The Metabolic Basis of Kidney Cancer

Kidney cancer, or renal cell carcinoma (RCC), affects nearly 300,000 individuals worldwide each year and is responsible for over 100,000 deaths annually. Although patients who present with localized or locally advanced disease have a 5-year survival rate of 20% to 95% depending on the extent of disease, patients with metastatic disease have a 0% to 10% 5-year survival rate (1). RCC has historically been considered a single disease; patients with renal tumors all underwent the same surgical procedures and patients with advanced disease were treated with similar drugs, none of which were effective. Although there would occasionally be a response, the systemic therapies available did not result in increased survival for patients with advanced disease. An increased understanding of RCC has shown that it consists of a number of different types of cancer that are characterized by different histologies, clinical courses, and responses to therapy and that are caused by different genes (Fig. 1). It is known that at least 17 different genes can cause RCC and that mutation of these genes can affect the cell's ability to respond to changes in oxygen, iron, nutrients, or, most notably in the case of mutations in genes for the tricarboxylic acid (TCA) cycle enzymes fumarate hydratase and succinate dehydrogenase (SDH), energy (Fig. 2).

Much of what is known about the genetic and metabolic basis of RCC has come from the study of patients with inherited RCC susceptibility syndromes, such as von Hippel–Lindau (VHL; VHL gene), hereditary papillary RCC (HPRC; MET gene), Birt–Hogg–Dubé (BHD; FLCN gene), and hereditary leiomyomatosis and RCC (HLRCC; FH gene). Studies of familial as well as sporadic (nonfamilial) disease have provided the foundation for the development of therapeutic approaches targeting the metabolic basis of RCC.

Clear-cell renal cell carcinoma (ccRCC) occurs in both a sporadic (nonfamilial) and a familial form. Patients affected with VHL disease are at risk for the development of tumors in a number of organs, including the kidneys. Patients with VHL have a lifetime risk for ccRCC which can be recurrent, bilateral, and multifocal. Clinical management of VHL-associated RCC involves active surveillance of small renal tumors until the largest tumor reaches 3 cm in size, at which time nephron-sparing surgery is recommended. Complete nephrectomy may be required for larger tumors or tumors invading the renal vasculature (2).

Genetic linkage analysis was performed in VHL families to localize the VHL gene to the short arm of chromosome 3 (3). Germline alteration of the VHL gene, including point mutation, splice-site mutation, and partial/complete gene deletions, has been identified in nearly 100% of VHL families (4). The VHL gene is a two-hit tumor suppressor gene in which either a germline (in patients with VHL) or sporadic alteration of VHL is associated with loss of chromosome 3p containing the second (wild-type) allele, resulting in complete inactivation of the gene. Multiple studies of the sporadic ccRCC tumors have shown nearly universal loss of chromosome 3p (91%–98%) in combination with either mutation (52%–87%) or promoter hypermethylation (7%–11%) of the remaining VHL allele (5–7).

The product of the VHL gene, pVHL, is a component of an E3 ubiquitin ligase complex with elongin C (encoded by TCEB1/ELOC), elongin B (encoded by TCEB2/ELOB), Cullin-2 (encoded by CUL2), and RBX1 (encoded by RBX1; Fig. 3; refs. 8–11). The VHL complex targets the hypoxia-inducible factors HIF1α and HIF2α, transcription factors for ubiquitin-mediated degradation in an oxygen-dependent fashion (12–14). In normoxia, the oxygen, iron, and α-ketoglutarate–dependent prolyl hydroxylase (PHD) enzymes transfer hydroxyl groups onto two proline residues in the oxygen-dependent domain of both the HIF1α and HIF2α proteins, enabling recognition by the VHL complex, ubiquitination, and subsequent proteasomal degradation. In oxygen or iron deprivation, loss of PHD enzymic activity results in the stabilization of the HIFα proteins, and the formation of heteromeric dimers with the aryl hydrocarbon receptor nuclear translocator protein (ARNT) that migrate to the nucleus and activate the transcription of genes carrying hypoxia response elements (HRE; refs. 12–15). Stabilized HIF1α and HIF2α regulate the activity of downstream genes and pathways that aid the cell in adapting to low oxygen/iron levels, resulting in alterations such as increased glycolysis via upregulation of the GLUT1 glucose transporter (encoded by SLC2A1), hexokinase 2 (encoded by HK2), and lactate dehydrogenase A (encoded by LDHA); suppression of oxygen usage by reducing pyruvate entry into the TCA cycle via upregulation of pyruvate dehydrogenase kinase 1 (encoded by PDK1); and increased red blood cell production by increased erythropoietin (encoded by EPO) production (Fig. 3). Activation of the hypoxia response is also advantageous in the growth and maintenance of a cancer, including increased angiogenesis via upregulation of vascular endothelial growth factor (encoded by VEGF), and increased cellular growth and transformation via upregulation of platelet-derived growth factor (encoded by PDGF) and transforming growth factor α (encoded by TGFA; Fig. 3; refs. 16, 17).

Activation of HIFα-dependent pathways is a canonical cellular response to low oxygen/iron that is moderated once normal oxygen/iron levels are achieved. The loss of pVHL function in ccRCCs results in the stability of the HIFα-dependent pathways independent of oxygen/iron levels in a state of “pseudohypoxia.” A subset of sporadic ccRCCs that lack VHL alterations have mutations in other components of the VHL E3 ubiquitin ligase complex, such as elongin C (TCEB1/ELOC), that function like VHL-loss mimics and result in the stabilizationof the HIFα proteins (7, 18). Although some genes are transcriptionally regulated by both HIF1α and HIF2α, each transcription factor has specific targets. In vitro and in vivo studies have indicated that HIF2α may be the more critical factor for tumorigenesis in ccRCC (19, 20). ccRCC tumors frequently lose a segment of chromosome 14q that encodes HIF1α (6, 21). These “HIF2α-only” ccRCCs have increased activity of c-MYC, a known HIF2α-specific transcriptional target (22).

ccRCC tumors lose VHL-dependent oxygen sensing, resulting in stabilized HIFα leading to metabolic reprogramming. Expression and metabolic analyses of ccRCCs have shown upregulated expression of GLUT1, hexokinase, and lactate dehydrogenase A and increased levels of glycolytic metabolites consistent with increased glucose uptake, a dependency on glycolysis, and a shift to aerobic glycolysis (23–25). At the same time, expression of the TCA cycle components is decreased, and entry of pyruvate into the TCA cycle is suppressed, consistent with impaired oxidative phosphorylation. Analysis of TCA cycle metabolites shows decreased levels of fumarate and malate and increased levels of succinate, isocitrate, and citrate (24, 25). These findings are consistent with the reductive carboxylation observed in VHL-deficient ccRCC cell lines and xenografts that converts α-ketoglutarate to citrate by partially reversing the TCA cycle (26–28). Reductive carboxylation can provide the citrate necessary to fuel the increased levels of fatty-acid synthesis present in ccRCC tumors with the required α-ketoglutarate predominantly derived from absorbed glutamine being converted to glutamate by the glutaminase enzymes (26–28). In addition, ccRCC tumors are characterized by increased expression of the oxidative pentose phosphate pathway that produces both the ribose sugars necessary for replication and the NADPH necessary to fuel the reductive carboxylation of isocitrate and maintenance of the reduced glutathione pool (6, 24, 25). Transcriptome-based analysis of ccRCC demonstrates that metabolic reprogramming, including decreased TCA cycle and AMPK complex gene expression and increased pentose phosphate pathway and fatty-acid synthesis gene expression, correlates with patient outcome, i.e., high-grade, high-stage, and low-survival disease (6).

The metabolic shift in ccRCC can be evaluated in vivo using PET imaging to assess the uptake of ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG). ¹⁸F-FDG–PET can assess metastatic disease and can quantitate the effect of therapies targeting glucose metabolism. A recent imaging study based on infusing ¹³C-labeled metabolic substrates, glucose, glutamine, and acetate, into patients with cancer demonstrated the flux of these substrates through the metabolic pathways in tumor cells in vivo (29). The patients with ccRCC infused with [U-¹³C]glucose showed that ¹³C-labeling of glycolytic intermediates in tumors was enhanced, whereas the TCA cycle components had significantly reduced ¹³C-labeling, consistent with aerobic glycolysis and the Warburg effect.

Although upregulation of the VHL/HIF pathway is central to the metabolic reprogramming of ccRCC tumors, other metabolic alterations may also play an important role. ccRCC tumors frequently have mutations in the chromatin-remodeling, chromosome 3p-encoded genes PBRM1, SETD2, and BAP1, as well as in other chromatin remodeling complexes, such as SWI/SNF, that alter gene expression and can affect many aspects of cellular metabolism (6, 7). Mutation of the PI3K/AKT/mTOR pathway, including PTEN, MTOR, PIK3CA, and gain of chromosome 5q are also frequent in ccRCC, with focal amplifications refining the region of interest to 5q35 that includes STSQM1, which encodes p62 that is involved in autophagy and the NRF2 antioxidant response pathway (6, 7, 30).

The metabolic reprogramming in ccRCC is further driven by the intratumoral genetic heterogeneity that characterizes ccRCC tumors (31). Although VHL loss is clonal in ccRCC, additional chromosomal changes and mutations in chromatin remodeling and PI3K pathway genes are often subclonal, potentially resulting in different levels of metabolic reprogramming in different regions of the primary tumor (31, 32). Recent multiregional analysis of primary ccRCC tumors using Dixon-based MRI to evaluate lipid accumulation demonstrated the heterogeneity of fat fraction within some ccRCC tumors (33). Subsequent analysis of these regions after surgical resection by mass spectrometry–based lipidomics and metabolomics confirmed this heterogeneity in metabolic profiles from different regions of the ccRCC tumor (33). A comprehensive evaluation of intratumoral heterogeneity in primary and metastatic tumors from 100 patients with metastatic ccRCC demonstrated that the metastatic sites are characterized by considerably less heterogeneity than primary tumors (34). If metastatic tumors are also found to be characterized by homogeneous metabolic profiles, this could provide unique insight into the most critical metabolic pathways to target in this disease.

Understanding the metabolic basis of the VHL/HIF pathway provided the foundation for the development of therapeutic approaches targeting this pathway. Since 2005, the FDA has approved nine agents targeting the VHL/HIF pathway in patients with advanced RCC. These therapies include a neutralizing antibody against VEGF, bevacizumab, that directly targets VEGF and six small molecule–based tyrosine kinase inhibitor therapies that target the VEGF receptor: pazopanib, axitinib, lenvatinib, cabozantinib, or a combination of the VEGF and PDGF receptors: sunitinib and sorafenib (Fig. 3; reviewed in ref. 35). Two additional agents, temsirolimus and everolimus, which inhibit the PI3K/AKT/mTOR pathway, lead to reduction of HIFα levels by inhibition of de novo protein translation (Fig. 3). Use of these agents to target metabolic reprogramming has been associated with increased disease-free progression and survival and prolonged disease stability.

A newly developed therapeutic approach using small-molecule agents, such as PT2977 and PT2399, that directly and specifically inhibit the interaction between HIF2α and its essential binding partner, ARNT, are currently being evaluated in preclinical models as well as in clinical trials in patients with both localized and advanced ccRCC (Fig. 3). In preclinical studies, PT2399 significantly reduced cellular proliferation and survival in ccRCC cell lines as well as in ccRCC patient-derived xenografts (36, 37). Phase II trials are currently under way to evaluate the effectiveness of agents directly targeting the HIF2α pathway in patients (38).

HPRC is an autosomal dominant hereditary renal cancer syndrome in which individuals are at risk for developing bilateral, multifocal type 1 papillary RCC (PRCC; ref. 39). The histology of both sporadic and hereditary type 1 PRCC tumors is characterized by papillary or tubular papillary architecture with slender papillae and delicate fibrovascular cores lined with small cells containing basophilic low-grade nuclei and scant amphophilic cytoplasm (40). The median age at diagnosis for patients affected with HPRC is 57 years and, although HPRC-associated type 1 PRCC tumors are malignant and can metastasize, their growth tends to be indolent. Management of patients with HPRC with type 1 PRCC involves active surveillance until the largest tumor reaches the 3 cm threshold, at which time surgical intervention is most often recommended (41).

Genetic linkage analysis in HPRC families localized the disease gene to chromosome 7q31, and germline mutations in the MET proto-oncogene were identified in affected family members (42). Mutations identified in HPRC families have all been missense, located in the tyrosine kinase domain of MET, and predicted to cause constitutive activation of the MET kinase in the absence of MET receptor ligand, hepatocyte growth factor (HGF; ref. 43). Type 1 PRCC tumors are typically characterized by the trisomy of chromosome 7. Duplication of the chromosome 7 bearing the mutant MET allele has been demonstrated in papillary type 1 tumors associated with HPRC, which is predicted to give the cancer cells a growth advantage (43).

Although MET mutations are found in only 13% to 18% of sporadic papillary type 1 RCC (43, 44), altered MET status (mutation, splice variant, or gene fusion) or increased chromosome 7 copy number was found in 81% of sporadic type 1 PRCC tumors evaluated in The Cancer Genome Atlas (TCGA) study (44), underscoring a central role for MET activation in this kidney tumor subtype.

Signaling by growth factors, cytokines, and nutrients through the HGF–MET axis is important for cell proliferation, motility, scattering, differentiation, and morphogenesis during normal embryogenesis and development. HGF binding to MET results in the receptor dimerization and phosphorylation of critical intracellular tyrosines, which recruit adaptor proteins that serve as docking platforms for multiple signal transducers to activate downstream signaling cascades including PI3K–AKT, RAS–RAF–MEK1/2–ERK1/2, JNKs, STAT3, and NF-κB (Figs. 1 and 4; ref. 45). Deregulated MET signaling in HPRC kidney cells due to activating germline MET mutations leads to inappropriate upregulation of cancer-promoting signaling pathways that drive proliferation, motility, and survival of tumor cells.

Agents that target the tyrosine kinase (TK) domain of MET would be predicted to be potentially efficacious in the treatment of HPRC (Fig. 4A). A phase II clinical trial with foretinib, a dual kinase inhibitor that targets the MET and VEGF receptors, was conducted in patients with type 1 papillary renal carcinoma. In patients with germline MET mutation, 50% of patients with HPRC had partial response, and the remaining 50% had stable disease. One of 5 patients with somatic MET mutations had a partial response, and no response was seen in patients with MET amplification (46). In a subsequent phase II clinical trial, treatment with crizotinib, a small-molecule MET TK inhibitor, produced a partial response in 50% (2/4) of patients with MET mutations (47). Eighteen percent of patients with MET-driven PRCC had a partial response in a trial involving another small-molecule MET TK inhibitor, savolitinib (48).

BHD syndrome is an autosomal dominant inherited cancer syndrome in which affected individuals are at risk for the development of benign cutaneous tumors (fibrofolliculomas), pulmonary cysts (often associated with pneumothorax), and kidney tumors. Cutaneous fibrofolliculomas are found in >85% of individuals affected with BHD over the age of 25, and BHD-associated pulmonary cysts occur in 70% to 84% of affected individuals (49, 50). Family members affected with BHD also have a 7-fold increased risk for developing renal tumors (49). BHD-associated renal tumors with variable histologies, including hybrid oncocytic tumors (50%) with features of both chromophobe RCC and oncocytoma, chromophobe RCC (34%), and ccRCC (9%), are found in up to one third of affected individuals (50, 51). Like VHL and HPRC, BHD-associated renal tumors are managed by active surveillance until the largest tumor reaches 3 cm, at which point nephron-sparing surgery is recommended (52). As patients with BHD are at risk for bilateral, multifocal renal tumors, preserving maximum kidney function is a priority.

Genetic linkage analysis in BHD families localized the BHD disease gene to chromosome 17p11, and germline mutations in a novel gene, folliculin (FLCN), were found in affected individuals (53). Insertion/deletion, nonsense, splice-site, and missense mutations, as well as partial deletions, have been identified across the FLCN coding region. FLCN is a tumor suppressor gene consistent with the Knudson two-hit model of tumorigenesis (54). Naturally occurring rat and canine models of BHD with germline Flcn mutations develop renal tumors with mutation or loss of the remaining wild-type Flcn allele. The FLCN-null renal tumor cell line UOK257, established from tumor material from a patient with BHD, is tumorigenic in immunocompromised mice and loses its oncogenicity upon the restoration of FLCN expression (50).

Early studies to investigate FLCN function identified two novel binding partners, folliculin interacting protein 1 (FNIP1; ref. 55) and folliculin interacting protein 2 (FNIP2; refs. 56, 57), which interact with the carboxy-terminus of FLCN and with AMPK, a critical energy sensor and negative regulator of mTORC1 (Fig. 4A; ref. 58). Biochemical analysis of the polycystic kidneys and cystic tumors that developed in mice with kidney-targeted Flcn inactivation demonstrated activation of mTORC1 (59–61), and renal tumors that developed in Flcn heterozygous mice subsequent to loss of the wild-type Flcn allele displayed enhanced activity of both mTORC1 and mTORC2, and AKT (62). Taken together, these data establish a role for FLCN as a negative regulator of the AKT–mTOR pathway (Fig. 4A). Evidence supporting the positive regulation of mTOR by FLCN in two other Flcn heterozygous mouse models has suggested that modulation of mTOR activity by FLCN may depend upon cell type or nutritional or energy status (63, 64). The mTOR inhibitor sirolimus (rapamycin) was partially effective in reducing the number and size of cysts and tumors that developed in the kidney-targeted Flcn-deficient mice and allograft tumors (59–61, 65).

mTORC1 is a master regulator of cell growth that responds to environmental cues including amino acids. mTORC1 activation by amino acids requires a lysosome-associated complex that consists of the vacuolar adenosine triphosphatase (v-ATPase), the Ragulator complex, and the Rag GTPases, which exist as obligate heterodimers RagA/B and RagC/D (Fig. 4B). Under conditions of amino acid sufficiency, Ragulator activates RagA/B through its guanine nucleotide exchange factor (GEF) activity resulting in GTP-loaded RagA/B and recruitment of mTORC1 to the lysosome, where it becomes activated by RAS-homolog expressed in brain (RHEB) that is tightly regulated by growth factor signaling and cellular energy status (66). Evidence from multiple laboratories supports a role for the FLCN–FNIP complex in coordinating cellular response to amino acid availability through regulation of the nucleotide states of the Rag heterodimers (Fig. 4B). FLCN in association with FNIP is recruited to the lysosome surface in response to amino acid starvation where the FLCN–FNIP complex acts as a GTPase-activating protein (GAP) toward RagC/D (67, 68), thereby providing the required GDP-loaded RagC for mTOR binding to the Rag heterodimer when amino acids are abundant. FLCN–FNIP localization to the lysosome requires RagA/B to be in the inactive GDP-bound state, which in turn is dependent upon the GAP activity of the GATOR1 (GAP activity toward Rag1) complex toward RagA/B (Fig. 4B; ref. 69). Recently, amino acid sensors for the mTORC1 pathway upstream of GATOR1 have been uncovered (70), but the mechanistic details of how sensing low amino acids facilitates the interaction between GDP-loaded RagA/B and the FLCN–FNIP proteins remain to be determined.

Early studies have suggested that FLCN inactivation results in metabolic reprogramming in which FLCN-deficient cells undergo a metabolic shift to favor aerobic glycolysis. Elevated HIF transcriptional activity and upregulation of HIF target gene expression were observed in FLCN-null UOK257 renal tumor cells, ACHN renal tumor cells with FLCN knockdown, and BHD-associated chromophobe RCC (71). In subsequent experiments by this group, Flcn-deficient mouse embryonic fibroblasts (Flcn^−/− MEF) displayed a 2-fold increase in HIF transcriptional activity and expression of HIF targets that correlated with increased glucose uptake, lactate production, and extracellular acidification, confirming a “Warburg effect” metabolic transformation in response to Flcn deficiency (72). HIF-dependent elevation of ATP levels in the Flcn^−/− MEFs correlated with enhanced mitochondrial respiration due to increased mitochondrial mass, which led to a significant rise in intracellular reactive oxygen species (ROS). Significantly, this study showed that intracellular ROS was driving the HIF transcriptional activation responsible for the metabolic reprogramming under Flcn deficiency (72). Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) is a well-known controller of mitochondrial biogenesis through transcriptional upregulation of genes responsible for mitochondrial biosynthesis, and active (phosphorylated) AMPK is known to directly phosphorylate and upregulate the expression of PGC1α (73). Previously, it was demonstrated that the FLCN–FNIP complex interacts directly with AMPK (55, 56), and loss of FLCN was shown to result in the upregulation of PGC1α expression and its transcriptional targets in BHD renal tumors and Flcn-deficient mouse kidney, muscle, and heart (50, 74, 75). In agreement with these previous reports, PGC1α mRNA and protein were increased 3-fold in the Flcn^−/− MEFs with a corresponding increase in PGC1α target genes and coactivators. In addition, AMPK was constitutively activated in Flcn-deficient cells, and upregulation of PGC1α was confirmed to be AMPK-dependent. Importantly, ROS production was shown to be PGC1α-dependent (72). These observations were recapitulated in the human FLCN-deficient UOK257 and FTC-133 cancer cell lines and were reversed by FLCN reexpression. It is notable that advanced VHL-deficient ccRCC and fumarate hydratase–deficient HLRCC-associated type 2 PRCC are both characterized by impaired mitochondrial metabolism (oxidative phosphorylation), although FLCN-deficient BHD-associated RCC is associated with increased oxidative metabolism and activated mitochondrial biogenesis. It is possible that this reflects a difference in the ability of distinct cells of origin to tolerate loss of mitochondrial activity, and/or it could explain the more mitigated aggressiveness of FLCN-deficient tumors in BHD.

Finally, increased mitochondrial mass, nuclear HIF1α staining, and expression of HIF target genes were seen in a BHD-associated chromophobe RCC (72). Taken together, these data support the concept that loss of FLCN constitutively activates AMPK, resulting in PGC1α-driven mitochondrial biogenesis and increased ROS production, leading to HIF transcriptional activity that drives Warburg metabolic reprogramming favoring aerobic glycolysis. Based on this paradigm, therapeutic agents that target the glycolytic pathway may show promise for the treatment of BHD-associated kidney cancer.

Two additional RCC susceptibility syndromes, Cowden syndrome and tuberous sclerosis complex (TSC), are associated with mutations in the PI3K/AKT/mTOR pathway (Fig. 4A).

TSC is an autosomal dominant disorder in which affected individuals are at risk for the development of hamartomas in the brain, cutaneous angiofibromas, cardiac rhabdomyomas, pulmonary lymphangioleiomyomatosis (LAM), and kidney neoplasia. TSC is caused by germline loss-of-function mutations in the TSC1 gene on chromosome 9q34, which encodes hamartin, or the TSC2 gene on chromosome 16p13, which encodes tuberin (76, 77). Although one third of TSC cases are thought to be familial, up to two thirds of patients have no family history of TSC, and these mutations are thought to be de novo mutations. Angiomyolipomas, consisting of smooth muscle, fat, and vascular components, are the most frequently observed kidney neoplasm in TSC and demonstrate loss of heterozygosity of the TSC1 or TSC2 allele. However, RCCs that harbor biallelic inactivation of TSC2 are also seen in patients with TSC, and can present with varying histologies, including clear-cell, papillary, and chromophobe RCC (77, 78).

Cowden syndrome is an autosomal dominant disorder characterized by an increased risk for manifestations in several organs, including tumors of the breast, thyroid, endometrium, and kidney (79). Cowden syndrome is associated with germline mutations of a number of genes, including mutation of the PTEN gene on chromosome 10q23 (79).

The PTEN gene protein product is a phosphatase that catalyzes the conversion of PIP3 to PIP2. In response to the stimulation of growth factor receptors, intracellular levels of PIP3 are increased and activate several downstream pathways, including the PI3K/AKT/mTOR pathway (80). To attenuate and control these pathways, PTEN converts PIP3 back to PIP2. In PTEN-deficient tumors, the increased levels of PIP3 remain constant, leading to continuous activation of AKT that phosphorylates and inhibits the TSC complex, resulting in the upregulation of mTOR (Fig. 4A; ref. 80). The protein products of TSC1 and TSC2, TSC1 and TSC2, are negative regulators of mTORC1 (Fig. 4A). Together, they form a heterodimeric complex with TBC1 domain family member 7 (TBC1D7) that acts on the small GTPase RHEB and stimulates its conversion from GTP-active state to GDP-inactive state, thereby inhibiting the phosphorylation and activation of mTORC1 by RHEB. The TSC1–TSC2 complex responds to nutrients, growth factors, and amino acids, triggering GAP activity toward RHEB. Additionally, in response to cellular energy deficit, TSC2 is phosphorylated and activated by AMPK, leading to suppression of mTORC1 activity. Loss of TSC1 or TSC2 function by mutation, or direct phosphorylation and inactivation of TSC2 by protein kinase B (AKT), leads to RHEB activation and mTOR-dependent phosphorylation of two downstream effectors, p70S6 kinase (which in turn phosphorylates ribosomal protein S6), and 4E-binding protein 1 (4EBP1), resulting in increased protein synthesis, cell growth, and proliferation (Fig. 4A; refs. 76, 80).

A preclinical model using epithelial-specific PTEN-deficient mice demonstrated that use of rapamycin to target the mTOR pathway promoted the rapid regression of advanced mucocutaneous lesions (81). Similarly, mTOR activation in TSC-associated LAM and angiomyolipomas has been successfully targeted therapeutically by treatment with rapalogs (i.e., sirolimus and everolimus; ref. 77). Although the effect on the central nervous system and pulmonary lesions is potentially quite remarkable and possibly paradigm-changing, rapalog treatment of renal lesions is likely cytostatic, because regrowth was observed upon discontinuation of sirolimus treatment (82).

MiT-RCCs are a subset of sporadic RCC driven by chromosomal rearrangements creating gene fusions of microphthalmia-associated transcription factor (MiT) family members TFE3, TFEB, and MITF (Fig. 5). MiT-RCCs make up approximately 1% to 5% of sporadic RCC tumors and are more common in children and young adults, representing 42% of all pediatric RCC cases (83). Recent reports suggest that adult-onset translocation RCCs may be underappreciated partly because they overlap morphologically with the more common papillary and clear-cell types of RCC (84). Among the cohort of adult RCCs described by the TCGA consortium, MiT-RCCs made up approximately 1% of ccRCCs and 12% of papillary type 2 RCCs (44, 85).

TFE3–RCCs are histologically heterogeneous tumors that frequently display a papillary architecture formed by clear and eosinophilic cells with occasional psammomatous calcifications (86, 87). They are relatively aggressive and have a propensity toward early metastasis to regional lymph nodes (84). The less common and generally less aggressive TFEB-fusion RCCs typically present with a biphasic microscopic architecture (84, 88, 89). Diagnosis of MiT-RCC is performed by IHC evaluating nuclear TFE3/TFEB immunoreactivity and “break-apart” FISH to detect chromosomal rearrangements involving TFE3 and TFEB (84). The treatment of localized disease is surgical and includes lymph-node resection, but the possibility of late-onset metastases makes a long clinical follow-up necessary (2). Multikinase inhibitors, such as VEGF pathway antagonists (e.g., sunitinib), as well as immunotherapy with cytokines, such as IL2 and IFNα, have shown limited response in patients with metastatic MiT-RCC. There is currently no known effective form of therapy for patients with advanced forms of these cancers.

Members of the basic helix–loop–helix and leucine zipper containing MiT transcription factor family share similar protein structures, recognize similar or identical DNA sequences upon homo- and heterodimerization among each other, and drive a transcriptional program involved in the development of melanocytes, osteoclasts, and mast cells. Under physiologic conditions, nuclear localization and transcriptional activity of MiT family proteins is tightly regulated. Chromosomal rearrangements leading to MiT-RCC cause the formation of a chimeric open reading frame with a 3′ part of TFE3/TFEB/MITF and a variable 5′ part of a fusion partner (Fig. 5). MiT fusion isoforms retain the wild-type C-terminus part of the MiT protein with functional DNA binding and dimerization domains. The fusion partners instead drive increased transcription and/or constitutive nuclear localization of the MiT fusion protein, which results in its dysregulated transcriptional activity that promotes carcinogenesis (90). Translocation of TFE3, located on chromosome Xp11, frequently results in fusion with recurrent partners [e.g., PRCC, ASPSCR1, SFPQ (also known as PSF), and NONO], but numerous less frequent fusion partners have been described (44, 90, 91). MiT-RCCs characterized by chromosomal translocations of TFEB (chromosome 6p21) mostly involve MALAT1, a nontranslated gene, and result in overexpression of the full-length TFEB protein. Only recently have TFEB translocations with other fusion partners and TFEB amplifications been described (44, 89, 91).

Carriers of a germline pathogenic variant of MITF (p.E318K) were shown to have a >5-fold increased risk to develop melanoma and RCC, as compared with the general population. The variant MITF protein is affected by impaired sumoylation, differentially regulates DNA binding, and drives enhanced transcriptional activity of genes involved in cell growth, proliferation, and inflammation (Fig. 5; refs. 92, 93). Somatic MITF gene fusions have been reported in RCC, suggesting a role for MITF fusions, similar to TFE3 and TFEB gene fusions, in renal tumorigenesis (91).

The transcriptional activity of MiT family members has been shown to affect a number of signaling pathways associated with increased cell proliferation and differentiation. TFE3 and TFEB bind to a 10-bp DNA motif (GTCACGTGAC) termed Coordinated Lysosomal Expression and Regulation (CLEAR) element, thereby increasing the transcription of lysosomal proteins and regulating lysosomal acidification, autophagy, and exo-, endo-, and phagocytosis in a MAPK-dependent mechanism (94). MiT family members are master regulators of lysosomal biogenesis and autophagy, involved in shaping the cell's metabolism and reaction to metabolic stress (95). After tumor initiation, autophagy may be an important mechanism for cancer cell survival in nutrient-deficient environments. A kidney-directed TFEB-overexpressing mouse model develops cystic kidney enlargement, neoplastic lesions, and liver metastasis in an autophagy-independent manner (96). Several genes involved in glucose metabolism, mitochondrial biogenesis (e.g., PPARGC1A encoding PGC1α), and lipid metabolism are under the regulation of a CLEAR element, and constitutive activation of MiT proteins may therefore increase oxidative metabolism (reviewed in ref. 97). A knock-in alveolar soft-part sarcoma mouse model expressing the ASPSCR1–TFE3 fusion gene was shown to express high levels of lactate importers, harbor abundant mitochondria, and metabolize lactate as a metabolic substrate, suggesting a role for lactate as a driver of MiT translocation tumors (98).

Under physiologic and nutrient-rich conditions, TFE3 and TFEB are recruited to lysosomes and phosphorylated by mTORC1 in a mechanism involving FLCN and FNIP (Fig. 5; ref. 67, 99). MiT phosphorylation creates a binding site for 14-3-3, which keeps the transcription factors localized to the cytosol and inactive. Amino acid or serum deprivation causes mTORC1 inactivation, release of TFE3/TFEB into the nucleus, and activation of a transcriptional program that leads to lysosomal biogenesis and autophagy, which initiates a metabolic adaptation of the cell (97). Activation of the PI3K/AKT/mTOR pathway, measured by increased phosphorylation of the S6 ribosomal protein, has been shown in TFE3-fusion cell lines and tumors, and occasional partial responses to mTORC1 inhibitors (e.g., rapamycin and temsirolimus) have been reported in patients with MiT-RCC (100–102). Recent preclinical data suggest promising antitumor effects of dual mTORC1/2 and combined AKT/mTOR inhibitors against translocation RCC cells in vitro and in vivo (2, 102). Such combination treatments may therefore represent a potential therapeutic approach to target a pathway that is constitutively upregulated in MiT-RCC.

HLRCC is an autosomal dominant familial cancer syndrome in which affected individuals are at risk for the development of cutaneous and uterine leiomyomas and an aggressive form of type 2 PRCC (103, 104). HLRCC-associated renal tumors can be early-onset (as early as 10 years of age) and have a propensity to metastasize when the primary tumor is small (as small as 0.5 cm; ref. 104). Early surgical intervention with wide surgical margins is recommended for HLRCC-associated renal tumors due to the aggressive and infiltrative nature of the tumor. Because the renal tumors have such an accelerated growth rate and can spread when the primary tumors are small, annual abdominal imaging is recommended for surveillance of at-risk individuals starting at 8 years of age (105, 106).

HLRCC is characterized by pathogenic germline variation of the fumarate hydratase (FH) gene (107). The FH gene encodes the gene for the TCA cycle enzyme fumarate hydratase, which catalyzes the interconversion of fumarate and L-malate. HLRCC-associated renal tumors are found to have somatic loss of the wild-type allele of the FH gene. These FH-deficient cells undergo a Warburg metabolic shift characterized by aerobic glycolysis and impaired oxidative phosphorylation (108–110). Stable isotope tracer studies of HLRCC tumor–derived cells, UOK262, demonstrate that very little glucose-derived carbon enters the TCA cycle, whereas glutamine-derived carbon readily enriched TCA cycle intermediates and contributed substantially to the overaccumulation of fumarate (Fig. 6A; ref. 26). The flow of glutamine-derived carbon into the TCA cycle in HLRCC cells presents a metabolic conundrum, as the lack of fumarase activity in the cells prevents the completion of multiple turns of the TCA cycle. HLRCC RCC cells have adapted to this deficiency by reducing their amount of mitochondrial oxidative phosphorylation, instead promoting the formation of citrate directly by reductive carboxylation of α-ketoglutarate (Fig. 6A; refs. 26, 110, 111). In this process, carbon dioxide is conjugated to glutamine-derived α-ketoglutarate by isocitrate dehydrogenase to form citrate, with the concomitant oxidation of NAD(P)H to NAD(P)⁺. Although it is currently unclear whether reductive carboxylation in HLRCC tumor cells occurs preferentially in the mitochondrion, the cytosol, or both, the robust production of cytosolic NADPH catalyzed by enhanced activity of the pentose phosphate pathway (PPP) observed in FH^−/− UOK262 cells suggests that the capacity for cytosolic reductive carboxylation of glutamine-derived α-ketoglutarate is substantial (110).

In addition to the direct metabolic alterations described above, abnormal fumarate accumulation in FH-deficient cells leads to additional alterations of carbon flow through central metabolic pathways. Frezza and colleagues showed that FH^−/− cells rely on the biosynthesis and degradation of heme and the secretion of the heme degradation product bilirubin for their survival, leading to synthetic lethality in cells in which both FH and the heme degradative enzyme heme oxygenase are inactivated (112). Zheng and colleagues found that FH^−/− cells exhibit reversed flux of substrates in the urea cycle enzyme argininosuccinate lyase (ASL) due to elevated fumarate levels, resulting in auxotrophy for arginine and suggesting that arginine depletion may be of therapeutic utility in HLRCC tumors (113).

As noted above, AMPK is an energy sensor that responds to cellular energy status by undergoing phosphorylation and increasing kinase activity in response to rises in the cytosolic AMP/ATP ratio. The result of increased AMPK activation is the suppression of downstream targets that include mediators of cellular proliferation, such as the mTOR pathway, and suppression of anabolic functions, including fatty-acid biosynthesis, via phosphorylation and inactivation of the acetyl-CoA carboxylases ACC1/2. Unexpectedly, AMPK phosphorylation levels were found to be decreased in FH^−/− UOK262 cells despite their strict reliance on glycolysis for production of ATP to fuel anabolic metabolism and cellular proliferation (114). The decreased AMPK activation in UOK262 cells leads to several additional metabolic alterations including decreased levels of p53 as well as the iron importer divalent metal transporter 1 (DMT1). Decreased DMT1 leads to cellular iron deficiency, which promotes the stabilization of HIF1α and the repression of HIF2α translation due to the presence of an iron-responsive element in the 5′ untranslated region of the HIF2A mRNA transcript in FH-deficient HLRCC RCC cells (114). Repression of AMPK phosphorylation along with activation of HIF1α in UOK262 cells promotes invasive and tumorigenic potential (114).

SDH, a tetrameric enzyme complex made up of the products of SDHA, SDHB, SHDC, and SDHD, is found in the inner mitochondrial membrane and is part of both the citric acid cycle and the electron transport chain. SDH catalyzes the oxidation of succinate to fumarate in a coupled reaction with the reduction of ubiquinone to ubiquinol. Patients with loss-of-function pathogenic variations of SDHB, SDHC, and SDHD are at risk for the development of tumors in several organs, including pheochromocytoma, paraganglioma, gastrointestinal stromal tumors, and RCC (115, 116). SDHB-deficient RCCs are characterized by an oncocytic histologic appearance, with eosinophilic cytoplasm consistent with mitochondrial accumulation (117, 118). Patients with germline pathogenic variations in the SDH subunit genes are at risk for the development of early-onset, bilateral, and multifocal RCC that have the propensity to spread when tumors are small. It is recommended that patients harboring germline SDH mutations be screened annually for renal lesions as well as other SDH-related neoplasms (116).

SDH-deficient renal tumor cells are characterized by a Warburg shift to aerobic glycolysis and near-complete impairment of oxidative phosphorylation (118). Isotope-resolved metabolic analysis of tumor-derived SDHB-deficient UOK269 tumor cells revealed robust lactic acid fermentation and very little entry of glucose into TCA cycle metabolites (118). However, glutamine readily labeled TCA cycle intermediates including succinate, which accumulated to extremely high levels in UOK269 cells. Labeling of UOK269 cells with 1-¹³C-glutamine demonstrated that reductive carboxylation of glutamine-derived α-ketoglutarate into citrate occurred in these SDH-deficient tumor cells (118). In a study evaluating SDH-deficient immortalized mouse renal cells, Cardaci and colleagues identified deficiency of additional respiratory chain components, further supporting the observation that oxidative TCA cycle metabolism is impaired in SDH-deficient cells (119).

Of the familial SDH mutations that have been found to cause renal tumors, mutations in SDHB are the most prevalent (116). SDHB missense mutations in patients with SDH tumor syndrome are enriched in regions of the SDHB protein that are crucial for ligation of the three Fe–S cluster cofactors that are essential for its function (118). The SDHB protein possesses two tripeptide L(I)YR motifs that are essential for the recruitment of the Fe–S cluster chaperone/cochaperone machinery responsible for insertion of Fe–S clusters into the nascent SDHB polypeptide during complex II assembly (120, 121). Missense mutations in these two L(I)YR motifs in SDHB occur frequently in patients who are at risk for the development of SDH-deficient tumors (118).

As outlined above, increased accumulations of fumarate and succinate occur in tumor cells upon loss of FH and SDH activities, respectively. These metabolite accumulations lead to profound alterations in metabolic cellular processes that extend far beyond intermediary metabolism. Succinate is the metabolic by-product of the PHD enzymes responsible for the degradation of HIF1/2α, the Jumonji domain family of histone lysine demethylases (JMJ-KDM), and the ten-eleven translocation (TET) family of hydroxylases that catalyze the hydroxylation of 5-methylcytosine residues in DNA to 5-hydroxymethylcytosine (Fig. 6B–D; ref. 122). The common features of these enzymes are that they are all iron-dependent dioxygenases that catalyze the hydroxylation of their substrates in an enzymatic reaction that consumes molecular oxygen with the concomitant oxidative decarboxylation of α-ketoglutarate, producing succinate and CO₂. Both succinate and fumarate can exhibit product-level inhibition of the enzymatic reactions of PHDs (123), KDMs (124, 125), and the TET enzymes (125). Among the results of inhibition of these dioxygenases are stabilization of HIF1α, histone, and CpG island hypermethylation, each of which has been shown in SDH- and FH-deficient tumors (Fig. 6B–D; refs. 122, 123, 126). Increased HIF levels are associated with increased VEGF and GLUT1 expression, potentially supplying increased vascularity and glucose transport to supply the needs of a rapidly growing cancer.

The oncometabolite 2-hydroxyglutarate (2HG) was recently reported to be elevated in some ccRCC tumors as compared with paired normal renal cortex samples (127). Although accumulation of the D-enantiomer of 2HG is associated with neomorphic gain-of-function mutations in isocitrate dehydrogenases IDH1 and IDH2 in IDH-mutant cancers, the more moderate accumulations of 2HG in the ccRCC tumors were not associated with IDH mutations and involved accumulation of the L-enantiomer of 2HG. The accumulation of L2HG in ccRCC was shown to be associated with decreased expression of the L-2-hydroxyglutarate dehydrogenase (L2HGDH) enzyme and decreased levels of 5-hydroxymethylcytosine in DNA, potentially linking this metabolite to epigenetic changes in ccRCC (127).

In addition to the inhibition of dioxygenases, elevated fumarate plays a role in the biology of FH-deficient tumors due to its propensity for covalent reactions with intracellular substrates, including the sulfhydryl groups present in proteins and small molecules (122). Fumarate can react with the sulfur atom of reduced glutathione to produce succinated glutathione (Fig. 6E), which inhibits glutathione function and results in increased oxidative stress in FH^−/− cells (128, 129). Fumarate also reacts with cysteine thiols in proteins to produce a covalent modification known as cysteine S-succination (Fig. 6F; ref. 130). Elevated S-succination of proteins in HLRCC tumors can be detected by specific antibodies (131), and is a useful biomarker for the IHC identification of FH-deficient tumors (132). Elevated fumarate has been found to activate the antioxidant response element pathway via the nuclear factor erythroid 2-related factor 2 (NRF2) transcription factor, an effect mediated by aberrant S-succination of cysteine residues in the Kelch-like ECH-associated protein 1 (KEAP1) and resulting in its inactivation (e.g., Fig. 6F; refs. 133, 134). Activation of NRF2 transcriptional targets in HLRCC tumor cells likely confers a cellular survival advantage in a setting of elevated oxidative stress by activating glutathione metabolism genes as well as other detoxification enzymes, including HMOX and NQO1 (133, 134). In a recent drug screen, the kinase inhibitor vandetanib emerged as a potent and specific inhibitor of FH-deficient tumor cell growth both in vitro and in vivo (135). The cytotoxic effects of vandetanib on HLRCC cells were associated with inhibition of ABL1 phosphorylation, which resulted in the repression of nuclear translocation of NRF2 and downregulation of antioxidant response genes in FH-deficient UOK262 tumor cells. It was also found that ABL1 stimulated mTOR phosphorylation and could promote aerobic glycolysis in FH^−/− cells through increased translation of HIF1α (135). Because of the synergistic effects of the repression of antioxidant defense and aerobic glycolysis, targeting ABL1 with vandetanib provides a potentially promising approach for therapy, and was found to be even more effective both in vitro and in vivo when combined with the AMPK activator metformin (135).

Several major questions about the biology and pathogenesis in the TCA cycle gene-mutant renal cell cancers remain unanswered. What confers the highly specialized tissue specificity in these disorders, and what are the specific cells of origin of FH- and SDH-deficient neoplasms? What are the initial mechanisms that lead to the aberrant proliferation of these cells upon loss of FH and SDH activity, and what, if any, additional alterations are required for progression to tumor formation, evasion of cellular senescence, and metastasis? Finally, although loss of FH and SDH activity in tumor cells has been shown to impair the flow of metabolites through the TCA cycle (26, 118), the precise mechanisms underlying the profound decreases in cellular respiration and electron transport chain function remain to be determined.

In summary, kidney cancer is fundamentally a metabolic disease. Each of the genes known to cause kidney cancer affects the cell's ability to respond to changes in oxygen, iron, nutrients, or, most notably in the TCA cycle gene mutation renal cancers, energy. The elucidation of these fundamental pathways will hopefully provide the foundation for the development of effective forms of management and therapy for patients affected with localized, locally advanced, as well as advanced RCC.

No potential conflicts of interest were disclosed.

This research was supported by the Intramural Research Program of the NIH, NCI, Center for Cancer Research. This project has been funded in part with federal funds from the Frederick National Laboratory for Cancer Research, NIH, under contract HHSN261200800001E.