Metabolic Adaptation to Nutritional Stress in Human Colorectal Cancer

Tumor cells respond to their microenvironment, which can include hypoxia and malnutrition, and adapt their metabolism to survive and grow. Some oncogenes are associated with cancer metabolism via regulation of the related enzymes or transporters. However, the importance of metabolism and precise metabolic effects of oncogenes in colorectal cancer remain unclear. We found that colorectal cancer cells survived under the condition of glucose depletion, and their resistance to such conditions depended on genomic alterations rather than on KRAS mutation alone. Metabolomic analysis demonstrated that those cells maintained tricarboxylic acid cycle activity and ATP production under such conditions. Furthermore, we identified pivotal roles of GLUD1 and SLC25A13 in nutritional stress. GLUD1 and SLC25A13 were associated with tumor aggressiveness and poorer prognosis of colorectal cancer. In conclusion, GLUD1 and SLC25A13 may serve as new targets in treating refractory colorectal cancer which survive in malnutritional microenvironments.

Scientific RepoRts | 6:38415 | DOI: 10.1038/srep38415 Furthermore, there is a non-canonical pathway of glutamine in pancreatic ductal adenocarcinoma cells that is regulated by the KRAS oncogene 12 . However, the importance of glutamine metabolism and precise metabolic effects of oncogenes in colorectal cancer cells remain unknown.
The aim of this study is to elucidate metabolic adaptation to nutritional stress and the role of the involved oncogenes in human colorectal cancer. The present study showed that the metabolism of colorectal cancer, distinct from that of pancreatic cancer, depended on genomic alterations, which previously have been uncharacterized and was not restricted to KRAS mutation alone. Colorectal cancer can survive under the condition of glucose depletion while retaining TCA cycle activity. The cells' survival relies on a delicate balance between energy and reactive oxygen species (ROS) production. Glutamate dehydrogenase 1 (GLUD1) and SLC25A13 have pivotal roles under glucose-deprived conditions and are associated with tumor aggressiveness and colorectal cancer prognosis.

Results
Survival of colorectal cancer cells under condition of glucose depletion. Glucose and glutamine are two of the most abundant nutrients in plasma, and together, they account for most of the carbon and nitrogen metabolism occurring in mammalian cells. Both nutrients are essential for growth of pancreatic ductal adenocarcinoma cells with KRAS mutation 12 . To assess the role of glucose and glutamine in colorectal cancer cells, a proliferation assay was performed under various media conditions ( Fig. 1A and Supplementary Fig. S1A). For the assay, we confirmed that DLD1 and HCT116 cells had a KRAS mutation at codon 13 involving a nucleotide change from GGC to GAC, and that HT29 and CaR1 cells did not have this KRAS mutation ( Fig. 1B and Supplementary Fig. S1B). Notably, DLD1, HCT116, and CaR1 cells could survive under the glucose-deprived conditions (Fig. 1C-E and Supplementary Fig. S1C-E). Furthermore, DLD1 cells that had strong resistance to the condition of glucose depletion were able to survive for 14 days ( Fig. 1F and G), and the passage of DLD1 cells was possible under that condition. The rate of apoptotic cells under the glucose-deprived conditions was lower in DLD1 cells than in HT29 cells (1.5% vs. 24.7%, respectively) (Fig. 1H). These findings show that colorectal cancer cells can survive under conditions of glucose depletion (glutamine sufficiency), which is profoundly different from pancreatic cancer cells in which both nutrients are indispensable.
Resistance to glucose-deprived conditions in colorectal cancer depends on genomic alterations not restricted to KRAS mutation alone. Many reports indicate that the oncogene KRAS has an important role not only in cellular transformation and metabolic reprograming during tumorigenesis but also in chemotherapy for colorectal cancer, considering that mutant KRAS transduces oncogenic signaling and is the most reliable predictor for cancer response to anti-EGFR monoclonal antibodies [11][12][13] . We were interested in how genomic alterations, including alteration of the KRAS oncogene, in colorectal cancer influence cell proliferation under glucose-deprived conditions. To this end, mouse transformed NIH3T3 cells (Cle-H3) by human colorectal cancer genome with KRAS mutations were used in a proliferation assay 14 . The results showed that the Cle-H3 cells survived under glucose-deprived conditions, but the parental NIH3T3 did not ( Fig. 2B-D), which suggested a role of the colorectal cancer genome in cell survival under glucose-deprived conditions. To assess whether these findings resulted solely from the oncogene KRAS, the proliferation assay using oncogenic KRAS-overexpressing NIH3T3 cells was similarly performed (Fig. 2E). The data showed that both KRAS-overexpressing NIH3T3 cells and control NIH3T3 cells did not survive under glucose-deprived conditions ( Fig. 2F and G). Furthermore, knockdown of KRAS in the Cle-H3 and DLD1 cells did not influence their survival under glucose-depleted conditions ( Supplementary Fig. S2A-F). These results demonstrated that resistance to the glucose-deprived conditions in colorectal cancer depended on uncharacterized genomic alterations that were not restricted to KRAS mutation alone.
Metabolism of colorectal cancer was not restricted to KRAS mutation alone. We hypothesized that KRAS mutation alone may not markedly influence colorectal cancer metabolism. In the metabolomic analysis using control and KRAS knockdown DLD1 cells, which can survive under glucose-deprived conditions, we found that the metabolism of colorectal cancer did not depend on KRAS mutation alone but depended on the medium condition as well ( Fig. 3A and B). In addition, k-means clustering and the Calinski criterion showed that k = 3 gave the optimal number of clusters, which indicated low dependence of metabolism of colorectal cancer on KRAS mutation alone (Supplementary Fig. S3A and B). Then, we measured the metabolites in plasma from wild type and KRAS mutation (Cre/LSL-KRASmut) mice. The principal component analysis showed no significant difference in cellular metabolites between two groups ( Supplementary Fig. S3C). Considering that a mutation of another oncogene BRAF plays a role in tumor aggressiveness of retractable colorectal cancer in down stream pathway of KRAS, a role of BRAF was studied. We used two kinds of BRAF specific compounds, PLX4032 and PLX4720, in colorectal cancer HT29 cells in culture, and measured whole metabolites by CE-MS analysis. The data indicated that BRAF inhibition resulted in the suppression of metabolites in TCA cycle (Supplementary Fig. S4A-G), confirming that series of our experiments demonstrated the concept that mutant KRAS status does not predict glucose dependence in culture, and that glutamine fuels oxidative phosphorylation in TCA cycle, as observed in a stress condition such as BRAF specific inhibition in colorectal cancer cells. This observation suggests that the survival of colorectal cancer cells depends on a metabolomic mechanism different from those of other cancers, including pancreatic cancer 12 . On the basis of previous reports that showed that metabolism was important in DNA methylation 15 , DNA methylation analysis was performed under various conditions ( Supplementary Fig. S3D). The data showed that DNA methylation frequency in colorectal cancer was not markedly influenced by depletion of glucose or glutamine.  cells but not in HT29 cells (Fig. 4). Under glucose-deprived conditions, the levels of glycerate 3-phosphate, 2-phosphoglycerate, and phosphoenolpyruvic acid increased in HT29 cells but not in DLD1 cells, which suggested that glucose depletion in HT29 cells induces gluconeogenesis. It is known that nicotinamide adenine dinucleotide consists of an oxidized form (NAD + ) and a reduced form (NADH) and has a pivotal role in the electron transport chain in oxidative phosphorylation 16 ; it is produced by glycolysis, the TCA cycle, and β -oxidation of fatty acids. The results showed that, in DLD1 cells, NADH decreased under glucose-deprived conditions compared with glucose-and glutamine-containing conditions, but NAD + increased ( Supplementary Fig. S5A). It is known that NADPH (reduced form) also has an oxidized form (NADP + ) and provides reducing equivalents for various reactions related to biosynthesis; the molecule is used for generation of glutathione-scavenging ROS. The data showed that NADPH increased in DLD1 cells under glucose-deprived conditions, whereas NADPH decreased remarkably and NADP + increased in HT29 cells ( Supplementary Fig. S5B). Under glucose-deprived conditions, DLD1 and HT29 cells had few metabolites in the pentose phosphate pathway, which is the main source of nicotinamide adenine dinucleotide phosphate ( Supplementary Fig. S5C), although other pathways, including those related to malic enzyme, isocitrate dehydrogenase, and nicotinamide nucleotide transhydrogenase, may be involved in its supply 17 . We performed the sphere formation assay, which is sensitive for intracellular redox status. Cells exhibiting high levels of ROS are susceptible to cell death and some cancer cells develop the redox pathway, which contributes to the cell survival and sphere formation. The results showed that the sphere formation frequency under the glucose depletion condition was greater in DLD1 cells when compared with HT29 cells, suggesting that cell survival was associated with the redox balance ( Supplementary Fig. S5D). Together, cells were cultured at 3 × 10 6 cells per 10-cm dish in 10 ml complete medium; this medium was replaced the following day with glucose(− ) and glutamine(− ) medium supplemented with 10% dialyzed fetal bovine serum. Glucose (10 mM) or glutamine (2 mM) was added, respectively. Cells incubated for 24 h were analyzed. Data are presented as mean ± SD for two independent experiments.
these results indicate that DLD1 cells retain TCA cycle activity under the condition of glucose depletion and resist increases in ROS.

Increase in amino acids levels in colorectal cancer cells under condition of glucose depletion.
In addition to glycolytic and TCA cycle metabolites, the levels of amino acids in DLD1 and HT29 cells were studied by metabolomic analysis based on a method described in a previous report 18 (Fig. 5). Compared with glucoseand glutamine-containing conditions, glucose-deprived conditions led to an increase in most of the amino acids DLD1 and HT29 cells were plated at 3 × 10 6 cells per 10-cm dish in 10 ml complete medium; the medium was replaced the following day with glucose(− ) and glutamine(− ) media supplemented with 10% dialyzed fetal bovine serum. Glucose (10 mM) or glutamine (2 mM) was added, respectively. Cells incubated for 24 h were analyzed. Data are presented as mean ± SD for two independent experiments. in the DLD1 and HT29 cells. In particular, aspartic acid and asparagine levels increased remarkably in DLD1 cells (25.3 and 13.4 times, respectively). The total level of amino acids increased in both cell lines.

Reliance of survival of DLD1 cells on the delicate balance between ROS and energy production.
To adapt to nutritional stress, cells exposed to starvation have reportedly used autophagy to maintain ATP energy production and macromolecular synthesis 19 . Indeed, glucose depletion caused a decline in ATP levels in DLD1 and HT29 cells (Supplementary Fig. S6A) and induced autophagy in HT29 cells but not in DLD1 cells, as shown by LC3-II levels ( Supplementary Fig. S6B). These findings suggest that the influence of autophagy on the increase in amino acids was limited to DLD1 cells (versus HT29 cells). Asparagine has reportedly been shown to suppress expression of C/EBP homologous protein (CHOP), which induces apoptosis 20 . In parallel with the effect of asparagine level, glucose depletion for 24 h induced CHOP in HT29 cells but not in DLD1 cells ( Supplementary Fig. S6B), which suggested that the resistance to glucose-deprived conditions in colorectal cancer was partially associated with the level of asparagine. Given that the NADPH level in HT29 cells was low under glucose-deprived conditions, we studied ROS levels. The data showed that the level of ROS in HT29 cells under glucose-deprived conditions was 5.3 times higher than the levels under glucose-and glutamine-containing conditions, but it was only 1.7 times higher in DLD1 cells (Supplementary Fig. S6C). We were interested in whether the cell survival in glucose deprivation may depend on oxidative phosphorylation in mitochondria. Metformin reportedly inhibits the respiratory chain complex I associated with ATP generation in mitochondria 21 . As expected, the addition of metformin in culture under glucose-deprived conditions resulted in inhibition of growth of DLD1 cells in a concentration-dependent manner (Supplementary Fig. S6D). Taken together, the data support the idea that retention of resistance to glucose-deprived conditions may rely on oxidative phosphorylation and ROS level control.
Significant association of GLUD1 with resistance to glucose-deprived conditions in colorectal cancer. Glutaminolysis has a profound influence on the maintenance of redox balance and energy production.
Under glucose-deprived conditions, its importance appears to further increase, considering that it may change according to the surrounding environment. To verify this hypothesis, gene microarray analysis was performed ( Supplementary Fig. S7A) to identify candidate enzymatic genes related to the metabolism of glutamate and glutamine. GLUD1 expression was greatly increased in DLD1 but not in HT29 cells. The results are consistent with those of a previous report in which GLUD1 was shown to have an important role under the condition of glucose deprivation in glioblastoma cells 22 . The changes in GLUD1 expression were confirmed by quantitative real-time PCR (Supplementary Fig. S7B). Changes in glutamate dehydrogenase activity related to expression of GLUD1 were also confirmed ( Supplementary Fig. S7C). Knockdown of GLUD1 in DLD1 cells under glucose-deprived conditions significantly decreased their growth ( Supplementary Fig. S7D and E). Taken together, the present study showed that GLUD1 was important in resistance to glucose-deprived conditions in colorectal cancer.
Combined expression of GLUD1 and SLC25A13 is significantly associated with prognosis in colorectal cancer. Cancer cells that have resistance to nutritional stress are expected to contribute to worse prognosis. Therefore, the published GSE17536 database was used for screening of genes related to the prognosis of colorectal cancer patients, as described previously 23,24 (Supplementary Table S1). SLC25A13 was identified by analyses of the GLUD1 gene and other genes involved in amino acid metabolism in all possible combinations as pairs. SLC25A13 codes for mitochondrial aspartate-glutamate carrier (AGC), which has a central role in the malate-aspartate shuttle 25 . The reducing equivalents of NADH plus H + through glycolysis are transferred from the cytosol to the mitochondria for the electron transport chain in oxidative phosphorylation by this shuttle, which is concerned with ROS and ATP production. In the DLD1 cells, the mRNA level of SLC25A13 decreased under glucose-deprived conditions relative to the levels in glucose-and glutamine-containing conditions but increased conversely in HT29 cells ( Supplementary Fig. S7F). Then we measured the enzyme activity of TCA related enzymes (succinate dehydrogenase and fumarase). The data showed that DLD1 had higher activity of succinate dehydrogenase, fumarase and malate dehydrogenase than HT29 cells did ( Supplementary Fig. S8A-C). These findings suggest that the resistance to glucose-deprived conditions in colorectal cancer may result from inhibition of ROS production via repressed AGC. Subsequently, the relevance of GLUD1 and SLC25A13 expression in colorectal cancer with respect to clinicopathological characteristics and prognosis was evaluated by immunohistochemistry (Fig. 6A). High GLUD1 expression or low SLC25A13 expression was found to be associated with tumor aggressiveness, including depth of tumor invasion, lymph node and distant metastasis, lymphatic and venous invasion, and stage (Supplementary Table S2). Notably, high GLUD1 expression combined with low SLC25A13 expression had a higher association with tumor aggressiveness than did individual expression of each (Table 1). This combination also mirrored the effect on prognosis (Fig. 6B-G). There was no significant association between the individual expression of GLUD1 and SLC25A13. Then we assess the association of GLUD1 and PKM2 protein expression; the PKM2 is a splicing form of pyruvate kinase gene. The experimental data indicated that the GLUD1 expression was independent of PKM2 expression ( Supplementary Fig. S8D), supporting the notion that both glucose metabolism and glutamine metabolism are important for the colorectal cancer. Taken together, colorectal cancer cells that can adapt to nutritional stress via regulation of GLUD1 and SLC25A13 contribute to tumor aggressiveness and result in worse prognosis (Fig. 6H).

Discussion
We found that the metabolism of colorectal cancer cells that could survive under glucose-deprived conditions was significantly different from that of pancreatic ductal adenocarcinoma cells. In addition, the metabolism depended on genomic alterations that are uncharacterized and not restricted to KRAS mutation alone. In pancreatic ductal adenocarcinoma cells, a non-canonical glutamine pathway mediated by the oncogene KRAS that  regulates glutamic-oxaloacetic transaminase 1 (GOT1) and GLUD1 has been described 12 . In non-small cell lung cancer, KRAS has a profound influence on glutamine metabolism through regulation of ME1 and GOT1, the high expression of which has been shown to correlate with poorer prognosis after radiotherapy 26 . A human breast carcinoma cell line that expresses an activated KRAS protein has been shown to have high glycolytic flux, low TCA cycle activity, and increased usage of glutamine for anabolic synthesis 13 . The solitary effects of oncogenic KRAS in metabolism are reportedly significant in various cancers; however, these effects were not observed for colorectal cancer in the present study. It would appear that in different cancers, metabolism varies, and the influence of oncogenes on metabolism is also diverse. Understanding the inherent metabolisms of different cancer types and how they are affected by oncogenes may lead to the development of specific cancer therapies.
In glutaminolysis, glutamine is first converted by glutaminase to glutamate and ammonia and then into α -ketoglutaric acid by either GLUD1 or less prominently by other transaminases. Jin et al. reported that expression of GLUD1 was elevated in breast and lung cancer tissue relative to that in normal tissue 27 . In addition, the mammalian target of rapamycin complex 1 the activity of which is dysregulated in many cancers, activates GLUD1 via inhibition of SIRT4 28 . Cancer cells enhance glutaminolysis via positive regulation of GLUD1 to respond to increased demand of glutamine, which is used as a carbon source for energy generation, a component of protein and nucleotide, and in production of glutathione and NADPH. On the basis of the results of immunohistochemistry for 104 colorectal cancer specimens, Liu et al. reported that GLUD1 expression was high and associated with poorer prognosis 29 . Our results showing an association between GLUD1 expression and prognosis in colorectal cancer were consistent with those of Liu et al. Enhanced glutaminolysis by increased expression and activity of GLUD1 under nutritional stress may contribute to these findings. A recent study investigated regulation of redox homeostasis by GLUD1. Fumarate, the level of which is controlled by GLUD1, has been shown to positively regulate a ROS-scavenging enzyme, glutathione peroxidase 1 27 . Our results showed that the increase in ROS levels under glucose-deprived condition was suppressed in DLD1 cells in which expression and activity of GLUD1 increased. Our results support those of the study by Jin et al. (Figs S5C, S6B and S6C).
A mutation in SLC25A13 coding for citrin has been shown to cause citrin deficiency that results in type-II citrullinemia and is characterized by hyperammonemia, steatohepatitis, and neuropsychiatric symptoms 25 . Although expression of SLC25A13 is known to be high in the liver, the significance of its expression in the colon has not been understood fully. In the present study, SLC25A13 expression was observed in the glandular cells of normal colorectal tissue and colorectal cancer tissue by immunohistochemistry (Fig. 6A). Mutations of SLC25A13 reportedly lead to hepatic overload and increase the risk of hepatocellular carcinoma 30 . Shohet et al. reported that higher expression of SLC25A13 correlated with a poorer outcome in neuroblastoma 31 . However, there are only a few studies that have examined the relevance of SLC25A13 to cancer. To the best of our knowledge, this is the first study to identify the importance of SLC25A13 in colorectal cancer. We showed that SLC25A13 was negatively associated with the depth of tumor invasion, extent of lymph node metastasis, distant metastasis, lymphatic invasion, and stage in colorectal cancer. Furthermore, SLC25A13 was profoundly correlated with colorectal cancer prognosis (Fig. 6D-G and Table S2). In addition, colorectal cancer cells were found to adapt to nutritional stress through regulation of SLC25A13. Therefore, SLC25A13 may have an important role in development of colorectal cancer.
In conclusion, we found that the metabolism of colorectal cancer, which is different from that of pancreatic cancer, depended on genomic alterations that are uncharacterized and not restricted to KRAS mutation alone. Colorectal cancer cells that have resistance to glucose depletion retain TCA cycle activity and rely on the delicate balance between ROS production and ATP production. GLUD1 and SLC25A13 have pivotal roles in nutritional stress and are associated with tumor aggressiveness and poorer prognosis of colorectal cancer. These proteins may serve as new targets in the treatment of refractory colorectal cancer.

Materials and Methods
Cell lines, culture and serum. The human colorectal cancer cell line HT29 was obtained from the ATCC (Manassas, VA, USA). Other colorectal cancer cell lines included DLD1, which was obtained from the Cell Resource Center for Biomedical Research Institute of Development, Aging, and Cancer (Tohoku University), and HCT116 and CaR1, which were obtained from the Japan Cancer Research Resources Bank (JCRB, Tokyo, Japan). Mouse embryonic fibroblast cell lines, parental NIH3T3, and transformed NIH3T3 cells by whole genomic DNA derived from a human colon cancer cell line (MA) were purchased from the RIKEN Cell Bank (Tsukuba, Japan). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM D6046; Sigma Aldrich, St. Louis, MO, USA) containing 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μ g/ml streptomycin (Life Technologies, Carlsbad, CA, USA) at 37 °C in a humidified incubator with 5% CO 2 . Mice serum were obtained under anesthesia from wild type and KRAS mutation (Cre/LSL-KRASmut) BL6 mice, which were purchased from the Jackson laboratory (Bar Harbor, ME, USA), and maintained under the ethical agreement of animal facility at Osaka University (24-122; approved by chairman professor Kaneda).
Cell proliferation assay. Cells were added to 24-well plates at 10,000-15,000 cells per well in 0.5 ml of medium that was replaced the following day with glucose-and glutamine-deficient medium (DMEM D5030; Sigma Aldrich) supplemented with 10% dialyzed FBS (#26400-044; Invitrogen, Carlsbad, CA). Glucose (Wako Pure Chemical Industries, Osaka, Japan) and glutamine (Nacalai Tesque, Kyoto, Japan) were added. After cells were fixed in 80% methanol and stained with 0.2% crystal violet, the relative cell proliferation was quantified by absorbance at 595 nm.

Apoptosis detection. Apoptotic cells were analyzed using an annexin V [fluorescein isothiocyanate
(FITC)-conjugated] apoptosis kit (K101-400; BioVision, Mountain View, CA, USA). In brief, cells growing on 6-cm dishes at 1-2 × 10 6 cells/dish for 24 h were loaded with 0.5 ml binding buffer, 5 μ l annexin V-FITC, and 5 μl propidium iodide. After incubation at room temperature for 5 min in the dark, annexin V-FITC binding and PI staining were analyzed by flow cytometry.
Transfection of vector. Cells overexpressing KRAS were generated using a pCMV6-Entry plasmid containing the KRAS G12V sequence with SgfI-MluI restriction sites (OriGene, Rockville, MD, USA) and Fugene6 (Roche Applied Science, Indianapolis, USA). KRAS knockdown cells were generated using the Lenti-vpak Packaging Kit (OriGene). A lentiviral shRNA construct, NM_033360 that targets human KRAS was obtained from the MISSION TRC-Hs1.0 library (Sigma). Control cells were transfected using the same procedure but with an empty control vector. DLD1 cells were selected using 2 μ g/ml puromycin to establish stable cell lines.
Metabolomic analysis. Cells were cultured at 3 × 10 6 cells per 10-cm dish in 10 ml complete medium; the medium was replaced the following day with glucose(− ) and glutamine(− ) medium supplemented with 10% dialyzed fetal bovine serum. Glucose (10 mM) or glutamine (2 mM) was added, respectively. For hypoxia treatment, cells were incubated for 24 h with 1% O 2 . After culture medium was aspirated from the dish, the cells were washed twice using 5% mannitol solution and treated with 800 μ l of methanol and 550 μ l of Milli-Q water containing internal standards [H3304-1002; Human Metabolome Technologies (HMT), Tsuruoka, Japan]. The metabolite extracts were centrifuged at 2,300 g at 4 °C for 5 min. Next, to remove macromolecules, 800 μ l of the upper aqueous layer was filtered by centrifugation using a Millipore 5-kDa cutoff filter at 9,100 g at 4 °C for 120 min, and resuspended in 50 μ l of Milli-Q water for metabolomic analysis. Metabolomic analysis was performed using a C-SCOPE package in HMT and capillary electrophoresis-time-of-flight mass spectrometry for cation analysis and capillary electrophoresis-tandem mass spectrometry for anion analysis 33,34 . To obtain information regarding the peak, including the m/z ratio, migration time (MT), and peak area, the peaks were identified using automatic integration software (MasterHands; Keio University, Tsuruoka, Japan and MassHunter Quantitative Analysis B.04.00; Agilent Technologies, Santa Clara, CA, respectively). Peak areas were normalized against those of the internal standards, and the relative area values were further normalized by sample amounts. Hierarchical cluster analysis and principal component analysis were performed using HMT's proprietary softwares, PeakStat and SampleStat, respectively. Immunohistochemical analysis. Immunohistochemical analysis was performed as previously described 32 . Colorectal cancer tissue samples were obtained from 151 consecutive patients who underwent surgery at the Osaka University Hospital between 2006 and 2009. None of the patients had received chemotherapy or radiotherapy before surgery. All patients provided their written informed consent for use of their clinical samples in this study, which was approved by the Institutional Review Board. GLUD1 antibody (ab166618; Abcam; 1:500 dilution), SLC25A13 antibody (PA5-21991; Thermo Fisher Scientific; 1:500 dilution) were used.
Statistical analysis. Data were indicated as the mean ± standard deviation (SD). We performed Student's t-test and Fisher's exact test to determine statistically significant differences using JMP Pro 10 software (SAS Institute, Cary, NC, USA). The Kaplan-Meier method was used to assess recurrence-free survival and overall survival. A P-value < 0.05 was considered to indicate statistical significance.