Glutamine synthase 2 (GLS2) is a key regulator of glutaminolysis and has been previously implicated in activities consistent with tumor suppression. Here we generated Gls2 knockout (KO) mice that develop late-occurring B-cell lymphomas and hepatocellular carcinomas (HCC). Further, Gls2 KO mice subjected to the hepatocarcinogenic Stelic Animal Model (STAM) protocol produce larger HCC tumors than seen in wild-type (WT) mice. GLS2 has been shown to promote ferroptosis, a form of cell death characterized by iron-dependent accumulation of lipid peroxides. In line with this, GLS2 deficiency, either in cells derived from Gls2 KO mice or in human cancer cells depleted of GLS2, conferred significant resistance to ferroptosis. Mechanistically, GLS2, but not GLS1, increased lipid reactive oxygen species (ROS) production by facilitating the conversion of glutamate to α-ketoglutarate (αKG), thereby promoting ferroptosis. Ectopic expression of WT GLS2 in a human hepatic adenocarcinoma xenograft model significantly reduced tumor size; this effect was nullified by either expressing a catalytically inactive form of GLS2 or by blocking ferroptosis. Furthermore, analysis of cancer patient datasets supported a role for GLS2-mediated regulation of ferroptosis in human tumor suppression. These data suggest that GLS2 is a bona fide tumor suppressor and that its ability to favor ferroptosis by regulating glutaminolysis contributes to its tumor suppressive function.

Significance:

This study demonstrates that the key regulator of glutaminolysis, GLS2, can limit HCC in vivo by promoting ferroptosis through αKG-dependent lipid ROS, which in turn might lay the foundation for a novel therapeutic approach.

Glutaminolysis, the stepwise process where glutaminase catalyzes the conversion of glutamine into glutamate (Glu; ref. 1), is an important pathway that intersects with several metabolic pathways such as glycolysis, tricarboxylic acid (TCA) cycle, redox homeostasis, along with lipid and amino acid homeostasis (1–3). There are two glutaminase enzymes: glutaminase 1 (GLS1) and glutaminase 2 (GLS2; ref. 4). While GLS1 is regulated by the oncogenes MYC (5), Rho GTPases (6), and Notch (7), we and another group previously reported that the p53 tumor-suppressor protein activates the transcription of GLS2, which positively regulates aerobic energy production in mitochondria and suppresses the simultaneously produced oxidative stress (8, 9). Since the two glutaminase enzymes appear to play opposite roles in tumorigenesis, switching between GLS1 and GLS2 is attracting attention as a new cancer target (10–15). While GLS1 is well validated as a cancer promoter (6, 7, 16, 17), the tumor-suppressor functions of GLS2 have not been studied extensively, although one study has reported that GLS2 can inhibit the small GTPase Rac1 by direct binding (18). We previously reported that the expression of GLS2 is reduced in liver tumors and that ectopically expressed GLS2 reduces the growth and colony forming ability of tumor cells (8). Consistent with our reports, overexpression of GLS2 has also been reported to suppress the malignant phenotype (8, 9, 19–21) and sensitize glioma cells to chemotherapeutic alkylating agents (20). Yet, in other studies it was reported that knockdown of GLS2 inhibits the growth of cervical cancer cells (22) and sensitizes human hepatoma and lung carcinoma to ionizing radiation (23). Thus, the tumor suppressor function of GLS2 appears to be complex and dependent on the cancer cell types used for different in vitro studies. Most relevantly, it is still unclear whether and how GLS2 acts as a tumor suppressor in vivo and what mechanisms enable it to do so.

Recently, Gao, and colleagues reported that GLS2 is involved in an iron and lipid peroxide-dependent cell death, termed ferroptosis (24), which has been suggested to be associated with cancer suppression (25–27). In fact, Jennis M. and colleagues (28) generated mice expressing a p53 variant that is defective in expression of Gls2, and such mice are deficient in undergoing ferroptosis. Ferroptosis, which can be initiated by small-molecule ferroptosis inducers (FIN) such as erastin and imidazole ketone erastin (IKE; a more potent form of erastin), involves the eventual inhibition of cellular antioxidant defenses such as glutathione peroxidase 4 (GPX4) and coenzyme Q10, leading to accumulation of lipid peroxides. This type of cell death is suppressed by lipophilic antioxidants such as ferrostatin-1 (Fer-1; refs. 29, 30). Recent studies have revealed that certain FINs inhibit tumor growth and can be combined with traditional methods of cancer therapy, especially for resistant cancers (25–27). Although the exact mechanism by which ferroptosis inhibits tumor growth is still unknown, different lines of evidence suggest that such FINs may be used to treat tumor cells and ultimately cancer.

In this study, we generated Gls2 knockout (KO) mice to determine whether these mice develop tumors and to investigate a possible association of GLS2 with ferroptosis in this setting. We have also discovered that failure of energy homeostasis via glutaminolysis is involved in tumorigenesis caused by lack of GLS2.

Animals

Gls2tm1a(EUCOMM)Wtsi ES cell lines were obtained from the International Knockout Mouse Consortium (KOMP; www.KOMP.org). The targeting vector, containing an IRES-lacZ trapping cassette and floxed neo cassette, was inserted into the first intron of the Gls2 gene, resulting in disruption of functional Gls2 transcription. The targeting vector was introduced into C57BL/6N ES cells and the mice bearing a null mutation of Gls2 alleles were generated as described previously (31).

Stelic Animal Model (STAM), a nonalcoholic steatohepatitis (NASH)–cirrhosis–hepatocarcinogenic model (Stelic Institute & Co.), was used for hepatocarcinogenetic experiments. Briefly, 2-day-old male pups were injected with streptozotocin (200 μg per mouse) and fed a high-fat diet (D12492, Research Diet Inc.) from the age of 4 weeks. This mouse model progressed from nonalcoholic fatty liver disease (NAFLD) to NASH at 8 weeks of age and developed hepatocellular carcinoma (HCC) at 16 to 20 weeks of age. The background liver histology of the STAM model was assessed by using the NASH score that includes fatty change, lobular inflammation, ballooning, sinus dilatation, and fibrosis. Livers including tumors were cut at 5- to 7-mm intervals. If there were no large tumors, six slices were made per whole liver. All sliced samples were examined histologically and every tumor was confirmed HCC or lymphoma by the pathologist.

The mice were maintained under specific pathogen-free conditions, on a 12-hour light–dark cycle, fed a normal diet (CE-2, Clea Japan Inc.), or high-fat diet (D12492, Research Diet Inc.) ad libitum. Only male mice were used for experiments. All animal studies were conducted in accordance with the International Guiding Principles for Biomedical Research Involving Animals and were approved by the Animal Care and Use Committees of Chiba University (Chiba, Japan) and the National Institute for Physiological Sciences in Japan (Okazaki, Japan).

Genotyping

Gls2-deficient mice were genotyped by PCR using P1 primer, 5′-cccatatctgttcagttctccagg-3′, and P2 primer, 5′-ttgctcaaggccaactcacagttc-3′. The PCR products are 365 bp for the wild-type (WT) allele, and 334 bp for the mutant allele.

Primary hepatocyte culture, cell lines, and reagents

Primary hepatocytes were prepared by perfusing the portal vein of the liver with 95 μg/mL EGTA, 150 U/mL collagenase solution, and enriched using a 36% Percoll (Sigma-Aldrich) gradient. Hepatocytes were cultured in Hepatocyte Culture Medium (COSMO BIO) on collagen-coated six-well plates. SKHep1 p53KO cells were generated using CRISPR/Cas9 genome editing as previously described (32). HepG2 (WT p53-positive HCC) and HepG3 (p53-null HCC) cell lines were purchased from ATCC (LCG Standards GmbH).

Probucol, α-ketoglutarate (αKG), Glu, amino-oxyacetate (AOA), and Fer-1 were purchased from Sigma-Aldrich. Erastin was purchased from Selleckchem. Liproxstatin-1 and deferoxamine were purchased from Funakoshi. IKE was synthesized in the lab of Dr. Brent Stockwell as in Larraufie and colleagues (33).

Plasmid construction, recombinant virus production, and transduction

Full-length human GLS2 (hGLS2; NM_013267; 1,809 bp) was subcloned into the p3 × FLAG-CMV10 vector (Sigma-Aldrich), pAcGFP1-N1 (Addgene), or CSII-EF-RfA-IRES2-Venus (RDB4389, Riken BioResource Center). We used PCR to generate a GLS2 glutaminase core domain deletion mutant, in which amino acids 177–468 are deleted from the full-length GLS2.

FLAG-tagged plasmids expressing WT hGLS2 (GLS2WT) or the 177–468 deletion mutant (GLS2del) in 3 × FLAG-CMV10, or 3 × FLAG-CMV10 (empty vector; Mock), or pAcGFP1-N1-GLS2 were transfected into SKHep1 p53KO cells using Lipofectamine 3000 (Invitrogen), as described in the manufacturer's protocol.

We produced recombinant lentiviruses and the lentivirus vector-containing culture supernatant as described previously (34). SKHep1 p53KO cells were infected with either CSII-EF-RfA-IRES2-Venus–empty vector (Mock) as well as GLS2WT- or GLS2del-containing CSII-EF-RfA-IRES2-Venus lentiviruses. 3 days after infection, GFP-positive lentivirus-infected cells were sorted with a FACSAria (Becton Dickinson).

To generate lentivirus-based short hairpin RNA (shRNA) constructs, a 19-bp shRNA-coding fragment with a 5′-ACGTGTGCTGTCCGT-3′loop was introduced into pENTR4-H1 digested with BgIII/XbaI. To insert the hGLS2 or hGLS1-shRNA into the lentivirus vector, we mixed the resulting pENTR4-H1-shRNA vector and CS-RfA-CG vector with Gateway LR clonase (Invitrogen). The GLS2 or GLS1 target sequences are CTCCATAAGCACCCTAGGC or AAGAGAAAGTGGAGATCGA, respectively. SKHep1 WT cells were infected with either empty vector (shCont) or shGLS2/shGLS1-containing CS-RfA-CG lentiviruses.

RNAi

21 nucleotide siRNA duplexes with 3′dXdY overhangs corresponding to hGLS2 mRNA (GLS2 RNAi, 5′-ATCAAGATGGACTGTAACAAA -3′) or hGLS1 mRNA (GLS1 RNAi, 5′-AAGAGAAAGTGGAGATCGAAA-3′) or firefly luciferase mRNA for control RNAi (luciferase, 5′-AACTTACGCTGAGTACTTCGA-3′) were synthesized by QIAGEN Inc. SKHep1 WT cells were transfected with the indicated siRNA oligonucleotide (40 nmol/L) using Lipofectamine 3000 (Thermo Fisher Scientific, Inc.) reagent according to the manufacturer's protocol.

Measurement of cell viability

Cell viability was measured by normalizing ATP levels using the Cell Titer-Glo regent as previously described (8). Cell viability was normalized to the corresponding DMSO-treated control cells and presented as a percentage of the control.

Measurement of reactive oxygen species and intracellular Fe2+

Cells were incubated with, C11-BODIPY (2 μmol/L), Mitosox (5 μmol/L), or 2′,7′-dichlorodihydro-fluorescein diacetate (DCF; 3 μmol/L) for lipid ROS, mitochondrial ROS, or cytosolic ROS respectively and measured in IncuCyte live-cell analysis system. All ROS reagents were purchased from Molecular Probes, Invitrogen. Fe2+ was assessed with the turn-on FeRhoNox−1 fluorescent probe as described by Hirayama and colleagues (35) and obtained from Goryo Chemical Inc.

qRT-PCR

RT-qPCR was performed as described (8). All gene-specific mRNA expression values were normalized against the internal housekeeping gene, 18S for mice and L32 for humans. Please see Supplementary information for primer sequences.

RNA sequencing analysis

Total RNA (10 μg) from each sample was prepared as previously described (36). The resulting size-fractionated cDNA was used for sequencing with an Illumina GAIIx. The generated sequence tags were mapped onto the human genome sequence (mm9 from the UCSC Genome Browser) using the Eland program (Illumina). Unmapped or redundantly mapped sequences were removed from the dataset, and only uniquely mapped sequences without any mismatches were used for analyses. Gene expression was quantified as fragments per kilobase million (FPKM; ref. 37).

Western blot analysis and antibodies

We performed Western blot analyses as previously described (8). The primary antibodies for mouse samples were as follows: Gls2 (Abcam), Gls1 (Abcam), Caspase-3 (Abcam), p53 Ab3 (Merck), and β-actin (Sigma-Aldrich). The primary antibodies for human samples were as follows: M2-FLAG (Sigma-Aldrich), p53 DO1 (Santa Cruz Biotechnology), and anti-actin (Sigma). A polyclonal antibody, hGLS2 (N14), was raised against hGLS2 protein (TaKaRa) using a synthetic peptide (corresponding to a sequence within the GLS2 N terminus: PHSHQPQHQDHDSS) conjugated to KLH. The titer of the crude rabbit sera was confirmed by ELISA and then the sera were subjected to affinity purification using recombinant hGLS2 protein as previously described (8).

Metabolomic analysis

Extracts were prepared from approximately 2 × 106 cells with methanol containing internal standard solution (Human Metabolome Technologies). Cationic compounds were measured in the positive mode of capillary electrophoresis–connected time-of-flight mass spectrometry (CE-TOFMS) and anionic compounds were measured in the positive and negative modes of CE-MS/MS.

Determination of Glu and glutamine concentrations

Concentrations of Glu and glutamine in the medium were determined using a glutamine/Glu determination kit (GLN-1; Sigma-Aldrich) as previously described (8).

Oxygen consumption assays

One day following transfection with p3 × FLAG-CMV10-hGLS2 constructs or p3 × FLAG-CMV10–empty vector (Mock), equal numbers of cells in KRH buffer (Krebs-Ringer Hepes buffer; 25 mmol/L Hepes, 130 m NaCl, 5 mmol/L KCl, 1.3 mmol/L CaCl2, and 1.3 mmol/L MgSO4, pH 7.4) supplemented with 2.5% BSA and 2 mmol/L sodium pyruvate were seeded in triplicate into BD Oxygen Biosensor System (BD Biosciences). Oxygen consumption was measured in a fluorescence plate reader at 485 nm excitation and 630 nm emission at 2-hour intervals for 24 hours. 

For measurement of mitochondrial respiration (Seahorse Biosciences), SKHep1 cells were cultured overnight on XFe 96 plates at a density of 20,000 cells per well. Then the medium was replaced 1 hour before the start of the measurement of mitochondrial respiration. Oligomycin (2 μmol/L), carbonilcyanide p-triflouromethoxyphenylhydrazone (FCCP; 0.5 μmol/L), antimycin (0.5 μmol/L), and rotenone (1 μmol/L; Sigma-Aldrich) were added to XFe 96 media, and the samples were loaded into the sensor cartridge. Oxygen consumption rate (OCR) was then recorded by using a XF24 extracellular flux analyzer.

Metabolic studies

For the oral glucose tolerance test (OGTT), insulin tolerance test and pyruvate tolerance test, 16 to 20 weeks old mice were fasted for 16 hours and then glucose (1 g/kg) was administered orally, insulin (0.75 IU/kg) and sodium pyruvate (1 g/kg) were administered intraperitoneally (38). Blood glucose concentrations were measured at the indicated time points using a glucometer (Glutestmint; Sanwa Kagaku Kenkyusho Co. Ltd.).

Xenograft model of tumor growth suppression assay in animals

For assessment of tumor growth suppression, a total of 1 × 105 of SKHep1 p53KO cells infected with CSII-EF-RfA-IRES2-Venus lentiviruses containing either GLS2wt/GLS2del or CSII-EF-RfA-IRES2-Venus–empty vector constructs (Mock) and SKHep1 cells with either CS-RfA-CG–empty vector (shCont) or shGLS2/shGLS1 constructs, were suspended in 50 μL of DMEM with 50 μL Matrigel (BD Biosciences). Subcutaneous injections of SKHep1 cells with indicated vector constructs were performed on the right and left flanks of the same SCID mice (5–6 weeks, male; Clea Japan Inc). 6 weeks after subcutaneous injections, tumors were dissected and GFP expressed from lentiviral vector was measured using the IVIS imaging system (Perkin Elmer) according to the manufacturer's protocol (Perkin Elmer). Although both the Institutional Animal Care and Use Committee protocols demand that the maximal tumor size should not exceed 2 cm in any dimension, none of the implanted tumors in our experiments exceeded this restriction.

Colony formation assay, transwell invasion assays, and wound-healing assays

Colony formation assays were performed as previously described (39) and invasion assays were performed using transwell invasion chambers coated with Matrigel (50 μL per filter; BD Biosciences) as described in the manufacturer's protocol. SKHep1 p53KO cells were transfected with either p3 × FLAG-CMV10-GLS2WT, p3 × FLAG-CMV10-GLS2del, or p3 × FLAG-CMV10–empty vector constructs (Mock) and cultured for 48 hours prior to being transferred onto the top of Matrigel-coated invasion chambers with 1% FBS DMEM (5 × 104 cells/well). DMEM containing 10% FBS was added to the lower chambers. After incubation for 24 hours at 37°C in an atmosphere containing 5% CO2, invaded cells on the lower surface were stained with crystal violet stain and counted under a light microscope. For wound scratch assays, transfected cells were cultured on collagen I (BD Biosciences) until confluence and then wounded using a blue 1 mL pipette tip (40). Media was removed, cells washed with PBS, and replenished with fresh media. Images were acquired immediately following media replacement, and after 24 hours via phase-contrast microscope at 10×. After exporting images, wound areas were measured using ImageJ.

Histologic examination

Hematoxylin and eosin staining was performed as previously described (39). 4-Hydroxynonenal (4-HNE), a highly toxic aldehyde product of lipid peroxidation, was evaluated using an anti–4-HNE mAb (bs-6313R, Bioss Inc.) and labeled DAKO EnVision + System-HRP Labelled Polymer Anti-Rabbit. Oil-Red O staining was performed in frozen liver sections as previously described (41). Sirius red staining for collagen deposition was performed as previously referenced (42). The primary antibodies for mouse samples were as follows: Gls2 (Sigma-Aldrich) and p53 Pab240 (Santa Cruz Biotechnology Inc). The slides were examined under a Keyence BZ-8100 microscope (Keyence Japan).

Methylation-specific PCR assay

The methylation-specific PCR assay for the Gls2 promoter was performed as previously described (43). Genomic DNA was purified from tumor tissues using the DNeasy Blood and Tissue Kit (Qiagen). For DNA bisulfite modification, we used the DNA modification kit (Epigentek Inc.) according to the manufacturer's protocol. Primers for qRT-PCR analyses of live mice samples for DNA bisulfite modification analyses were as follows: mGls2 methylated oligonucleotides, 5′-gatttaagaatttgggttttatgac-3′ and 5′-caatacacacttacaaacaaacgat-3′; mGls2 unmethylated oligonucleotides, 5′-tgatttaagaatttgggttttatgat-3′ and 5′-aaccaatacacacttacaaacaaaca-3′. Amplified total DNA was subjected to electrophoresis on a 2% agarose gel, and then visualized by ethidium bromide staining.

Statistics

Results are expressed as means ± SEM. Differences between two groups were assessed using the unpaired two-tailed Student t test. P < 0.05 was considered to be statistically significant.

Data availability

The data generated in this study are available upon request from the corresponding authors.

Gls2 KO mice exhibit markedly a increased incidence of HCC

Since GLS2 proteins share a high degree of homology (98%) between human and mice (Supplementary Fig. S1A), we generated Gls2 KO mice to further study the functional roles of GLS2 in vivo. The gene targeting strategy for the generation of the KO mice is shown in Supplementary Fig. S1B. Briefly, a gene trap vector was inserted into introns 1 and 7 to delete exons 2 to 7 after Cre recombinase (Cre)-mediated gene recombination. The targeting vector was electroporated into ES cells, and upon establishing cells that had undergone homologous recombination, KO mice were successfully generated (Supplementary Fig. S1C). In these KO mice, Gls2 protein and mRNA expression were ablated while levels of Gls1 protein remained unchanged (Fig. 1A; Supplementary Fig. S1D and S1E). These expression changes were assessed in the livers of mice, since the liver possesses the highest expression of GLS2 amongst all the organs examined in both mice and humans (Supplementary Fig. S2A and S2B). The KO mice were born at Mendelian ratios (Supplementary Fig. S2C), with no external malformations. When WT and KO mice were sacrificed over time, however, we found that by 120 weeks of age, all 7 KO mice examined had developed tumors—4 of these exhibited B-cell lymphomas, 2 had HCCs, and 1 had both HCC and B-cell lymphoma (Fig. 1B and C). By contrast, none of the 6 WT mice examined exhibited any tumors at 120 weeks of age (Fig. 1B and D). When the HCC in the KO mice were segregated into size-based categories (under 1 mm, 1–3 mm, and above 3 mm pathologically; Fig. 1E), 2 out of the 3 HCCs in the KO mice were relatively large tumors that exceeded 3 mm pathologically (Fig. 1F). This indicates that the absence of GLS2 significantly increases the onset of liver tumors albeit at a later stage in murine life, and further enables the tumors to grow in size.

Figure 1.

Progression of tumorigenesis in the absence of Gls2. A, Immunoblot showing Gls2 and Gls1 protein levels in the livers of KO (Gls2-/-), heterozygous (Hetero; Gls2±), and WT (Gls2+/+) mice. Actin was used as the loading control. B, Summary of tumorigenesis in WT and KO mice after 65 weeks of age. C and D, Macroscopic findings of tumorigenesis at 120 weeks of age in KO mice (C) and in WT mice (D). E, Hematoxylin and eosin stain of liver sections indicating HCC ≤1 mm, HCC 1 mm to 3 mm, and HCC >3 mm. F, The average number of liver tumors and the relative size distribution of HCC in two KO mice at 120 weeks was determined based on the analysis of the hematoxylin and eosin stain (classified as >3 mm, 1 mm–3 mm, ≤3 mm) shown in E. G, Diagram showing generation and pathologic analysis of STAM, a model for NASH and HCC. HFD, high-fat diet. H, Macroscopic findings of tumorigenesis in the three experimental groups of STAM: WT-STAM, KO-STAM, and KO mice fed with high-fat diets mixed with 1% probucol (KO-STAM with probucol). Arrowheads, HCCs. I, The average number of tumors and the relative size distribution of HCC from the liver of WT-STAM, KO-STAM, and KO-STAM with probucol at 15 weeks of age were determined based on the analysis of hematoxylin and eosin staining (classified as ≤1 mm, 1 mm–3 mm, >3 mm). J, The IHC appearance of 4-HNE (brown) in normal liver and HCCs of WT-STAM and KO-STAM. Scale bars, 100 μm. K, Expression levels of Ptgs2 (mRNA) in normal liver and HCCs of WT-STAM and KO-STAM were determined by the comparative threshold cycle method and then normalized to 18S expression. Values are the means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant; w, weeks.

Figure 1.

Progression of tumorigenesis in the absence of Gls2. A, Immunoblot showing Gls2 and Gls1 protein levels in the livers of KO (Gls2-/-), heterozygous (Hetero; Gls2±), and WT (Gls2+/+) mice. Actin was used as the loading control. B, Summary of tumorigenesis in WT and KO mice after 65 weeks of age. C and D, Macroscopic findings of tumorigenesis at 120 weeks of age in KO mice (C) and in WT mice (D). E, Hematoxylin and eosin stain of liver sections indicating HCC ≤1 mm, HCC 1 mm to 3 mm, and HCC >3 mm. F, The average number of liver tumors and the relative size distribution of HCC in two KO mice at 120 weeks was determined based on the analysis of the hematoxylin and eosin stain (classified as >3 mm, 1 mm–3 mm, ≤3 mm) shown in E. G, Diagram showing generation and pathologic analysis of STAM, a model for NASH and HCC. HFD, high-fat diet. H, Macroscopic findings of tumorigenesis in the three experimental groups of STAM: WT-STAM, KO-STAM, and KO mice fed with high-fat diets mixed with 1% probucol (KO-STAM with probucol). Arrowheads, HCCs. I, The average number of tumors and the relative size distribution of HCC from the liver of WT-STAM, KO-STAM, and KO-STAM with probucol at 15 weeks of age were determined based on the analysis of hematoxylin and eosin staining (classified as ≤1 mm, 1 mm–3 mm, >3 mm). J, The IHC appearance of 4-HNE (brown) in normal liver and HCCs of WT-STAM and KO-STAM. Scale bars, 100 μm. K, Expression levels of Ptgs2 (mRNA) in normal liver and HCCs of WT-STAM and KO-STAM were determined by the comparative threshold cycle method and then normalized to 18S expression. Values are the means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant; w, weeks.

Close modal

Next, to confirm that GLS2 loss was involved in the development of liver cancer, we created STAM mice (44), which is a model for NASH leading to HCC. STAM mice (C57BL/6 strain) were established by using a combination of chemical (subcutaneous injection of low-dose streptozotocin) and subjecting mice to a high-fat diet (Fig. 1G). These interventions resulted in marked hyperglycemia with low insulin secretion, hypercholesterolemia, and high alanine aminotransferase (ALT) levels (Supplementary Fig. S3A–S3D). The protocol also resulted in progression from fatty liver to NASH by 8 weeks of age, followed by cirrhosis and finally, development of HCC in all the mice by 20 weeks of age (Fig. 1G; Supplementary Fig. S3E). At 15 weeks of age, macroscopically clear liver masses were already observed in the Gls2 KO-STAM mice but not in WT-STAM mice (Fig. 1H). Subsequent pathologic analysis revealed that the KO-STAM mice also showed an increase in number of large HCCs (above 3 mm) when compared with WT-STAM mice (Fig. 1I). Neither the presence of small HCCs (below 3 mm) nor NASH scores differed between the two groups (Supplementary Fig. S3F). Taken together, we surmise that the loss of GLS2 contributes to an earlier onset of tumors, be it spontaneously or under stress; further our data also suggest that GLS2 loss signficantly promotes progression of liver tumors.

HCCs exhibit a weak ferroptotic response in the absence of GLS2

Since we previously reported that GLS2 can provide an antioxidant effect in vitro, we hypothesized that probucol, a drug with antioxidant properties (45) and reported to have a protective effect on liver stenosis and NASH (46, 47), might suppress the development of HCC in KO-STAM mice. Unexpectedly, and contrary to our original hypothesis, the KO-STAM mice with probucol mixed into their high-fat diets (KO-STAM with probucol) showed exacerbated HCC (Fig. 1H and I). This counterintuitive result suggested that the formation of HCC in the KO mice could not simply be a result of increased oxidative stress. As GLS2 was previously shown to favor ferroptosis (24, 48, 49), and as probucol has been recently shown be an inhibitor of ferroptosis (50), we hypothesized that failure to induce ferroptosis in our KO mice might be involved in the onset of hepatocellular carcinoma (51).

For the rest of this study we have employed several experimental approaches to examine ferroptosis: First, since reagents such as antibodies that can identify cells undergoing ferroptosis are not easily available, where appropriate we have evaluated the expression of established ferroptosis markers such as 4-HNE, Ptsg2, and Chac1 (52, 53). Second, we have used both primary hepatocytes from Gls2 KO mice and the human cell lines (SKHep1 WT, HepG2, and Hep3B) that were initially derived from human hepatocellular carcinomas in order to examine the interplay between GLS2 and ferroptosis. Finally, we have used the two different classes of ferroptosis inducers (erastin/IKE and RSL3) to cause cell death as measured by an ATP-based cell viability assay and we have confirmed that this death is ferroptosis by demonstrating that it can be inhibited by the ferroptosis specific inhibitors (ferrostatin, liproxstatin, and deferoxamine; refs. 48, 49) as well as by confirming the associated increase in lipid ROS prior to the start of cell death.

We observed that ferroptosis markers, 4-HNE and Ptgs2, were increased in liver tumors from WT-STAM compared with tumors from KO-STAM or nontumors (Fig. 1J and K). We next performed the Malondialdehyde assay to measure the end products of lipid peroxidation, the inducer of ferroptotic death, and showed that Malondialdehyde was significantly decreased in the livers from KO compared with WT mice (Fig. 2A). In addition, primary hepatocytes from the KO mice also showed decreased sensitivity to inducers of ferroptosis when compared with hepatocytes derived from WT mice (Fig. 2B and C). These altered levels of cell death in the hepatocytes were fully attenuated by the ferroptosis inhibitor, Fer-1 (Fig. 2B and C) and were not associated with changes in the levels of the apoptosis marker, cleaved caspase-3 (Supplementary Fig. S4A), thereby confirming that the cell death seen was in fact due to ferroptosis. Since the HCC lesions in WT-STAM mice were more resistant to ferroptosis compared with the non-HCC lesions (Supplementary Fig. S4B and S4C), this supports the possibility that the livers of KO mice have properties similar to HCC. Further, the ferroptosis resistance observed in primary hepatocytes from KO mice was attenuated by overexpression of WT GLS2, supporting our conclusion that GLS2 is essential for complete ferroptosis induction in primary hepatocytes (Fig. 2D and E).

Figure 2.

Loss of Gls2 results in HCC lesions with increased resistance to ferroptosis. A, The levels of malondialdehyde (MDA), the end product of lipid peroxidation, in the livers from KO (22 weeks, n = 4) and WT (22 weeks, n = 4) mice. Values are the means ± SEM. Primary hepatocytes from WT (20 weeks, n = 9) or KO mice (20 weeks, n = 6) were treated with indicated ferroptosis inducers (erastin, 20 μmol/L; IKE, 10 μmol/L) for 72 hours in the absence or presence of the ferroptosis inhibitor Fer-1 (20 μmol/L). Representative visualization of the treated primary hepatocytes from WT and KO mice are shown in B (×20 magnification). ATP-based cell viability was assayed after these treatments and the corresponding data in C are presented as a percentage of the control (DMSO). Values are the means ± SEM. **P < 0.01. D and E, Cell viability in D and Ptgs2 gene expression in E were assayed in mice primary hepatocytes from KO mice (20 weeks) transfected with p3 × FLAG-CMV10—empty vector (Mock, n = 4) or p3 × FLAG-CMV10–hGLS2 vector (GLS2wt, n = 4). F, RNA levels of the ferroptosis marker Ptgs2 were obtained from RNA-seq analysis performed in liver tissues of WT with non-HCC (20 weeks, n = 1), KO with non-HCC (20 weeks, n = 3), and KO with HCC (120 weeks, n = 1). *, P < 0.05; **, P < 0.01.

Figure 2.

Loss of Gls2 results in HCC lesions with increased resistance to ferroptosis. A, The levels of malondialdehyde (MDA), the end product of lipid peroxidation, in the livers from KO (22 weeks, n = 4) and WT (22 weeks, n = 4) mice. Values are the means ± SEM. Primary hepatocytes from WT (20 weeks, n = 9) or KO mice (20 weeks, n = 6) were treated with indicated ferroptosis inducers (erastin, 20 μmol/L; IKE, 10 μmol/L) for 72 hours in the absence or presence of the ferroptosis inhibitor Fer-1 (20 μmol/L). Representative visualization of the treated primary hepatocytes from WT and KO mice are shown in B (×20 magnification). ATP-based cell viability was assayed after these treatments and the corresponding data in C are presented as a percentage of the control (DMSO). Values are the means ± SEM. **P < 0.01. D and E, Cell viability in D and Ptgs2 gene expression in E were assayed in mice primary hepatocytes from KO mice (20 weeks) transfected with p3 × FLAG-CMV10—empty vector (Mock, n = 4) or p3 × FLAG-CMV10–hGLS2 vector (GLS2wt, n = 4). F, RNA levels of the ferroptosis marker Ptgs2 were obtained from RNA-seq analysis performed in liver tissues of WT with non-HCC (20 weeks, n = 1), KO with non-HCC (20 weeks, n = 3), and KO with HCC (120 weeks, n = 1). *, P < 0.05; **, P < 0.01.

Close modal

To further evaluate the relationship between tumorigenesis and ferroptosis as well as to other forms of programmed cell death such as autophagy and apoptosis in KO mice, RNA sequencing (RNA-seq) analysis was performed in non-HCC liver tissues of WT mice and KO mice as well as in HCCs derived from 120-week–old KO mice (Supplementary Fig. S4D). Expression of several genes related to autophagy and apoptosis was generally increased in the non-HCC samples from the KO compared with WT mice. In the HCCs from the KO mice, expression of autophagy-related genes tended to decrease, while apoptosis-related genes showed varying trends that differed with the gene that was assayed. Note that the ferroptosis-related gene, Ptgs2 was expressed in the WT mice, while Ptgs2 expression was completely abolished in both non-HCC and HCC from KO mice. (Fig. 2F). Given that autophagy has also been shown to accompany death due to ferroptosis (54, 55), these data together demonstrate that ferroptosis induction was downregulated in HCCs from the KO mice. Finally, we confirmed that a high dose of probucol was able to substantially suppress erastin-induced ferroptosis in human SKHep1 cells in agreement with a published report (Supplementary Fig. S5A and S5B; ref. 50). Together these in vitro data suggest that probucol exacerbates HCC in the KO-STAM mice most likely by further inhibiting ferroptosis in vivo. Taken together, our data indicate that ferroptosis is involved in the onset and progression of HCC in Gls2 KO mice.

GLS2 but not GLS1 can facilitate ferroptosis by promoting lipid ROS formation

To investigate the mechanism by which GLS2 might potentiate ferroptosis we used human SK-HEP-1 cells from which either GLS1 or GLS2 were depleted using siRNAs (Fig. 3A). When these cells were treated with erastin, ablation of GLS1 did not significantly reduce the degree of cell death over 24 hours compared with control siRNA (Fig. 3B; Supplementary Fig. S6A and S6B). By contrast, when GLS2 was silenced in these cells, ferroptotic death was reduced as evidenced by the time of death onset, the lethal concentration of erastin, and maximum amount of cell death achieved at the higher erastin concentration; with the most significant differences observed at 18 and 24 hours after erastin treatment (Fig 3B; Supplementary Fig. S6B). Conversely, under normal conditions, GLS2-silenced cells actually showed a decrease in viability as measured by ATP levels (Fig. 3B; right panel).

Figure 3.

GLS2 promotes an increase in lipid ROS and a concomitant increase in death due to erastin treatment. A, SKHep1 WT cells were transfected with luciferase RNAi (siLuci), hGLS1 RNAi (siGLS1), or hGLS2 RNAi (siGLS2) for 48 hours, followed by immunoblot analysis to detect GLS1 and GLS2 or actin as indicated. B, Cell viability of SKHep1 WT cells that were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with erastin (0, 5, or 10 μmol/L). The viability was assayed at 18 hours (left) and 24 hours (right) post erastin treatment. Values are the means ± SEM (n = 4). *, P < 0.05 versus siLuci. C and D, SKHep1 WT cells were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with vehicle (DMSO) or erastin (5 μmol/L) for 12 hours (n = 6) or 18 hours (n = 10). Lipid ROS was detected by C11-BODIPY. The ratio of oxidized form (green) to the nonoxidized form (red) are presented. Bar graph in C, means ± SEM. *, P < 0.05; ***, P < 0.001 versus siLuci. Representative images are shown in D. E and F, Intracellular Fe2+ was detected by FeRhoNox-1, a fluorescent probe that measures labile iron. The green fluorescence signal from the probe is shown in representative images in F. The bar graph in E, means ± SEM (n = 6). Scale bars in D and F, 100 μm. G and H, Cell viability in G and lipid ROS in H were measured in HepG2 cells that were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with vehicle (DMSO), erastin (10 μmol/L), or RSL3 (3 μmol/L). The cells were treated for 30 hours (n = 6) in G and for 24 hours (n = 4–6) in H. Bar graphs, means ± SEM. *, P < 0.05; **, P < 0.01 versus siLuci. h, hours.

Figure 3.

GLS2 promotes an increase in lipid ROS and a concomitant increase in death due to erastin treatment. A, SKHep1 WT cells were transfected with luciferase RNAi (siLuci), hGLS1 RNAi (siGLS1), or hGLS2 RNAi (siGLS2) for 48 hours, followed by immunoblot analysis to detect GLS1 and GLS2 or actin as indicated. B, Cell viability of SKHep1 WT cells that were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with erastin (0, 5, or 10 μmol/L). The viability was assayed at 18 hours (left) and 24 hours (right) post erastin treatment. Values are the means ± SEM (n = 4). *, P < 0.05 versus siLuci. C and D, SKHep1 WT cells were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with vehicle (DMSO) or erastin (5 μmol/L) for 12 hours (n = 6) or 18 hours (n = 10). Lipid ROS was detected by C11-BODIPY. The ratio of oxidized form (green) to the nonoxidized form (red) are presented. Bar graph in C, means ± SEM. *, P < 0.05; ***, P < 0.001 versus siLuci. Representative images are shown in D. E and F, Intracellular Fe2+ was detected by FeRhoNox-1, a fluorescent probe that measures labile iron. The green fluorescence signal from the probe is shown in representative images in F. The bar graph in E, means ± SEM (n = 6). Scale bars in D and F, 100 μm. G and H, Cell viability in G and lipid ROS in H were measured in HepG2 cells that were transfected with siLuci, siGLS1, or siGLS2 for 36 hours and then treated with vehicle (DMSO), erastin (10 μmol/L), or RSL3 (3 μmol/L). The cells were treated for 30 hours (n = 6) in G and for 24 hours (n = 4–6) in H. Bar graphs, means ± SEM. *, P < 0.05; **, P < 0.01 versus siLuci. h, hours.

Close modal

We then analyzed the effect of GLS2 on lipid ROS and intracellular Fe2+ levels, which are the main ferroptosis mediators. We used C11-BODIPY to calculate the ratio of oxidized (green) to nonoxidized (red) forms of lipids in order to assess lipid ROS (56). Ablation of GLS2 in the SKHep1 cells caused a decrease in lipid ROS at 12 hours posttreatment with erastin, even though at this time point ferroptotic death was still not pronounced. The decrease in lipid ROS became even more significant at 18 hours after erastin treatment when there was observable death in the control cells (Fig. 3C and D). No such decrease in lipid ROS was observed in GLS1-silenced cells compared with the control. On the other hand, GLS2 silencing did not affect intracellular Fe2+ levels after erastin treatment, suggesting that GLS2 functions donwstream of labile iron production or release (Fig. 3E and F). We extended these findings to a second human HCC cell line (HepG2), where ferroptotic death and lipid ROS were partially rescued by GLS2 silencing after treatment with both class I and class II ferroptosis inducers (erastin and RSL3; Fig. 3G and H).

We previously reported that GLS2 exerts antioxidant activity via GSH (8), and here we found that silencing of GLS2 modestly reduced the GSH/GSSG ratio in SKHep1 cells (Supplementary Fig. S6C). However, when erastin was administered to induce ferroptosis, the GSH/GSSG ratio was markedly depleted as previously reported (Supplementary Fig. S6D; ref. 52). Note that the depletion of GSH under erastin was much more extensive than that seen with the knockdown of GLS2 (Supplementary Fig. S6C and S6D). Although we had found that reducing GLS2 leads to GSH depletion, lowering GLS2 also inhibited ferroptosis, causing a concomitant decrease in lipid ROS. As an explanation, we believe that the antioxidant function of GLS2 via enhancing GSH levels is not of much consequence in the context of ferroptosis. Instead, we hypothesize that the role of GLS2 in promoting the accumulation of lipid ROS is dominant in that setting. We propose that there is a shift in the roles of GLS2 when ferroptosis is induced and we provide support for this theory below.

GLS2 promotes αKG production through Glu and thereby increases lipid ROS during ferroptosis

Next, to clarify the role of GLS2 in regulation of lipid ROS, we asked whether GLS2 promotes aerobic energy production through the TCA cycle, which has been reported to influence ferroptosis through mitochondrial membrane hyperpolarization (57).

To approach this experimentally, we used p53 KO clones that were derived from SKHep1 cells by using CRISPR/Cas9 methodology (SKHep1 p53KO; ref. 58) as they were more resistant to ferroptosis compared with SKHep1 WT cells (Fig. 4A) as reported in previous papers (32, 59–61). These cells are expected to be unable to induce the expression of GLS2 in response to stressors owing to the absence of p53 (8). In line with this, we found that GLS2 protein (Fig. 4B; top panel) and mRNA (Fig. 4B; bottom panel) levels were increased in response to ferroptosis in SKHep1 WT but not in p53 KO cells, suggesting that insufficient induction of GLS2 in the p53 KO cells can contribute to this resistance. The ferroptosis markers, PTGS2 and CHAC1 were induced these cell lines, consistent with the induction of ferroptosis in both WT and p53 KO cells (Fig. 4C). These data indicate that GLS2 might respond to the induction of ferroptotic stress also only in the presence of p53. Therefore, we ectopically expressed GLS2 in SKHep1 p53KO cells (Fig. 4D) and ablated GLS2 using siRNA in SKHep1 WT cells (Fig. 3A) to analyze associated cellular changes and response to ferroptosis induction.

Figure 4.

GLS2 mediates ferroptosis sensitivity via αKG. A, Cell viability in SKHep1 WT cells or SKHep1 p53 SKHep1 p53KO cells in response to treatment with erastin (5 μmol/L) in the presence or absence of Fer-1 (5 μmol/L) was measured at the indicated time points. Values are the means ± SEM (n = 4). B, Top, immunoblot analysis of p53 and GLS2 expression in SKHep1 WT or SKHep1 p53KO cells following treatment with erastin (0, 2.5, or 5 μmol/L) for 12 hours. Bottom, change in mRNA levels of the indicated p53 target genes along with GLS1 after DMSO or erastin (5 μmol/L) treatment for 12 hours in SKHep1 WT or SKHep1 p53KO cells. C, Change in mRNA levels of ferroptosis markers after DMSO or erastin (5 μmol/L) treatment for 12 hours in SKHep1 WT or SKHep1 p53KO cells. D, SKHep1 p53KO cells were transfected with p3 × FLAG-CMV10–empty vector (Mock), p3 × FLAG-CMV10–hGLS1 (GLS1) or p3 × FLAG-CMV10–hGLS2 (GLS2) constructs for 48 hours prior to lysis and processing for immunoblotting with anti-GLS2 antibody. E, Intracellular colocalization of GLS2 (PAcGFPN1-hGLS2) and mitochondria (MitoTracker) in SKHep1 KO cells. F, Relative OCR in SKHep1 p53KO cells transfected with Mock or GLS2 (left). Relative OCR in SKHep1 WT cells transfected with luciferase RNAi (siLuci; left) or hGLS2 RNAi (siGLS2; right). *, P < 0.05 versus Mock or siLuci at each time point. G, Intracellular metabolite levels were quantified by CE-MS analysis and normalized to the number of SKHep1 WT cells. Then ratios of αKG to Glu in SKHep1 WT cells transfected with siLuci (gray) or siGLS2 (red; for 48 hours) were calculated. SKHep1 WT cells were transfected with either control (siLuci) or siGLS2 for 36 hours and then treated with erastin (5 μmol/L) for 18 hours. H and I, αKG (10 mmol/L) or Glu (500 μmol/L) was added to culture medium at the same time as erastin and then cell morphology (H) and viability (I) were determined. Bar graph in I, means ± SEM (n = 4). *, P < 0.05 versus siGLS2. J, Cell viability of HepG2 cells that were transfected with either control (siLuci) or siGLS2 for 36 hours and then treated with erastin (10 μmol/L) for 30 hours. αKG (10 mmol/L) or Glu (500 μmol/L) was added to culture medium at the same time as erastin. Bar graph in J, means ± SEM (n = 6). **, P < 0.01 versus siGLS2. SK Hep1 p53KO cells were transfected with Mock or GLS2 constructs for 24 hours and then cells were treated with DMSO, erastin (5 μmol/L or 10 μmol/L) alone, or erastin (5 μmol/L or 10 μmol/L) with AOA (5 mmol/L) for 24 hours. K and L, Cell morphology (K) and cell viability (L) were assayed for these treated cells. Scale bar, 100 μm (H and K). Bar graph in L, means ± SEM (n = 4). *, P < 0.05 versus GLS2. M and N, Cell viability of Hep3B cells that were transfected with Mock or GLS2 constructs for 24 hours and then cells were treated for 24 hours with either DMSO or ferroptosis inducers (erastin 10 μmol/L or RSL3 3 μmol/L) or indicated ferroptosis inducer with AOA (10 mmol/L; M) or erastin (10 μmol/L) in the absence or presence of the ferroptosis inhibitor (Fer-1 10 μmol/L, liproxstatin 2 μmol/L, or deferoxamine 50 μmol/L; N). Bar graphs in M and N, means ± SEM (n = 6). *, P < 0.05; **, P < 0.01 versus GLS2 for M or versus erastin in N. DFO, deferoxamine; h, hours.

Figure 4.

GLS2 mediates ferroptosis sensitivity via αKG. A, Cell viability in SKHep1 WT cells or SKHep1 p53 SKHep1 p53KO cells in response to treatment with erastin (5 μmol/L) in the presence or absence of Fer-1 (5 μmol/L) was measured at the indicated time points. Values are the means ± SEM (n = 4). B, Top, immunoblot analysis of p53 and GLS2 expression in SKHep1 WT or SKHep1 p53KO cells following treatment with erastin (0, 2.5, or 5 μmol/L) for 12 hours. Bottom, change in mRNA levels of the indicated p53 target genes along with GLS1 after DMSO or erastin (5 μmol/L) treatment for 12 hours in SKHep1 WT or SKHep1 p53KO cells. C, Change in mRNA levels of ferroptosis markers after DMSO or erastin (5 μmol/L) treatment for 12 hours in SKHep1 WT or SKHep1 p53KO cells. D, SKHep1 p53KO cells were transfected with p3 × FLAG-CMV10–empty vector (Mock), p3 × FLAG-CMV10–hGLS1 (GLS1) or p3 × FLAG-CMV10–hGLS2 (GLS2) constructs for 48 hours prior to lysis and processing for immunoblotting with anti-GLS2 antibody. E, Intracellular colocalization of GLS2 (PAcGFPN1-hGLS2) and mitochondria (MitoTracker) in SKHep1 KO cells. F, Relative OCR in SKHep1 p53KO cells transfected with Mock or GLS2 (left). Relative OCR in SKHep1 WT cells transfected with luciferase RNAi (siLuci; left) or hGLS2 RNAi (siGLS2; right). *, P < 0.05 versus Mock or siLuci at each time point. G, Intracellular metabolite levels were quantified by CE-MS analysis and normalized to the number of SKHep1 WT cells. Then ratios of αKG to Glu in SKHep1 WT cells transfected with siLuci (gray) or siGLS2 (red; for 48 hours) were calculated. SKHep1 WT cells were transfected with either control (siLuci) or siGLS2 for 36 hours and then treated with erastin (5 μmol/L) for 18 hours. H and I, αKG (10 mmol/L) or Glu (500 μmol/L) was added to culture medium at the same time as erastin and then cell morphology (H) and viability (I) were determined. Bar graph in I, means ± SEM (n = 4). *, P < 0.05 versus siGLS2. J, Cell viability of HepG2 cells that were transfected with either control (siLuci) or siGLS2 for 36 hours and then treated with erastin (10 μmol/L) for 30 hours. αKG (10 mmol/L) or Glu (500 μmol/L) was added to culture medium at the same time as erastin. Bar graph in J, means ± SEM (n = 6). **, P < 0.01 versus siGLS2. SK Hep1 p53KO cells were transfected with Mock or GLS2 constructs for 24 hours and then cells were treated with DMSO, erastin (5 μmol/L or 10 μmol/L) alone, or erastin (5 μmol/L or 10 μmol/L) with AOA (5 mmol/L) for 24 hours. K and L, Cell morphology (K) and cell viability (L) were assayed for these treated cells. Scale bar, 100 μm (H and K). Bar graph in L, means ± SEM (n = 4). *, P < 0.05 versus GLS2. M and N, Cell viability of Hep3B cells that were transfected with Mock or GLS2 constructs for 24 hours and then cells were treated for 24 hours with either DMSO or ferroptosis inducers (erastin 10 μmol/L or RSL3 3 μmol/L) or indicated ferroptosis inducer with AOA (10 mmol/L; M) or erastin (10 μmol/L) in the absence or presence of the ferroptosis inhibitor (Fer-1 10 μmol/L, liproxstatin 2 μmol/L, or deferoxamine 50 μmol/L; N). Bar graphs in M and N, means ± SEM (n = 6). *, P < 0.05; **, P < 0.01 versus GLS2 for M or versus erastin in N. DFO, deferoxamine; h, hours.

Close modal

This exogenous GLS2 displayed pronounced colocalization with that of the mitochondrial-specific stain, MitoTracker, indicating that it resembled the functioning of endogenous GLS2 (Fig. 4E). Furthermore, overexpression of GLS2 modestly but significantly increased oxygen consumption while the silencing of GLS2 decreased oxygen consumption, which in turn reflects mitochondrial activity through the TCA cycle and electron transport (Fig. 4F). Thus, we hypothesized that GLS2 might be able to control the TCA cycle and thereby influence ferroptosis via αKG production that is catalyzed by GLS2. In line with this, when we performed metabolomic analysis with the GLS2-silenced SKHep1 cells that showed increased ferroptosis resistance (Fig. 3) had lower levels of αKG, and the ensuing ratio of αKG to glutamate was also diminished (Fig. 4G; Supplementary Fig. S7), even though changes in the levels of Glu and glutamine were subtle (Supplementary Fig. S7).

In line with the above hypothesis, under normal conditions, the observed modest decrease in viability of cells due to GLS2 silencing (Fig. 3B and G) was abolished by the addition of αKG or Glu, without affecting cellular morphology in both SKHep1 WT (Fig. 4H and I) and HepG2 cells (Fig. 4J; Supplementary Fig. S8A). The increased resistance to erastin-induced ferroptosis by GLS2 silencing in these cells was also lost upon the coaddition of either αKG or glutamate (Fig. 4HJ). While the addition of AOA, αKG, or glutamate also accordingly altered the lipid ROS levels in these cells (Supplementary Fig. S8B and S8C), the intracellular Fe2+ levels were not affected under any of these conditions (Supplementary Fig. S8D).

On the other hand, the overexpression of GLS2 in two human cancer cell lines lacking p53 expression (SKHep1 p53KO cells and p53-null hepatocellular carcinoma Hep3B cells) enhanced their sensitivity to ferroptosis at 24 hours after erastin treatment with a concomitant increase in the levels of lipid ROS (Fig. 4KM; Supplementary Fig. S8B and S8E–S8G). These effects of GLS2 overexpression were reduced by the administration of ferroptosis inhibitors Fer-1, Liproxstatin-1, deferoxamine (Fig. 4N), as well as AOA (62), which inhibits the conversion of glutamate to αKG (Fig. 4KM; Supplementary Fig. S8E and S8F). Note that while the overall cell survival even with the ferroptosis inhibitors is lower under GLS2 overexpression, these inhibitors provide a similar degree of increase in viability under both control and GLS2 overexpression conditions and the overall reduction in cell viability is probably a function of the stoichiometric relationship of the inhibitors with the death. In contrast, under normal conditions, GLS2-overexpressed cells showed a small but significant increase in viability as measured by ATP levels, but this increase was also nullified by the administration of AOA treatment (Fig. 4L and M). Note that the ability of iron chelator, deferoxamine, to suppress the effects of GLS2 overexpression similar to inhibitors of lipid peroxidation (Fer-1, Liproxstatin-1) further suggests that GLS2 affects molecular pathways involved ferroptosis that are downstream of the production or release of labile iron.

Based on these findings that the ferroptosis response of human cancer cells can be altered by modulating the levels of GLS2 and αKG, we propose that GLS2 induces lipid ROS-dependent ferroptosis by regulating mitochondrial function through production of αKG.

GLS2 responds to the induction of ferroptosis, modulates cell death, and suppresses tumorigenesis via its glutaminase core domain

The GLS2 protein consists of 603 amino acids and its glutaminase core domain is located between positions 177–463 (Supplementary Fig. S9A). To determine whether the glutaminase core domain is important for the ability of GLS2 to promote ferroptosis and tumor suppression in SKHep1 p53KO cells we constructed a mutant form of GLS2 that lacks residues 177–463 (GLS2del; Fig. 5A). When full-length GLS2wt or GLS2del were ectopically expressed in SKHep1 p53KO cells, as expected, only GLS2wt increased glutamine consumption and Glu production in the medium (Supplementary Fig. S9B and S9C). When we examined the OCR using the flux analyzer, the increase in oxygen consumption due to the expression of GLS2wt in SKHep1 p53KO cells was completely abolished when GLS2del was expressed instead (Fig. 5B). Notably, GLS2wt but not GLS2del suppressed colony formation in soft agar, cell invasion, and cell migration (Supplementary Fig. S9D–S9F). Thus, the glutaminase core domain is essential for the tumor suppressive functions of GLS2. Importantly as well, expression of GLS2del was unable to mimic the ability of GLS2wt expression to enhance erastin-induced ferroptosis (Fig. 5C; Supplementary Fig. S9G).

Figure 5.

GLS2 requires its core domain to promote ferroptosis and tumor suppression. A, Western blot analysis of GLS2 expression in SKHep1 p53KO cells transfected with p3 × FLAG-CMV10–empty vector (Mock) or p3 × FLAG-CMV10–hGLS2 vector (GLS2wt) or p3×FLAG-CMV10-hGLS2 177–463 deletion mutant vector (GLS2del). B, SKHep1 p53KO cells were transfected with Mock, GLS2wt, or GLS2del constructs for 48 hours and then OCR as recorded using a flux analyzer. The OCR was measured at baseline and after treatment with oligomycin, FCCP, and a mixture of antimycin and rotenone. C, SKHep1 p53KO cells were transfected with Mock, GLS2wt, or GLS2del constructs for 24 hours and then treated with erastin (5 μmol/L) for 24 hours to assay changes in cell viability. Bar graph depicts means ± SEM (n = 4). *, P < 0.05 versus GLS2wt. D, Xenograft tumors were obtained by subcutaneously injecting SKHep1 p53 KO cells that were transfected with either the GLS2wt lentivirus vector or CSII-EF-RfA-IRES2-Venus–empty vector (Mock) into SCID mice. Injections of indicated lentiviral vectors were performed on the right and left flanks of the same SCID mice. GFP expressed by the indicated lentivirus vectors was measured using the IVIS imaging system. E, Top, representative macroscopic findings of the results of the procedure shown in D. Bottom, volume of subcutaneous tumors obtained 6 weeks after injection of SKHep1 p53KO cells treated with Mock lentivirus vector on the left flank and either GLS2wt (left) or hGLS2del (right) on the right flank in SCID mice (n = 3). F, qRT-PCR analysis of ferroptosis markers Ptgs2 and Chac1 expression in subcutaneous tumors shown in E. G, Representative macroscopic findings and volume measurements of subcutaneous tumors in SCID mice (n = 4) that were obtained 6 weeks after injection of SKHep1 p53 KO cells treated with GLS2wt (left flank) and GLS2wt with ferroptosis inhibitor Fer-1 (right flank; left panel) and Mock lentivirus vector (left flank) and Mock with Fer-1 (right flank; right panel). H, qRT-PCR analysis of ferroptosis marker Ptgs2 expression in subcutaneous tumors shown in G. I, Representative macroscopic findings and volume measurements of subcutaneous tumors obtained 6 weeks after injection of SKHep1 cells treated with shGLS2 or shGLS1 (right flank; n = 4) and shCont lentivirus vector (left flank; n = 5). Bar graph in C and E to I, means ± SEM. *, P < 0.05.

Figure 5.

GLS2 requires its core domain to promote ferroptosis and tumor suppression. A, Western blot analysis of GLS2 expression in SKHep1 p53KO cells transfected with p3 × FLAG-CMV10–empty vector (Mock) or p3 × FLAG-CMV10–hGLS2 vector (GLS2wt) or p3×FLAG-CMV10-hGLS2 177–463 deletion mutant vector (GLS2del). B, SKHep1 p53KO cells were transfected with Mock, GLS2wt, or GLS2del constructs for 48 hours and then OCR as recorded using a flux analyzer. The OCR was measured at baseline and after treatment with oligomycin, FCCP, and a mixture of antimycin and rotenone. C, SKHep1 p53KO cells were transfected with Mock, GLS2wt, or GLS2del constructs for 24 hours and then treated with erastin (5 μmol/L) for 24 hours to assay changes in cell viability. Bar graph depicts means ± SEM (n = 4). *, P < 0.05 versus GLS2wt. D, Xenograft tumors were obtained by subcutaneously injecting SKHep1 p53 KO cells that were transfected with either the GLS2wt lentivirus vector or CSII-EF-RfA-IRES2-Venus–empty vector (Mock) into SCID mice. Injections of indicated lentiviral vectors were performed on the right and left flanks of the same SCID mice. GFP expressed by the indicated lentivirus vectors was measured using the IVIS imaging system. E, Top, representative macroscopic findings of the results of the procedure shown in D. Bottom, volume of subcutaneous tumors obtained 6 weeks after injection of SKHep1 p53KO cells treated with Mock lentivirus vector on the left flank and either GLS2wt (left) or hGLS2del (right) on the right flank in SCID mice (n = 3). F, qRT-PCR analysis of ferroptosis markers Ptgs2 and Chac1 expression in subcutaneous tumors shown in E. G, Representative macroscopic findings and volume measurements of subcutaneous tumors in SCID mice (n = 4) that were obtained 6 weeks after injection of SKHep1 p53 KO cells treated with GLS2wt (left flank) and GLS2wt with ferroptosis inhibitor Fer-1 (right flank; left panel) and Mock lentivirus vector (left flank) and Mock with Fer-1 (right flank; right panel). H, qRT-PCR analysis of ferroptosis marker Ptgs2 expression in subcutaneous tumors shown in G. I, Representative macroscopic findings and volume measurements of subcutaneous tumors obtained 6 weeks after injection of SKHep1 cells treated with shGLS2 or shGLS1 (right flank; n = 4) and shCont lentivirus vector (left flank; n = 5). Bar graph in C and E to I, means ± SEM. *, P < 0.05.

Close modal

We then created a xenograft model with the SKHEP1 p53KO cells to express the two variants of GLS2 using lentiviral vectors (Fig. 5D). In this setting, GLS2wt-expressing cells resulted in significantly smaller tumors at 6 weeks after subcutaneous injection into SCID mice when compared with injections of either control (Mock) cells or cells expressing GLS2del (Fig. 5E and F). The tumor growth inhibitory effects of GLS2wt were accompanied by an increase in the expression of ferroptosis markers in SCID mice, which was lost if replaced by GLS2del (Fig. 5F). Of key importance, the growth suppressive abilities of GLS2wt cells as well as the induction of ferroptosis markers were abolished by treatment with the ferroptosis inhibitor Fer-1 (Fig. 5G and H). In addition, we confirmed that shGLS2-expressing SKHep1 cells (also created using lentivral constructs) resulted in larger tumors in SCID mice, while cells expressing shGLS1 actually resulted in smaller tumors compared with the control (shCont; Fig. 5I). These data strongly support the likelihood that GLS2 exerts its antitumor effect through its ability to promote ferroptosis and this requires the GLS2 glutaminase core domain.

HCC downregulate GLS2 levels, which are correlated with the survival rates of patients with HCC

To further address the physiologic role of GLS2 in HCC and infer clinical relationships, we examined its expression profiles in livers of STAM mice that had developed HCC, which revealed that Gls2 mRNA and protein levels were actually decreased at this stage (Fig. 6A and B). Reduced expression of Gls2 was likely due to the hypermethylation of CpG islands in the Gls2 promoter that we assessed, and which in turn likely resulted in the downregulation of Gls2 expression in mouse HCC (Supplementary Fig. S10A and S10B). Immunohistologic analysis demonstrated that Gls2 protein levels were also decreased in the HCC samples compared to the non-HCC samples even though p53 levels themselves were increased in HCC (Fig. 6B and C). It is possible that the increased levels of p53 protein resulted from the combination of streptozotocin and high-fat diet to which the mice were subjected.

Figure 6.

GLS2 expression levels are decreased in both mice and human HCC and these levels are correlated with malignancy and poor prognosis. A, qRT-PCR analysis of Gls2 expression in liver tissues from HCC stages (22 weeks, n = 6) compared with normal stage (6 weeks, n = 5) in WT-STAM. Bar graph, means ± SEM. *, P < 0.05 versus the normal stage. B, Western blot analysis of Gls2 or p53 (Ab3) protein expression in the liver from HCC and normal stages in WT-STAM mice. C, IHC analysis of Gls2 and p53 (Pab240) in liver from WT-STAM, mice containing HCC and non-HCC regions. D,GLS2, PTGS2, GLS1, or p53 mRNA expression levels in human HCC samples (HCC) compared with normal liver tissues (Normal) were assessed using TCGA database and the GDC portal (https://portal.gdc.cancer.gov/). E, The numbers of HCC or normal samples from patients in TCGA database were categorized based on GLS2 expression levels as GLS2-low (below the median) or GLS2-high (over the median). F,PTGS2 mRNA expression level was assessed in GLS2 low or GLS2 patients. G and H, Clinical (age at initial pathological diagnosis; G) and pathologic data (histological grades; G1, G2, G3, or G4; H) in human HCC according to GLS2-low (n = 205) and GLS2-high (n = 205) were extracted from TCGA. I, Kaplan—Meier analyses were performed using KM-plotter database (kmplot.com/analysis/). The graph represents survival curves of patients stratified according to GLS2-low (black) and GLS2-high (red) in human HCC, lung cancer, and breast cancer samples. J, COSMIC analysis in 3,256 samples of human HCC. The mutation subtypes of GLS2 from the COSMIC database are shown. Number of mutations for each subtype is shown in parentheses. NS, not significant.

Figure 6.

GLS2 expression levels are decreased in both mice and human HCC and these levels are correlated with malignancy and poor prognosis. A, qRT-PCR analysis of Gls2 expression in liver tissues from HCC stages (22 weeks, n = 6) compared with normal stage (6 weeks, n = 5) in WT-STAM. Bar graph, means ± SEM. *, P < 0.05 versus the normal stage. B, Western blot analysis of Gls2 or p53 (Ab3) protein expression in the liver from HCC and normal stages in WT-STAM mice. C, IHC analysis of Gls2 and p53 (Pab240) in liver from WT-STAM, mice containing HCC and non-HCC regions. D,GLS2, PTGS2, GLS1, or p53 mRNA expression levels in human HCC samples (HCC) compared with normal liver tissues (Normal) were assessed using TCGA database and the GDC portal (https://portal.gdc.cancer.gov/). E, The numbers of HCC or normal samples from patients in TCGA database were categorized based on GLS2 expression levels as GLS2-low (below the median) or GLS2-high (over the median). F,PTGS2 mRNA expression level was assessed in GLS2 low or GLS2 patients. G and H, Clinical (age at initial pathological diagnosis; G) and pathologic data (histological grades; G1, G2, G3, or G4; H) in human HCC according to GLS2-low (n = 205) and GLS2-high (n = 205) were extracted from TCGA. I, Kaplan—Meier analyses were performed using KM-plotter database (kmplot.com/analysis/). The graph represents survival curves of patients stratified according to GLS2-low (black) and GLS2-high (red) in human HCC, lung cancer, and breast cancer samples. J, COSMIC analysis in 3,256 samples of human HCC. The mutation subtypes of GLS2 from the COSMIC database are shown. Number of mutations for each subtype is shown in parentheses. NS, not significant.

Close modal

In agreement with these results obtained from mice, examination of The Cancer Genome Atlas (TCGA) database revealed that the mRNA levels of both GLS2 and the ferroptosis marker PTGS2 are significantly lower in human patients with HCC compared with normal patients, whereas GLS1 mRNA levels are slightly higher in HCC, even though levels of p53 mRNA were similar between the two groups (Fig. 6D). To examine the relationship between GLS2 expression and clinical characteristics, we divided the patients into two groups based on the median of GLS2 expression: i.e., a low group (below the median) and a high group (over the median). The group with low GLS2 expression contains only patients with HCC (Fig. 6E) and shows slight but significant reduction in PTGS2 expression compared with the group of patients with high GLS2 expression (Fig. 6F). Among patients with HCC, there was no significant difference in age at initial pathologic diagnosis between groups with low and high GLS2 levels (Fig. 6G). On the other hand, the G3 and G4 histologic grades of HCC (where G1–G4 extend from lower to higher, respectively) were significantly higher in the group with low GLS2 (Fig. 6H). Importantly as well, Kaplan–Meier survival analysis using the KM-plotter database demonstrated that low expression levels of GLS2 are significantly correlated with poorer survival outcome in patients with not only HCC but also those with lung and breast cancers (Fig. 6I).

Furthermore, using the cBioPortal Meta analysis tool, we performed cross-cancer analysis of GLS2 gene for amplifications, deletions, and mutations, which indicated that while a large number of tumor types mainly harbored amplifications, the deletions were also present in breast cancer and hepatobiliary cancer (Supplementary Fig. S11). Besides, using the Catalogue Of Somatic Mutations In Cancer (COSMIC) database showed we determined that GLS2 harbors a relatively high rate of loss-of-function mutations (nonsense, inframe insertion, frameshift insertion, inframe deletion, and frameshift deletion) in human HCCs (Fig. 6J).

Taken together, these results indicate that GLS2 is downregulated in HCC and this is associated with increased histological malignancy and decreased survival.

In this study, we have used Gls2 KO mice and human cancer cells to evaluate the possiblity that GLS2 is a suppressor of HCC, and that protection by GLS2 from such tumorigenesis involves its role in promoting ferroptosis. We show that Gls2 KO mice have a marked propensity to develop late HCCs. Further, by subjecting mice to the STAM diet, ensuing HCC is exacerbated in the absence of GLS2. In concert with this and extending previous studies linking GLS2 to ferroptosis (24), hepatocytes from Gls2 KO mice are more resistant to experimentally induced ferroptosis. In fact, not only cancer cells but also hepatocytes from mice display reduced ferroptosis when GLS2 is not expressed. These findings extend to human cancer cells derived from HCC where we demonstrate that modulation of GLS2 levels leads to correspondingly altered extents of ferroptosis. Most importantly, ferroptosis is required to suppress tumors derived from cells expressing WT GLS2 from forming in mice. Finally we show that human HCC cancer patients’ tumors as well as HCCs from mice display reduced levels of GLS2. Taken together these data strongly support a role for GLS2 in ferroptosis and consequent tumor suppression.

Increased oxidative stress or involvement of ferroptosis has been previously postulated as a mechanism of HCC development in Gls2 KO mice (63–65), and our data indicate that ferroptosis may indeed be involved in preventing cancer onset and progression. Although Gao and colleagues (24) first reported that glutaminolysis and transferrin regulate ferroptosis, the mechanism has not been fully elucidated. We show here that of the two major glutaminases, GLS2 but not GLS1 can modulate ferroptosis. While there is only low homology between the N- and C-terminal regions of these two proteins, there is a high degree of homology in their respective glutaminase core domains, and we have found the GLS2 core domain to be necessary for the role of GLS2 in ferroptosis. A possible explanation for this difference might be that GLS2 and GLS1 have different subcellular localizations; GLS2 is found in mitochondria while GLS1 is present in the cytoplasm (66). Interestingly, however, in previous reports (51), mitochondrial ROS does not increase during ferroptosis induction, even though mitochondria is the central organelle where significant amounts of cellular ROS are generated in response to stress.

As another explanation, the mitochondria might be important through contributing to the production of lipid peroxides or for the formation of microsomal membranes from membrane phospholipids that are involved in ferroptotic death. Several mitochondrial genes have been found to be associated with ferroptosis, and it is possible that peroxidation of cardiolipin, a mitochondria-specific phospholipid, links mitochondrial lipid peroxidation to ferroptosis (30, 67, 68). Conversely, there are also reports that the removal of mitochondrial DNA (51) or the removal of mitochondria itself (69) does not prevent ferroptosis. By way of reconciling these disparate roles, we speculate that since energy metabolic pathways are all interconnected, energy metabolism in mitochondria may be compensated in some contexts but not others.

We propose that in the context that we have studied, the primary role of mitochondria might be involved in the energy homeostasis that is associated with ferroptosis. In accordance with our hypothesis, a recent report provided evidence that TCA metabolites that are downstream of αKG, such as succinate, fumarate, and malate, can cause the hyperpolarization of the mitochondrial membrane, which is associated with cysteine deprivation–induced lipid ROS accumulation and ferroptosis (57). We suggest that GLS2 is involved in lipid ROS accumulation through αKG-dependent activation of the TCA cycle and electron transport in mitochondia, where most ROS are produced and phospholipids are also abundant, and this is important for ferroptosis induction.

As GLS2 expression was reduced in HCC through hypermethylation of its promoter region, a chemical modulator of αKG or glutaminolysis might have a strong therapeutic potential via ferroptosis induction for these cancers. This is in line with the finding that combining the ferroptosis inducer, erastin with chemotherapeutic drugs such as cytarabine/ara-C, cisplatin, doxorubicin/adriamycin, and temozolomide have shown a remarkable synergistic effect in antitumor activities (27).

In Gls2 KO mice, the induction of ferroptosis was reduced in cells and tissues related to the liver that we studied and mice were afflicted primarily with HCC and lymphomas. But we do not know if other tissues were also defective in undergoing ferroptosis. As GLS2 is highly expressed in liver, this presents the possiblity that there is a somewhat unique dependence on this protein for suppression of cancer through ferroptosis in this tissue.

However, while here we have mainly focused on hepatocellular carcinoma, other cancers that might potentially be suppressed due to ferroptosis, such as lymphoid tumors (53, 70–72), which might also have a requirement for GLS2 in driving ferroptosis sensitivity. Furthermore, while it is intriguing that Gls2 KO mice develop liver cancer at a late age of 120 weeks, whether GLS2 can also be involved in other ferroptosis-associated diseases such as reperfusion injury and neuronal death (24, 73) is of considerable interest for future studies.

B.R. Stockwell reports grants from NCI/NIH; and grants from NCI/NIH during the conduct of the study; personal fees from Inzen Therapeutics, Exarta Therapeutics, ProJenX Inc., Weatherwax Biotechnologies Corporation; and personal fees from Akin Gump Strauss Hauer & Feld LLP outside the submitted work; in addition, B.R. Stockwell has a patent for Ferroptosis and drug discovery pending and issued. C. Prives reports grants from NCI/NIH and personal fees from Aileron Therapeutics and CPRIT. No disclosures were reported by the other authors.

S. Suzuki: Conceptualization, formal analysis, investigation, methodology, writing–original draft. D. Venkatesh: Conceptualization, supervision, project administration, writing–review and editing. H. Kanda: Investigation, dr kanda performed extensive pathological analysis of the gls2 ko mice. A. Nakayama: Formal analysis. H. Hosokawa: Investigation. E. Lee: Investigation. T. Miki: Investigation. B.R. Stockwell: Supervision, writing–review and editing. K. Yokote: Conceptualization. T. Tanaka: Conceptualization, funding acquisition. C. Prives: Conceptualization, supervision, funding acquisition, writing–review and editing.

The authors are grateful to Yuko Yamagata, Erika Sugawara, and Masanori Fujimoto of Chiba University as well as Ella Freulich at Columbia University (New York, NY) for their technical assistance. The authors also want to thank Ichiro Tatsuno of Chiba Prefectural University of Health Sciences, Atsushi Kaneda of Chiba University and Satoshi Inoue of Saitama Medical University (Saitama, Japan) for his academic advice.

This research was supported by Grants-in-Aid for Scientific Research (C) 26430105 and 17K09875, International Joint Research Program no. 16KK0197, the Takeda Science Foundation, the Kowa Life Science Foundation, the Banyu Foundation, and the Kato Memorial Bioscience Foundation. This work was supported by JSPS KAKENHI (grant number JP 16H06276; AdAMS) as well as CA87497 to C. Prives and B.R. Stockwell.

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.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Yang
L
,
Venneti
S
,
Nagrath
D
.
Glutaminolysis: a hallmark of cancer metabolism
.
Annu Rev Biomed Eng
2017
;
19
:
163
94
.
2.
Lamonte
G
,
Tang
X
,
Chen
JL
,
Wu
J
,
Ding
CK
,
Keenan
MM
, et al
.
Acidosis induces reprogramming of cellular metabolism to mitigate oxidative stress
.
Cancer Metab
2013
;
1
:
23
.
3.
Araujo
L
,
Khim
P
,
Mkhikian
H
,
Mortales
CL
,
Demetriou
M
.
Glycolysis and glutaminolysis cooperatively control T cell function by limiting metabolite supply to N-glycosylation
.
Elife
2017
;
6
:
e21330
.
4.
Chung-Bok
MI
,
Vincent
N
,
Jhala
U
,
Watford
M
.
Rat hepatic glutaminase: identification of the full coding sequence and characterization of a functional promoter
.
Biochem J
1997
;
324
(
Pt 1
):
193
200
.
5.
Gao
P
,
Tchernyshyov
I
,
Chang
TC
,
Lee
YS
,
Kita
K
,
Ochi
T
, et al
.
c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism
.
Nature
2009
;
458
:
762
5
.
6.
Wang
JB
,
Erickson
JW
,
Fuji
R
,
Ramachandran
S
,
Gao
P
,
Dinavahi
R
, et al
.
Targeting mitochondrial glutaminase activity inhibits oncogenic transformation
.
Cancer Cell
2010
;
18
:
207
19
.
7.
Kahlert
UD
,
Cheng
M
,
Koch
K
,
Marchionni
L
,
Fan
X
,
Raabe
EH
, et al
.
Alterations in cellular metabolome after pharmacological inhibition of Notch in glioblastoma cells
.
Int J Cancer
2016
;
138
:
1246
55
.
8.
Suzuki
S
,
Tanaka
T
,
Poyurovsky
MV
,
Nagano
H
,
Mayama
T
,
Ohkubo
S
, et al
.
Phosphate-activated glutaminase (GLS2), a p53-inducible regulator of glutamine metabolism and reactive oxygen species
.
Proc Natl Acad Sci U S A
2010
;
107
:
7461
6
.
9.
Hu
W
,
Zhang
C
,
Wu
R
,
Sun
Y
,
Levine
A
,
Feng
Z
.
Glutaminase 2, a novel p53 target gene regulating energy metabolism and antioxidant function
.
Proc Natl Acad Sci U S A
2010
;
107
:
7455
60
.
10.
Luengo
A
,
Gui
DY
,
Vander Heiden
MG
.
Targeting metabolism for cancer therapy
.
Cell Chem Biol
2017
;
24
:
1161
80
.
11.
Gross
MI
,
Demo
SD
,
Dennison
JB
,
Chen
L
,
Chernov-Rogan
T
,
Goyal
B
, et al
.
Antitumor activity of the glutaminase inhibitor CB-839 in triple-negative breast cancer
.
Mol Cancer Ther
2014
;
13
:
890
901
.
12.
Jacque
N
,
Ronchetti
AM
,
Larrue
C
,
Meunier
G
,
Birsen
R
,
Willems
L
, et al
.
Targeting glutaminolysis has antileukemic activity in acute myeloid leukemia and synergizes with BCL-2 inhibition
.
Blood
2015
;
126
:
1346
56
.
13.
Choi
YK
,
Park
KG
.
Targeting glutamine metabolism for cancer treatment
.
Biomol Ther
2018
;
26
:
19
28
.
14.
Sheikh
TN
,
Patwardhan
PP
,
Cremers
S
,
Schwartz
GK
.
Targeted inhibition of glutaminase as a potential new approach for the treatment of NF1 associated soft tissue malignancies
.
Oncotarget
2017
;
8
:
94054
68
.
15.
Mates
JM
,
Campos-Sandoval
JA
,
de
L
Santos-Jimenez
J
,
Segura
JA
,
Alonso
FJ
, et al
.
Metabolic reprogramming of cancer by chemicals that target glutaminase isoenzymes
.
Curr Med Chem
2019
;
27
:
5317
-
5339
.
16.
Le
A
,
Lane
AN
,
Hamaker
M
,
Bose
S
,
Gouw
A
,
Barbi
J
, et al
.
Glucose-independent glutamine metabolism via TCA cycling for proliferation and survival in B cells
.
Cell Metab
2012
;
15
:
110
21
.
17.
Lobo
C
,
Ruiz-Bellido
MA
,
Aledo
JC
,
Marquez
J
,
Nunez De Castro
I
,
Alonso
FJ
.
Inhibition of glutaminase expression by antisense mRNA decreases growth and tumourigenicity of tumour cells
.
Biochem J
2000
;
348 Pt 2
:
257
61
.
18.
Zhang
C
,
Liu
J
,
Zhao
Y
,
Yue
X
,
Zhu
Y
,
Wang
X
, et al
.
Glutaminase 2 is a novel negative regulator of small GTPase Rac1 and mediates p53 function in suppressing metastasis
.
Elife
2016
;
5
:
e10727
.
19.
Liu
J
,
Zhang
C
,
Lin
M
,
Zhu
W
,
Liang
Y
,
Hong
X
, et al
.
Glutaminase 2 negatively regulates the PI3K/AKT signaling and shows tumor suppression activity in human hepatocellular carcinoma
.
Oncotarget
2014
;
5
:
2635
47
.
20.
Szeliga
M
,
Zgrzywa
A
,
Obara-Michlewska
M
,
Albrecht
J
.
Transfection of a human glioblastoma cell line with liver-type glutaminase (LGA) down-regulates the expression of DNA-repair gene MGMT and sensitizes the cells to alkylating agents
.
J Neurochem
2012
;
123
:
428
36
.
21.
Martin-Rufian
M
,
Nascimento-Gomes
R
,
Higuero
A
,
Crisma
AR
,
Campos-Sandoval
JA
,
Gomez-Garcia
MC
, et al
.
Both GLS silencing and GLS2 overexpression synergize with oxidative stress against proliferation of glioma cells
.
J Mol Med
2014
;
92
:
277
90
.
22.
Lee
YZ
,
Yang
CW
,
Chang
HY
,
Hsu
HY
,
Chen
IS
,
Chang
HS
, et al
.
Discovery of selective inhibitors of Glutaminase-2, which inhibit mTORC1, activate autophagy and inhibit proliferation in cancer cells
.
Oncotarget
2014
;
5
:
6087
101
.
23.
Xiang
L
,
Xie
G
,
Liu
C
,
Zhou
J
,
Chen
J
,
Yu
S
, et al
.
Knock-down of glutaminase 2 expression decreases glutathione, NADH, and sensitizes cervical cancer to ionizing radiation
.
Biochim Biophys Acta
2013
;
1833
:
2996
3005
.
24.
Gao
M
,
Monian
P
,
Quadri
N
,
Ramasamy
R
,
Jiang
X
.
Glutaminolysis and transferrin regulate ferroptosis
.
Mol Cell
2015
;
59
:
298
308
.
25.
Shen
Z
,
Song
J
,
Yung
BC
,
Zhou
Z
,
Wu
A
,
Chen
X
.
Emerging strategies of cancer therapy based on ferroptosis
.
Adv Mater
2018
;
30
:
e1704007
.
26.
Lu
B
,
Chen
XB
,
Ying
MD
,
He
QJ
,
Cao
J
,
Yang
B
.
The role of ferroptosis in cancer development and treatment response
.
Front Pharmacol
2017
;
8
:
992
.
27.
Mou
Y
,
Wang
J
,
Wu
J
,
He
D
,
Zhang
C
,
Duan
C
, et al
.
Ferroptosis, a new form of cell death: opportunities and challenges in cancer
.
J Hematol Oncol
2019
;
12
:
34
.
28.
Jennis
M
,
Kung
CP
,
Basu
S
,
Budina-Kolomets
A
,
Leu
JI
,
Khaku
S
, et al
.
An African-specific polymorphism in the TP53 gene impairs p53 tumor suppressor function in a mouse model
.
Genes Dev
2016
;
30
:
918
30
.
29.
Yang
WS
,
Stockwell
BR
.
Ferroptosis: death by lipid peroxidation
.
Trends Cell Biol
2016
;
26
:
165
76
.
30.
Xie
Y
,
Hou
W
,
Song
X
,
Yu
Y
,
Huang
J
,
Sun
X
, et al
.
Ferroptosis: process and function
.
Cell Death Differ
2016
;
23
:
369
79
.
31.
Skarnes
WC
,
Rosen
B
,
West
AP
,
Koutsourakis
M
,
Bushell
W
,
Iyer
V
, et al
.
A conditional knockout resource for the genome-wide study of mouse gene function
.
Nature
2011
;
474
:
337
42
.
32.
Venkatesh
D
,
O'Brien
NA
,
Zandkarimi
F
,
Tong
DR
,
Stokes
ME
,
Dunn
DE
, et al
.
MDM2 and MDMX promote ferroptosis by PPARα-mediated lipid remodeling
.
Genes Dev
2020
;
34
:
526
43
.
33.
Larraufie
MH
,
Yang
WS
,
Jiang
E
,
Thomas
AG
,
Slusher
BS
,
Stockwell
BR
.
Incorporation of metabolically stable ketones into a small molecule probe to increase potency and water solubility
.
Bioorg Med Chem Lett
2015
;
25
:
4787
92
.
34.
Honda
A
,
Hirose
M
,
Hatori
M
,
Matoba
S
,
Miyoshi
H
,
Inoue
K
, et al
.
Generation of induced pluripotent stem cells in rabbits: potential experimental models for human regenerative medicine
.
J Biol Chem
2010
;
285
:
31362
9
.
35.
Hirayama
T
.
Fluorescent probes for the detection of catalytic Fe(II) ion. free radic
Biol Med
2019
;
133
:
38
45
.
36.
Kanai
A
,
Suzuki
K
,
Tanimoto
K
,
Mizushima-Sugano
J
,
Suzuki
Y
,
Sugano
S
.
Characterization of STAT6 target genes in human B cells and lung epithelial cells
.
DNA Res
2011
;
18
:
379
92
.
37.
Mortazavi
A
,
Williams
BA
,
McCue
K
,
Schaeffer
L
,
Wold
B
.
Mapping and quantifying mammalian transcriptomes by RNA-Seq
.
Nat Methods
2008
;
5
:
621
8
.
38.
Miki
T
,
Nagashima
K
,
Tashiro
F
,
Kotake
K
,
Yoshitomi
H
,
Tamamoto
A
, et al
.
Defective insulin secretion and enhanced insulin action in KATP channel-deficient mice
.
Proc Natl Acad Sci U S A
1998
;
95
:
10402
6
.
39.
Yoshimoto
S
,
Loo
TM
,
Atarashi
K
,
Kanda
H
,
Sato
S
,
Oyadomari
S
, et al
.
Obesity-induced gut microbial metabolite promotes liver cancer through senescence secretome
.
Nature
2013
;
499
:
97
101
.
40.
Chen
Y
.
Scratch wound healing assay
.
Bio-protocol.
2012
;
2
:
e100
.
41.
Loo
TM
,
Kamachi
F
,
Watanabe
Y
,
Yoshimoto
S
,
Kanda
H
,
Arai
Y
, et al
.
Gut microbiota promotes obesity-associated liver cancer through PGE2-mediated suppression of antitumor immunity
.
Cancer Discov
2017
;
7
:
522
38
.
42.
Tokunaga
Y
,
Osawa
Y
,
Ohtsuki
T
,
Hayashi
Y
,
Yamaji
K
,
Yamane
D
, et al
.
Selective inhibitor of Wnt/beta-catenin/CBP signaling ameliorates hepatitis C virus-induced liver fibrosis in mouse model
.
Sci Rep
2017
;
7
:
325
.
43.
Sakuma
I
,
Higuchi
S
,
Fujimoto
M
,
Takiguchi
T
,
Nakayama
A
,
Tamura
A
, et al
.
Cushing syndrome due to ACTH-secreting pheochromocytoma, aggravated by glucocorticoid-driven positive-feedback loop
.
J Clin Endocrinol Metab
2016
;
101
:
841
6
.
44.
Fujii
M
,
Shibazaki
Y
,
Wakamatsu
K
,
Honda
Y
,
Kawauchi
Y
,
Suzuki
K
, et al
.
A murine model for non-alcoholic steatohepatitis showing evidence of association between diabetes and hepatocellular carcinoma
.
Med Mol Morphol
2013
;
46
:
141
52
.
45.
Siveski-Iliskovic
N
,
Kaul
N
,
Singal
PK
.
Probucol promotes endogenous antioxidants and provides protection against adriamycin-induced cardiomyopathy in rats
.
Circulation
1994
;
89
:
2829
35
.
46.
Zhang
X
,
Li
Z
,
Liu
D
,
Xu
X
,
Shen
W
,
Mei
Z
.
Effects of probucol on hepatic tumor necrosis factor-alpha, interleukin-6 and adiponectin receptor-2 expression in diabetic rats
.
J Gastroenterol Hepatol
2009
;
24
:
1058
63
.
47.
Merat
S
,
Malekzadeh
R
,
Sohrabi
MR
,
Hormazdi
M
,
Naserimoghadam
S
,
Mikaeli
J
, et al
.
Probucol in the treatment of nonalcoholic steatohepatitis: an open-labeled study
.
J Clin Gastroenterol
2003
;
36
:
266
8
.
48.
Stockwell
BR
,
Friedmann Angeli
JP
,
Bayir
H
,
Bush
AI
,
Conrad
M
,
Dixon
SJ
, et al
.
Ferroptosis: a regulated cell death nexus linking metabolism, redox biology, and disease
.
Cell.
2017
;
171
:
273
85
.
49.
Hirschhorn
T
,
Stockwell
BR
.
The development of the concept of ferroptosis
.
Free Radic Biol Med
2019
;
133
:
130
43
.
50.
Bueno
DC
,
Canto
RFS
,
de Souza
V
,
Andreguetti
RR
,
Barbosa
FAR
,
Naime
AA
, et al
.
New probucol analogues inhibit ferroptosis, improve mitochondrial parameters, and induce glutathione peroxidase in HT22 cells
.
Mol Neurobiol
2020
;
57
:
3273
90
.
51.
Dixon
SJ
,
Lemberg
KM
,
Lamprecht
MR
,
Skouta
R
,
Zaitsev
EM
,
Gleason
CE
, et al
.
Ferroptosis: an iron-dependent form of nonapoptotic cell death
.
Cell
2012
;
149
:
1060
72
.
52.
Dixon
SJ
,
Patel
DN
,
Welsch
M
,
Skouta
R
,
Lee
ED
,
Hayano
M
, et al
.
Pharmacological inhibition of cystine-glutamate exchange induces endoplasmic reticulum stress and ferroptosis
.
Elife
2014
;
3
:
e02523
.
53.
Yang
WS
,
SriRamaratnam
R
,
Welsch
ME
,
Shimada
K
,
Skouta
R
,
Viswanathan
VS
, et al
.
Regulation of ferroptotic cancer cell death by GPX4
.
Cell
2014
;
156
:
317
31
.
54.
Gao
M
,
Monian
P
,
Pan
Q
,
Zhang
W
,
Xiang
J
,
Jiang
X
.
Ferroptosis is an autophagic cell death process
.
Cell Res
2016
;
26
:
1021
32
.
55.
Liu
J
,
Kuang
F
,
Kroemer
G
,
Klionsky
DJ
,
Kang
R
,
Tang
D
.
Autophagy-dependent ferroptosis: machinery and regulation
.
Cell Chem Biol
2020
;
27
:
420
35
.
56.
Drummen
GP
,
van Liebergen
LC
,
Op den Kamp
JA
,
Post
JA
.
C11-BODIPY(581/591), an oxidation-sensitive fluorescent lipid peroxidation probe: (micro)spectroscopic characterization and validation of methodology
.
Free Radic Biol Med
2002
;
33
:
473
90
.
57.
Gao
M
,
Yi
J
,
Zhu
J
,
Minikes
AM
,
Monian
P
,
Thompson
CB
, et al
.
Role of mitochondria in ferroptosis
.
Mol Cell
2019
;
73
:
354
63
.
58.
Moon
SH
,
Huang
CH
,
Houlihan
SL
,
Regunath
K
,
Freed-Pastor
WA
,
Morris
JPt
, et al
.
p53 represses the mevalonate pathway to mediate tumor suppression
.
Cell
2019
;
176
:
564
80
.
59.
Jiang
L
,
Kon
N
,
Li
T
,
Wang
SJ
,
Su
T
,
Hibshoosh
H
, et al
.
Ferroptosis as a p53-mediated activity during tumour suppression
.
Nature
2015
;
520
:
57
62
.
60.
Wang
SJ
,
Li
D
,
Ou
Y
,
Jiang
L
,
Chen
Y
,
Zhao
Y
, et al
.
Acetylation is crucial for p53-mediated ferroptosis and tumor suppression
.
Cell Rep
2016
;
17
:
366
73
.
61.
Murphy
ME
.
Ironing out how p53 regulates ferroptosis
.
Proc Natl Acad Sci U S A
2016
;
113
:
12350
2
.
62.
Zhang
K
,
Wu
L
,
Zhang
P
,
Luo
M
,
Du
J
,
Gao
T
, et al
.
miR-9 regulates ferroptosis by targeting glutamic-oxaloacetic transaminase GOT1 in melanoma
.
Mol Carcinog
2018
;
57
:
1566
76
.
63.
Nie
J
,
Lin
B
,
Zhou
M
,
Wu
L
,
Zheng
T
.
Role of ferroptosis in hepatocellular carcinoma
.
J Cancer Res Clin Oncol
2018
;
144
:
2329
37
.
64.
Xia
X
,
Fan
X
,
Zhao
M
,
Zhu
P
.
The relationship between ferroptosis and tumors: a novel landscape for therapeutic approach
.
Curr Gene Ther
2019
;
19
:
117
24
.
65.
Yu
H
,
Guo
P
,
Xie
X
,
Wang
Y
,
Chen
G
.
Ferroptosis, a new form of cell death, and its relationships with tumourous diseases
.
J Cell Mol Med
2017
;
21
:
648
57
.
66.
Cassago
A
,
Ferreira
AP
,
Ferreira
IM
,
Fornezari
C
,
Gomes
ER
,
Greene
KS
, et al
.
Mitochondrial localization and structure-based phosphate activation mechanism of Glutaminase C with implications for cancer metabolism
.
Proc Natl Acad Sci U S A
2012
;
109
:
1092
7
.
67.
Krainz
T
,
Gaschler
MM
,
Lim
C
,
Sacher
JR
,
Stockwell
BR
,
Wipf
P
.
A mitochondrial-targeted nitroxide is a potent inhibitor of ferroptosis
.
ACS Cent Sci
2016
;
2
:
653
9
.
68.
Ji
J
,
Baart
S
,
Vikulina
AS
,
Clark
RS
,
Anthonymuthu
TS
,
Tyurin
VA
, et al
.
Deciphering of mitochondrial cardiolipin oxidative signaling in cerebral ischemia-reperfusion
.
J Cereb Blood Flow Metab
2015
;
35
:
319
28
.
69.
Gaschler
MM
,
Hu
F
,
Feng
H
,
Linkermann
A
,
Min
W
,
Stockwell
BR
.
Determination of the subcellular localization and mechanism of action of ferrostatins in suppressing ferroptosis
.
ACS Chem Biol
2018
;
13
:
1013
20
.
70.
Iglehart
JK
,
York
RM
,
Modest
AP
,
Lazarus
H
,
Livingston
DM
.
Cystine requirement of continuous human lymphoid cell lines of normal and leukemic origin
.
J Biol Chem
1977
;
252
:
7184
91
.
71.
Gout
PW
,
Buckley
AR
,
Simms
CR
,
Bruchovsky
N
.
Sulfasalazine, a potent suppressor of lymphoma growth by inhibition of the x(c)- cystine transporter: a new action for an old drug
.
Leukemia
2001
;
15
:
1633
40
.
72.
Gout
PW
,
Simms
CR
,
Robertson
MC
.
In vitro studies on the lymphoma growth-inhibitory activity of sulfasalazine
.
Anticancer Drugs
2003
;
14
:
21
9
.
73.
Do Van
B
,
Gouel
F
,
Jonneaux
A
,
Timmerman
K
,
Gele
P
,
Petrault
M
, et al
.
Ferroptosis, a newly characterized form of cell death in Parkinson's disease that is regulated by PKC
.
Neurobiol Dis
2016
;
94
:
169
78
.

Supplementary data