Decreased Expression of Fructose-1,6- Bisphosphatase

Published OnlineFirst April 6, 2016; DOI: 10.1158/0008-5472.CAN-15-2601

Cancer Molecular and Cellular Pathobiology Research

Decreased Expression of Fructose-1,6- bisphosphatase Associates with Glucose Metabolism and Tumor Progression in Hepatocellular Carcinoma Hidenari Hirata1,2,3, Keishi Sugimachi1, Hisateru Komatsu1, Masami Ueda1,Takaaki Masuda1, Ryutaro Uchi1, Shotaro Sakimura1, Sho Nambara1, Tomoko Saito1, Yoshiaki Shinden1, Tomohiro Iguchi1, Hidetoshi Eguchi1, Shuhei Ito1, Kotaro Terashima2, Katsumi Sakamoto2, Masakazu Hirakawa2, Hiroshi Honda3, and Koshi Mimori1

Abstract

Fructose-1,6-bisphosphatase (FBP1), the rate-limiting enzyme promoter methylation and copy-number loss of FBP1 were inde- in gluconeogenesis, is reduced in expression in certain cancers pendently associated with decreased FBP1 expression. Similarly, where it has been hypothesized to act as a tumor suppressor, FBP1 downregulation in HCC cell lines was also associated with including in hepatocellular carcinoma (HCC). Here, we report copy-number loss. HCC specimens exhibiting low expression of functional evidence supporting this hypothesis, providing a pre- FBP1 had a highly malignant phenotype, including large tumor clinical rationale to develop FBP1 as a therapeutic target for HCC size, poor differentiation, impaired gluconeogenesis, and enhanced treatment. Three independent cohorts totaling 594 cases of HCC aerobic glycolysis. The effects of FBP1 expression on prognosis and were analyzed to address clinical significance. Lower FBP1 expres- glucose metabolism were confirmed by gene set enrichment anal- sion associated with advanced tumor stage, poor overall survival, ysis. Overall, our findings established that FBP1 downregulation in and higher tumor recurrence rates. In HCC cell lines, where HCC contributed to tumor progression and poor prognosis by endogenous FBP1 expression is low, engineering its ectopic over- altering glucose metabolism, and they rationalize further study of expression inhibited tumor growth and intracellular glucose FBP1 as a prognostic biomarker and therapeutic target in HCC uptake by reducing aerobic glycolysis. In patient specimens, patients. Cancer Res; 76(11); 3265–76. 2016 AACR.

Introduction cytes (3). A recent study reported the loss of gluconeogenic capacity in HCC (4). Moreover, the switch from aerobic glycolysis Hepatocellular carcinoma (HCC) is one of the most prevalent to gluconeogenesis may be an effective treatment in patients with cancers worldwide and is associated with a high mortality rate (1). HCC (5). Nevertheless, the role of impaired gluconeogenesis in Therefore, identiﬁcation of useful biomarkers predicting clinical HCC progression is still poorly understood. outcomes and molecular targets for treatment is required. Altered The tumor-suppressive function of the fructose-1,6-bisphospha- glucose metabolism, particularly aerobic glycolysis, is one of the tase (FBP1) gene, which encodes a rate-limiting gluconeogenic hallmarks of cancers, including HCC (2), supporting the notion enzyme, has been reported in certain cancers. Loss of FBP1 that the molecules involved in altered glucose metabolism may be expression has been shown to promote tumor progression by potential biomarkers and therapeutic targets. Many studies have enhancing aerobic glycolysis, thereby resulting in poor prognosis examined aerobic glycolysis in various cancers; however, gluco- in patients with clear cell renal cell carcinoma (ccRCC; ref. 6) and neogenesis has not been studied extensively. Gluconeogenesis is breast cancer (7). With regard to HCC, in vitro experiments have also thought to play an important role in tumor progression in shown that FBP1 expression is silenced by promoter methylation HCC because gluconeogenesis mainly occurs in normal hepato- and that ectopic FBP1 expression inhibits anchorage-dependent growth of a cell line through cell-cycle arrest (8). However, the role 1Department of Surgery, Kyushu University Beppu Hospital, Beppu, of FBP1 in altered glucose metabolism has not been assessed in Japan. 2Department of Radiology, Kyushu University Beppu Hospital, HCC. Furthermore, no studies have investigated whether aberrant 3 Beppu, Japan. Department of Clinical Radiology, Graduate School of FBP1 expression affects clinical outcomes in patients with HCC. Medical Sciences, Kyushu University, Fukuoka, Japan. Therefore, the objective of this study was to examine the Note: Supplementary data for this article are available at Cancer Research function and clinical signiﬁcance of FBP1 expression in HCC. Online (http://cancerres.aacrjournals.org/). Through analysis of three independent large cohorts of HCC Corresponding Author: Koshi Mimori, Kyushu University Beppu Hospital, 4546 cases, the present study evaluated whether aberrant FBP1 expres- Tsurumihara, Beppu 8740838, Japan. Phone: 81-977-27-1650; Fax: 81-977-27- sion was associated with tumor progression and prognosis in 1651, E-mail: [email protected] patients with HCC. We also elucidated the genetic and epigenetic doi: 10.1158/0008-5472.CAN-15-2601 mechanisms regulating FBP1 expression and investigated the role 2016 American Association for Cancer Research. of FBP1 expression in altered glucose metabolism.

www.aacrjournals.org 3265

Hirata et al.

Materials and Methods Encyclopedia (CCLE; ref. 10) from cBioPortal For Cancer Geno- mics (http://www.cbioportal.org/index.do; ref. 11). Of the 967 Patients and sample collection human cancer cell lines in the CCLE database, FBP1 expression One hundred eighteen patients with HCC who underwent profiles were available for 27 liver cancer cell lines. A z-score hepatic resection at Kyushu University Beppu Hospital and affil- threshold of 2 was used as the cutoff value. For the analysis of iated hospitals between 2000 and 2004 were enrolled in this copy-number alterations, data were available for 23 liver cancer study. Resected tumor tissues and paired normal liver tissues were cell lines (Supplementary Table S1). Moreover, FBP1 promoter immediately stored in RNAlater (Ambion), frozen in liquid methylation status was determined using the results of methyl- nitrogen, and kept at –80C until RNA extraction. A 5-year ation-specific PCR, as described in a previous report (8), or follow-up was conducted after operation. The median follow-up methylation array data from GEO database accession numbers for the 118 patients was 59.5 months (range, 3.0–60 months). GSE60753 (12) and GSE67485 (13). Data on promoter methyl- Patients were staged according to the seventh edition of the ation profiles were available for 7 liver cancer cell lines (Supple- International Union against Cancer TNM classification system. mentary Table S2). Of the 118 HCC patients, paired normal liver tissues were available for 115 patients. All protocols were approved by the Ethics Western blotting analysis and Indications Committee of Kyushu University. Written Proteins were detected using antibodies against FBP1 informed consent was obtained from all patients. (SAB1405798; Sigma-Aldrich), thioredoxin-interacting protein (TXNIP; K0204-3; Medical & Biological Laboratories), and b-actin Cell culture (Santa Cruz Biotechnology). The methods are described in detail HuH7 and HepG2 cells authenticated by short tandem repeat in the Supplementary Materials and Methods. profiling using the PowerPlex 1.2 System (Promega) were obtained from the cell bank of RIKEN BioResource Center (Tsu- Sphere formation assays kuba, Japan) in 2015. Cells were used within 6 months of Detailed methods are provided in the Supplementary Materials culturing after resuscitation. Details are provided in the Supple- and Methods. mentary Materials and Methods. Xenograft studies Construction of an FBP1 expression lentiviral vector Five-week-old female BALB/c nu/nu mice were obtained from FBP1 fi A full-length cDNA insert of human was ampli ed by SLC, Inc. and maintained under specific pathogen-free condi- PCR. Lentiviruses were generated by transfection of HEK293T cells tions. All animal procedures were performed in compliance with with pCMV-VSV-G-RSV-Rev, pCAG-HIVgp, and either CSII-CMV- the Guidelines for the Care and Use of Experimental Animals 0 Nhe 0 FBP1 or CSII-CMV-MCS (empty) plasmid DNAs (5 I and 3 established by the Committee for Animal Experimentation of Xba I sites) using Lipofectamine 2000 (Invitrogen) following the Kyushu University. A total of 5 106 cells (empty vector control manufacturer's protocol. Forty-eight hours after transfection, the cells or HuH7 cells stably expressing FBP1) in 150 mL of phos- lentivirus-containing supernatant was collected and passed phate-buffered saline were injected subcutaneously into both m fi through a 0.45- m lter. Infections were then carried out by flanks of the mice. Tumor size was measured every 3 to 4 days incubating cells with medium containing the lentiviral superna- with a digital caliper and calculated using the following formula: tant for 48 hours. tumor volume ¼ length width2 0.5.

RNA preparation and quantitative real-time RT-PCR Quantification of metabolites Detailed methods are provided in the Supplementary Materials Intracellular glucose uptake was determined by 2-(N-[7-nitro- and Methods. benz-2-oxa-1,3-diazol-4-yl]amino)-2-deoxy-D-glucose (2-NBDG), a fluorescently-tagged glucose derivative, using a Glucose Uptake Analysis of public clinical datasets and gene set enrichment Cell-Based Assay Kit (Cayman Chemical). Lactate secretion was analysis examined using a Lactate Assay Kit (BioVision). Isotopomer dis- We obtained paired gene-expression and survival profiles of tribution analysis was performed by F-SCOPE package of Human 242 HCC samples from the National Cancer for Biotechnology Metabolome Technologies using capillary electrophoresis time-of- Information Gene Expression Omnibus (GEO) database (acces- flight mass spectrometry (14–16). Cells were incubated in DMEM sion codes GSE14520; ref. 9). Expression profiles of 228 paired with glucose replaced by 0.1% [U-13C]glucose (Sigma-Aldrich) for normal liver samples were available. 4 hours. All methods are described in detail in the Supplementary Moreover, we obtained paired RNA sequencing and survival Materials and Methods. data of 234 HCC cases in The Cancer Genome Atlas (TCGA) from the Broad Institute's Firehose (http://gdac.broadinstitute.org/ Statistical analysis runs/stddata__2014_10_17/data/LIHC/20141017/). Expression For continuous variables, statistical analyses were performed profiles of 48 paired normal liver samples were available. Data using Student t tests or Welch t tests, and the degree of linearity was from methylation arrays, SNP arrays, and whole-exome sequenc- estimated by Pearson's correlation coefficient. Differences among ing were also obtained. Details are provided in the Supplementary multiple groups were examined by one-way ANOVA followed by Materials and Methods. Tukey–Kramer tests. Categorical variables were compared using c2 tests or Fisher's exact tests. Overall survival (OS) and recur- Analysis of the Cancer Cell Line Encyclopedia database rence-free survival (RFS) were estimated using the Kaplan–Meier We obtained normalized mRNA expression and DNA copy- method, and survival curves were compared using log-rank tests. number data of human cancer cell lines in the Cancer Cell Line Multivariable analysis was performed using the Cox proportional

3266 Cancer Res; 76(11) June 1, 2016 Cancer Research

FBP1 and Glucose Metabolism in HCC

hazards model to determine the prognostic value of FBP1 expres- adenosine triphosphate (ATP) and adenosine diphosphate were sion. Differences or correlations with P values of less than 0.05 decreased in FBP1-overexpressing cells (Supplementary Fig. S1F were considered significant. Data analyses were performed using R and S1G), suggesting that FBP1 suppressed de novo nucleic syn- software version 3.1.1 (The R Foundation). thesis through the pentose phosphate pathway (PPP) in HCC cells (five 13C atoms were preserved through the PPP). Total ATP synthesis was also decreased when FBP1 was overexpressed (Sup- Results plementary Fig. S1H), consistent with the results of reduced FBP1 repressed tumor progression and improved altered aerobic glycolysis and TCA cycle in FBP1-overexpressing cells. glucose metabolism in HCC cells Overall, the results of functional analysis suggested that FBP1 acted Initially, we performed functional analysis using HCC cell lines as a tumor suppressor through inhibition of glucose uptake and to elucidate whether FBP1 suppressed tumor progression and aerobic glycolysis, which encouraged us to further explore the affected glucose metabolism in HCC. By estimating the FBP1 significance of FBP1 in HCC. expression profile across 967 cancer cell lines from the CCLE (10), we found that none of the 27 available liver cancer cell lines Lower FBP1 expression promoted tumor progression in HCC showed high expression of FBP1 (Fig. 1A). Notably, 35% of liver cases cancer cell lines from the CCLE database harbored FBP1 copy- Next, to clarify the clinical significance of FBP1 expression in number loss, with markedly reduced expression of FBP1 (Fig. 1B; HCC, we analyzed FBP1 expression in tumor tissues and Supplementary Table S1). Copy-number loss, promoter methyl- normal liver tissues from three independent cohorts of HCC ation, or both contributed to downregulation of FBP1 expression cases (Fig. 3A). For the analysis of patients who underwent (Supplementary Table S2). Therefore, because FBP1 expression resection of primary HCC at our institution (Kyushu HCC set), was low in liver cancer cell lines, we used overexpression experi- FBP1 expression in tumors was significantly lower than that in ments rather than knockdown experiments. We established HCC normal liver tissues (P < 0.001). Given the observation that cell lines (HuH7 and HepG2) stably expressing FBP1 or empty FBP1 was expressed at low levels in both tumor tissues and liver vector control. Overexpression of FBP1 was confirmed by RT-PCR cancer cell lines, these data implied that downregulation of and Western blotting (Fig. 1C). Consistent with previous reports FBP1 expression was essential for HCC tumorigenesis. Further- in ccRCC cells (6) and breast cancer cells (7), FBP1 inhibited more, decreased FBP1 expression was significantly associated anchorage-independent tumor cell growth in vitro (Fig. 1D) and with advanced tumor stage (stage I vs. stage II, P ¼ 0.017; stage I xenograft tumor growth (Fig. 1E) in HCC cells. vs. stage III, P ¼ 0.005). The median expression levels of FBP1 Because the FBP1 gene encodes a rate-limiting gluconeogenic in each tumor stage (stage I, stage II, and stage III) relative to enzyme, we further investigated the effects of FBP1 on glucose those in normal liver tissues were 0.427, 0.179, and 0.059, metabolism in HCC cells. Intracellular glucose uptake and lactate respectively. To confirm the results of our series, we analyzed secretion were significantly decreased in FBP1-overexpressing HCC the public datasets of HCC cases from the GSE14520 and the cells as compared with those in vector control cells (Fig. 2A and B). TCGA datasets. In accordance with our cases, FBP1 expression We also examined glucose-induced TXNIP expression, which is in tumors was significantly lower than that in normal liver commonly used as an intracellular glucose sensor (7, 17–19). tissues (P < 0.001 and P < 0.001, respectively). Moreover, the TXNIP was robustly induced in vector control cells following significant association between lower FBP1 expression and glucose stimulation. In contrast, the induction of TXNIP was advanced tumor stage was validated in the GSE14520 (stage suppressed in FBP1-overexpressing HCC cells (Fig. 2C), supporting I vs. stage III, P < 0.001; stage II vs. stage III, P < 0.009) and the the results showing that FBP1 inhibited intracellular glucose TCGA (stage I vs. stage II, P < 0.001; stage I vs. stage III, P < uptake in HCC cells. Interestingly, the expression levels of hexoki- 0.001) datasets. The median expression levels of FBP1 in each nase2 (HK2; refs. 20, 21) and phosphofructokinase-1 (PFK1,also tumor stage (stage I, stage II, and stage III) relative to those in known as PFKM; refs. 22, 23), key regulators of aerobic glycolysis in normal liver tissues were 0.262, 0.202, and 0.090, respectively, cancer cells, were significantly decreased in FBP1-overexpressing in the GSE14520 dataset and 0.280, 0.143, and 0.153, respec- HCC cells as compared with those in vector control cells after tively, in the TCGA dataset. These results suggested that the loss glucose stimulation (Fig. 2D). To further estimate the effect of FBP1 of FBP1 expression was associated with tumor progression in on aerobic glycolysis, we performed isotopomer distribution anal- HCC cases. ysis using [U-13C]glucose, which directly produces glycolytic intermediates containing six or three 13C atoms (M6/M3 species), and Promoter methylation and copy-number loss inhibited FBP1 the first turn of the tricarboxylic acid (TCA) cycle intermediates expression in HCC cases containing two 13C atoms (M2 species; Supplementary Fig. S1A). To investigate the mechanisms suppressing FBP1 expression in M3 enrichment of lactate was significantly inhibited when HCC cases, promoter methylation and copy-number profiles of FBP1was overexpressed (Fig. 2E), and glycolytic intermediates such FBP1 were analyzed in the TCGA dataset (Fig. 3B). We found that as fructose-1,6-biphosphate (F-1,6-BP; M6 species; Fig. 2F), dihy- the FBP1 promoter exhibited significantly higher methylation in droxyacetone phosphate (M3 species; Supplementary Fig. S1B), tumors than in normal liver tissues (P < 0.001; Fig. 3C). In and glucose-6-phosphate (Supplementary Fig. S1C) were addition, FBP1 expression was inversely correlated with the degree decreased in FBP1-overexpressing cells, supporting the view that of FBP1 methylation in tumors (Pearson correlation coefficient ¼ FBP1 suppressed glucose uptake via inhibition of aerobic glycol- 0.437; P < 0.001; Fig. 3D), indicating that FBP1 promoter ysis. Similar to ccRCC cells (6), ectopic FBP1 also tended to inhibit methylation repressed the transcriptional activity of FBP1 in HCC. M2 enrichment of TCA cycle intermediates such as malate and Similarly, copy-number loss of FBP1 was also significantly asso- citrate (Fig. 2G; Supplementary Fig. S1D and S1E). Furthermore, ciated with lower FBP1 expression (the median expression level in isotopomer distribution analysis showed that M5 enrichment of HCCs harboring copy-number loss relative to that in HCCs

www.aacrjournals.org Cancer Res; 76(11) June 1, 2016 3267

Hirata et al.

Figure 1. FBP1 was expressed at low levels and repressed tumor progression in HCC cells. A, FBP1 expression profiles across 967 cell lines from the CCLE. Labeled cell lines used for overexpression experiments in the present and previous studies are highlighted in black. Labeled cell lines used for FBP1-knockdown experiments in a previous study are highlighted in gray. The horizontal dotted-line represents the z-score threshold. Note that there were no liver cancer cell lines expressing high levels of FBP1. The numbers in parentheses next to the labeled cell lines represent the reference numbers. B, box plots of FBP1 expression based on FBP1 copy-number alterations in liver cancer cell lines from CCLE. Data on copy-number profiles were available for 23 of 27 liver cancer cell lines. C, overexpression of FBP1 was verified by RT-PCR and Western blotting analysis. D, the number of spheres was significantly decreased when FBP1 was overexpressed. E, xenograft tumor growth of HuH7 cells with or without ectopic FBP1 expression. Values represent the means SDs of triplicate experiments. RCC, renal cell carcinoma; EV, empty vector control cells.

harboring copy-number neutral/gain was 0.519; Fig. 3E). For (P < 0.001; Fig. 3B), respectively. These findings suggested that an integrated assessment of the influence of both promoter promoter methylation and copy-number loss were independently methylation and copy-number alterations on FBP1 expression associated with downregulation of FBP1 expression. We did not levels, multiple linear regression analysis was performed. The find any correlations between FBP1 mutations and FBP1 expres- standardized effect sizes (t values) for promoter methylation sion because only two nonsynonymous mutations were observed and copy-number alterations were 7.55 (P < 0.001) and 3.48 in the TCGA dataset (Supplementary Fig. S2).

3268 Cancer Res; 76(11) June 1, 2016 Cancer Research

FBP1 and Glucose Metabolism in HCC

Figure 2. FBP1 improved altered glucose metabolism in HCC cells. A, intracellular glucose uptake was evaluated by 2-NBDG, a fluorescently tagged glucose derivative, in HCC cells with or without ectopic FBP1 expression. B, lactate secretion was significantly decreased when FBP1 was overexpressed. C, the expression of glucose-induced TXNIP was used to sense intracellular glucose uptake. Cells were incubated in glucose-free medium for 12 hours, followed by glucose stimulation for an additional 3 hours. D, the expression levels of genes encoding enzymes involved in aerobic glycolysis, that is, HK2, PFK1,andPKM2, were assessed by RT-PCR after glucose stimulation, as described in C. E–G, isotopomer distribution analysis in vector control or FBP1-overexpressing HuH7 cells labeled with [U-13C]glucose for 4 hours. M3 enrichment of lactate (E), M6 enrichment of 1,6-diphosphate (F-1,6-BP; F), and M2 enrichment of malate and citrate (G) were quantified. a, in the analysis of F-1,6-BP, the metabolite concentrations of one experiment in the vector control and all three experiments in FBP1-expressing cells were below the detection limit. P values were calculated using Welch's t test. Values represent the means SDs of triplicate experiments. EV, empty vector control cells; ND, not detectable.

www.aacrjournals.org Cancer Res; 76(11) June 1, 2016 3269

Hirata et al.

Figure 3. Downregulation of FBP1 expression via promoter methylation and copy-number loss was associated with tumor progression in HCC cases. A, box plots of FBP1 expression in normal liver and tumor tissues grouped into stages I to III in three independent cohorts of HCC cases. Information on tumor stage was available for 114 of 118 patients from the Kyushu HCC set (left), 225 of 242 cases from GSE14520 (middle), and 214 of 234 cases from TCGA (right). Five patients with stage IV cancer were excluded from the analysis of TCGA data. B, an integrative view of expression, promoter methylation, and copy-number proﬁles of FBP1 across 234 HCC cases in TCGA. The samples are sorted according to FBP1 expression levels. C, box plots of FBP1 methylation status in normal liver and tumor tissues from TCGA (data were available for 46/48 normal tissues), D FBP1 expression levels in tumor samples were inversely correlated with the degree of FBP1 methylation in TCGA. E, box plots of FBP1 expression based on FBP1 copy-number alterations in TCGA (data were available for 229/234 cases). F, box plots of FBP1 expression based on tumor size in GSE14520 (data were available for 241/242 cases). Information on tumor size was not available in TCGA. G, box plots of FBP1 expression based on tumor differentiation in TCGA (data were available for 230/234 cases, and one anaplastic HCC was excluded). Information on tumor differentiation was not available in GSE14520. NS, not signiﬁcant.

3270 Cancer Res; 76(11) June 1, 2016 Cancer Research

FBP1 and Glucose Metabolism in HCC

Table 1. FBP1 expression and clinicopathologic factors in patients in the Kyushu larger tumor size (GSE15420: 5cmvs.> 5cm;P < 0.001; Fig. HCC set (n ¼ 118) 3F) and poorly differentiated HCC (TCGA: poorly vs. well; P ¼ High expression Low expression 0.003; Fig. 3G), suggesting that highly malignant HCC (n ¼ 88) (n ¼ 30) FBP1 P expressed low levels of .FortheanalysisofOS(Fig. Factors Number (%) Number (%) FBP1 fi Age (y; mean SD) 66.3 7.9 66.4 9.4 0.956 4A), the low expression group had signi cantly poorer FBP1 P ¼ Gender 0.955 OS than the high expression group ( 0.005). Consis- Male 65 (73.9) 22 (73.3) tent with our cohort, lower FBP1 expression was significantly Female 23 (26.1) 8 (26.7) associated with poorer OS in the GSE14520 (P < 0.001) and HBV 0.661 TCGA (P ¼ 0.010) datasets. In the multivariate analysis that (þ) 20 (22.7) 8 (26.7) excluded tumor stage as a confounding variable, the association ( ) 68 (77.3) 22 (73.3) FBP1 HCV 0.711 between lower expression and poorer OS remained sig- fi fi (þ) 59 (67.0) 19 (63.3) ni cant in the Kyushu HCC [HR, 2.49; 95% con dence interval () 29 (33.0) 11 (36.7) (CI), 1.22–5.06; P ¼ 0.012) and TCGA (HR, 2.84; 95% CI, Child-Pugh grade 0.684 1.22–6.64; P ¼ 0.016) datasets (Table 2). A 79 (89.8) 26 (86.7) To validate the effects of FBP1 expression on survival in patients B 8 (9.1) 4 (13.3) with HCC, we applied gene set enrichment analysis (GSEA; ref. 25) C 1 (1.1) 0 (0.0) Tumor size (cm, mean SD) 3.72 2.91 5.42 3.50 0.010b on the GSE14520 and TCGA datasets using two gene sets accord- Number of tumors 0.527 ing to OS following resection of HCC (26). GSEA revealed that Single 64 (72.7) 20 (66.7) upregulation of FBP1 was significantly correlated with better Multiple 24 (27.3) 10 (33.3) survival in the GSEA14520 (P < 0.001) and TCGA (P < 0.001) a b Tumor differentiation <0.001 datasets (Fig. 4B). Conversely, downregulation of FBP1 was Well/moderately 79 (89.8) 19 (63.3) significantly correlated with worse survival in the GSEA14520 Poorly/undifferentiated 6 (6.8) 10 (33.3) P < P < NA 3 (3.4) 1 (3.3) ( 0.001) and TCGA ( 0.001) dataset (Fig. 4C). Collectively, – Capsular formation 0.517 survival analysis with Kaplan Meier curves and GSEA demon- (þ) 58 (65.9) 21 (70.0) strated that lower FBP1 expression was significantly associated () 30 (34.1) 8 (26.7) with poor survival in patients with HCC. NA 0 (0.0) 1 (3.3) Portal vein invasion 0.061 Higher FBP1 expression was associated with a lower risk of (þ) 37 (42.0) 18 (60.0) () 51 (58.0) 11 (36.7) recurrence in HCC NA 0 (0.0) 1 (3.3) Subsequently, we assessed the effects of FBP1 expression on Hepatic vein invasion 0.869 recurrence following resection of primary HCC (Fig. 4D). In our (þ) 25 (28.4) 9 (30.0) analysis of the Kyushu HCC set, RFS in the high FBP1 expression () 60 (68.2) 20 (66.7) group was significantly better than that in the low FBP1 expression NA 3 (3.4) 1 (3.3) group (P ¼ 0.016). Consistent with the results of our series, higher Bile duct invasion 0.063 FBP1 fi (þ) 3 (3.4) 4 (13.3) expression was also signi cantly associated with better RFS () 85 (96.6) 25 (83.3) in the available public data (GSE14520; P < 0.001). The associ- NA 0 (0.0) 1 (3.3) ation between lower FBP1 expression and poorer RFS remained Abbreviations: HBV, hepatitis B virus; HCV, hepatitis C virus; NA, not available. significant after excluding tumor stage as a confounding variable aWHO grades 1/2 or Edmondson-Steiner grades 1/2 were classified as well/ in the GSE14520 dataset (HR, 1.49; 95% CI, 1.01–2.20; P ¼ moderately; WHO grades 3/4 or Edmondson-Steiner grades 3/4 were classified 0.045; Table 2). as poorly/undifferentiated. We then performed GSEA to determine whether the gene sets bP < 0.05. associated with HCC recurrence were differentially regulated depending on the expression levels of FBP1 in HCC. GSEA showed FBP1 expression levels predicted survival in patients with HCC that higher FBP1 expression was significantly correlated with a To further investigate the clinical significance of FBP1 expres- lower risk of early recurrence (27) in the GSE14520 and TCGA sion, we assessed survival rates in three independent datasets of datasets (P < 0.001 and P < 0.001, respectively; Fig. 4E), and HCC. On the basis of the FBP1 expression level, cases were with that of late recurrence (28) in the GSE14520 and TCGA divided into two groups using the minimum P value approach, datasets (P < 0.001 and P < 0.001, respectively; Fig. 4F). These which is a comprehensive methodtoidentifytheoptimalrisk results demonstrated that higher FBP1 expression was associated separation cutoff point in continuous gene-expression mea- with a lower risk of recurrence. surements for survival analysis in multiple datasets (24). The cutoff values for high and low FBP1 expression groups were 0.6 FBP1 expression levels reflected altered glucose (FBP1/GAPDH expression) in the Kyushu HCC set, 6.8 (log2 metabolism in HCC expression) in the GSE14520 dataset, and 13.0 (log2 expres- To further investigate the role of FBP1 in HCC progression, we sion) in the TCGA dataset. For the analysis of 118 patients in focused on glucose metabolism. First, FBP1 expression and glu- the Kyushu HCC set, clinicopathologic factors were compared coneogenic activity were compared using genome-wide gene- between the groups (Table 1). Notably, the low FBP1 expres- expression profiles of the GSE14520 and TCGA datasets. To sion group exhibited larger tumor size (P ¼ 0.010) and poorer comprehensively analyze gluconeogenic activity, we defined the differentiation grade (P < 0.001) than the high FBP1 expression gluconeogenic module composed of genes specific to gluconeo- group. These findings were validated with the available public genesis (4–6). This module included genes encoding enzymes data. Lower FBP1 expression was significantly associated with involved in irreversible steps of gluconeogenesis, for example,

www.aacrjournals.org Cancer Res; 76(11) June 1, 2016 3271

Hirata et al.

Figure 4. FBP1 expression levels predicted survival and recurrence in patients with HCC. A, Kaplan–Meier OS curve based on FBP1 expression in three independent datasets of HCC. Left, Kyushu HCC set; middle, GSE14520; right, TCGA. B and C, gene set enrichment analysis of the GSE14520 and TCGA datasets showed that higher FBP1 expression was significantly correlated with better survival in HCC (B). In contrast, lower FBP1 expression was significantly correlated with poorer survival in HCC (C). D, Kaplan–Meier recurrence-free survival in patients with HCC based on FBP1 expression. Left, Kyushu HCC set; middle, GSE14520; right, information on recurrence status was not available in TCGA. E and F, gene set enrichment analysis of the GSE14520 and TCGA datasets revealed that upregulation of FBP1 expression was significantly correlated with a lower risk of early (E) and late (F) recurrence. ES, enrichment score.

3272 Cancer Res; 76(11) June 1, 2016 Cancer Research

FBP1 and Glucose Metabolism in HCC

Table 2. Effect of FBP1 expression on OS and recurrence-free survival in the multivariate analysis that excluded tumor stage as a confounding variable Univariate analysis Multivariate analysis Factor HR 95% CI P HR 95% CI P OS Kyushu HCC set FBP1 expression (low/high) 2.47 (1.28–4.77) 0.007b 2.49 (1.22–5.06) 0.012b TNM stage (II/I) 0.97 (0.45–2.10) 0.939 0.96 (0.44–2.07) 0.910 TNM stage (III/I) 2.12 (0.75–5.95) 0.156 1.37 (0.45–4.07) 0.577 GSE14520 FBP1 expression (low/high) 2.08 (1.37–3.14) <0.001b 1.45 (0.91–2.29) 0.116 TNM stage (II/I) 2.09 (1.20–3.64) 0.009b 1.99 (1.14–3.47) 0.016a TNM stage (III/I) 5.30 (3.05–9.20) <0.001b 4.64 (2.60–8.29) <0.001b TCGA FBP1 expression (low/high) 2.68 (1.23–5.85) 0.013b 2.84 (1.22–6.64) 0.016b TNM stage (II/I) 0.64 (0.31–1.29) 0.210 0.55 (0.27–1.13) 0.102 TNM stage (III/I) 1.44 (0.83–2.50) 0.196 1.35 (0.77–2.34) 0.292 Recurrence-free survivala Kyushu HCC set FBP1 expression (low/high) 1.79 (1.11–2.88) 0.017b 1.48 (0.87–2.51) 0.150 TNM stage (II/I) 1.67 (0.94–2.96) 0.081 1.657 (0.93–2.94) 0.084 TNM stage (III/I) 3.27 (1.49–7.18) 0.003b 2.74 (1.20–6.25) 0.017b GSE14520 FBP1 expression (low/high) 1.90 (1.34–2.70) <0.001b 1.49 (1.01–2.20) 0.045b TNM stage (II/I) 1.96 (1.29–2.99) <0.001b 1.86 (1.22–2.84) 0.004b TNM stage (III/I) 3.08 (1.95–4.86) <0.001b 2.62 (1.61–4.26) <0.001b Abbreviation: CI, conﬁdential interval. aInformation on recurrence status was not available in the TCGA dataset. bP < 0.05.

glucose-6-phosphatase, catalytic subunit (G6PC); phosphoenol- follows: KEGG_GLYCOLYSIS_GLUCONEOGENESIS in the pyruvate carboxykinase 1; phosphoenolpyruvate carboxykinase GES14520 (P < 0.001) and TCGA (P < 0.001) datasets (Fig. 2; and pyruvate carboxylase (Fig. 5A). Peroxisome proliferator- 5D); and REACTOME_GLUCOSE_METABOLISM in the activated receptor gamma, coactivator 1 alpha (PGC-1a), a tran- GES14520 (P < 0.001) and TCGA (P ¼ 0.023) datasets (Fig. scription factor regulating the expression of gluconeogenic genes 5E). The results of GSEA indicated that normal glucose metabo- (29), was also contained in this module. Consistent with a lism was significantly impaired in HCCs expressing low levels of previous report (4), all of these genes were expressed at lower FBP1 compared with that in HCCs expressing high levels of FBP1. levels in tumors than normal liver tissues supporting the obser- Finally, we assessed the effects of the other genes involved in vation that hepatic gluconeogenesis was reduced in HCC (Sup- gluconeogenesis/aerobic glycolysis on prognosis in patients plementary Fig. S3A–S3E). Moreover, we found a significant with HCC. Lower G6PC and PGC-1a expression and higher positive correlation between FBP1 expression levels and gluco- PKM2 expression were significantly associated with poor OS in neogenic module activity in the GSE14520 (Pearson correlation both the GSEA14520 and TCGA datasets (Supplementary Table coefficient ¼ 0.719; P < 0.001) and TCGA (Pearson correlation S3). Consistent with these observations, higher expression coefficient ¼ 0.716; P < 0.001) datasets (Fig. 5B). Our findings levels of genes specific to gluconeogenesis, particularly those suggested that hepatic gluconeogenesis was impaired in HCCs involved in maintenance of normal glucose metabolism (ele- expressing low levels of FBP1. vated gluconeogenesis and decreased aerobic glycolysis), were Subsequently, we evaluated aerobic glycolysis in HCC cases. significantly associated with better OS (Supplementary Fig. Three enzymes are involved in catalyzing the irreversible steps of S4A–S4C), suggesting that the glucose metabolic shift affected glycolysis: hexokinase, phosphofructokinase, and pyruvate kinase prognosis. Together with the results of functional analysis and (Fig. 5A). Among these three enzymes, we focused on the HK2 survival analysis, our results showed that FBP1 was not only (20, 21), PFK1 (22, 23), and pyruvate kinase M2 (PKM2;refs.30,31) one of the key regulators of HCC progression but a promising genes, which have been reported to play a pivotal role in promoting biomarker-predicting prognosis in patients with HCC. Among aerobic glycolysis in cancer cells; therefore, an aerobic glycolytic genes involved in gluconeogenesis, downregulation of FBP1 module containing these three genes was defined. Contrary with may contribute to a vicious cycle of altered glucose metabolism gluconeogenesis, significant inverse correlations were observed in HCC considering the observation that loss of FBP1 directly between FBP1 expression and aerobic glycolytic module activity enhanced tumor progression and aerobic glycolysis. Overall, in the GSE14520 (Pearson correlation coefficient ¼0.458; we found that lower FBP1 expression was significantly associ- P < 0.001) and TCGA (Pearson correlation coefficient ¼0.536; ated with altered glucose metabolism, which contributed to P < 0.001) datasets (Fig. 5C), indicating that lower FBP1 expression tumor progression in HCC. was associated with high aerobic glycolysis in HCC cases. To confirm whether FBP1 expression levels were associated with altered glucose metabolism in HCC, GSEA was performed. Discussion Predefined gene sets involved in normal glucose metabolism were We demonstrated the function and clinical significance of FBP1 significantly enriched in HCCs expressing high levels of FBP1,as expression in HCC. Promoter methylation and copy-number loss

www.aacrjournals.org Cancer Res; 76(11) June 1, 2016 3273

Hirata et al.

Figure 5. FBP1 expression levels reflected altered glucose metabolism in HCC cases. A, illustration of the glucose metabolic pathway. Irreversible gluconeogenic enzyme genes, that is, G6PC; phosphoenolpyruvate carboxykinase 1 (PEPCK1); phosphoenolpyruvate carboxykinase 2 (PEPCK2); pyruvate carboxylase (PC); and peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PGC-1a), a transcription factor regulating G6PC and PEPCK1, are highlighted in red. Enzymes catalyzing three irreversible steps of glycolysis and isoforms involved in aerobic glycolysis, that is, HK2, PFK1,andPKM2, are highlighted in blue. B, correlations between FBP1 expression level and gluconeogenic activity in the GSE14520 and TCGA datasets are shown as heat maps. The module activity of gluconeogenesis was calculated for each case using the average of the normalized expression of genes specific to gluconeogenesis. C, correlations between FBP1 expression levels and the activity of aerobic glycolysis in the GSE14520 and TCGA datasets are shown as heat maps. The module activity of aerobic glycolysis was calculated for each case using the average of the normalized expression of HK2, PFK1, and PKM2. D and E, gene set enrichment analysis demonstrated that two predefined gene sets involved in normal glucose metabolism were significantly enriched in HCC cases exhibiting high FBP1 expression, suggesting that normal glucose metabolism was preserved in these cases. ES, enrichment score.

were responsible for the decreased expression of FBP1, which In the present study, we validated the prognostic impact of contributed to tumor progression, poor prognosis, and altered FBP1 expression using three independent large cohorts, and the glucose metabolism. Our data indicated that FBP1 acted as a results were reproducible. For either OS or RFS in each cohort, this tumor suppressor by restoring altered glucose metabolism, and prognostic impact persisted even after adjusting for tumor stage. implied that FBP1 was a more central regulator of the glucose Moreover, our data indicated that decreased expression of FBP1 metabolic pathway in HCC. promoted tumor progression, thereby affecting the clinical

3274 Cancer Res; 76(11) June 1, 2016 Cancer Research

FBP1 and Glucose Metabolism in HCC

outcomes and clinicopathologic characteristics of patients with In summary, we demonstrated the clinical significance of FBP1 HCC. Therefore, FBP1 expression could be a dependable bio- expression in HCC using three independent clinical datasets. marker for predicting prognosis and provide additional prognos- Downregulation of FBP1 via promoter methylation and copy- tic information beyond tumor stage in patients with HCC. number loss promoted tumor progression by altering glucose The proposed mechanism underlying the tumor-suppressive metabolism. FBP1 could be a useful biomarker for predicting the properties of FBP1 involves the inhibition of aerobic glycolysis prognosis of patients with HCC and may represent a novel (6, 7). Our results were consistent with this proposal and sug- therapeutic target for the treatment of HCC. gested that the reduced transcriptional activity of HK2 and PFK1 induced by ectopic FBP1 played a role in the decreased glucose-6- Disclosure of Potential Conflicts of Interest phosphate and F-1,6-BP levels. One of the limitations of this No potential conflicts of interest were disclosed. study was that glucose metabolism was mainly examined by fi fi transcriptional pro ling of genes speci c to gluconeogenesis or Authors' Contributions glycolysis in clinical cases. Enzyme activity can also be regulated Conception and design: H. Hirata, K. Sugimachi, K. Mimori by protein structure and expression. Indeed, FBP1 is involved in Development of methodology: H. Hirata, H. Komatsu, M. Ueda, T. Masuda, the posttranslational modification of PKM2 in the glycolytic R. Uchi pathway (30, 31). The loss of FBP1 directly increases the level of Acquisition of data (provided animals, acquired and managed patients, F-1,6-BP, which enhances PKM2 activity through dynamic struc- provided facilities, etc.): H. Hirata, K. Sugimachi, H. Komatsu, H. Eguchi, S. Ito tural switching from the less active dimer to the active tetramer. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Hirata, K. Sugimachi, H. Komatsu, M. Ueda, Furthermore, the effect of FBP1 knockdown on altered glucose FBP1 T. Masuda, R. Uchi, S. Sakimura, S. Nambara, T. Saito, M. Hirakawa metabolism was not evaluated because was expressed at low Writing, review, and/or revision of the manuscript: H. Hirata, K. Sugimachi, levels in all available liver cancer cell lines from the CCLE. Despite T. Iguchi, K. Mimori these limitations, our findings provide a reasonable explanation Administrative, technical, or material support (i.e., reporting or organizing for the effects of FBP1 expression on altered glucose metabolism data, constructing databases): K. Terashima, K. Sakamoto in HCC. In addition to the enzymatic activity of FBP1 in glucose Study supervision: Y. Shinden, H. Honda, K. Mimori metabolism, Li and colleagues (6) reported that the direct binding of FBP1 to hypoxia-inducible factor (HIF) inhibits ccRCC pro- Acknowledgments gression by reducing HIF activity. Our data showed that the effects This research used the supercomputing resource provided by the Human Genome Center, Institute of Medical Science, the University of Tokyo (http://sc. of FBP1 on malate, citrate, and other metabolites were not as hgc.jp/shirokane.html). The authors thank Dr. A. Niida for providing bioin- robust as that seen in ccRCC cells, likely due to the fact that ccRCC formatics tools, K. Oda, M. Kasagi, T. Kawano, and J. Takano for technical cells have constitutive HIF activity and HCC cells do not in assistance, Hiroshima Red Cross Hospital, Iizuka Hospital, and Oita Red Cross normoxic conditions. Considering these diverse functions of Hospital for providing clinical samples. The authors also appreciate the tech- FBP1, further studies are needed to clarify the detailed mechan- nical assistance from The Research Support Center, Research Center for Human isms through which FBP1 suppresses cancer progression in HCC. Disease Modeling, Kyushu University. In HCC, FBP1 promoter methylation has been reported only in cell lines and a few clinical cases (8). However, in our larger Grant Support analysis in this study, we found that copy-number loss and This work was supported in part by the Japan Society for the Promotion of fi promoter methylation could explain the decreased expression of Science Grant-in-Aid for Scienti c Research (grant numbers 24592005, FBP1 FBP1 FBP1 26861003, 15K10168, and 15H04921). . promoter methylation downregulated expres- The costs of publication of this article were defrayed in part by the sion in patients with basal-like breast cancer (7), gastric cancer payment of page charges. This article must therefore be hereby marked (32), and small intestinal neuroendocrine tumor (33). Copy- advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate number loss of FBP1 was also observed in ccRCC cases (34). Thus, this fact. our data indicated that genomic alterations as well as epigenetic modification were responsible for inhibiting FBP1 expression in Received September 23, 2015; revised March 28, 2016; accepted March 31, HCC. 2016; published OnlineFirst April 6, 2016.

References 1. Farazi PA, DePinho RA. Hepatocellular carcinoma pathogenesis: from 6. Li B, Qiu B, Lee DS, Walton ZE, Ochocki JD, Mathew LK, et al. Fructose-1,6- genes to environment. Nat Rev Cancer 2006;6:674–87. bisphosphatase opposes renal carcinoma progression. Nature 2014;513: 2. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the War- 251–5. burg effect: the metabolic requirements of cell proliferation. Science 7. Dong C, Yuan T, Wu Y, Wang Y, Fan TW, Miriyala S, et al. Loss of FBP1 by 2009;324:1029–33. Snail-mediated repression provides metabolic advantages in basal-like 3. Lemaigre FP, Rousseau GG. Transcriptional control of genes that regulate breast cancer. Cancer Cell 2013;23:316–31. glycolysis and gluconeogenesis in adult liver. Biochem J 1994;303:1–14. 8. Chen M, Zhang J, Li N, Qian Z, Zhu M, Li Q, et al. Promoter hypermethyla- 4. Wang B, Hsu SH, Frankel W, Ghoshal K, Jacob ST. Stat3-mediated activation mediated downregulation of FBP1 in human hepatocellular carcino- tion of microRNA-23a suppresses gluconeogenesis in hepatocellular car- ma and colon cancer. PLoS ONE 2011;6:e25564. cinoma by down-regulating glucose-6-phosphatase and peroxisome pro- 9. Roessler S, Jia HL, Budhu A, Forgues M, Ye QH, Lee JS, et al. A unique liferator-activated receptor gamma, coactivator 1 alpha. Hepatology 2012; metastasis gene signature enables prediction of tumor relapse in early-stage 56:186–97. hepatocellular carcinoma patients. Cancer Res 2010;70:10202–12. 5. Ma R, Zhang W, Tang K, Zhang H, Zhang Y, Li D, et al. Switch of glycolysis to 10. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, gluconeogenesis by dexamethasone for treatment of hepatocarcinoma. Nat et al. The Cancer Cell Line Encyclopedia enables predictive modelling of Commun 2013;4:2508. anticancer drug sensitivity. Nature 2012;483:603–7.

www.aacrjournals.org Cancer Res; 76(11) June 1, 2016 3275

Hirata et al.

11. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. 23. Moreno-Sanchez R, Marin-Hernandez A, Gallardo-Perez JC, Quezada H, Integrative analysis of complex cancer genomics and clinical profiles using Encalada R, Rodriguez-Enriquez S, et al. Phosphofructokinase type 1 the cBioPortal. Sci Signal 2013;6:pl1. kinetics, isoform expression, and gene polymorphisms in cancer cells. 12. Hlady RA, Tiedemann RL, Puszyk W, Zendejas I, Roberts LR, Choi JH, et al. J Cell Biochem 2012;113:1692–703. Epigenetic signatures of alcohol abuse and hepatitis infection during 24. Mizuno H, Kitada K, Nakai K, Sarai A. PrognoScan: a new database for human hepatocarcinogenesis. Oncotarget 2014;5:9425–43. meta-analysis of the prognostic value of genes. BMC Med Genomics 2009; 13. Habano W, Kawamura K, Iizuka N, Terashima J, Sugai T, Ozawa S. Analysis 2:18. of DNA methylation landscape reveals the roles of DNA methylation in the 25. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, regulation of drug metabolizing enzymes. Clin Epigenetics 2015;7:105. et al. Gene set enrichment analysis: a knowledge-based approach for 14. Ohashi Y, Hirayama A, Ishikawa T, Nakamura S, Shimizu K, Ueno Y, et al. interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A Depiction of metabolome changes in histidine-starved Escherichia coli by 2005;102:15545–50. CE-TOFMS. Mol Biosyst 2008;4:135–47. 26. Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, et al. Classification and 15. Ooga T, Sato H, Nagashima A, Sasaki K, Tomita M, Soga T, et al. Meta- prediction of survival in hepatocellular carcinoma by gene expression bolomic anatomy of an animal model revealing homeostatic imbalances profiling. Hepatology 2004;40:667–76. in dyslipidaemia. Mol Biosyst 2011;7:1217–23. 27. Woo HG, Park ES, Cheon JH, Kim JH, Lee JS, Park BJ, et al. Gene expression- 16. Sugimoto M, Wong DT, Hirayama A, Soga T, Tomita M. Capillary electro- based recurrence prediction of hepatitis B virus-related human hepatocel- phoresis mass spectrometry-based saliva metabolomics identified oral, lular carcinoma. Clin Cancer Res 2008;14:2056–64. breast and pancreatic cancer-specificprofiles. Metabolomics 2010;6:78–95. 28. Hoshida Y, Villanueva A, Kobayashi M, Peix J, Chiang DY, Camargo A, et al. 17. Muoio DM.TXNIP links redox circuitry to glucose control. Cell Metab Gene expression in fixed tissues and outcome in hepatocellular carcinoma. 2007;5:412–4. N Engl J Med 2008;359:1995–2004. 18. Stoltzman CA, Peterson CW, Breen KT, Muoio DM, Billin AN, Ayer DE. 29. Yoon JC, Puigserver P, Chen G, Donovan J, Wu Z, Rhee J, et al. Control of Glucose sensing by MondoA:Mlx complexes: a role for hexokinases and hepatic gluconeogenesis through the transcriptional coactivator PGC-1. direct regulation of thioredoxin-interacting protein expression. Proc Natl Nature 2001;413:131–8. Acad Sci U S A 2008;105:6912–7. 30. Christofk HR, Vander Heiden MG, Harris MH, Ramanathan A, Gerszten RE, 19. Peterson CW, Stoltzman CA, Sighinolfi MP, Han KS, Ayer DE. Glucose Wei R, et al. The M2 splice isoform of pyruvate kinase is important for controls nuclear accumulation, promoter binding, and transcriptional cancer metabolism and tumour growth. Nature 2008;452:230–3. activity of the MondoA-Mlx heterodimer. Mol Cell Biol 2010;30:2887–95. 31. Yang W, Lu Z. Regulation and function of pyruvate kinase M2 in cancer. 20. Mathupala SP, Rempel A, Pedersen PL. Glucose catabolism in cancer cells: Cancer Lett 2013;339:153–8. identification and characterization of a marked activation response of the 32. Liu X, Wang X, Zhang J, Lam EK, Shin VY, Cheng AS, et al. Warburg effect type II hexokinase gene to hypoxic conditions. J Biol Chem 2001;276: revisited: an epigenetic link between glycolysis and gastric carcinogenesis. 43407–12. Oncogene 2010;29:442–50. 21. Patra KC, Wang Q, Bhaskar PT, Miller L, Wang Z, Wheaton W, et al. 33. Karpathakis A, Dibra H, Pipinikas C, Feber A, Morris T, Francis JM, et al. Hexokinase 2 is required for tumor initiation and maintenance and its Prognostic impact of novel molecular subtypes of small intestinal neuro- systemic deletion is therapeutic in mouse models of cancer. Cancer Cell endocrine tumour. Clin Cancer Res 2016;22;250–8. 2013;24:213–28. 34. Dondeti VR, Wubbenhorst B, Lal P, Gordan JD, D'Andrea K, Attiyeh EF, 22. Yi W, Clark PM, Mason DE, Keenan MC, Hill C, Goddard WAIII, , et al. et al. et al. Integrative genomic analyses of sporadic clear cell renal cell carcinoma Phosphofructokinase 1 glycosylation regulates cell growth and metabo- define disease subtypes and potential new therapeutic targets. Cancer Res lism. Science 2012;337:975–80. 2012;72:112–21.

3276 Cancer Res; 76(11) June 1, 2016 Cancer Research

Decreased Expression of Fructose-1,6-bisphosphatase Associates with Glucose Metabolism and Tumor Progression in Hepatocellular Carcinoma

Hidenari Hirata, Keishi Sugimachi, Hisateru Komatsu, et al.

Cancer Res 2016;76:3265-3276. Published OnlineFirst April 6, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-15-2601

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2016/04/06/0008-5472.CAN-15-2601.DC1

Cited articles This article cites 34 articles, 12 of which you can access for free at: http://cancerres.aacrjournals.org/content/76/11/3265.full#ref-list-1

Citing articles This article has been cited by 12 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/76/11/3265.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/76/11/3265. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.