Biochemical and Epigenetic Insights into L-2-Hydroxyglutarate, a Potential Therapeutic Target in Renal Cancer

Cancer-associated mutations in metabolic enzymes have led to the identification of oncometabolites including fumarate, succinate, and D-2-hydroxyglutarate (D-2-HG; refs. 1–7). A unifying theme among these oncometabolites is their ability to competitively inhibit α-ketoglutarate (α-KG)-dependent dioxygenases (αKGD; refs. 8–15). Members of the αKGDs include the TET (ten eleven translocation) enzymes (TETs 1–3) that hydroxylate 5-methylcytosine (5-mC) to 5-hydroxymethylcytosine (5-hmC), which can promote DNA demethylation (16), the Jumonji family of histone lysine demethylases, and prolyl hydroxylases (PHDs 1–3), which regulate hypoxia-inducible factors 1α and 2α (HIF-1α, HIF-2α). In leukemia, bioinformatic and functional studies indicate that a relevant target of D-2-HG is TET2 (12, 17). However, the relevant target(s) of D-2-HG, fumarate, and succinate in other malignancies as well as the contribution to tumorigenesis remains less well characterized. For example, the in vivo efficacy of mutant IDH inhibitors for glioma has yielded conflicting results (18–20).

Elevations of the l-enantiomer of 2-HG (L-2-HG) are present in clear cell renal cell carcinoma (ccRCC), the most common histology (21, 22). Recent studies have also demonstrated increases of L-2-HG in the setting of hypoxia, mitochondrial dysfunction, and acidic conditions (23–29). Similar to D-2-HG, L-2-HG can competitively inhibit dioxygenases due to structural similarity to α-KG. Correspondingly, increased DNA and/or histone methylation has been found in the setting of raised L-2-HG (28–30). However, the underlying biochemistry, relevant targets, and contribution to tumor growth of this small molecule in renal cancer remain poorly understood. Elevations of L-2-HG in RCC are due to reduced expression of L-2-HG dehydrogenase (L2HGDH). L2HGDH is considered an enzyme of metabolite repair to counter the “off-target” activity of intermediary metabolism enzymes including lactate dehydrogenase (LDH) and malate dehydrogenases (MDH), particularly under acidic conditions (23, 26, 31–33). Inheritance of biallelic mutant L2HGDH results in a rare inborn error of metabolism referred to as L-2-HG aciduria that has been linked with brain tumors (34–39). The concept that small molecules can promote malignancy has now presented novel therapeutic opportunities. In the case of IDH gain-of-function mutations in leukemia, which result in raised D-2-HG levels, specific inhibitors that selectively inhibit mutant IDH (as opposed to wild-type IDH) are now approved for use. On the basis of these data, we determined the contribution of L-2-HG to the malignant phenotype of RCC. We also outline the biochemical axis that promotes L-2-HG accumulation as well as epigenetic targets. Collectively, our data demonstrate that L-2-HG promotes tumor growth in RCC, and therefore, identifies potential strategies for RCCs with raised levels of this small molecule.

Renal cell lines were acquired from ATCC except RXF-393 (NCI) and OSRC-2 (Riken). Cells acquired from ATCC, NCI, and Riken were characterized via short tandem repeat profiling. As cells were passaged for less than 3 months after resuscitation and were periodically screened for Mycoplasma using a PCR-based assay, no further authentication was performed. HEK293T and HK-2 cells were maintained with DMEM containing 10% FBS and penicillin/streptomycin. A498 cells were cultured in MEM containing 10% FBS and penicillin/streptomycin. RXF-393, OS-RC-2, and 769-P cells were maintained in RPMI containing 10% FBS and penicillin/streptomycin. For L-2-HG isotopologue analysis, cells were incubated in media without l-glutamine for 4 hours and then incubated for 6 hours in media containing 2 mmol/L l-Glutamine-13C₅ (Sigma).

Glutaminase inhibitor (CB-839, Calithera Biosciences) was dissolved in DMSO and cells were treated with CB-839 (1 μmol/L) for 48–72 hours with replacement of CB-839 for every 24 hours. MDH inhibitor, a derivative of benzylpaullones (4k), was dissolved in DMSO and cells were treated with 1 μmol/L for 48 hours (40). L-2-HG ester was generated as reported previously (21). For effects on phenotypes in HK-2 renal epithelial cells, cells were continuously cultured in the indicated concentration of ester for 7–10 days followed by wound-healing assays.

L2HGDH and L2HGDH A241G cDNA was inserted into LV2606 (kindly provided by John Kappes, University of Alabama at Birmingham, Birmingham, AL). For L2HGDH or MDH knockdown, the following shRNAs in pLKO.1 vector (Sigma) were used: shL2HGDH (sh3, TRCN0000064323), shL2HGD (sh4, TRCN0000064324), and shL2HGDH (sh5, TRCN0000064325), shMDH1 (TRCN0000275199), shMDH1 (TRCN0000028484), and shMDH2 (TRCN0000028485). For EZH2 knockdown, shRNA lentiviral vectors for human EZH2 gene (EZH2 shRNA) and control shRNA were kindly provided by Soory Varambally (UAB, University of Alabama at Birmingham, Birmingham, AL). To generate stable cell lines, lentiviral plasmids were transfected with packaging vectors into HEK293T cells using calcium phosphate method. Supernatants from transfected HEK293T cells were collected after 72 hours, filtered, and applied to cells. Viral transduced cells were selected in culture medium containing puromycin. All transduced cells represent polyclonal populations.

A498 cells stably expressing L2HGDH were treated with 30 nmol/L of control siRNA (#D-001800-01), KDM6A siRNA#1 (#J-014140-10), and KDM6A siRNA#2 (# J-014140-12) ON-TARGET plus siRNA (Dharmacon) using Lipofectamine RNAiMAX (Invitrogen, catalog no. 13778030) transfection reagent for 72 hours. Cells were again treated with 30 nmol/L of control siRNA, KDM6A siRNA#1, and KDM6A siRNA#2 for 48 hours.

2-HG enantiomer analysis (i.e., d- and L-2-HG quantification) of samples was performed as described previously (21). For total 2-HG (D-2-HG + L-2-HG) measurement of samples from in vitro studies, cell pellets were washed in 1× cold PBS three times. Metabolites were extracted with 10% cold trichloroacetic acid (TCA) and the precipitate was removed by centrifugation. TCA in the supernatant was removed by vortexing with 4 volumes of 1,1,2-trichlorotrifluoroethane (FREON)-trioctylamine (Sigma) mixture and the upper aqueous layer was collected for analysis after centrifugation. Samples were analyzed by ion chromatography coupled with negative electrospray mass spectrometry (RFIC-MS; Dionex) and 2-HG was determined by selected ion monitoring (SIM 147.1). 2-HG in cell extracts was quantified by using a calibration curve of 2-HG and normalized to protein content.

Immunodeficient nude (Nu/Nu) mice were obtained from Jackson Laboratory or Charles River and fed with standard chow diet. Cells were mixed with equal volume of Matrigel Matrix (Corning, #354230) and injected (RXF-393 = 1 × 10⁶, A498 = 2 × 10⁶ and OSRC-2 = 1.5 × 10⁶ cells/injection) subcutaneously in the flanks of 5–6 weeks-old nude (Nu/Nu) mice. Caliper measurements of growing tumors were taken periodically and the tumor volume was calculated. For glutaminase inhibition study, Caki-1 human clear cell renal cell carcinoma cells were mixed 1:1 with Matrigel and implanted subcutaneously in the flanks into female scid/bg mice (Charles River; 2.5 × 10⁶ cells/mouse). On day 16 postimplant, mice were randomized to the following two groups (n = 10 mice per group): (i) vehicle (25% hydroxyl-propyl-β-cyclodextrin) orally twice daily by gavage; or (ii) CB-839 at 200 mg/kg orally twice daily by gavage. For all in vivo studies, tumors were measured with calipers and tumor volume was calculated using the formula = (a × b²/2) where “b” is the smallest diameter and “a” is the largest perpendicular diameter.

Total RNA from cultured cells were extracted using TRIzol reagent (Invitrogen). cDNA was synthesized using QuantiTect reverse transcription kit (Qiagen). qRT-PCR was done using TaqMan expression assay probes (Invitrogen) for human L2HGDH (Hs00227575), SPOCK2 (Hs00360339), and SHISA2 (Hs01590823) in QuantStudio 6K Flex Real-Time PCR System (Applied Biosystems). RPLP0 probe (Hs99999902) was used as internal control, and ΔΔC_t method was used to calculate relative mRNA levels.

α-L2HGDH (Proteintech, #15707-1-AP, GeneTex, #GTX32695), α-5hmc (Active Motif, #39791), α-KDM6A (GeneTex, #GTX120873), α-MDH2 (Abcam, #ab96193), α-H3K27Me3 (Millipore, #07-449), α-H3 (Cell Signaling Technology, #9715), α-EZH2 (Cell Signaling Technology, #5246), and β-actin (Abcam, #ab20272) were used as per manufacturers’ instructions.

Blotting of genomic DNA was performed as described previously (21). Membranes were then blocked with 5% milk in TBST for 30 minutes and were incubated with anti-5-hmC antibody (Active Motif) overnight at 4°C and horseradish peroxidase–conjugated anti-rabbit IgG secondary antibody for 1 hour at room temperature. After washing three times with TBST, the membrane was treated with ECL and scanned.

Confluent cell monolayers in 6-well culture dishes were wounded by making a scratch using a pipette tip. Phase-contrast/bright field pictures were taken of the wounded area at day 0 and wound healing on the time point as mentioned in the respective figures. Distances between the wounded area was measured using ImageJ software and percentage migration/wound healing was calculated with a percentage of distance coverage by migrating cells into the scratch wound area.

Cells were harvested by trypsinization, washed with PBS, and resuspended into serum-free culture medium. A total of 5–10 × 10³ cells/100–200 μL in serum-free medium was loaded into moistened top chamber and 600 μL of complete culture medium containing 10% FBS was loaded into bottom chamber of Transwell (Corning) plate. Cells were incubated for 16 hours. Nonmigrated cells (top chamber) were scrapped off using cotton swab and the migrated cells trapped into the membrane were stained with Diff-Quik Stain kit (Siemens). Bright field images of the migrated cells were capture and quantified per mm² of the membrane.

Total RNA was isolated from A498 cells stably expressing control vector and L2HGDH and purity was assessed by gel electrophoresis (Agilent 2100 Bioanalyzer). Transcriptional profiling experiments were then carried out using the Illumina HumanHT-12 V4 Array. Data were analyzed using GeneSpring GX software. Differences between treatment groups were determined by an unpaired t test and the multiple testing correction method of Benjamini Hochberg. A corrected P value (q value) was then calculated to correct for false-positive discoveries. Heatmaps were created with R using the gplot heatmap.2 package.

Clinical and RNASeq (V2) raw data for normal and tumor samples (Kidney renal clear cell carcinoma) were downloaded from TCGA website (https://tcga-data.nci.nih.gov). In-house developed parser was used to process and extract clinical information and gene expression levels of L2HGDH (http://ualcan.path.uab.edu; ref. 41). For survival analysis, patients were ordered by expression level and split into top 50% and bottom 50%. Using this information, a two condition (high expression vs. low expression) Kaplan–Meier survivability plot was generated using the survival library in R.

Statistical analyses were carried out using GraphPad Prism6 software. Comparisons between groups for statistical significance were performed with a two-tailed unpaired t test. P < 0.05 was considered statistically significant in all cases.

All animal studies were approved by the Institutional Animal Care and Use Committee.

Prior studies demonstrating an increased prevalence of brain tumors in the context of L-2-HG aciduria suggest that L2HGDH may have tumor suppressor properties. We therefore examined the effects of reduced L2HGDH expression in HK-2 immortalized renal epithelial cell that have relatively high basal L2HGDH expression compared with RCC cells. Using shRNA, we created 3 distinct knockdown pools that demonstrated reduced expression of L2HGDH at the mRNA and protein levels (Fig. 1A; Supplementary Fig. S1A). Consistent with the role of L2HGDH in L-2-HG metabolism, knockdown clones demonstrated elevated cellular levels of 2-HG as determined by gas chromatography–mass spectroscopy (Supplementary Fig. S1B). As a functional readout of raised 2-HG levels, we assessed the effects of L2HGDH knockdown on DNA 5-hyroxymethylcytosine (5-hmC) levels as prior studies demonstrate that L-2-HG can competitively inhibit α-KG dioxygenases including the TET enzymes. Consistent with these findings, L2HGDH knockdown led to reduced DNA 5-hmC levels in renal epithelial cells (Supplementary Fig. S1C). L2HGDH knockdown resulted in a modest increase in proliferation in HK-2 cells (Supplementary Fig. S1D). We investigated additional in vitro phenotypes and found that knockdown of L2HGDH promoted a migratory phenotype as assessed by a wound-healing scratch assay (Fig. 1B and C). To assess whether these effects were related to raised cellular levels of L-2-HG, we treated HK-2 cells with exogenous L-2-HG octyl ester which expectedly raised cellular levels of 2-HG (Supplementary Fig. S1E). L-2-HG ester treatment also promoted a migratory phenotype in HK-2 cells phenocopying the results of L2HGDH knockdown (Fig. 1D; Supplementary Fig. S1F).

On the basis of these data, we assessed the effects of L2HGDH reconstitution on RCC migratory phenotypes. Consistent with prior data from our group (21), we identified multiple RCC lines with reduced L2HGDH expression and raised L-2-HG levels relative to HK-2 cells (Supplementary Fig. S2). We stably transduced RCC cells with WT L2HGDH cDNA or control vector. As an additional control, we stably transduced RCC cells with a L2HGDH point mutant (A241G) that results in a substitution of lysine with glutamate (K81E) previously reported in patients with L-2-HG aciduria and found to lack catalytic activity (39, 42). Protein expression was confirmed by immunoblotting (Supplementary Fig. S3A). In multiple lines tested, WT L2HGDH was able to significantly lower L-2-HG levels, whereas only slight reductions in L-2-HG were noted with the A241G mutant (Fig. 1E; Supplementary Fig. S3B and S3C). Consistent with increased TET activity due to reduced L-2-HG, 5-hmC levels were increased in WT cells relative to control vector or A241G-transduced cells (Supplementary Fig. S3D-E). Similar to our loss-of-function studies, L2HGDH restoration in RCC cells reduced migration as determined by scratch assay (Fig. 1F and G; Supplementary Fig. S4A–S4C). As an orthogonal assay, we assessed the effects of L2HGDH on migration via a Boyden chamber assay. In multiple RCC lines tested, WT L2HGDH suppressed Boyden chamber migration relative to either control or A241-mutant cells (Fig. 1H and I; Supplementary Fig. S4D and S4E). These data demonstrate that L2HGDH's catalytic activity can suppress RCC migration.

These data prompted us to examine the effects of L2HGDH in vivo. We stably expressed L2HGDH in RXF-393 RCC cells, which we previously demonstrated to express low levels of L2HGDH and had high levels of L-2-HG. Notably, L2HGDH restoration suppressed tumor growth (Fig. 2A and B). End-of-study analysis of tumor explants confirmed continued expression of L2HGDH in RCC cells (Supplementary Fig. S5). We next assessed whether L2HGDH's catalytic activity was required for effects on growth suppression. In both OSRC-2 and A498 RCC lines, tumors with WT L2HGDH grew more slowly than tumors expressing the A241G mutant (Fig. 2C–E). End-of-study tumors for WT tumors were of less weight than A241G-mutant tumors (Fig. 2F and G). Collectively, these data demonstrate that L2HGDH suppression of in vivo tumor growth is dependent on catalytic activity.

These data prompted us to delineate the biochemical mechanisms that promote L-2-HG accumulation in the setting of reduced L2HGDH expression. Prior studies demonstrate that α-ketoglutarate (α-KG) is the source for L-2-HG (31, 43). Two potential exogenous sources for α-KG include glucose and glutamine. Glutamine has previously been shown to be the predominant source of L-2-HG in hypoxic cancer cells (SF188 gliobastoma line) as well as activated T cells (23, 28). Culture of RCC cells with elevated basal L-2-HG levels in the absence of glucose resulted in a modest reduction in total 2-HG levels (Fig. 3A). In contrast, culture of cells in the absence of glutamine resulted in a dramatic reduction of total 2-HG levels (Fig. 3A). Enantiomer resolution of 2-HG demonstrated that glutamine deprivation markedly reduced L-2-HG levels in RCC cells (Supplementary Fig. S6A and S6B). Reductions of D-2-HG were also evident in cells cultured in the absence of glutamine. We next determined whether glutamine is a direct carbon source for L-2-HG in RCC cells. Isotopologue studies with glutamine fully labeled with C13 (U-¹³C₅) was used to determine whether and how many carbons from glutamine are incorporated into L-2-HG. In both A498 and RXF-393 RCC cells, the predominant form of L-2-HG derived from U-¹³C₅ glutamine is the m+5 fraction (C13 label incorporated into all 5 molecules of L-2-HG; Fig. 3B; Supplementary Fig. S6C). These data are consistent with the working model that L-2-HG is generated by the direct reduction of α-KG to L-2-HG. The next most predominant isotopologue identified in A498 RCC lines is the m+3 fraction (C13 label incorporation into 3 carbons). These data would be consistent with α-KG going through a round of the TCA cycle, which would result in the loss of 2 labeled carbons. L2HGDH reexpression in A498 cells led to reduced accumulation of labeled L-2-HG following incubation with U-¹³C₅ as demonstrated by reduced levels of m+5 and m+3 L-2-HG isotopologues (Fig. 3C).

The contribution of glutamine to L-2-HG in RCC cells led us to consider whether inhibition of glutamine metabolism could impact L-2-HG levels in RCC cells. A key step in glutamine utilization by cells is the conversion to glutamate by glutaminase. Recent studies demonstrate that glutaminase can be pharmacologically targeted (44). We therefore examined the effects of the glutaminase inhibitor CB-839 on L-2-HG metabolism in RCC cells. CB-839 treatment of RCC cells under glutamine containing conditions led to a significant reduction of cellular glutamate consistent with glutaminase inhibition (Supplementary Fig. S6D). Glutaminase inhibition led to a significant reduction of L-2-HG levels in multiple RCC lines (Fig. 3D–F). Small molecules including both D- and L-2-HG have been shown to competitively inhibit αKG dioxygenase including the TETs (1–3), which catalyze the conversion of DNA 5-methylC to 5-hydroxymethyl cytosine (5-hmC). Given the significant reductions of L-2-HG following glutaminase inhibition, we assessed for effects on DNA 5-hmC. In both A498 and RXF-393 cells, glutaminase inhibition resulted in increased DNA 5-hmC levels as determined via dot blot assay indicating that lowering of L-2-HG levels results in activation of TET activity (Fig. 3G). Given that CB-839 effectively lowers L-2-HG levels in RCC cells, we next assessed for effects on cellular migration. In Caki-1 RCC cells, CB-839 reduced cellular migration as determined by transwell migration assay (Fig. 3H and I) as well as scratch assay (Fig. 3J and K). Similar results on migration as determined by scratch assay were found in OSCR-2 and RXF-393 cells (Supplementary Fig. S7A–S7D). Moreover, for both assays evaluated, L-2-HG ester treatment was able to rescue the effects of glutaminase inhibition indicating that the effects of CB-839 on migration were in part mediated by lowering L-2-HG levels. Furthermore, glutaminase inhibition was able to suppress in vivo tumor growth in RCC cells (Fig. 3L). Collectively, these data demonstrate that glutamine metabolism promotes L-2-HG accumulation and that inhibition of this pathway can suppress tumor phenotypes.

Prior studies indicate that L-2-HG is produced by the promiscuous activity of dehydrogenases including malate dehydrogenase (MDH1 and 2; ref. 31). We identified shRNA constructs that knockdown expression of both cytosolic MDH (MDH1) and mitochondrial MDH (MDH2; Fig. 4A; Supplementary Fig. S8A). While knockdown of MDH1 reduced L-2-HG levels in RXF-393 cells (Supplementary Fig. S8B), MDH2 knockdown had a more pronounced effect on L-2-HG levels (Fig. 4B). In line with these data, pharmacologic MDH inhibition with a recently described inhibitor referred to as 4K also reduced 2-HG levels in RCC cells (Fig. 4C; Supplementary Fig. S8C). Similarly, MDH2 knockdown significantly reduced L-2-HG levels in OSRC-2 cells (Fig. 4D). We also observed that MDH2 knockdown reduced proliferation in OSRC-2 cells (Fig. 4E). On the basis of prior studies, we assessed effects of MDH2 knockdown on migration. We assessed migration at 16 hours to exclude effects on proliferation (Fig. 4F and G). Similar to prior manipulations that lowered L-2-HG levels, MDH2 knockdown reduced cell migration. However, the reduced migration could be rescued via treatment with L-2-HG ester. Collectively, these data demonstrate that MDH2 promotes L-2-HG accumulation and that inhibition of MDH2 can suppress in vitro tumor phenotypes in RCC cells with raised L-2-HG.

Small molecules including both D- and L-2-HG as well as fumarate and succinate have been shown to inhibit enzymes involved in histone and DNA methylation and therefore could impact gene expression. For deeper insight into L-2-HG's effects, we performed comparative studies between A498/CV (high L-2-HG cells) and A498/L2HGDH (low L-2-HG cells). Gene expression array analysis comparing A498/CV and A498/L2HGDH cells reveals that genes reexpressed upon lowering L-2-HG cells are enriched for targets of polycomb repressor complex 2 (PRC2) and/or H3K27me3 target genes (Fig. 5A and B). PRC2 lays down the repressive H3K27me3 mark. Consistent with these data, WT L2HGDH lowered H3K27me3 levels in RCC cells, whereas the A241G mutant did not (Fig. 5C). We validated that WT L2HGDH could increase expression of these target genes relative to control vector or A241G transduced cells (Fig. 5D). Given that L-2-HG can competitively inhibit lysine histone demethylases (KDM), these data led us to consider KDMs that demethylate H3K27me3. Recurring mutations of KDM6A (also referred to as UTX), which encodes a H3K27 demethylase, have been identified in renal cancer (45, 46). We therefore utilized siRNA to knock down the expression of KDM6A in A498 cells transduced with WT L2HGDH (A498/L2HGDH), which has reduced migration relative to either control or A241G-mutant cells (Fig. 1G and H). Immunoblotting confirmed knockdown of KDM6A at the protein level (Fig. 5E). KDM6A knockdown in A498/L2HGDH cells enhanced migration and therefore phenocopied the enhanced migration of high L-2-HG cells (Fig. 5F and G). On the basis of these data, we assessed the effects of PRC2 inhibition in parental A498 cells, which have high basal levels of L-2-HG. The catalytic subunit of PRC2, EZH2 (enhancer of zeste homologue 2), is required for H3K27 methyltransferase activity. Knockdown of EZH2 in A498 cells, confirmed by immunoblotting, resulted in reduced H3K27me3 levels (Fig. 5H). Notably, EZH2 knockdown resulted in reduced cell migration as determined by transwell chamber assay (Fig. 5I and J).

Given these data, we next examined the translational significance of the L-2-HG/L2HGDH axis in the context of RCC. Using the TCGA dataset, we examined the outcomes of patients with RCC as a function of L2HGDH expression. Notably, patient with tumors in the lower 50% of L2HGDH expression had reduced survival (Fig. 6A). Correspondingly, lower L2HGDH mRNA expression was associated with higher tumor stage and grade (Fig. 6B and C). A primary mechanism by which L2HGDH expression is reduced is via copy loss of L2HGDH. We previously demonstrated using the TCGA dataset that copy loss was associated with reduced expression (21). The L2HGDH gene is located at 14q, a commonly deleted region in ccRCC. Chromosomal losses of 14q resulting in loss of heterozygosity (LOH) has previously been shown to be associated with worsened prognosis in patients with ccRCC (47, 48). We therefore analyzed the relationship between 14q LOH and L2HGDH expression based on a recent data set that examined both the genomic and transcriptomic landscape of 100 ccRCCs (49). 14q LOH was identified in 42 of 100 ccRCC tumors (Supplementary Table S1). Consistent with prior studies, deletion was the primary mechanism for LOH (as opposed to alternate mechanism such as copy neutral LOH). LOH at the L2HGDH locus was present in 38 of these 42 cases (Supplementary Table S1). Moreover, 14q LOH tumors demonstrate reduced L2HGDH expression relative to tumors without 14q copy number alterations (Fig. 6D). These data further support copy loss as a mechanism of reduced L2HGDH expression and that the L2HGDH locus is commonly included in 14q losses present in ccRCC. Consistent with these data, LOH at the L2HGDH locus was associated with worsened survival in the TCGA dataset (Fig. 6E). Death from ccRCC is most commonly due to progression of disease that results in metastasis. We performed unbiased metabolomic profiling in normal kidney, primary tumors, and metastatic RCC tissues (manuscript in preparation). Consistent with prior studies, we identified elevations of total 2-HG levels in primary tumors relative to normal kidney (Fig. 6F). However, metastatic tissues had even higher total 2-HG levels compared with either primary tumors or normal kidney (Fig. 6F). We confirmed that the predominant enantiomer contributing to the total 2-HG pool in metastatic tissues was the L(S) enantiomer (Fig. 6G). Collectively, our translational studies support the biological relevance of the L-2-HG/L2HGDH axis to renal carcinogenesis identified by the in vitro and in vivo data presented.

Here we provide clear evidence that the L-2-HG/L2HGDH axis has biological significance to renal carcinogenesis. In particular, our in vivo studies demonstrate that restoration of L2HGDH can suppress tumor growth. Moreover, our studies demonstrate that suppression of tumor growth is related to L2HGDH's catalytic activity thereby bolstering the notion that L-2-HG is a bona fide oncometabolite that contributes to renal tumorigenesis. These data are in line with data from patients with L-2-HG aciduria in which L-2-HG levels are markedly elevated. A notable finding from this rare disease is multiple reports of patients with brain tumors suggesting that L2HGDH has tumor suppressor functions consistent with our data in RCC (34, 37, 38, 50, 51). In addition to brain tumors, a Wilms’ tumor of the kidney has also been reported in a L-2-HG aciduria patient (52). Wilms’ tumor is a common renal tumor affecting individuals in the pediatric population. To date, clear cell renal tumors have not been reported in these patients. However, patients with this disease process have significant neurological sequelae that may shorten lifespan relative to patients typically affected by RCC who are usually diagnosed in the seventh decade and beyond.

Although the concept of oncometabolites is well established, we would like to highlight that these are among the only data to demonstrate that lowering levels of a metabolite via a genetic approach (i.e., L2HGDH restoration) can impact tumor growth in vivo. In vivo studies on the role of fumarate and succinate are lacking. Rohle and colleagues reported that IDH1 knockdown via shRNA in IDH1-mutant (heterozygous) TS603 glioma cells could suppress tumor growth (18). One caveat to this study is that the IDH knockdown approach could not exclude effects on tumor growth mediated by the wild-type IDH1 allele present in these cells. This is particularly relevant as recent studies by Clavert and colleagues demonstrate that nonmutated IDH1 is overexpressed in glioblastoma and that knockdown of wild type IDH1 can suppress glioma growth in vivo. Studies on the use of IDH inhibitors for glioma have yielded conflicting results in vivo (18, 19, 53). Moreover, recent transgenic models have demonstrated conflicting results on the role of mutant IDH in glioma progression (54, 55). In contrast, preclinical and clinical studies, particularly in leukemia, have provided clear rationale for the use of mutant IDH enzyme inhibitors to the point that they are now approved for use in the setting of IDH mutation (56–58). However, the effects are likely related to differentiation as in vitro studies in leukemia demonstrate that pharmacologic inhibition of mutant IDH may actually increase proliferation (58, 59). Our data are also the first to demonstrate that raised L-2-HG can promote a migratory phenotype. Prior studies indicate that D-2-HG can promote mesenchymal phenotypes in breast and colon cancer cells (60–62). However, L-2-HG failed to induce this behavior in colon cancer cells (61). Collectively, these data indicate that context is important with regard to the effects of either 2-HG enantiomer on tumor biology/phenotypes.

Bioinformatic analyses demonstrate that copy loss of L2HGDH is linked with 14q losses in ccRCC. Loss of 14q is associated with worsened outcomes in patients with RCC (48, 63). Despite these data, the genes on 14q whose loss contributes to this clinical phenotype remain to be characterized. Our data indicated that L2HGDH is one of the relevant genes. Collectively, our data support a model in which 14q loss results in a cellular milieu that leads to elevation of L-2-HG in RCC. Our studies demonstrate possible strategies to lower L-2-HG levels in RCC including targeting the source of L-2-HG in RCC via the glutamine axis or via inhibition of MDH (Fig. 6H).

To date, studies on both 2-HGs have demonstrated that multiple histone marks can be impacted. However, a major challenge has been identifying the relevant histone demethylases. Our transcriptomic studies indicate that KDM6A is a target of L-2-HG in RCC and are in agreement with mounting evidence demonstrating the importance of the H3K27 methylation axis in renal cancer. As noted before, mutations of KDM6A (also referred to as UTX), which encodes the enzyme that demethylates H3K27, have been reported in renal cancer (45, 46). Notably, multiple studies have now reported that KDMs are HIF target genes, which is particularly relevant in ccRCC in which inactivating mutations of VHL and ensuing HIF stabilization is a common event (64–68). A recent study by Chakraborty and colleagues indicated that VHL-defective RCC cells are dependent on the H3K27 methyltransferase activity of EZH1 for cell survival to counter the HIF-mediated induction of KDMs (69). Multiple studies indicate the importance of EZH2 in RCC phenotypes including in vivo tumor growth and therapy resistance (70, 71). Our data, therefore, add another layer of regulation of this axis in RCC via metabolism due to L-2-HG's ability to act as a competitive inhibitor.

Our data may have biomarker implications for ccRCC, a major gap in the kidney cancer field. As previously noted, the L2HGDH gene is located at 14q, a region whose loss is associated with poorer outcomes. We previously established the connection between reduced L2HGDH expression and elevated L-2-HG levels in ccRCC. Given that reduced L2HGDH expression is associated with worsened outcomes and increasing stage/grade, L-2-HG levels in a tumor could serve as a biomarker that could dictate therapy and/or render prognostic information.

Finally, our data add to the growing body of evidence that inhibition of glutamine metabolism may have therapeutic relevance in renal cancer. These data have particular relevance given our data demonstrating the role of glutamine metabolism in maintaining L-2-HG levels in multiple RCC lines tested as well as the in vivo efficacy of a glutaminase inhibitor in RCC cells with elevated L-2-HG levels. Recent studies in the context of clinical trials indicate that glutaminase inhibition may have efficacy in a subset of RCC patients (72). Efficacy of glutaminase inhibition in RCC in vivo has been linked with PARP sensitivity and oxidative stress (73, 74). Therefore, despite our rescue experiments with esterified L-2-HG in vitro, we cannot exclude effects on these alternate pathways in vivo. Nevertheless, these data warrant future studies to explore whether L-2-HG could be utilized as a biomarker for therapeutics that target glutamine metabolism.

In summary, our studies demonstrate that L2HGDH has tumor-suppressive effects and therefore support the concept of L-2-HG as an oncometabolite in kidney cancer. By delineating the metabolic events that promote L-2-HG accumulation and the downstream sequelae of this buildup, we provide new opportunities for therapeutic intervention in patients with L-2-HG–driven kidney tumors.

E. Emberley is a senior scientist and has ownership interest (including stock, patents, etc.) at Calithera Biosciences. No potential conflicts of interest were disclosed by other authors.

Conception and design: E.-H. Shim, G.J. Brinkley, S. Sudarshan

Development of methodology: S. Shelar, E.-H. Shim, G.J. Brinkley, S. Dutta, S. E. Velu, E. Emberley, A. Becker, C. Kunick, S. Sudarshan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Shelar, E.-H. Shim, G.J. Brinkley, A. Kundu, F. Carobbio, T. Poston, J. Tan, V. Parekh, D. Benson, D. Rakheja, Y. Sato, S. Ogawa, S. E. Velu, A. Pan, J. Chen, T. Haung, D. Absher, S. Sudarshan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Shelar, E.-H. Shim, G.J. Brinkley, F. Carobbio, T. Poston, D. K. Crossman, P. J. Buckhaults, D. Rakheja, S. Dutta, S. E. Velu, T. Haung, D. Absher, S. Sudarshan

Writing, review, and/or revision of the manuscript: S. Shelar, E.-H. Shim, D. K. Crossman, D. Absher, S. Sudarshan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. K. Crossman, R. Kirkman, A. Becker, C. Kunick

Study supervision: E. Emberley, S. Sudarshan

Other (grew and provided cancer cells that were used for in vivo experiments): E. Emberley

The research reported in this article was supported by R01CA200653, 5I01BX002930, the UAB-HudsonAlpha Center for Genomic Medicine (to S. Sudarshan) and in part by the UAB Comprehensive Cancer Center (P30CA013148).

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.