Reznik et al., p. 1879.
This article is highlighted in the In This Issue feature, p. 1861
Renal cell carcinoma (RCC) is one of the top ten cancers diagnosed globally, with more than 200,000 new cases worldwide each year (1, 2). RCC currently results in approximately 74,000 cancer cases and 15,000 mortalities annually in the United States (3, 4), and, according to the Surveillance, Epidemiology, and End Results (SEER) database, 16% of patients with RCC have distant metastatic disease, with a 5-year survival rate of only 11.6% (5).
The most common subtype of renal cancer is clear cell renal cell carcinoma (ccRCC), which accounts for more than 70% of all kidney cancers (6). ccRCC tumors almost uniformly have alterations in the von Hippel–Lindau (VHL) gene, including mutations, deletions, or hypermethylation, rendering the VHL protein (pVHL) functionally inactive (7, 8). In normal oxygen conditions (normoxia), the hypoxia-inducible factors (HIF) are hydroxylated by one of three prolyl hydroxylase domain (PHD) enzymes, leading to recognition and binding to pVHL. pVHL, which functions as an E3 ubiquitin ligase, subsequently targets HIFs for proteasomal degradation (9). In contrast, during hypoxia, or in ccRCC with a dysfunctional pVHL, HIF1α and HIF2α are not degraded and accumulate in cells. The HIFα subunits (predominantly HIF2α in ccRCC) then translocate into nucleus, bind to hypoxia-responsive elements (HRE) and drive transcription of a host of hypoxia-responsive adaptation genes (10). Some of the HIF targets include vascular endothelial growth factor (VEGF), glycolysis genes, and cyclin D1 (CCND1), which contribute to the angiogenesis, Warburg metabolism, and hyperproliferative phenotypes of ccRCC.
The “clear cell” type of RCC is characterized by high lipid and glycogen content (11). Lipid homeostasis in cells is a tightly regulated process to maintain normal cellular functions, including lipid synthesis, uptake, and subcellular distribution or compartmentalization (12). Of all, the storage of fatty acid (FA) in lipid droplets, which is the histologically identifiable phenotype of ccRCC, is an important equilibrating mechanism to prevent lipotoxicity. FAs can be used as building blocks for complex lipids and can also be used as substrates for beta-oxidation (β-oxidation) in mitochondria to yield energy (13). Dysregulated metabolism is a well-established hallmark of cancer and an enticing target for cancer management. In ccRCC, reports have shown that lipid metabolism contributes to metabolic reprogramming and unbridled cell growth (14). Yet, the exact mechanism and the significance of the lipid-storage phenotype remain poorly understood. The potential to exploit a metabolic vulnerability as a therapeutic strategy in lipid-associated cancers also remains largely unexplored.
In the past few decades, cases of obesity and obesity-related disorders have been increasing steadily in both developed and developing countries, leading to the declaration by the World Health Organization of a global obesity epidemic in 1999 (15). Evidence has also established that tumorigenesis is escalated by adiposity, showing a direct link between metabolism and cancer. In fact, obesity (body mass index at least 30 kg/m2) is a known risk factor for ccRCC, and perinephric fat invasion has been associated with poor disease outcome in ccRCC (10, 16). Hypertrophic obesity (increase in adipocyte size), hyperplastic obesity (increase in adipocyte number), and increased intracellular lipids can contribute to an increase in circulating proinflammatory adipokines that act on peripheral tissues and stimulate key cellular biological processes, recently found to include tumorigenesis (17). Realizing the intimate interplay between adiposity and cancer, we hypothesized that adipokines could play detrimental roles in driving tumorigenesis in lipid-laden ccRCC. Notably, the roles and mechanisms of obesity-related factors in increasing cancer risk remain unclear.
In the current study, by combining single-cell resolution transcriptomic data, metabolomic analyses, and multiple ccRCC model systems, we discovered that the adipokine chemerin, which is encoded by the retinoic acid receptor responder 2 (RARRES2) gene, is overexpressed in ccRCC due to both an autocrine, tumor cell–dependent mechanism and obesity-dependent paracrine production, and it plays important roles in regulating lipid metabolism and tumorigenesis. We found that chemerin depletion results in dramatic metabolic rewiring in which excessive lipid oxidation leads to mitochondrial dysfunction and ferroptosis. Chemerin controls expression of over 300 genes, most notably HIF1α and HIF2α, through a FA-dependent mechanism. The secretory nature of chemerin makes its elevation amenable to detection in the plasma of patients with ccRCC and further to interruption through the use of a monoclonal antibody. Accordingly, we find targeting chemerin to be effective in controlling orthotopic xenograft tumors in vivo. Together, the data provide a biological link among obesity, suppression of fatty acid oxidation (FAO), and tumor growth in lipid-laden ccRCC, suggesting a potential molecular marker and therapeutic target to improve standard therapeutic approaches.
Chemerin is Elevated in Patients with ccRCC through Obesity- and Tumor-Dependent Processes
In analyses of bulk tumors, ccRCC cells show evidence of transdifferentiation into adipocytes (18), which are known to secrete a variety of factors that both maintain their own adipocyte functions and promote adipocyte differentiation of distant cells. To validate transdifferentiation in ccRCC tumor cells, we utilized single-cell RNA sequencing (scRNA-seq) of tumors (n = 3) and normal kidney tissues (n = 5; ref. 19). By comparing the tumor cell and normal kidney epithelium clusters (Supplementary Figs. S1 and S2A–S2E; Supplementary Table S1), we verified that ccRCC tumor cells display a reduced renal epithelial gene signature in favor of an adipogenic signature (Fig.1A). Tumor cells also demonstrated dedifferentiation and epithelial-to-mesenchymal transition, in agreement with the literature (Fig.1A; ref. 20).
In an effort to identify a potential autocrine adipokine that could be critical for promoting and maintaining ccRCC, we performed in silico data mining using The Cancer Genome Atlas (TCGA) and Oncomine databases, and we examined the association of 100 known adipokines with ccRCC, some of which are shown in Fig.1B. The adipokines were ranked according to the number of studies showing an overexpression profile in ccRCC and the correlation with patient survival; the adipokine chemerin stood out as a candidate. From scRNA-seq data, we determined that tumor expression of chemerin arises from ccRCC tumor cells rather than other cells in the tumor microenvironment, agreeing with a possible autocrine pathway (Fig.1C; Supplementary Table S2). In TCGA, chemerin mRNA expression is higher in ccRCC clinical samples (n = 533) as compared with adjacent normal kidney tissues (n = 72) (Fig.1D and E), and higher chemerin expression portends a negative prognosis in patients with ccRCC (upper third vs. lower third; Fig.1F). In contrast, chemerin expression is not associated with outcome in patients with either papillary renal cell carcinoma (pRCC; Supplementary Fig. S3A) or chromophobe renal cell carcinoma (Supplementary Fig. S3B) when considering upper and lower thirds, or it may even define a better outcome when considering upper and lower fifths (21).
Because chemerin is notably overexpressed in ccRCC tumors, we next measured plasma chemerin levels and investigated the possibility of using chemerin as a biomarker for patients with ccRCC. We found that chemerin protein levels are significantly elevated in patient plasma relative to that of healthy individuals (Fig.1G). The area under the curve (AUC) was 0.99 at the cutoff value of 121.6 ng/mL, demonstrating a sensitivity of 98.31% and specificity of 95.83% (Supplementary Fig. S3C). Because previous studies have indicated that chemerin is positively correlated with adiposity (22), we considered patient body mass index (BMI) among the patient samples and found that obese patients (BMI ≥ 30) had significantly higher circulating chemerin levels than overweight patients (25 ≤ BMI < 30) or patients with low BMI (BMI < 25; Fig.1H). This confirmed the notion that the paracrine chemerin production in ccRCC is obesity-dependent. Importantly, plasma chemerin remains elevated when comparing low-BMI patients with ccRCC to healthy individuals with low BMI, as well as high-BMI patients with ccRCC to controls with high BMI (Fig.1H). We then assessed chemerin production from tumors, normal adjacent kidney tissue, tumor-adjacent fat, and fat distant to the tumor by performing qRT-PCR to determine the greatest contributor of chemerin. We again found significantly elevated RARRES2 mRNA in tumor samples versus normal tissue and that adjacent and distant fat produced equally higher levels than normal kidney, but not as much as the tumor itself (Supplementary Fig. S3D). We validated the expression changes with immunohistochemistry staining of chemerin protein using a tumor microarray (TMA) with 30 ccRCC samples and associated adjacent normal kidney sections (Supplementary Fig. S3E and S3F). Although chemerin was clearly higher in tumors, we noticed somewhat higher than expected signals in the “normal” regions, possibly due to possible diffusion of the soluble protein from nearby tumor. We then stained fresh-frozen tumor samples with the lipophilic dye Oil Red O (ORO) and compared ORO uptake with plasma chemerin. We found a significant positive correlation, suggesting an association between plasma chemerin, regardless of source, and lipid levels in ccRCC tumors (Fig.1I; Supplementary Fig. S3G). Furthermore, the plasma chemerin levels are also correlated with tumor size independently of BMI (Fig.1J). Again, we assessed chemerin in a non–clear cell subtype (pRCC), and, although circulating levels were elevated over controls, they remained significantly lower than in ccRCC (Supplementary Fig. S3H). Together, the data validate that chemerin is a potentially relevant adipokine in ccRCC.
Chemerin Sustains ccRCC Tumor Growth
To functionally dissect a contribution of chemerin in ccRCC, three different short hairpin RNAs (shRNA) were used to stably knock down the chemerin gene RARRES2 in four different ccRCC cell lines: 786-O, 769-P, UOK101, and A-498 (Fig.2A–D, top panels; Supplementary Fig. S4A–S4C). All of these ccRCC cell lines contain detectable amounts of chemerin protein (Supplementary Fig. S4D). All ccRCC isogenic cell lines demonstrated impaired cell proliferation capacity after chemerin knockdown compared with shGFP controls (Fig.2A–D, bottom panels), but not HK-2 cells, which are immortalized proximal tubule epithelial cells (Supplementary Fig. S4E and S4F). Interestingly, the magnitude of inhibition of proliferation appears to correlate with overall chemerin expression levels (i.e., greatest effect in UOK101 and 769-P). We also observed a reduction in colony-forming ability in 786-O cells after chemerin knockdown, another surrogate marker for reduced proliferation ability (Supplementary Fig. S4G). To eliminate potential off-target effects of shRNAs, we also utilized CRISPR/Cas9 technology to knock out chemerin, and the cells showed an even more pronounced impairment of proliferative capacity, where almost all cells died by day 9 of the assay (Fig.2E). To investigate whether chemerin is important for ccRCC tumorigenesisin vivo, the 786-O and 769-P cell models infected with either shGFP control or shRARRES2 viruses were transplanted subcutaneously into the flanks of nude mice. Similarly, two unique CRISPR/Cas9 clones of 786-O cells with chemerin knocked out and a polyclonal 786-O population with control single-guide RNA (sgRNA) were also implanted subcutaneously into the flanks of nude mice. Tumors with chemerin inhibition uniformly showed dramatically impaired tumor formation capability in both the knockdown (Fig.2F; Supplementary Fig. S4H) and knockout systems (Fig.2G; Supplementary Fig. S4I), as well as reduced Ki-67 staining in the knockout system (Supplementary Fig. S4J).
Next, we interrogated the cellular changes that contribute to the functional phenotypes observed by measuring DNA synthesis as a surrogate for proliferation and by assessing Annexin V/propidium iodide (PI) staining for programmed cell death. Knockdown of chemerin in both 786-O and 769-P cells reduced DNA synthesis, derived from reduced 5-ethynyl-2′-deoxyuridine (EdU) incorporation (Fig.2H), and Annexin V/PI staining showed increased cell death in the cells after chemerin knockdown (Fig.2I). Thus, chemerin expression in ccRCC tumor cells is functionally critical for sustaining tumor cell survival.
Inhibition of Chemerin Reprograms Lipid Metabolism in ccRCC and Promotes FAO
To understand whether the chemerin-mediated survival in ccRCC is driven by a change in lipid phenotype, we investigated the effect of chemerin on lipid droplets. To this end, chemerin expression was depleted using shRNAs, and the cells were stained with the lipophilic dye BODIPY 493/503 (4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene). Chemerin silencing significantly reduced lipid deposition in the cells, an effect that was partially rescued by incubation of the cells with 50 nmol/L of recombinant chemerin protein for 18 hours (Fig.3A). The reduction in lipid deposition was also seen in chemerin knockout tumors measured by ORO staining (Fig.3B). RNA-seq of 786-O shGFP control and shRARRES2 cells revealed that chemerin regulates numerous lipid metabolism genes (Fig.3C–E; Supplementary Fig. S5A). According to gene set enrichment analysis (GSEA) of the transcriptomes, several lipid metabolism pathways were downregulated, including the lysophospholipid pathway, steroid hormone biosynthesis, and ether lipid metabolism (Fig.3D; Supplementary Fig. S5B). Some of the genes were then confirmed by qRT-PCR. The genes tested included the FAO enzymes CPT1A, ACAD9, DBI, and ACOT7 (Fig.3E), which were upregulated after chemerin silencing in ccRCC and suggested increased FAO. In contrast, the transcription of the lipid uptake protein FABP7 and the lipid droplet component protein PLIN4 were reduced after chemerin knockdown (Fig.3E). Consistent with the gene expression changes, there was a reduction in fatty acids measured by mass spectrometry (Fig.3F), especially the saturated FAs (SFA) and monounsaturated FAs (MUFA). These results confirm dramatic lipid metabolism alterations after chemerin inhibition, highlighting an important role of chemerin in regulating the ccRCC cellular lipid profile.
To determine if decreased tumor cell growth (and possibly tumorigenesis) from chemerin suppression is due to the enhanced mitochondrial β-oxidation of lipids, we treated chemerin-inhibited cells with etomoxir, an inhibitor of the FA mitochondrial transporter CPT1. We found that etomoxir treatment rescued the reduction in DNA synthesis (Fig.3G; Supplementary Fig. S5C) and proliferation defect (Supplementary Fig. S5D) from chemerin silencing. This observation suggests that chemerin-mediated suppression of FAO is necessary to maintain cell proliferation in ccRCC.
Chemerin Inhibition Increases Oxidized Lipid and Enhances Ferroptosis Susceptibility
To further characterize the role of chemerin in lipid metabolism, we utilized an untargeted lipidomics approach (Fig.4A and B; Supplementary Fig. S6A and S6B). The lipid species in the analysis included phosphatidylcholine (PC), plasmanyl-PC, phosphatidylethanolamine (PE), phosphatidic acid (PA), ceramides, coenzyme Q (CoQ), acetoacetate (AcAc), phosphatidylglycerol (PG), sphingomyelin, sphingosine, oxidized triacylglycerols, oxidized PE, oxidized PC, monogalactosyldiacylglycerol (MGDG), monoglyceride (MG), lysophosphatidylethanolamine (LPE), lysophosphatidylcholine (LPC), oxidized LPC, and diacylglcerols (DG; Supplementary Fig. S6C and S6D). Consistent with transcriptomic findings demonstrating increased FAO, we observed an overall increase in oxidized or breakdown lipids and reduction of intact lipids in 786-O cells transfected with shRNAs targeting chemerin compared with shGFP control (Fig.4C). For example, during mitochondrial FAO, carnitine is added to fatty acyl-CoA by CPT1 to form acylcarnitines, which are then transported across mitochondrial inner membrane (8). Notably, the fatty acylcarnitine species that feed the downstream FAO cycle were significantly reduced after knockdown of chemerin, validating an increase in oxidation of acylcarnitine species (Fig.4D). Pathway analysis of the metabolome also revealed upregulated mitochondrial β-oxidation; taurine and hypotaurine metabolism, which promotes lipid metabolism (23); and amino acid pathways that regulate redox balance (24) in the cells after chemerin knockdown (Fig.4E).
The significant reduction in glycerophospholipid (GPL) species including PA, PC, and PG (Fig.4F) observed in chemerin-deficient cells is accompanied with a complementary increase in corresponding oxidized GPL products (Fig.4G). Concurrently, LPC, which is produced after removal of a single acyl group from standard phospholipids with two tails, was increased (Fig.4H; Supplementary Fig. S6E). This suggests that chemerin-knockdown cells scavenge FA tails from most of the GPL species as substrates for β-oxidation. In addition to oxidation in GPL, other lipid classes were also observed to have undergone oxidation or breakdown. In sphingolipids, intact sphingomyelin was reduced after chemerin knockdown, and the breakdown product sphingosine was significantly increased (Supplementary Fig. S6F). Another product in sphingolipid metabolism, ceramide, also known to induce cell death, was increased after chemerin knockdown (Supplementary Fig. S6F). A similar trend was observed in glycerolipids (Supplementary Fig. S6G).
As the accumulation of oxidized lipid products is known to induce ferroptosis in cells (25), we next sought to investigate if the chemerin knockdown–induced cell death could be explained by ferroptosis. We found that polyunsaturated FA (PUFA) species, potent ferroptosis inducers, increased after chemerin inhibition (Fig.4I, orange bars). The increase in PUFA was observed in peroxidized glycerophospholipids, particularly oxidized PC containing PUFA chains, including arachidonic acid (C20:4) and docosahexaenoic acid (C22:6). On the other hand, MUFA, which reduces PUFA incorporation into phospholipids and reduces ferroptosis, was decreased (Fig.3F). In addition, the increase in LPC is a lipidomic signature of ferroptosis when oxidized PUFA is cleaved from the sn2 position of phospholipid (Fig.4H; Supplementary S6E). PUFAs have been reported to be tumoricidal by enhancing free radical generation (26). This was also evidenced in cells with chemerin knockdown, where we observed an increase in the production of lipid reactive oxygen species (ROS) measured by C11-BODIPY 581/591 (Fig.4J). As a proof of concept, the treatment of chemerin-deficient cells with ferroptosis inhibitors, 1 μmol/L ferrostatin-1 and 0.5 μmol/L liproxstatin-1, rescued the proliferation defect (Fig.4K). Collectively, the results suggest that ferroptosis is induced in ccRCC after chemerin inhibition by peroxidized lipid species with increased PUFAs and elevated lipid ROS.
Reduction of OXPHOS Complex IV after Chemerin Inhibition Enhances ROS Production
As part of the lipid reprogramming in ccRCC cells after chemerin inhibition, we found that CoQ10, the major form of CoQ found in humans, was also decreased (Fig.5A). Because CoQ is an electron carrier of the mitochondrial respiratory chain (MRC), we hypothesized that chemerin silencing results in mitochondrial dysfunction that could promote ferroptosis. We then analyzed the other components of the MRC and oxidative phosphorylation system (OXPHOS) upon chemerin knockdown. We observed that the level of complex IV (CIV), the fourth and last enzyme of the MRC, was also decreased (Fig.5B and C; Supplementary Fig. S7A). It is relevant to note that CIV contains heme A as a cofactor (27). Heme A is essential for CIV enzymatic activity and is also required for CIV biogenesis, as defects in heme A synthesis severely compromise CIV holoenzyme formation (28). In cells, heme A is present exclusively in CIV and differs from heme B (the more common type of heme in the cell), as it has a farnesyl group at the C2 position of the porphyrin ring. Lipid profiling confirmed that farnesyl-diphosphate, which is the farnesyl group donor for heme A biosynthesis, was reduced after chemerin silencing (Fig.5A), suggesting a possible defect in heme A and, consequently, CIV levels. Furthermore, several phospholipids were shown to intimately interact with CIV and were proposed to play a role in its function and stability (29). The reduction of phospholipid after chemerin inhibition (Fig.4F) could also contribute to the observed decrease in CIV.
Due to the alteration in the MRC components CoQ and CIV, we next sought to investigate whether the mitochondrial respiratory function was affected by chemerin inhibition by polarographic analysis of cellular respiration. Consistently, we observed a defect in respiratory capacities, including endogenous, coupled, and maximal respiration (Fig.5D), which was similarly observed in the Seahorse Mito Stress Test assay (Supplementary Fig. S7B). This was associated with a reduction in adenosine triphosphate (ATP) production (Fig.5E). In order for β-oxidation to support ATP synthesis, cells require a functional MRC so that the electrons from lipid oxidation can be transferred to MRC complexes by the reducing equivalents NADH and FADH2. However, chemerin knockdown cells have high FA β-oxidation but low respiratory capacity. This is likely to create an unbalance between the high electron influx from the β-oxidation pathway and the downstream slow electron flux in the MRC. A large input of electrons in a CIV-compromised MRC represents a favorable condition for the premature electron escape from MRC CI and CIII to molecular oxygen, generating superoxide radicals. The increase in ROS could in turn contribute to promote ferroptosis (Fig.4J), consistent with previous observations (30). It is important also to note that there was no structural change in the filamentous mitochondrial network nor the number of mitochondria after silencing of chemerin (Supplementary Fig. S7C).
Chemerin Expression Is VHL-Dependent and Regulates HIF in ccRCC
We next sought to investigate underlying signaling mechanism of how chemerin regulates FAO and mitochondrial functions. Because most of the ccRCC tumors are VHL-null, we first examined whether chemerin expression is VHL-dependent. By interrogating the scRNA-seq data, we found that chemerin was detected at higher expression only in ccRCC with VHL mutation, but generally not in rare ccRCC that retains wild-type VHL or in normal kidney epithelium (Fig.6A; Supplementary Fig. S8A). To validate this observation, 786-O cells with VHL mutation were transduced with an adenovirus overexpressing VHL protein (Fig.6B) to restore VHL function of HIF regulation. When VHL was expressed, we observed destabilization of HIF2α and a subsequent reduction in both the chemerin protein and mRNA level (Fig.6B; Supplementary Fig. S8B). HIF2α silencing also transcriptionally reduced chemerin expression (Fig.6C; Supplementary Fig. S8C). Thus, chemerin expression in ccRCC is VHL-dependent. In the case of wild-type VHL and HIF2α dependence, hypoxia could therefore lead to elevated expression as well.
We noted from the literature the recent description of the zinc finger DNA-binding transcription factor KLF6 as a target of the VHL–HIF2 axis in ccRCC, and after interrogating the list of KLF6-responsive genes we found chemerin (Supplementary Fig. S8D; ref. 31). Accordingly, knockdown of KLF6 leads to reduced chemerin expression (Fig.6D; Supplementary Fig. S8E and S8F). Furthermore, RNA-seq data for multiple ccRCC cell lines from the Cancer Dependency Map (www.depmap.org) demonstrate a positive correlation between KLF6 and RARRES2 expression (R2 = 0.8344; P = 0.0302; Supplementary Fig. S8G). Together, the data suggest that VHL alterations, through deregulation of HIF2α and KLF6, mediate tumor-derived chemerin expression in ccRCC.
Strikingly, several FAO genes in our analysis that were regulated by chemerin have been reported as HIF target genes (14, 32), and chemerin inhibition reduced transcription of both HIF1A and EPAS1 in cells (Fig.6E). HIF2α protein levels were also significantly decreased by chemerin silencing in both 786-O and 769-P cells (Fig.6F), where HIF1α protein is not detectable. GSEA of the RNA-seq data confirmed that, in addition to the lipid pathways noted in Fig.3D, the hypoxia pathway was one of the most downregulated pathways after chemerin knockdown (Fig.6G; Supplementary Fig. S8H). We validated reduced expression of the HIF target genes VEGF, LOX, and IGFBP3 by qRT-PCR (Fig.6H). Interestingly, as a potential link between lipid metabolism and HIFα mRNA expression, Yoo and colleagues demonstrated previously that HepG2 cells induce HIF1α expression as a protective measure against FA-induced toxicity (33). Exogenous addition of palmitate, for example, led to a robust induction of HIF1A mRNA. We thus tested whether levels of free FA in ccRCC cells could promote HIF1α or HIF2α expression. We found that palmitate (an SFA), opposite pattern as seen in linoleic acid (a PUFA), increases mRNA (Fig.6I) and protein levels of HIF2α and subsequently chemerin in ccRCC cells (Supplementary Fig. S8I); thus, we propose that through maintenance of low FA metabolism, chemerin participates in a positive feedback loop to maintain high levels of HIF (Fig.6J).
Monoclonal Antibody against Chemerin Reduces ccRCC Tumor Growth
Given the significant clinical correlates of chemerin in ccRCC, we investigated the potential translational impact of inhibiting chemerin with a monoclonal antibody (mAb). Treatment of ccRCC 786-O cells showed a dose-dependent reduction in cell viability with increasing concentrations of chemerin mAb (Fig.7A, top left panel). In contrast, cell viability was not affected when HK-2 cells were treated with a similar dose range of the antibody (Fig.7A, top right panel). Likewise, neither cell line was affected by control immunoglobulin G (IgG) mAb treatment at the same doses (Fig.7A, bottom panels). Importantly, 786-O cells treated with chemerin mAb also showed a decrease in lipid droplets (Fig.7B) and an increase in lipid ROS (Fig.7C) as compared with treatment with IgG.
Next, we examined the potential for therapeutic targeting of chemerin using the antibody in an orthotopic ccRCC model. 786-O cells that were engineered to express firefly luciferase were transplanted under the left kidney capsules in nude mice, and tumor growth was monitored using bioluminescence imaging (BLI). The mice were randomly assigned to two different treatment arms: one receiving 20 mg/kg chemerin mAb twice weekly, and the other receiving IgG control mAb of the same dose. The doses of antibody used in this study did not result in significant changes in animal weight, suggesting minimal toxicity (Supplementary Fig. S9A). Consistent with the cell culture results, in vivo 786-O tumor growth was significantly reduced by treatment of antibody targeting chemerin (Fig.7D and E). The mice treated with chemerin mAb also had lower tumor weights than those receiving IgG mAb upon sacrifice (Fig.7F and G). Collectively, the results suggest that targeting chemerin therapeutically in ccRCC represents a promising avenue that could have translational potential.
In the current report, we utilized a multidisciplinary approach including scRNA-seq, functional genomics, untargeted lipidomics, and metabolomics to provide new insights into the role of the adipokine chemerin in lipid metabolism and tumorigenesis in ccRCC. The identification of HIF-dependent chemerin expression to prevent FAO and escape from ferroptosis highlights a critical metabolic dependency of ccRCC on suppression of lipid metabolism. The findings have several direct implications, including the linkage of an obesity-driven systemic factor with tumor promotion, as well as a novel regulatory mechanism of HIF signaling. The soluble nature of chemerin suggests that both biomarker and therapeutic opportunities may be useful in devising new strategies to treat kidney cancer.
Obesity has been recognized as a global public health issue, and it correlates with higher incidence of cancer. With more than half of the U.S. population currently overweight (BMI more than 25), it is thought that increased body mass is contributing to the rise in kidney cancer incidence (34). The hypertrophied adipose tissues in obesity can trigger inflammation, hypoxia, and angiogenesis, which subsequently may promote tumor formation (35). Although efforts are under way to elucidate mechanisms explaining the link between obesity and cancer, the dysregulated secretion of signaling molecules by adipose tissue (adipokines) is a likely culprit (36). Over the last decade, many reports have demonstrated that adipokines play roles in cancer development. For example, some adipokines, including leptin, have proinflammatory roles and can stimulate cancer stem cells to become more tumorigenic (35). Chemerin, a recently discovered multifunctional adipokine, has been shown to regulate adipogenesis, inflammation, angiogenesis, and energy metabolism (17). Chemerin expression increases when preadipocytes differentiate into adipocytes, and circulating chemerin has been correlated with BMI of patients (17). However, the underlying mechanisms of adipokines on cancer pathogenesis and lipid metabolism remain enigmatic. Because normal kidney epithelium undergoes adipogenic transdifferentiation when becoming ccRCC, we explored the dependency of autocrine- and paracrine-derived chemerin in lipid metabolism and tumor survival.
Although we found consistently that chemerin expression is elevated at both the mRNA (from database analyses) and the protein level (from the TMA) in ccRCC as compared with normal adjacent tumors, it is interesting to note that the magnitude of change tended to be less in the protein staining. Considering the extracellular role of chemerin, it is not unreasonable to hypothesize that tumor-derived protein diffuses into the adjacent tissue, as has been shown for VEGF (37), much like chemerin production from systemic or adjacent fat can diffuse into the tumor or a premalignant lesion and contribute to ccRCC biology. The nature of a secreted factor thus holds promise of both a detectable biomarker and an actionable target pathway. One significant question is the relative contributions of systemic chemerin (potentially related to obesity) versus tumor-derived chemerin to ccRCC. Our data support a model where obesity elevates chemerin in a supporting manner, which is superseded by tumor production to maintain tumor growth. Considering the levels in our cohort of patients, it appears that approximately 25% of the circulating chemerin could be attributable to obesity, whereas 75% could be associated with the presence of a tumor. One could also argue that chemerin is likely not driving tumorigenesis as an oncogene but rather promoting expansion and survival of an incipient VHL-deficient cell as it rewires its metabolism via constitutive HIF activation. An obesity environment may therefore be tumor-promoting.
Critically, although obesity is recognized as a risk factor of ccRCC, it is commonly accepted that higher BMI is associated with improved survival (38–40). It remains unclear as to whether this phenomenon, known as the “obesity paradox,” is due to the nature of tumors in obese patients, reverse causation of cancer cachexia, or detection bias. Nevertheless, how chemerin would fit into a model as an obesity-dependent driver of ccRCC and yet be prognostic of worse outcome could be explained by several possibilities. First, it is unlikely that chemerin is the only adipokine playing a role in ccRCC. Indeed, in our initial informatics screening process, we identified other potential proteins that associate with poor outcome and are also overexpressed in tumors. We did not look for potentially counteracting molecules that could explain better outcomes, but it is possible that they exist. Second, our studies have not attempted as of yet to determine how chemerin affects responses to standard-of-care approaches such as tyrosine kinase inhibitors. It is possible that obesity contributes to tumor formation, but also results in better drug distribution or responses. In fact, there are reports that obesity associates with more highly angiogenic tumors, and chemerin itself has noted roles in recruiting endothelial progenitor cells through a proangiogenic role (41, 42). Thus, further work will be required to delineate all of the implications of obesity in ccRCC, and specifically the functions of chemerin beyond metabolic rewiring.
Accumulation of lipids in lipid droplets is a well-established phenomenon in ccRCC, but the significance of lipid storage is still ambiguous. Some have suggested that the significance is to prevent lipotoxicity, and others have proposed a role in suppressing endoplasmic reticulum stress (32, 43). The observation that chemerin prevents ferroptosis by suppressing FAO aligns more with the evasion of lipotoxicity theory, but the fact that reduced FAO also supports elevated HIF expression adds a new layer of benefit to ccRCC tumors. Ferroptosis is a recently described form of regulated cell death that results from iron-dependent lipid peroxidation and is a promising strategy to induce cancer cell death (44). It was reported by Yang and colleagues (45) that RCC cells are more prone to ferroptosis if they are grown at low cell density (<50% confluency) as compared with high cell density. This could be due to the inadequate production of a secreted factor, such as chemerin, to support cell survival. Interestingly, we also noted in a prior work that induction of lipid droplets in vitro is density-dependent (14). Thus, it is intriguing to think that the link between ferroptosis evasion and lipid droplet production is the common secretion of chemerin.
Although our current work demonstrated an important role of chemerin in ccRCC tumor biology, the question remains whether or not a receptor is involved for the action of this circulating adipokine. In the literature, chemerin acts as a natural ligand for chemokine-like receptor 1 (CMKLR1, also known as ChemR23), G protein–coupled receptor 1 (GPR1) and C-C chemokine receptor-like 2 (CCRL2). Among these, CMKLR1 is the most well-studied in cancers, and chemerin/CMKLR1 has been shown to be involved in gastric cancer (46), hepatocellular carcinoma metastasis (47), and immune cell trafficking (48). In our case, after knocking out CMKLR1 in 769-P (Supplementary Fig. S10A and S10B), we observed a decrease in cell proliferation ability (Supplementary Fig. S10C) and lipid deposition (Supplementary Fig. S10D), similar to the effect after chemerin silencing. Some of the lipid metabolism genes regulated by chemerin (CPT1A and PLIN4) were also downregulated by CMKLR1 knockout (Supplementary Fig. S10E). This suggests the involvement of CMKLR1 receptor in lipid biology of chemerin in ccRCC. Yet more work is necessary to fully elucidate the functional role of CMKLR1 in the ferroptosis pathway. In addition, chemerin binds to GPR1 with an affinity similar to that of CMKLR1, and GPR1 has been demonstrated to be involved in glucose homeostasis in obesity (49), breast cancer (50), and adipogenesis (50). On the other hand, CCRL2 is not involved in signal transduction, and its role in human biology is still poorly understood. Hence, it is of great interest to clearly delineate the roles of these receptors in ccRCC biology, as they may also be useful therapeutic targets.
Chemerin has a physiologic role as an inflammatory mediator in homeostasis, as well as recently described pathologic inflammatory roles in multiple cancers (51). Thus, the effect of chemerin on tumor growth could also be driven by its function on inflammatory cells. Chemerin has been reported to recruit immune effector cells and increase T cell–mediated cytotoxicity in breast cancer (51), prostate cancer (52), and sarcoma models (53). However, these cancers, in contrast to ccRCC, have lower chemerin expression in tumor tissues than in healthy individuals, and higher chemerin expression is correlated with better survival outcome (51). It was demonstrated in multiple reports that chemerin increases natural killer cell and CD8+ T-cell infiltration in orthotopic syngeneic models; thus, these tumors downregulate chemerin expression as a means of immune escape (51). On the other hand, chemerin has also been shown to promote regulatory T-cell differentiation and promotes polarization of M2 macrophages in pRCC, both of which are tumor-promoting (54). Interestingly, it was also noted that chemerin may play a role in increasing lipid deposition in macrophages (51), showing the interplay of chemerin in immune cell modulation, lipid deposition, and tumor progression. In the study of pRCC, elevated chemerin did associate with worse survival (21). Our studies in ccRCC thus far have considered only chemerin in an immunodeficient setting, thereby excluding potential additional roles of inflammatory cell modulation. Future experiments are planned to delineate the broader roles of chemerin in ccRCC in an immunocompetent model and to address a potential interaction with efficacy of immunotherapy.
The classic triad symptoms of RCC are hematuria, abdominal pain, and a palpable mass, yet most patients are asymptomatic until incidental detection by radiographic imaging (34). Most of the cases, therefore, have advanced to higher stages or even metastasized when diagnosed (34). Hence, there is a need to have a circulating biomarker that could be useful for early detection. Several adipokines have been described as possible biomarker candidates. For example, a large case–control study in 2013 reported reduced serum adiponectin in patients with RCC, but one of the biggest limitations of the study is that only males were included (55). Moreover, the correlations of adipokines (including adiponectin and leptin) and RCC risks are inconsistent in multiple reports, further weakening their potential as biomarkers (56, 57). In a recent report by Lu and colleagues (58), serum chemerin was higher in patients with oral squamous cell carcinoma than healthy individuals. In the current report, we detected an elevated chemerin level in ccRCC patient plasma, which is also correlated with BMI status. Despite the potential of chemerin as plasma biomarker based on current data, more analyses in a larger cohort or a case–control study are needed to reduce the potential confounding factors such as other inflammatory states. Nonetheless, it is possible that plasma chemerin can be informative when combined with existing practice tools for early detection and follow-up of patients with ccRCC.
The implications of an adipokine regulating lipid metabolism and ferroptosis in ccRCC could have broad impact, as obesity is increasingly recognized as a risk factor for a multitude of cancers. Before the recent advances in immunotherapy, the primary focus of cancer therapy had traditionally been on tumor and tumor microenvironmental phenotypes. Consideration of systemic mediators of transformation and progression has remained an overlooked attribute. Although obesity-driven metabolic mediators may be particularly relevant for highly lipid-laden cancers such as ccRCC and clear cell ovarian tumors, metabolic alterations in cancers are essentially universal. Targeting chemerin to block its downstream effects on FAO and ferroptosis is therefore a conceptually novel therapeutic approach that possibly could be modeled in additional diseases.
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Scott M. Welford (email@example.com). This study did not generate new unique reagents.
Experimental Model and Subject Details
786-O, UOK101, A498, and HEK293T cells were cultured in DMEM (10-013-CV; Corning) containing 10% FBS (89510-186; VWR). 769-P cells were cultured in Roswell Park Memorial Institute (RPMI) medium (11875-093, 31800-022; Gibco) containing 10% FBS. HK-2 cells were cultured in Keratinocyte serum-free medium (K-SFM; 10724011; Gibco) supplemented with bovine pituitary extract and human recombinant epidermal growth factor. 769-P, A-498, HEK293T, and HK-2 cells were obtained from ATCC. The 786-O and UOK-101 cells were gifts of Dr. Sandra Turcotte and Dr. W. Marston Linehan (National Cancer Institute), respectively. Cells were used for experiments within 10 to 15 passages from thawing. All of the cells were tested to ensure that they are Mycoplasma free, and they were authenticated by short tandem repeat profiling by Genetica DNA Laboratories.
All animal experiments were in compliance with National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee of the University of Miami Miller School of Medicine. Six-week-old female athymic nude mice (Charles River Laboratories) were used for xenograft studies. Mice were housed (3–5 mice per cage) on a 12-hour light/dark cycle in individually ventilated cages and with ad libitum access to food and water. The experimental holding room had temperature (21.5°C setpoint) and humidity control (40% setpoint), and it was equipped with a high-efficiency particulate air filter. For the subcutaneous tumor growth model, cells were pelleted and resuspended in a phosphate-buffered saline (PBS)/Matrigel Matrix (356234; Corning) mix at 1:1 ratio. Then, 2 × 106 cells in a 100-μL solution were injected subcutaneously into each flank. When palpable tumors had been established, digital calipers were used to measure tumor size twice weekly. Tumor volume was calculated using the equation V = ½(length × width2). For orthotopic tumor growth assays, soft collagen pellets containing 1 × 106 cells were implanted orthotopically under the capsules of the left kidney. Tumors were palpated and monitored twice weekly using IVIS (PerkinElmer) bioluminescence imaging. At the end of the experiments, blood was collected, and both kidneys were harvested.
Fresh normal–tumor tissues were obtained from nephrectomy cases at the University of Miami Hospital. Patients' whole blood was also obtained and processed in the laboratory to obtain plasma. The samples were obtained under approvals from institutional research ethics review committees and with written informed patient consent under the institutional review board protocol 20140385.
RNA Isolation and qRT-PCR.
Total RNA was extracted using TRIzol reagent (15596018; Invitrogen). The concentration and quality of RNA were determined using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific). RNA (500 ng) was used to synthesize cDNA using the Quantabio qScript cDNA SuperMix (95048-100; Quantabio) according to the manufacturer's recommendations. The cDNA was diluted by 10-fold before being used for qRT-PCR gene expression analysis. qRT-PCR was performed using Power SYBR Green PCR Master Mix (4367659; Applied Biosystems) and SYBR Green primers (listed in Supplementary Table S3). The cycle threshold (Ct) values of the gene of interest were normalized to the Ct value of the housekeeping control, β-actin. The fold change was then calculated using the 2–ΔΔCt method.
For shRNA knockdown models, HEK293T cells were transfected with a mixture of the lentiviral transfer plasmid of interest (Supplementary Table S4), psPAX2, and PVSVG using Lipofectamine 2000 Transfection Reagent (11668-019; Invitrogen). shRNA GFP was used as a control. The media containing the viral particles was collected 24 hours post-transfection and filtered through 0.45-μm polyvinylidene difluoride (PVDF) sterile syringe filter (Santa Cruz Biotechnology). Cells were infected with the lentiviral supernatant in the presence of 10 μg/mL of polybrene (H9268; Sigma-Aldrich). RARRES2 knockout cell lines were generated by cloning sgRNAs (Supplementary Table S5) into the lentiviral vector lentiCrisprv2 (Addgene), followed by lentivirus production in HEK293T. After puromycin selection, single cell clones were generated by limiting dilution in 96-well plates. Single-cell clones were expanded and lysates were collected to verify knockout by Western blotting.
Protein Extraction and Western Blotting.
The cells were washed with 1× ice-cold PBS (SH30256.01; Cytiva) once before they were lysed in radioimmunoprecipitation assay buffer (RIPA) lysis buffer (50 mmol/L Tris-HCl, pH 7.5; 150 mmol/L NaCl; 0.5% NP-40; 50 mmol/L NaF) and scraped from the plate on ice. The RIPA lysis buffer contained 1:1,000 1× protease inhibitor cocktail (Sigma-Aldrich) and 1:100 1× phosphatase inhibitor (Sigma-Aldrich). The lysates were passed through a 16-gauge needle 15 times before they were centrifuged at 4°C for 20 minutes at 20,000g. The protein lysates were quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer's protocols, and absorbance was measured using a GloMax Discover Microplate Reader (Promega). Between 50 and 100 μg of protein samples were mixed with 5× sodium dodecyl sulfate (SDS) Laemmli Loading Buffer and boiled at 95°C for 5 minutes. Proteins were separated using 4% to 20% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gels (Bio-Rad Laboratories) and transferred onto PVDF membrane (Millipore). The membrane was blocked with Tris-buffered saline with 0.1% Tween 20 detergent (TBS-T) supplemented with 5% nonfat dry milk primary antibody for 45 minutes at room temperature rocking before blotting with primary antibodies at 4°C overnight. The membrane was washed in TBS-T three times for 5 minutes each and blotted in secondary antibodies at room temperature for 2 hours. An Azure 300 Chemiluminescent Western Blot Imaging System (Azure Biosystems) was used to visualize the membrane. The antibodies used are listed in Supplementary Table S6.
Enzyme-Linked Immunosorbent Assay for Plasma Chemerin.
Plasma were diluted 1:1,000 before being incubated in a high-binding microplate (Santa Cruz Biotechnology) coated with capture antibody according to the manufacturer's instructions (DY2324; R&D Systems). The samples were washed, and biotin-conjugated detection antibody was added into the plate wells for further incubation. Next, horseradish peroxidase (HRP)–conjugated streptavidin was added, followed by substrate A/B 1:1 mixture. Finally, the catalytic reaction was terminated by adding the stop solution, and the absorbance was measured at 450 nm using the GloMax Discover Microplate Reader.
Cell Proliferation Assay.
Cells were plated at a density of 100,000 cells in triplicate wells of a six-well plate, supplemented with DMEM containing 10% FBS. Cells were trypsinized and were counted using the Countess II cell counter (Life Technologies) at days 3, 6, 9, and 12.
Cell Viability Assay.
After treatment with monoclonal antibodies, 786-O cells were incubated with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenylterazolium bromide (MTT; Sigma-Aldrich) to assess cell survival. Briefly, the cells were plated at a density of 1,000 cells/plate in 96-well plates and treated with different concentrations of monoclonal antibodies (IgG or anti-chemerin) for the indicated incubation times (2 days, 4 days, and 6 days). The MTT solution was diluted 5-fold in DMEM and incubated at 37°C for 4 hours. The media were carefully removed, and the precipitate was dissolved in dimethyl sulfoxide. The solution was then quantified at 450 nm using the GloMax Discover Microplate Reader.
EdU Incorporation Assay.
Cells were plated on coverslips at 1 × 104 cells in 12-well plates and incubated with 10 μmol/L 5-ethynyl-2′-deoxyuridine (EdU) solution overnight at 37°C and 5% CO2. After incubation, the cells were fixed with 3.7% formaldehyde/PBS for 15 minutes at room temperature and washed with 1 mL of 3% bovine serum albumin (BSA)/PBS twice. Then, 1 mL of 0.5% Triton X-100 in PBS was added for 20 minutes at room temperature to permeabilize the cells. Next, cells were incubated with 0.5 mL of Click-iT reaction cocktail (C10337; Invitrogen), prepared according to the manufacturer's instructions, at room temperature for 30 minutes and away from direct light; 4′,6-diamidino-2-phenylindole (DAPI; 268298; Millipore) was added later for DNA staining. Coverslips were removed from wells and visualized using the z-stack of an Olympus FV1000 confocal scanning microscope. The images were analyzed using Fiji software (ImageJ).
Clonogenic Survival Assay.
The cells were plated in triplicate in 6-cm plates (100 cells/well) with complete DMEM to measure survival by colony-forming units (CFU). The colonies were counted after 2 weeks, and the experiment was repeated twice.
In one tube, the stock siRNA was diluted in serum-free medium, and DharmaFECT transfection reagent (Horizon) was added to another tube. The contents of each tube was gently mixed and incubated for 5 minutes at room temperature. The contents of both tubes were mixed and incubated for 20 minutes at room temperature. The culture media were removed from the plate, and 2.5 mL of serum-free medium was added together with the transfection medium. The cells were incubated at 37°C in 5% CO2 for 5 hours before changing the media to complete media with 10% FBS. Nontargeting control siRNA (siRISC) and KLF6 targeting siRNA (Horizon) were used.
Cells were stained with fluorescein isothiocyanate–conjugated annexin V and propidium iodide (ChemCruz sc-3541; Santa Cruz Biotechnology) using the Alexa Fluor 488 Annexin V Kit (V13245; Invitrogen) according to the manufacturer's protocol. The cells were later analyzed using an LSR II cytometer (BD Biosciences) and BD FACSDiva 8.1 software (BD Biosciences). The percentage of dead cells was measured by combining annexin V–positive and annexin V/PI–double-positive cells.
BODIPY Lipid Droplet Imaging.
Cells were plated on round glass coverslips of 12-well plates and incubated with 1 mL of DMEM supplemented with 10% FBS. The medium was removed and cells were washed with room-temperature PBS, fixed with 4% formaldehyde for 30 minutes, and incubated with 300 μL of 1-μmol/L BODIPY 493/503 (D3922; Life Technologies) for 15 minutes, protected from direct light. Then, 1 μg/mL of DAPI was added for nucleus staining. The z-stack images were acquired using the Olympus FV1000 confocal scanning microscope and analyzed using ImageJ.
Lipid Reactive Oxygen Species.
The cells were plated at 50,000 cells per well in 24-well plates, and 10 μmol/L C11-BODIPY-581/591 (D3861; Invitrogen) and 1 μg/mL DAPI were added to the media and incubated for 1 hour at 37°C at 5% CO2. Cells were imaged using the Olympus FV1000 confocal scanning microscope and analyzed using ImageJ. Nine randomly selected images per condition were used for analysis.
ORO Lipid Droplet Staining.
For tissue section staining, the tissues were cryoprotected by incubating in 30% sucrose/PBS solution. After the tissues sank to the bottom of the vial, the tissues were embedded in optimal cutting temperature (OCT) compound (Thermo Fisher Scientific). Embedded tissue blocks were sectioned at 10-μm thickness and stored at −80°C until ready for staining. Before ORO staining, the frozen sections were air-dried and fixed in 10% neutral buffered formalin (HT501128; Sigma-Aldrich). After fixing in 100% propylene glycol (0575; VWR) twice, sections were stained with freshly made ORO solution (O0625; Sigma-Aldrich) followed by staining for hematoxylin. Then, the slides were washed with distilled water and mounted using glycerin jelly (Sigma-Aldrich). The sections were visualized using an Olympus VS120 Virtual Slide Microscope scanner. For cell staining, the cells that had been plated in 12-well plates in triplicate were rinsed with PBS twice and fixed with 10% formaldehyde for 1 hour. They were then rinsed with 60% isopropanol for 5 minutes, followed by ORO staining (3 mg/mL) for 4 minutes, and the ORO solution was removed. After the cells were air-dried, 250 μL of isopropanol was added to the wells, and the eluted solution was measured at 510 nm. Next, the cells were stained with sulforhodamine B as previously described and measured at 564 nm using the GloMax Discover Microplate Reader, as previously described (59).
Blue Native Polyacrylamide Gel Electrophoresis Analysis.
Blue native polyacrylamide gel electrophoresis (BN-PAGE) was performed as previously described (60). Briefly, 2 × 106 shGFP control and shRARRES2 cells were collected, resuspended in PBS, and permeabilized with 2 mg/mL digitonin. Permeabilized cells were pelleted and resuspended in 100 μL buffer containing 1.5 mol/L aminocaproic acid and 50 mmol/L Bis-Tris, pH 7.0. Mitochondrial proteins were extracted in the presence of 1% lauryl maltoside (LM), followed by a clarification spin at 22,000g for 30 minutes at 4°C. The NativePAGE Novex 3%–12% Bis-Tris gel (Life Technologies) was loaded with 60 to 80 μg of total LM cell extracts. After electrophoresis, proteins were transferred to PDVF membranes and used for immunoblotting. Western blot was performed using primary antibodies raised against the following human OXPHOS subunits: NDUFA9 (ab14713; Abcam), CORE2 (UQCRC2, ab14745; Abcam), COX1 (ab14705; Abcam), SDHA (14865-1-AP; Proteintech), and ATP5 (ATP5A, ab14748; Abcam). In parallel, 60-μg aliquots of the same samples were resolved in a denaturing 12% SDS-PAGE gel and probed with a primary antibody against the voltage-dependent anion channel (VDAC) as loading control. Peroxidase-conjugated anti-mouse and anti-rabbit IgGs were used as a secondary antibodies (Rockland Immunochemicals).
Polarographic Analysis of Cellular Respiration.
Endogenous cell respiration was measured polarographically at 37°C using a Clark-type electrode from Hansatech Instruments. Substrate-driven respiration was assayed in digitonin-permeabilized cultured cells as reported previously (61). Briefly, trypsinized cells were washed with permeabilized-cell respiration buffer (PRB) containing 0.3 mol/L mannitol, 10 mmol/L KCl, 5 mmol/L MgCl2, 0.5 mmol/L ethylenediaminetetraacetic acid (EDTA), 0.5 mmol/L ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 1 mg/mL BSA, and 10 mmol/L KH3PO4, pH 7.4. The cells were resuspended at ∼4 × 106 cells/mL in 0.5 mL of the same buffer air-equilibrated at 37°C and supplemented with 10 units of hexokinase and 2 mmol/L adenosine diphosphate (ADP). The cell suspension was immediately placed into the polarographic chamber to measure endogenous respiration. Cells were then permeabilized with digitonin (3.4 μg/106 cells), and the complex I substrates glutamate and malate (to 2 mmol/L each final concentration) were added to measure state 3 or ADP-stimulated (coupled) respiration. Oligomycin (to 1 μmol/L final concentration) was successively added to inhibit ATP synthesis, and nonphosphorylating condition (state 4 or leak respiration) was measured. Mitochondrial respiration was uncoupled by 1-μL incremental addition of 0.1 mol/L carbonyl cyanide m-chlorophenyl hydrazone (CCCP) to reach maximal oxygen consumption. Finally, 3 μL of an 80-mmol/L KCN solution was added to assess the specificity of the assay. Oxygen consumption rates were measured as nanomoles O2 consumed per minute, normalized by the number of cells, and expressed as a percentage of the shGFP control endogenous respiration.
The relative cellular ATP concentration was measured using the Cell Titer-Glo Luminescent reagent (Promega) according to the manufacturer's protocol. Briefly, an equal volume of the single-one-step reagent was added to each well of cells and incubated at room temperature with 15 minutes of rocking. The luminescence was quantified using the GloMax Discover Microplate Reader. The ATP level was normalized to cell count from an additional plate with same setup.
Extracellular Flux Analysis.
Prior to the Seahorse XF96 Cell Mito Stress Test assay, the optimal cell density was determined to be 2,000 cells per well. A day before the experiment, the cells were plated in the Seahorse XF96 Cell Culture Microplate (102601-100; Agilent) using DMEM. The sensor cartridge was hydrated in Seahorse XF Calibrant at 37°C in a non-CO2 incubator overnight. The medium (103575-100; Agilent) was supplemented with 1 mmol/L pyruvate, 2 mmol/L glutamine, and 10 mmol/L glucose before adjusting the pH to be 7.4. Then, 20 μL of 15-μmol/L stock oligomycin, 22 μL of 5-μmol/L stock 2-[2-[4-(trifluoromethoxy)phenyl]hydrazinylidene]-propanedinitrile (FCCP), and 25 μL of 5-μmol/L stock rotenone/antimycin A were added to ports A, B, and C of the microplates, respectively. After replacing the cell media in the cell culture microplate with warmed assay media using a multichannel pipette, the microplate was placed in a 37°C non-CO2 incubator for 45 minutes. The assay was run using the Agilent Seahorse XF96 Analyzer, and the assay parameters (including oxygen consumption rate) were exported from the Seahorse XF Stress Test Report Generator.
The paraffin-embedded ccRCC tumor microarray (KD601, US Biolab) contained ccRCC tissue with matched cancer adjacent kidney tissue, two cores per case. The slides were baked for 60 minutes at 60°C, deparaffinized in xylene, and rehydrated through graded concentrations of ethanol in water. The slides were then subjected to antigen retrieval. Immunohistochemistry staining was performed using the BOND-III Fully Automated IHC and ISH Stainer according to the manufacturer's instructions (Leica Biosystems). Anti-rabbit polyclonal human chemerin (H-002-52, Phoenix Pharmaceuticals) was used at 1:200 dilution (29–66), and other reagents for IHC were provided by a Bond Polymer Refine Detection kit (DS9800; Leica Biosystems). The microarray sections were left to air-dry and were then mounted with permanent mounting medium and coverslipped. The brightfield images were acquired by an automated VS200 Research Slide Scanner (Olympus) at 40× magnification. The images were subsequently viewed using OlyVIA software (Olympus). The intensity of the staining was quantified using Image J (62).
Sodium Palmitate Preparation and Supplementation.
To prepare BSA–palmitate conjugate, sodium palmitate (P0500; Thermo Fisher Scientific) was added to a NaCl solution and heated in a water bath while stirring. The hot palmitate solution was mixed with BSA solution, 5 mL at a time. The final volume was adjusted with 150 mmol/L NaCl for 1 mmol/L palmitate solution, and pH was adjusted to 7.4. Sodium palmitate solution was added to the cells at a final concentration 500 μmol/L, and they were incubated for 24 hours.
Total RNA was extracted using the RNeasy Mini Kit (74104; QIAGEN) according to the manufacturer's instructions. Before library construction, all samples were assessed for RNA purity via a 260/280 ratio using a NanoDrop spectrophotometer, for RNA integrity and potential DNA contamination using agarose gel electrophoresis, and for RNA integrity again using the Agilent Bioanalyzer 2100. mRNA was purified from total RNA using poly-T oligo-attached magnetic beads, and the mRNA was fragmented randomly by the addition of fragmentation buffer. For New England Biolabs (NEB) library preparation using the NEBNext Ultra RNA Library Prep Kit for Illumina, briefly, the first strand cDNA was synthesized using random hexamer primer and Moloney Murine Leukemia Virus Reverse Transcriptase, RNase H Minus, and second-strand cDNA synthesis was subsequently performed using DNA polymerase I and RNase H. Thedouble-stranded cDNA was then purified using AMPure XP beads (Beckman Coulter), and the remaining overhangs of the purified double-stranded cDNA were converted into blunt ends via exonuclease/polymerase activities. After adenylation of the 3′ ends of the DNA fragments, the NEBNext Adaptor with hairpin loop structure was ligated to prepare for hybridization. In order to select cDNA fragments 150 to ∼200 bp in length, the library fragments were purified with the AMPure XP system (Beckman Coulter). Finally, the final library was obtained by PCR amplification and purification of PCR products using AMPure XP beads. The libraries were pooled in equimolar amounts and sequenced in paired-end 150-bp reactions on the Illumina NovaSeq 6000.
Untargeted Lipidomics and Metabolomics.
Cell samples were extracted using the modified Folch biphasic extraction procedure (63). Briefly, for metabolomics, 20 μL of internal standard mixture was added to cell pellets of 10 million cells. Protein quantification was done for all the samples for prenormalization. Ice-cold methanol (80%) was used for global metabolome extraction. The supernatant was collected after centrifugation, transferred to a new tube, and dried under nitrogen. The dried sample was reconstituted in 0.1% formic acid in water (70 μL). For lipidomics, 20 μL of 10× diluted internal standard mixture (stock solution of 50 ppm, w:v) was added. Samples were extracted using ice-cold 4:2:1 chloroform:methanol:water (v:v:v), and the organic phase was collected, dried down under nitrogen flow, and reconstituted in 75 μL of isopropanol plus 1 μL of injection standard mixture (100 ppm, w:v). They were used for global lipidome extraction. Metabolomics and lipidomics samples were run separately, and for each sequence solvent blanks, extraction blanks (without internal standard), neat quality controls (blanks with internal standards), and pooled sample quality controls were also prepared for evaluation of extraction and data collection efficiency. High-performance liquid chromatography coupled with high-resolution tandem mass spectrometry (LC-HRMS/MS) was used for data collection. Chromatographic separation was achieved using reverse-phase chromatography (Thermo Scientific Dionex UltiMate 3000 RS UHLPC system; Thermo Fisher Scientific) with an Ace C18-PFP column (100 × 2.1 mm, 2 μmol/L) for metabolomics and with an Acquity UPLC BEH C18 column (Waters) maintained at 30°C (2.1 × 100 mm, 1.7 μmol/L particle size) for lipidomics. In the case of metabolomics, the gradient consisted of solvent A (0.1% FA in H2O) and solvent B (acetonitrile), both with 10 mmol/L ammonium formate and 0.1% formic acid. The flow rate was 350 μL/min. The column temperature was maintained at 25°C. In the case of lipidomics, the gradient consisted of solvent A (60:40 acetonitrile:water) and solvent B (90:8:2 isopropanol:acetonitrile:water), both with 10 mmol/L ammonium formate and 0.1% formic acid. The flow rate was 500 μL/min. The column temperature was maintained at 50°C. Samples were analyzed in positive and negative electrospray ionization on a Thermo Scientific Q Exactive Orbitrap Mass Spectrometer (Thermo Fisher Scientific). Data-dependent (ddMS2-top5) MS/MS data were obtained on pooled samples per group for identification purposes. In addition, full-scan data were acquired for all the samples without MS/MS for comparing metabolite or lipid intensities across groups.
Quantification and Statistical Analysis
Unless indicated in the figure legend, all experiments were performed at least in triplicate and results are presented as mean ± standard error of the mean (SEM) of absolute values or percentages of control. All of the statistical analyses performed are described in each figure legend. Statistical P values were obtained by application of the appropriate statistical tests using the Prism 8 program. P values lower than 0.05 were considered significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant).
RNA Sequencing Data Analysis.
The raw FASTQ data were processed through an in-house bioinformatics pipeline including quality and adapter trimming with TrimGalore 0.6.1 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), alignment to the human genome hg38/GRCh38, and gene quantification with the STAR 2.5.2 algorithm (64). The differential expression analysis was performed using the quasi-likelihood function implemented in edgeR (65). GSEA was performed using the fully unfiltered gene list in the GSEA desktop application (http://software.broadinstitute.org/gsea/downloads.jsp; refs. 66, 67).
scRNA-seq Data Analysis.
scRNA-seq data were assessed from Young and colleagues (19) and analyzed as previously described. Briefly, in order to remove low-quality cells, only single cell libraries with at least 1,000 features with greater than 1 count per million were retained, and cells with high fractions of counts from mitochondrial genes (indicating dying cells) were removed. In addition, genes with expression less than 1 count per million in at least two cells per tumor were excluded from analysis. After concatenating cell-by-gene count matrices of all samples into a single matrix and normalized, the relationships among cells were visualized in two dimensions through Uniform Manifold Approximation and Projection (UMAP) plots. Each cluster was assigned to respective cell identity by marker differential gene expressions described in previous literatures. The data supplement from Young and colleagues (19) defined tissues of origin and was used to stratify during several different analyses. Our analysis followed the primary workflow of the Seurat pipeline (68, 69). According to the authors, the data were normalized by dividing the total number of unique molecular identifiers (UMI) in each cell to obtain the sequencing depth before transforming to a log scale for each cell using the Seurat (http://satijalab.org/seurat/) NormalizeData function (19):
where g is gene, c is cell, F is the Seurat scale.factor (default value, 10,000), and x is the UMI count matrix.
Next, COMBAT (19) was used in a batch correction step to regress out the variability introduced by individual 10× channels. All of the entries that were 0 before batch correction were maintained to be at 0 after batch correction to prevent imputing expression from genes with no expression. The data were renormalized so that they were consistent with being derived from an expression vector with the sum to 1 after such correction:
The threshold for mitochondrial genes was set at 5%.
Lipidomics and Metabolomics Data Analysis.
For lipidomics data analysis, LipidMatch Flow was used for file conversion, peak picking by implementing MZMine 2 (70), blank filtration (71, 72), lipid annotation (73), and combining positive and negative datasets. LipidMatch Flow software and tutorials (including video tutorials) can be found at http://secim.ufl.edu/secim-tools/. LipidMatch was used to annotate ions using data-dependent MS/MS analysis. For metabolomics data analysis, metabolite identification was performed with MZmine 2.0 and matching metabolite retention time and m/z value to an internal library of more than 1,000 metabolites. Metabolic pathway analysis was conducted using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database by matching metabolite sets with human metabolome (https://www.genome.jp/kegg/pathway.html). Metabolite set enrichment (fold enrichment) was investigated using MetaboAnalyst (open-source R package).
Data and Code Availability
This study did not generate new unique reagents. The RNA-sequencing data were deposited in GEO database GSE159716. The lipidomics and metabolomics data were deposited in MetaboLights MTBLS2435.
No potential conflicts of interest were disclosed.
S.K. Tan: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. I. Mahmud: Data curation, software, formal analysis, investigation, visualization, methodology, writing–review and editing. F. Fontanesi: Formal analysis, visualization, methodology, writing–review and editing. M. Puchowicz: Resources, data curation, investigation, methodology, writing–review and editing. C.K. Neumann: Data curation, software, formal analysis, writing–review and editing. A.J. Griswold: Software, formal analysis. R. Patel: Validation, investigation. M. Dispagna: Investigation. H.H. Ahmed: Software. M.L. Gonzalgo: Resources. J.M. Brown: Software, funding acquisition. T.J. Garrett: Resources, data curation, software, supervision, funding acquisition, methodology, writing–review and editing. S.M. Welford: Conceptualization, resources, data curation, formal analysis, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.
We thank M. Linehan for the UOK101 cells. This work was supported in part by grants from the National Institutes of Health (R01CA254409, P30CA240139, R01DK120679, P50AA024333, and P01HL147823; U24DK097209 to Southeast Center for Integrated Metabolomics at the University of Florida).
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.