Open Full Page

Published OnlineFirst January 27, 2016; DOI: 10.1158/1078-0432.CCR-15-1987

Biology of Human Tumors Clinical Cancer Research NDUFA4L2 Fine-tunes Oxidative Stress in Hepatocellular Carcinoma Robin Kit-Ho Lai1, Iris Ming-Jing Xu1, David Kung-Chun Chiu1, Aki Pui-Wah Tse1, Larry Lai Wei1, Cheuk-Ting Law1, Derek Lee1, Chun-Ming Wong1,2, Maria Pik Wong1, Irene Oi-Lin Ng1,2, and Carmen Chak-Lui Wong1,2

Abstract

Purpose: Hepatocellular carcinoma (HCC) lacks effective cura- Results: NDUFA4L2 was drastically overexpressed in human tive therapy. Hypoxia is commonly found in HCC. Hypoxia elicits HCC and induced by hypoxia. NDUFA4L2 overexpression was a series of protumorigenic responses through hypoxia-inducible closely associated with tumor microsatellite formation, absence factor-1 (HIF1). Better understanding of the metabolic adapta- of tumor encapsulation, and poor overall survival in HCC tions of HCC cells during hypoxia is essential to the design of new patients. We confirmed that NDUFA4L2 was HIF1-regulated in therapeutic regimen. HCC cells. Inactivation of HIF1/NDUFA4L2 increased mitochon- Experimental Design: Expressions of genes involved in the drial activity and oxygen consumption, resulting in ROS accu- electron transport chain (ETC) in HCC cell lines (20% and 1% O2) mulation and apoptosis. Knockdown of NDUFA4L2 markedly and human HCC samples were analyzed by transcriptome suppressed HCC growth and metastasis in vivo. HIF inhibitor, sequencing. Expression of NDUFA4L2, a less active subunit in digoxin, significantly suppressed growth of tumors that expressed complex I of the ETC, in 100 pairs of HCC and nontumorous liver high level of NDUFA4L2. tissues were analyzed by qRT-PCR. Student t test and Kaplan– Conclusions: Our study has provided the first clinical relevance Meier analyses were used for clinicopathologic correlation and of NDUFA4L2 in human cancer and suggested that HCC patients survival studies. Orthotopic HCC implantation model was used with NDUFA4L2 overexpression may be suitable candidates for to evaluate the efficiency of HIF inhibitor. HIF inhibitor treatment. Clin Cancer Res; 22(12); 3105–17. 2016 AACR.

Introduction accompanied by the loss of normal metabolic functions in the liver and the acquisition of new metabolic functions which Hepatocellular carcinoma (HCC), a malignancy derived favor cancer growth. Exploration of the molecular contexts from hepatocytes, accounts for 90% of primary liver cancer. associated with these metabolic changes will help to identify It is the ﬁfthmostcommoncancerandthethirdleadingcause novel targets for HCC treatment. of cancer deaths in the world (1). Majority of deaths in HCC are Hypoxia, or oxygen (O ) deprivation, is frequently found in attributed to its asymptomatic nature that delays diagnosis and 2 regions of HCC that are distant from functional vasculature. treatment. Most HCC patients are not suitable for the only Palliative therapies, such as transcatheter arterial (chemo) promising curative therapies, surgical resection, and liver trans- embolization (TAE/TACE) and hepatic artery ligation, that plantation. Sorafenib, an oral multikinase inhibitor and the involve restriction of blood supply to the tumors adversely only FDA-approved drug for advanced HCC patients, could induce hypoxia (4). To overcome the shortage of O , cells adapt modestly prolong the survival of patients for 3 months (2, 3). 2 to hypoxia through HIFs which are composed of the constitu- Liver is an organ responsible for many important metabolic tively expressed HIF1b subunit and the oxygen labile subunit functionsinthebody,suchastheCori-cycle,glycogenmetab- HIF1/2a (5). In the presence of O ,HIF1/2a is hydroxylated by olism, and blood glucose homeostasis. Hepatocarcinogenesis is 2 prolyl hydroxylases (6), facilitating binding of von Hippel– Lindau (VHL), which polyubiquitinates HIF1/2a for proteaso- mal degradation (7). In the absence of O2, stabilized HIF1/2a 1Department of Pathology, The University of Hong Kong, Hong Kong. dimerizes with HIF1b to initiate transcription of genes related 2State Key Laboratory for Liver Research, The University of Hong to hypoxia adaptive responses that advantage cancer develop- Kong, Hong Kong. ment (8). Although it is known that upregulation of HIF1 is Note: Supplementary data for this article are available at Clinical Cancer closely associated with poor clinical outcome in HCC patients Research Online (http://clincancerres.aacrjournals.org/). (4), the detailed molecular mechanisms by which HIF1 pro- Corresponding Authors: Carmen Chak-Lui Wong, The University of Hong Kong, motes HCC progression remain poorly understood. L-704 Laboratory Block, LKS Faculty of Medicine, 21 Sassoon Road, Pokfulam, HIF1a transcriptionally activates many metabolic genes, Hong Kong, Hong Kong. Phone: 852-39179014; Fax: 852-28195375l; E-mail: allowing the cells to adapt to hypoxia, by shunting the glucose [email protected]; Irene Oi-Lin Ng, [email protected] intermediates into glycolysis instead of the tricarboxylic acid doi: 10.1158/1078-0432.CCR-15-1987 (TCA) cycle (9). These metabolic genes include glucose trans- 2016 American Association for Cancer Research. porter (GLUT1), hexokinase 2 (HK2), lactate dehydrogenase A

www.aacrjournals.org 3105

Lai et al.

alpha subcomplex 4–like 2 (NDUFA4L2), to reduce complex I Translational Relevance activity and ROS production (23). Hepatocellular carcinoma (HCC) frequently experiences Currently, knowledge about NDUFA4L2, particularly on its hypoxia. Hypoxia results in an inefficient transfer of electrons roles in cancer development, is scarce. NDUFA4L2 was found to during oxidative phosphorylation leading to increased oxida- be one of the seven biomarkers that distinguish medullary tive stress. In this study, we demonstrated that a less active thyroid carcinoma from head and neck paraganglioma (PGL; complex I subunit in the electron transport chain, NDUFA4L2, ref. 24). Earlier study showed that NDUFA4L2 is overexpressed was significantly overexpressed in HCC and other human in PGL tumors that lack VHL (25). Nonetheless, its clinical cancer types. Overexpression of NDUFA4L2 was associated implications and in vivo roles in cancers, particularly in HCC, with aggressive clinical features in human HCC and shorter have never been thoroughly studied. This study uncovers the overall survival in HCC patients. A series of in vitro and in vivo clinical relevance and roles of NDUFA4L2 in REDOX homeo- assays converged to show that NDUFA4L2 reduced ROS- stasis in HCC. Furthermore, we have successfully demonstrated mediated apoptosis to confer HCC cells growth advantages. thattargetingHIF1/NDUFA4L2pathwaybyHIFinhibitor As NDUFA4L2 is a direct transcriptional target of HIF, we represents a novel therapeutic strategy for HCC. found that HIF inhibitor, digoxin, profoundly inhibited growth of tumors that expressed high level of NDUFA4L2 in orthotopic model. Our findings suggested that cancer patients Materials and Methods with NDUFA4L2 overexpression may be suitable candidates Patient samples for HIF inhibitor treatment. Human HCC and the corresponding paired nontumorous liver tissues were collected at Queen Mary Hospital, the University of Hong Kong during surgical resection. Human lung squamous cell carcinoma (SCC) and the corresponding paired nontumorous lung tissues samples were kindly provided by Dr. Maria P. Wong (LDHA), and pyruvate dehydrogenase kinase 1 (PDK1). GLUT1 (the University of Hong Kong). Use of human samples was facilitates glucose uptake (10). HK2 catalyzes the phosphory- approved by the Institutional Review Board of the University of lation of glucose to glucose-6-phosphate, the first step of glycol- Hong Kong/Hospital Authority Hong Kong West Cluster. ysis (11, 12). LDHA converts pyruvate to lactate (13, 14). PDK1 inactivates pyruvate dehydrogenase to prevent pyruvate conver- Cell lines sion to acetyl CoA. These processes restrict the entrance of glucose Human HCC cell line, PLC/PRF/5, and human cervical cancer intermediates into the TCA cycle, thereby reducing the activity of cell line, HeLa, were obtained from the American Type Culture oxidative phosphorylation (OXPHOS) in the mitochondria (9). Collect (ATCC) and cultured according to ATCC recommenda- Interestingly, all these metabolic genes are involved in cancer tions. Metastatic human HCC cell line, MHCC97L, was a gift from progression. Dr. Z.Y. Tang (Fudan University of Shanghai). All the cell lines The OXPHOS system comprises four electron transport chain used were authenticated by the AuthentiFiler PCR Amplification (ETC) complexes, including complex I (NADH–ubiquinone oxi- Kit (Applied Biosystems) on September 1st, 2014. All cell lines doreductase), complex II (succinate:ubiquinone oxidoreductase), used in this article were thawed from the authenticated cell stock complex III (ubiquinol–cytochrome c reductase), and complex IV and used within four passages. (cytochrome c oxidase). Most ATP in the cells is produced when electrons transfer through these complexes to the ultimate elec- Clinicopathologic correlation and patients' survival analysis tron acceptor, O2. Complex I, a 1 MDa complex of 45 subunits, is the first step where OXPHOS takes place. Complex I catalyzes the The clinicopathologic features of HCC patients were ana- transfer of electrons from NADH to flavoprotein through eight lyzed by pathologist as we previously described (26). The iron–sulfur clusters, and finally to ubiquinone. Complex I is a parameters included age, gender, tumor size, cellular differen- major ROS-producing site. Low level of ROS has been shown to tiation by Edmondson grading, direct liver invasion, absence of activate signaling pathways, such as MAPK-ERK, JNK, p38 MAPK, tumor encapsulation, presence of tumor microsatellite forma- and FAK, to promote cancer cell survival and metastasis (15), tion, venous invasion, and background liver disease. Overall whereas excessive ROS accumulation suppresses cancer cell survival was calculated from the date of surgical resection to the fi growth through induction of G –M cell-cycle arrest and apoptosis date of death or last follow-up. The prognostic signi cance of 2 – (16–18). Most of the chemotherapeutic and radiotherapeutic NDUFA4L2 overexpression was determined by the Kaplan strategies against cancer cells are mediated through ROS induc- Meier method followed by the log-rank test, as we previously tion (19, 20). described (27, 28). All statistical tests were performed by HIF1a regulates several ETC components to minimize ROS SPSS20.0. The demographic data of HCC patients were sum- production and optimize mitochondrial respiration. HIF1a marized in Supplementary Table S1. induces the switching of COX subunit 4 isoforms from COX4- 1 to COX4-2 in complex IV to maximize the efficiency of OXPHOS Fluorescence imaging and flow cytometry analysis under hypoxia (21). HIF1a induces expression of a miRNA, miR- Cellsweregrownonglasscoverslipsin6-wellcultureplate – 210, which suppresses iron sulfur cluster assembly proteins andculturedin20%and1%O2 for 24 hours. Cells were stained (ISCU1/2) in complexes I and III to reduce oxygen consumption, with 2 mmol/L 5,5',6,6'-tetrachloro-1,1',3,3' -tetraethyl-benzi- ROS production, and apoptosis during hypoxic stress (22). Main- midazolylcarbocyanine chloride (JC-1; Invitrogen). Images ly demonstrated in mouse embryonic fibroblasts, HIF1a induced were captured and scanned under 20 magnification with LSM ETC complex I subunit, NADH dehydrogenase (ubiquinone) 1 510 Meta laser scanning microscope (Carl Zeiss) connected to

3106 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

Axiocam microscope camera (Carl Zeiss). For flow cytometry described above. For subcutaneous tumor model, 1 106 analysis, cells were stained with 2 mmol/L JC-1. MHCC97L cells were injected subcutaneously to the flanks of nude mice. Tumor volume was calculated with the formula: 3 Oxygen consumption assay length (mm) width (mm) depth (mm) 0.52 (mm ). For Cells were grown on 96-well culture plate and cultured in pharmacologic studies, drug administration began a week after orthotopic or subcutaneous injection. Mice were administered 20% and 1% O2 for 24 hours. Cells were stained with oxygen- with 1.2 mg/kg/day digoxin or vehicles (saline) by intraperi- sensitive probe MitoXpress-Intra (NanO2) and washed according to the manufacturer's instruction (Luxcel Bioscience). Phos- toneal injection for 21 consecutive days. phorescence was measured by a multilabel reader Victor2 (Perkin-Elmer Life Sciences) with 340 nm excitation and 642 Transcriptome sequencing nm emission filters. Transcriptome sequencing was performed in 16 pairs of HCC tissues and corresponding NT liver tissues or MHCC97L cells exposed to 20% and 1% O for 24 hours. TruSeq standard ROS measurement 2 mRNA sample Prep kit (Illumina) was used for polyA þ mRNA Trypsinized cells were washed with PBS and stained with 2 0 0 library preparation. Illumina HiSeq2000 was employed for mmol/L general ROS indicator chloromethyl-2 ,7 -dichlorodihy- 100 bp paired-end sequencing (Axeq Technologies), and data drofluorescein diacetate (CM-H DCFDA; Life Technologies) or 5 2 were analyzed by TopHat-Cufflinks pipeline (29). Values were mmol/L mitochondrial ROS indicator MitoSOX (Molecular indicated by FPKM (fragments per kilobase transcript sequence Probes) for 5 and 10 minutes, respectively, and analyzed by flow per million mapped reads). Pathway analysis was performed by cytometer FACSCanto II Analyzer (BD Biosciences). For MitoSOX the Database for Annotation, Visualization and Integrated staining, cells were washed with staining buffer 3 times before Discovery (DAVID). flow cytometry analysis. Data from flow cytometry studies were analyzed by FlowJo software. For antioxidant treatment, 0.5, 2, or 5 mmol/L of N-acetyl-L-cysteine (NAC) or L-ascorbic acid (Sigma Statistical analysis t Aldrich) were incubated at 37C for 24 hours prior to ROS Statistical analysis was performed by two-tailed Student test measurement. For in vivo ROS measurement, tumors harvested or Wilcoxon's signed rank test using GraphPad Prism 5.0 from mice were dissociated with gentleMACS dissociator, and (GraphPad Software Inc.). All functional assays are represen- single-cell suspensions were stained with 2 mmol/L CM- tative of 3 independent experiments and expressed as mean SEM. P value less than 0.05 was considered to be statistically H2DCFDA for flow cytometry analysis. significant. Apoptosis and cell proliferation assays Quantitative real-time PCR, establishment of knockdown For apoptosis measurement, 2.5 105 cells were cultured in and knockout HCC cells, in vitro assays, and TCGA/ 20% and 1% O for 24 hours. Apoptosis was determined by flow 2 Oncomine data cytometry after staining with propidium iodide (Calbiochem) The following are described in Supplementary Materials and Annexin V (MBL International Corporation). For antioxidant and Methods: quantitative real-time PCR (qRT-PCR), chromatin treatment, 0.1 mmol/L glutathione ethyl ester (GSH-EE; Cayman immunoprecipitation (ChIP) assay, establishment of HIF or Chemical) was added to MHCC97L-shL2 cells and cultured in 1% NDUFA4L2 knockdown and HIF knockout HCC cells, mitochon- O for 40 hours. For cell proliferation measured by cell counting, 2 2 drial membrane potential, mitochondrial mass, migration assay, 104 MHCC97L-NTC and -shL2 cells were grown on 12-well histology, and The Cancer Genome Atlas (TCGA)/Oncomine culture plate and cultured in 20% and 1% O for 4 days. Cells were 2 data. trypsinized and counted with Z1 coulter particle counter (Beck- man Coulter) every 24 hours. For cell proliferation measured by 5- bromo-20-deoxyuridine (BrdUrd) assay, 2.5 104 MHCC97L- Results NTC and -shL2 cells were grown on 96-well culture plate and Mitochondrial NDUFA4L2 is frequently overexpressed in cultured in 20% and 1% O2 for 24 hours. BrdUrd labeling and human HCC and other human solid cancers colorimetric measurement were performed according to the man- To better understand the hypoxia-adaptation system in ufacturer's instructions (Roche Life Science). HCC, we performed transcriptome sequencing to compare the gene expression profiles of a HCC cell line, MHCC97L, that Animal studies were exposed to normoxia (20% O2) and hypoxia (0.1% and All animal studies were approved by the Committee on the 1% O2). Among the upregulated genes, mitochondrial NDU- Use of Live Animals in Teaching and Research, the University of FA4L2 was one of the five most abundant genes in hypoxic Hong Kong, and followed under the Animals (Control of conditions, indicating the biologic relevance of this gene in Experiments) Ordinance of HongKong.Fororthotopicliver human HCC. Other well-characterized HIF target genes, tumor injection in nude mice, 1 106 luciferase-labeled including ANPTGL4 (30), PLOD2 (31), VEGF (32), CA9 MHCC97L cells were injected into the left lobes of the livers (33), and P4HA2 (34), were on the top 20 of the list. Intrigu- of 6- to 8-week-old BALB/c nude mice. Six weeks after injection, ingly, when we interrogated the expression of genes that are mice were administered with 100 mg/kg D-luciferin by peri- involved in the mitochondrial complex I, only the subunit toneal injection 5 minutes before bioluminescent imaging NDUFA4L2 was distinctly and dramatically upregulated in (IVIS 100 Imaging System; Xenogen). Livers and lungs were hypoxia (both 0.1% and 1% O2), whereas all the other sub- harvested for ex-vivo imaging and histologic analysis. Livers units remained unaffected or downregulated (Fig. 1A). We were harvested for ex-vivo imaging and histologic analysis as further examined the transcriptome sequencing data on 16

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3107

Lai et al.

3108 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

Figure 2. NDUFA4L2 is regulated by HIF1a but not HIF2a in HCC. A, mRNA and protein expression of NDUFA4L2 in MHCC97L-EV, -shHIF1a, and -shHIF2a cells / cultured in 20% and 1% O2 for 24 and 48 hours, respectively. B, NDUFA4L2 protein expression in MHCC97L-WT (parental) and -HIF1a (KO) cells cultured in 20% and 1% O2 for 48 hours. C, NDUFA4L2 mRNA and protein expression in MHCC97L cells treated with HIF inhibitor digoxin at the indicated doses. D, putative HREs are mapped in the promoter of NDUFA4L2. Core HIF binding sequences (50-RCGTG-30) are highlighted as bold with underlines, while ancillary sequence is highlighted as bold. E, ChIP assay was performed with IgG and HIF1a antibodies in MHCC97L cells cultured in 20% and 1% O2. qRT-PCR was performed after immunoprecipitation with anti-IgG or anti-HIF1a antibodies, represented as fold enrichment normalized to 20% O2 (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.05; , P < 0.01; , P < 0.001 as indicated).

Figure 1. NDUFA4L2 is overexpressed in human HCC and other human solid cancers. A, transcriptome sequencing data comparing FPKM values of genes encoding mitochondrial complex I subunits in MHCC97L cells cultured in (left) 0.1% O2 or (middle) 1% O2 relative to 20% O2 and (right) 16 pairs of human HCC and the corresponding NT tissues. B, mRNA expression levels of NDUFA4L2 normalized with hypoxanthine-guanine phosphoribosyltransferase in an expanded cohort of 100 cases of paired human HCC, and corresponding NT tissues were determined by qRT-PCR. Waterfall plot shows that NDUFA4L2 is overexpressed in 71% of human HCC samples by at least 2-fold. C, clinicopathologic analysis shows that NDUFA4L2 overexpression is closely associated with more aggressive features in HCC, including the presence of tumor microsatellite formation and the absence of tumor encapsulation. HCC patients with high NDUFA4L2 expression (= median) tend to have lower 5-year overall survival rate (Student t test; P ¼ 0.086). D, mRNA expression levels of NDUFA4L2 obtained from TCGA database of 49 paired human HCC and corresponding NT tissues (RSEM, RNA-Seq by Expectation Maximization). E, mRNA expression levels of NDUFA4L2 normalized with b-actin in 31 cases of paired human lung SCC, and corresponding NT tissues were determined by qRT-PCR. Waterfall plot shows that NDUFA4L2 is overexpressed in 65% of human lung SCC samples by at least 2-fold. (Wilcoxon's signed rank test; , P < 0.001 for B, D, and E. Student t test; , P < 0.05; , P < 0.01 for C).

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3109

Lai et al.

3110 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

primary HCCs and their corresponding nontumorous (NT) antibody was used, as compared with the IgG control (Fig. 2E livers and found that NDUFA4L2 was substantially overex- and Supplementary Fig. S3). Intriguingly, we consistently pressed in HCC tissues (Fig. 1A). Moreover, overexpression of detected a band beneath the expected protein size of NDU- NDUFA4L2 mRNA was validated by qRT-PCR in a separate, FA4L2. This protein was only expressed in normoxia but not in larger cohort of 100 pairs of HCC samples (Fig. 1B). Further- hypoxia (Figs. 2A, B, and C, and 3A and B). NDUFA4L2 shared more, overexpression of NDUFA4L2 in HCC tumors was sig- 67% amino acid sequence with its paralogue NDUFA4, which nificantly associated with aggressive pathologic features, is 6 amino acids shorter than NDUFA4L2. To confirm the including presence of tumor microsatellite formation (Student identity of the unknown protein beneath NDUFA4L2, we t test; P ¼ 0.002) and absence of tumor encapsulation (Student knocked down NDUFA4 (-shA4-30 and -shA4-88) in MHCC97L t test; P ¼ 0.039; Fig. 1C and Supplementary Table S4). More cells and confirmed that the lower band was NDUFA4 (Sup- importantly, overexpression of NDUFA4L2 tended to be asso- plementary Fig. S4A and S4B). These results demonstrated that ciated with poorer overall survival of HCC patients (Student t NDUFA4L2 was preferentially expressed under hypoxia solely test; P ¼ 0.086; Fig. 1C). TCGA database, which includes through HIF1a, whereas NDUFA4 protein was preferentially transcriptome sequencing data of 49 pairs of HCC and NT expressed under normoxia. liver tissues from an independent cohort, echoed with our in- house data (Fig. 1D). Intriguingly, when we studied the micro- NDUFA4L2 reduces mitochondrial activity by decreasing array data available in the Oncomine database, we noticed that oxygen consumption and prevents excessive ROS production NDUFA4L2 was also overexpressed in other human solid under hypoxia cancers, including renal cell carcinoma (RCC), lung SCC, as To investigate the functions of hypoxia-induced NDUFA4L2 well as colorectal carcinoma (Supplementary Fig. S1). The in mitochondria in vitro, we generated NDUFA4L2 loss-of- overexpression of NDUFA4L2 mRNA in lung SCC was further function and gain-of-function HCC cell models. For NDU- confirmed in our 31 pairs of lung SCC tissues and their normal FA4L2 loss-of-function HCC cell model, shRNA against NDU- counterparts by qRT-PCR (Fig. 1E). Collectively, these results FA4L2 (shL2) or nontarget control (NTC) was stably expressed demonstrated the clinical relevance of NDUFA4L2 in human in MHCC97L cells by lentiviral transfection approach (Fig. 3A). HCC and other human solid cancers. For NDUFA4L2 gain-of-function HCC cell model, NDUFA4L2 and empty vector (EV) were stably expressed in another HIF1a, but not HIF2a, positively regulates NDUFA4L2 in HCC cell line, which expressed a lower level of NDUFA4L2, hypoxia PLC/PRF/5 (PLC; Fig. 3B). As NDUFA4L2 is located in the To investigate whether NDUFA4L2 expression was mediated firstcomplexoftheETC,weaskedifNDUFA4L2wouldaffect by HIFs under hypoxia, we generated HCC cells, MHCC97L, the mitochondrial activity by staining the cells with a probe, that stably expressed shRNA against HIF1a (-shHIF1a)or JC-1, which accumulates in active mitochondria to indicate HIF2a (-shHIF2a; Supplementary Fig. S2A). The hypoxia- mitochondrial potential. By fluorescence imaging, we showed induced NDUFA4L2 expression was markedly abolished when that the mitochondrial membrane potential was decreased HIF1a, but not HIF2a, was knocked down (Fig. 2A). We further in hypoxia, but was elevated when NDUFA4L2 was knocked established HIF1a knockout cells by Transcription Activator-like down (Fig. 3C). Flow cytometry analysis further confirmed the Effector Nuclease approach and found that hypoxia-induced result (Fig. 3C). Similar observation was obtained with tetra- NDUFA4L2 expression was drastically abrogated (Fig. 2B and methylrhodamine ethyl ester, another fluorescent probe that Supplementary Fig. S2B). Digoxin inhibited HIF1 protein syn- measures the mitochondrial membrane potential (Supple- thesis and profoundly inhibited hypoxia-induced NDUFA4L2 mentary Fig. S5A). Consistently, mitochondrial mass was expression in a dose-dependent manner (Fig. 2C). In addition, increased in NDUFA4L2 knockdown cells as indicated by the by in silico analysis, we identified two potential hypoxia re- nonyl acridine orange (NAO) staining (Fig. 3D). Consistently, sponse elements (HRE) containing the core HIF-binding we found that oxygen consumption in the mitochondria was sequence motif 50-RCGTG-30 at the promoter of NDUFA4L2 significantly elevated when NDUFA4L2 was knocked down (Fig. 2D). ChIP assay in cell lines of different origins, MHCC97L under hypoxia (Fig. 3E). HIF1a knockout mirrored the effect and HeLa cells, indicated that DNA was enriched when HIF1a of NDUFA4L2 knockdown in oxygen consumption (Fig. 3E),

Figure 3. NDUFA4L2 reduces mitochondrial activity, oxygen consumption, and ROS production in hypoxia. A, mRNA and protein expression of NDUFA4L2 in MHCC97L- NTC and -shL2 cells exposed to 20% and 1% O2 for 24 and 48 hours, respectively (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.001 as indicated). B, mRNA and protein expression of NDUFA4L2 in PLC-EV and -NDUFA4L2 cells cultured in 20% O2 for 24 and 48 hours, respectively (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.001 as indicated). C, Left, representative fluorescentimagesofJC-1stainedMHCC97L-NTCand -shL2 cells that were exposed to 20% and 1% O2. Right, mitochondrial membrane potential was further determined with JC-1 staining by flow cytometry analysis (results represent mean SEM. Student t test; N ¼ 5independent experiments: , P < 0.001 as indicated). Scale, 50 mm. D, mitochondrial mass was measured with NAO staining by flow cytometry analysis (results represent mean SEM. Student t test; N ¼ 5 independent experiments: , P < 0.01 as indicated). E, oxygen consumption was measured with / MitoXpress -Intra (NanO2)inMHCC97L-NTCand-shL2cellsorMHCC97L-WTand-HIF1a (KO) cells exposed to 20% and 1% O2 (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.05; , P < 0.01 as indicated). F, ROS level was measured with CM-H2DCFDA or MitoSox staining by flow cytometry analysis in (i and ii) MHCC97L-NTC and -shL2 cells cultured in 20% and 1% O2, (iii) PLC-EV and -NDUFA4L2 cells cultured in 20% O2, and (iv) PLC-EV and -CA5 cells cultured in 20% O2 (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.01 as indicated). Cells for all metabolic assays were cultured for 24 hours. Values obtained in mitochondrial membrane potential, mass, and ROS were normalized to 20% O2 MHCC97L-NTC or PLC-EV.

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3111

Lai et al.

3112 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

further supporting the regulatory role of HIF1a on NDUFA4L2. 4D and Supplementary Fig. S6D). More interestingly, cell Hypoxia triggers an imbalanced electron flow through the proliferation was restored by NAC in the NDUFA4L2 knock- ETC leading to generation of ROS in the mitochondria. Hence, downHCCcells(Fig.4E),suggestingthatimpairmentofcell we evaluated the intracellular ROS level by the general ROS proliferation was associated with ROS. All these data con- indicator, chloromethyl-2',7'-dichlorodihydrofluorescein dia- verged to show that NDUFA4L2 was responsible to reduce cetate (CM-H2DCFDA), and mitochondrial ROS indicator, ROS under hypoxia, thereby maintaining cell survival and MitoSOX. ROS level was elevated when cells were exposed to promoting cell proliferation. hypoxia and was further induced when NDUFA4L2 was knocked down especially under hypoxic condition (Fig. 3F). NDUFA4L2 promotes tumor growth, reduces oxidative stress, Reversely, a significant reduction of ROS was found in NDU- and enhances lung metastasis FA4L2-overexpressing HCC cells (Fig. 3F). Similar trend was To demonstrate the functions of NDUFA4L2 in tumor observed in HCC cells that expressed the constitutively active growth in vivo, we orthotopically injected the luciferase-labeled form of HIF1a (CA5; Fig. 3F). To eliminate off-target effect, MHCC97L-NTC and -shL2 cells into nude mice. MHCC97L- siRNAs targeting different sequences of NDUFA4L2 were trans- NTC tumors grew substantially larger than the MHCC97L-shL2 fected into MHCC97L cells (Supplementary Fig. S5B), and tumors (Fig. 5A and 5B). To evaluate the proliferative output of similar metabolic consequences were found as compared with MHCC97L-NTC and -shL2 cells in vivo,westainedthetumors stable knockdown of NDUFA4L2 (Supplementary Fig. S5C with proliferative marker (Ki67). We found that knockdown of and S5D). NDUFA4L2 markedly reduced Ki67 staining (Fig. 5C). We further dissociated the tumors and measured the ROS level of NDUFA4L2 maintains redox homeostasis and promotes cell the tumors derived from MHCC97L-NTC and -shL2 cell lines. proliferation and survival Consistent with our in vitro data, knockdown of NDUFA4L2 Knockdown of NDUFA4L2 resulted in a dramatic increase elevated ROS in vivo (Fig. 5D). Moreover, knockdown of NDU- of ROS, which we speculated would be harmful to HCC cells. FA4L2 markedly repressed growth of metastatic lesions in the To demonstrate the roles of ROS in HCC cells, we employed lungs of the tumor-bearing mice as indicated by Xenogen two antioxidants, NAC and L-ascorbic acid. ROS induced by imaging (4 of 6 in MHCC97L-NTC and 1 of 6 in -shL2 NDUFA4L2 knockdown was rescued dose dependently by group; Fig. 5E). We further found that knockdown of NDU- both antioxidants (Fig. 4A). To further demonstrate the reg- FA4L2 significantly reduced cell-migratory ability of HCC cells ulation of HIF1a on NDUFA4L2 in ROS scavenging, we re- in Transwell assay. Interestingly, the migratory ability of HCC expressed NDUFA4L2 in HIF1a knockdown PLC cells. ROS cells was inversely correlated with the amount of intracellular was dramatically induced in HIF1a knockdown cells under ROS (Supplementary Fig. S7A and Fig. 3E), suggesting that ROS hypoxia and was significantly rescued upon re-expression of may also affect cell motility. NDUFA4L2 (Fig. 4B). Excessive accumulation of ROS can cause cell senescence and apoptosis, eventually leading to cell Pharmacologic inhibition of HIF suppresses progression of death (16–18). We therefore evaluated apoptosis by Annexin tumors that express high level of NDUFA4L2 V–propidium iodide (PI) staining in MHCC97L-NTC and So far, we found that hypoxia-induced NDUFA4L2 tightly -shL2 cells. MHCC97L-NTC cells showed higher Annexin V regulated REDOX homeostasis and survival in HCC cells, and and PI staining under hypoxic than normoxic condition, HIF1a is the central regulator of NDUFA4L2. We therefore while knockdown of NDUFA4L2 further increased apoptosis reasoned that HIF inhibitor could repress the tumors that (Fig. 4C), which could be rescued by extracellular GSH-EE, a express high levels of NDUFA4L2 by damaging HCC cells modified and more permeable form of glutathione, an anti- through elevating ROS. We evaluated the ROS levels of oxidant that counteracts mitochondrial ROS (Supplementary MHCC97L-NTC and -shL2 cells treated with an HIF inhibitor, Fig. S6A and S6B). To further confirm that increased apoptosis digoxin, under hypoxic condition. We observed that digoxin was caused by ROS, we exogenously treated the cells with H2 was more efficient in elevating the ROS level in MHCC97L-NTC O2, a common source of ROS (35). Apoptosis was elicited by cells than -shL2 cells (Fig. 6A). Next, we evaluated the effect of H2O2 drastically in MHCC97L-NTC cells and was further digoxin in vivo. We performed orthotopic implantation of the elevated in -shL2 cells (Supplementary Fig. S6C). Consistently, luciferase-labeled MHCC97L-NTC and -shL2 cells in nude mice. both the cell counting and BrdUrd assays showed that cell One week after implantation, mice were administered with proliferation was substantially reduced when NDUFA4L2 1.2 mg/kg/day digoxin or vehicle control (saline) through i.p. was knocked down especially under hypoxic condition (Fig. injection for 21 days. Digoxin markedly repressed growth of

Figure 4. NDUFA4L2 reduces intracellular ROS to enhance HCC cell survival. A, CM-H2DCFDA staining showing the ROS levels of MHCC97L-NTC and -shL2 cells exposed to 1% O2 for 24 hours in the presence of indicated concentrations of NAC and ascorbic acid. Top, representative ﬂow cytometry analysis showing the ROS levels in the indicated subclones. Bottom, histograms summarizing the ROS levels in the indicated subclones. The ROS level was relative to 1% O2 MHCC97L-NTC.B,CM-H2DCFDA staining showing the ROS levels of PLC-EV, -shHIF1a,andHIF1a knockdown with NDUFA4L2 overexpressing (-shHIF1a þ NDUFA4L2) cells exposed to 1% O2 for 24 hours. The ROS level was relative to 20% O2 PLC-EV. Top, representative ﬂow cytometry analysis showing the ROS levels in the indicated subclones. Bottom, histograms summarizing the ROS levels in the indicated subclones. C, Annexin V and PI staining showing the percentage of apoptotic cells in MHCC97L-NTC and -shL2 cells exposed to 20% and 1% O2 for 24 hours. D, cell proliferation of MHCC97L-NTC and -shL2 cells exposed to 20% and 1% O2 for 96 hours. E, BrdUrd assay of MHCC97L-NTC and -shL2 cells treated with indicated concentrations of NAC was exposed to 1% O2 for 24 hours (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.05; , P < 0.01; , P < 0.001 as indicated).

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3113

Lai et al.

Figure 5. Knockdown of NDUFA4L2 suppresses tumor growth and induces oxidative stress in vivo. A, bioluminescence and (B) tumor volume of the orthotopic tumors derived from MHCC97L-NTC and -shL2 cells. Scale, 1 cm. C, Left, representative IHC pictures of Ki67 positivity in tumors. Cells were scored from 0 to 3 based on the staining intensity as represented by the arrows. Scale, 100 mm. Right, more than 1,000 cells were scored in each animal, and average percentage was calculated for each score. D, workﬂow of the measurement of intracellular ROS level in mouse model. Nude mice were orthotopically injected with MHCC97L-NTC and -shL2 cells. CM-H2DCFDA staining showing the ROS levels of the dissociated orthotopic tumors derived from MHCC97L-NTC and -shL2. E, bioluminescence of the lungs harvested from orthotopic tumor-bearing mice implanted with MHCC97L-NTC and -shL2 cells (Student t test; N ¼ 6 per group: , P < 0.05; , P < 0.01).

tumors with a more prominent effect on the MHCC97L-NTC derived from MHCC97L-EV cells more signiﬁcantly as com- group as compared with MHCC97L-shL2 group (Fig. 6B). pared with MHCC97L-shHIF1a cells (Supplementary Fig. S7B). Similarly, digoxin decreased the growth of subcutaneous tumors Meanwhile, body weights of the animals were not affected by

3114 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

Figure 6. HIF inhibitors suppress growth of tumors which express high level of NDUFA4L2. A, CM-H2DCFDA staining showing the ROS levels of MHCC97L-NTC and -shL2 cells exposed to 1% O2 for 24 hours in the presence of indicated concentrations of HIF inhibitor digoxin (results represent mean SEM. Student t test; N ¼ 3 independent experiments: , P < 0.01; , P < 0.001 as indicated). B, tumor volumes of orthotopic tumors derived from MHCC97L-NTC and -shL2 cells in mice that were administered with digoxin (1.2 mg/kg/day, i.p.) or vehicle (saline) for 21 consecutive days (Student t test; N ¼ 6 per group: , P < 0.05; , P < 0.001 as indicated). Scale, 1 cm. C, regulation and role of NDUFA4L2 in human HCC. i, hypoxia elicits an imbalanced electron ﬂow through the ETC, leading to excessive ROS accumulation which is detrimental to HCC cells. ii, HIF1a overcomes by transcriptionally activating NDUFA4L2 in the mitochondrial complex I selectively to reduce the mitochondrial activity and the electron ﬂux through ETC, thereby reducing ROS and ROS-induced apoptosis. This metabolic adaptation of oxidative stress mediated by HIF1a under hypoxia conferred survival advantage to the rapidly growing HCC cells which frequently experience hypoxia.

digoxin (Supplementary Fig. S7C). These results suggested that FA4L2 may regulate multiple biologic processes, including mem- HIF inhibitor blocks HCC progression at least partially through brane potential and membrane depolarization (Supplementary HIF1a/NDUFA4L2. Table S7). Of note, NRF2, a transcription factor that activates antioxidant genes, was found to be slightly induced (FPKM Potential gene targets of NDUFA4L2 29.246 in MHCC97L-NTC vs. FPKM 34.151 in -shL2), suggesting To explore novel functions of NDUFA4L2, we performed that a compensatory anti–oxidant-producing mechanism might transcriptome sequencing to compare the global gene expression be involved in response to the high ROS level upon NDUFA4L2 proﬁles of MHCC97L-NTC and MHCC97L-shL2 cells. The expres- knockdown. sions of 629 genes were found to be decreased in MHCC97L-shL2 cells for more than 0.5-fold as compared with MHCC97L-NTC (Supplementary Table S5). The expressions of 683 genes were Discussion found to be increased for more than 2-fold in MHCC97L-shL2 ROS are fundamentally produced from O2 when it is con- cells relative to MHCC97L-NTC (Supplementary Table S6). sumed in different metabolic reactions. These reactions take Intriguingly, pathway analysis by DAVID suggested that NDU- place in the mitochondria at the ETC for ATP generation,

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3115

Lai et al.

endoplasmic reticulum during disulfide bond formation for REDOX homeostasis can be achieved by two mechanisms: (i) protein folding, and peroxisomes during b-oxidation of fatty reduction of ROS production and (ii) elevation of antioxidant acids. Under hypoxia, these metabolic events cannot be carried production. Our study showed that NDUFA4L2 directly out properly, leading to the generation of O2-containing free decreased the generation of ROS in the mitochondria of HCC radicals. In the mitochondria that experiences shortage of the cells. Meanwhile, genes that are associated with the production ultimate electron acceptor, O2, electrons cannot be efficiently and utilization of antioxidants, including NADPH, glutathione, pass through the ETC, leading to rapid accumulation of ROS in and thioredoxin, have been implicated in HCC. Glucose 6 complexes I and III. Hypoxia is a common finding in HCC, and phosphate dehydrogenase, a critical enzyme in the pentose our study revealed the molecular mechanisms by which hyp- phosphate pathway generating NADPH, has been shown to oxic HCC cells modulate ROS level in the mitochondria. Our be overexpressed in HCC in a PTEN-dependent manner (41). study showed that HCC cells, by utilizing NDUFA4L2 in the Overexpression of thioredoxin and thioredoxin-related REDOX complex I of the ETC, reduced the activity of the mitochondria molecules, glutaredoxin and peroxiredoxin, was documented to prevent ROS accumulation and apoptosis (Fig. 6C). Not in HCC and was associated with increased proliferative output, only have we confirmed that this was dependent on HIF1a in aggressive clinicopathologic parameters, presence of metastasis, HCC cells, more importantly, our study unprecedentedly and poor prognosis in HCC patients (42, 43). Elevation of revealed the clinical relevance of NDUFA4L2 in human HCC. cysteine/glutamate transporter, xCT, which is important for Apart from HCC, we also showed that NDUFA4L2 was over- glutathione production, was found to be associated with poor expressed in multiple solid cancers, suggesting that our find- clinical outcome in HCC, and disruption of xCT in HCC cells ings do not limit to HCC but applies to other solid cancer increased ROS-mediated autophagic cell death (44). A recent models that frequently experience hypoxia. Notably, we and study has elegantly demonstrated that cotreatment of inhibitors Tello and colleagues. consistently found that NDUFA4 and blocking different pathways that synthesize antioxidants could NDUFA4L2 have opposite expression patterns at normoxia synergistically prevent tumor growth (45). Along similar lines, and hypoxia in different cell models (23). Studies have docu- our current study suggested that raising ROS level through HIF mented that HIF1a could induce expression of siah E3 ubi- inhibitor, which suppressed HIF1a/NDUFA4L2 pathway, quitin protein ligase 2 (SIAH2) and mitochondrial LON pep- represents an attractive therapeutic strategy for HCC treatment. tidase to degrade mitochondrial proteins such as a-ketogluta- In the coming decade, combination therapies targeting multi- rate dehydrogenase and COX4-1 to reduce mitochondrial flux ple machineries that buffer ROS represent an exciting transla- under hypoxia, respectively (21, 36). Whether NDUFA4 is tion research direction. degraded in a similar fashion remains an interesting topic to be addressed. Disclosure of Potential Conflicts of Interest Prenatal lethality in NDUFA4L2 knockout mice has greatly No potential conflicts of interest were disclosed. limited the expansion of knowledge on the in vivo functions of this complex I subunit (23). Our study, using orthotopic liver Authors' Contributions cancer model, has provided the first in vivo evidence to confirm Conception and design: R.K.-H. Lai, I.O.-L. Ng, C.C.-L. Wong the protumorigenic functions of NDUFA4L2. Our study also Development of methodology: R.K.-H. Lai, C.C.-L. Wong demonstrated that digoxin profoundly blocked growth of Acquisition of data (provided animals, acquired and managed patients, orthotopic tumors that expressed high level of HIF1a/NDU- provided facilities, etc.): R.K.-H. Lai, I. M.-J. Xu, D.K.-C. Chiu, A.P.-W. Tse, a M.P. Wong, C.C.-L. Wong FA4L2. Digoxin has been shown to inhibit HIF1 synthesis and Analysis and interpretation of data (e.g., statistical analysis, biostatistics, is well-tolerated in patients (30). Digoxin has been used for the computational analysis): R.K.-H. Lai, D.K.-C. Chiu, A.P.-W. Tse, C.-T. Law, treatment of congestive heart failure and is currently under C.-M. Wong, C.C.-L. Wong clinical trial for the treatment of breast cancer (ClinicalTrials. Writing, review, and/or revision of the manuscript: R.K.-H. Lai, D. Lee, gov identifier: NCT01763931). Our study suggested that digox- M.P. Wong, I.O.-L. Ng, C.C.-L. Wong in may be the most suitable candidate for patients whose Administrative, technical, or material support (i.e., reporting or organizing a data, constructing databases): L.L. Wei, C.-T. Law, I.O.-L. Ng, C.C.-L. Wong primary HCCs express high levels of HIF1 or NDUFA4L2. Study supervision: I.O.-L. Ng, C.C.-L. Wong Worth mentioning, complex I inhibitors including metformin have promising antitumor effects in the clinical setting. Metfor- Acknowledgments min has been the most frequently prescribed antidiabetic med- pQCXIN HIF1a (CA5) construct is a generous gift from Professor Gregg L. ication in the world. By reducing the ATP production in the ETC, Semenza. The authors thank the Core Facility of LKS Faculty of Medicine for metformin activates tumor suppressor 50-AMP-activated protein their technical support. They also thank Dr. Wilson Yick-Pang Ching and Dr. kinase (AMPK; ref. 37). Reports have indicated that metformin Yuan Zhou for the valuable advice and technical help. reduced cancer incidence (38, 39). Metformin also reduced proliferative output of breast cancer cells in nondiabetic cancer Grant Support patients (40). These interesting findings highlight an interesting This work was supported by the Small Project Funding of the University of dilemma—is energy or REDOX homeostasis more important for Hong Kong 2012 and the Hong Kong Research Grants Council General Research Fund (HKU 781213M). cancer survival? Accumulating studies suggested that ATP is not The costs of publication of this article were defrayed in part by the the limiting factor while excessive ROS is cell destructive. payment of page charges. This article must therefore be hereby marked Unquestionably, cancer cells could not survive without energy; advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate therefore, cancer cells need to acquire a tightly controlled homeo- this fact. static balance for the ETC, a place that generates both ATP and ROS, to sustain tumor growth. Drugs that could tip this balance Received August 19, 2015; revised December 15, 2015; accepted January 4, merits exploration for their efficiency in cancer treatment. 2016; published OnlineFirst January 27, 2016.

3116 Clin Cancer Res; 22(12) June 15, 2016 Clinical Cancer Research

NDUFA4L2 in HCC Development

References 1. Forner A, Llovet JM, Bruix J. Hepatocellular carcinoma. Lancet 2012;379: 25. Favier J, Briere JJ, Burnichon N, Riviere J, Vescovo L, Benit P, et al. The 1245–55. Warburg effect is genetically determined in inherited pheochromocyto- 2. Cheng AL, Kang YK, Chen Z, Tsao CJ, Qin S, Kim JS, et al. Efficacy and safety mas. PloS one 2009;4:e7094. of sorafenib in patients in the Asia-Pacific region with advanced hepato- 26. Ng IO, Lai EC, Fan ST, Ng MM, So MK. Prognostic significance of pathologic cellular carcinoma: a phase III randomised, double-blind, placebo-con- features of hepatocellular carcinoma. A multivariate analysis of 278 trolled trial. Lancet Oncol 2009;10:25–34. patients. Cancer 1995;76:2443–8. 3. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib 27. Wong CC, Au SL, Tse AP, Xu IM, Lai RK, Chiu DK, et al. Switching of in advanced hepatocellular carcinoma. N Engl J Med 2008;359:378–90. pyruvate kinase isoform L to m2 promotes metabolic reprogramming in 4. Wong CC, Kai AK, Ng IO. The impact of hypoxia in hepatocellular hepatocarcinogenesis. PloS one 2014;9:e115036. carcinoma metastasis. Front Med. 2014;8:33–41. 28. Wong CC, Wong CM, Tung EK, Au SL, Lee JM, Poon RT, et al. The 5. Semenza GL. Hypoxia-inducible factors in physiology and medicine. Cell microRNA miR-139 suppresses metastasis and progression of hepato- 2012;148:399–408. cellular carcinoma by down-regulating Rho-kinase 2. Gastroenterology 6. Epstein AC, Gleadle JM, McNeill LA, Hewitson KS, O'Rourke J, Mole DR, 2011;140:322–31. et al. C. elegans EGL-9 and mammalian homologs define a family of 29. Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, dioxygenases that regulate HIF by prolyl hydroxylation. Cell 2001; et al. Transcript assembly and quantification by RNA-Seq reveals unan- 107:43–54. notated transcripts and isoform switching during cell differentiation. Nat 7. Kaelin WGJr, Ratcliffe PJ. Oxygen sensing by metazoans: the central Biotechnol 2010;28:511–5. role of the HIF hydroxylase pathway. Mol Cell 2008;30:393–402. 30. Zhang H, Wong CC, Wei H, Gilkes DM, Korangath P, Chaturvedi P, et al. 8. Wang GL, Jiang BH, Rue EA, Semenza GL. Hypoxia-inducible factor 1 is a HIF-1-dependent expression of angiopoietin-like 4 and L1CAM mediates basic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension. vascular metastasis of hypoxic breast cancer cells to the lungs. Oncogene Proc Natl Acad Sci U S A 1995;92:5510–4. 2012;31:1757–70. 9. Kim JW, Tchernyshyov I, Semenza GL, Dang CV. HIF-1-mediated expres- 31. Gilkes DM, Bajpai S, Wong CC, Chaturvedi P, Hubbi ME, Wirtz D, et al. sion of pyruvate dehydrogenase kinase: a metabolic switch required for Procollagen lysyl hydroxylase 2 is essential for hypoxia-induced breast cellular adaptation to hypoxia. Cell Metab 2006;3:177–85. cancer metastasis. Mol Cancer Res 2013;11:456–66. 10. Iyer NV, Kotch LE, Agani F, Leung SW, Laughner E, Wenger RH, et al. 32. Forsythe JA, Jiang BH, Iyer NV, Agani F, Leung SW, Koos RD, et al. Cellular and developmental control of O2 homeostasis by hypoxia-induc- Activation of vascular endothelial growth factor gene transcription by ible factor 1 alpha. Genes Dev 1998;12:149–62. hypoxia-inducible factor 1. Mol Cell Biol 1996;16:4604–13. 11. Kim JW, Gao P, Liu YC, Semenza GL, Dang CV. Hypoxia-inducible factor 1 33. Giatromanolaki A, Koukourakis MI, Sivridis E, Pastorek J, Wykoff CC, and dysregulated c-Myc cooperatively induce vascular endothelial growth Gatter KC, et al. Expression of hypoxia-inducible carbonic anhydrase-9 factor and metabolic switches hexokinase 2 and pyruvate dehydrogenase relates to angiogenic pathways and independently to poor outcome in kinase 1. Mol Cell Biol 2007;27:7381–93. non-small cell lung cancer. Cancer Res 2001;61:7992–8. 12. Semenza GL, Roth PH, Fang HM, Wang GL. Transcriptional regulation of 34. Chaturvedi P, Gilkes DM, Wong CC, Luo W, Zhang H, Wei H, et al. genes encoding glycolytic enzymes by hypoxia-inducible factor 1. J Biol Hypoxia-inducible factor-dependent breast cancer-mesenchymal stem Chem 1994;269:23757–63. cell bidirectional signaling promotes metastasis. J Clin Invest 2013; 13. Fantin VR, St-Pierre J, Leder P. Attenuation of LDH-A expression uncovers a 123:189–205. link between glycolysis, mitochondrial physiology, and tumor mainte- 35. Murphy MP. How mitochondria produce reactive oxygen species. Biochem nance. Cancer Cell 2006;9:425–34. J 2009;417:1–13. 14. Semenza GL, Jiang BH, Leung SW, Passantino R, Concordet JP, Maire P, 36. Ma B, Chen Y, Chen L, Cheng H, Mu C, Li J, et al. Hypoxia regulates et al. Hypoxia response elements in the aldolase A, enolase 1, and lactate Hippo signalling through the SIAH2 ubiquitin E3 ligase. Nat Cell Bio dehydrogenase A gene promoters contain essential binding sites for hyp- 2015;17:95–103. oxia-inducible factor 1. J Biol Chem 1996;271:32529–37. 37. Pernicova I, Korbonits M. Metformin–mode of action and clinical implica- 15. Tochhawng L, Deng S, Pervaiz S, Yap CT. Redox regulation of cancer cell tions for diabetes and cancer. Nat Rev Endocrinol 2014;10:143–56. migration and invasion. Mitochondrion 2013;13:246–53. 38. Lee JH, Kim TI, Jeon SM, Hong SP, Cheon JH, Kim WH. The effects of 16. Ames BN. Dietary carcinogens and anticarcinogens. Oxygen radicals and metformin on the survival of colorectal cancer patients with diabetes degenerative diseases. Science 1983;221:1256–64. mellitus. Int J Cancer 2012;131:752–9. 17. Hampton MB, Orrenius S. Dual regulation of caspase activity by hydrogen 39. He X, Esteva FJ, Ensor J, Hortobagyi GN, Lee MH, Yeung SC. Metformin and peroxide: implications for apoptosis. FEBS Lett 1997;414:552–6. thiazolidinediones are associated with improved breast cancer-specific 18. Sena LA, Chandel NS. Physiological roles of mitochondrial reactive oxygen survival of diabetic women with HER2þ breast cancer. Ann Oncol 2012; species. Mol Cell 2012;48:158–67. 23:1771–80. 19. Renschler MF. The emerging role of reactive oxygen species in cancer 40. Hadad S, Iwamoto T, Jordan L, Purdie C, Bray S, Baker L, et al. Evidence for therapy. Eur J Cancer 2004;40:1934–40. biological effects of metformin in operable breast cancer: a pre-operative, 20. Toler SM, Noe D, Sharma A. Selective enhancement of cellular oxidative window-of-opportunity, randomized trial. Breast Cancer Res Treat 2011; stress by chloroquine: implications for the treatment of glioblastoma 128:783–94. multiforme. Neurosurg Focus 2006;21:E10. 41. Hong X, Song R, Song H, Zheng T, Wang J, Liang Y, et al. PTEN antagonises 21. Fukuda R, Zhang H, Kim JW, Shimoda L, Dang CV, Semenza GL. HIF-1 Tcl1/hnRNPK-mediated G6PD pre-mRNA splicing which contributes to regulates cytochrome oxidase subunits to optimize efficiency of respiration hepatocarcinogenesis. Gut 2014;63:1635–47. in hypoxic cells. Cell 2007;129:111–22. 42. Mollbrink A, Jawad R, Vlamis-Gardikas A, Edenvik P, Isaksson B, 22. Chan SY, Zhang YY, Hemann C, Mahoney CE, Zweier JL, Loscalzo J. Danielsson O, et al. Expression of thioredoxins and glutaredoxins in MicroRNA-210 controls mitochondrial metabolism during hypoxia by human hepatocellular carcinoma: correlation to cell proliferation, repressing the iron-sulfur cluster assembly proteins ISCU1/2. Cell Metab tumor size and metabolic syndrome. Int J Immunopathol Pharmacol 2009;10:273–84. 2014;27:169–83. 23. Tello D, Balsa E, Acosta-Iborra B, Fuertes-Yebra E, Elorza A, Ordonez A, 43. Sun QK, Zhu JY, Wang W, Lv Y, Zhou HC, Yu JH, et al. Diagnostic and et al. Induction of the mitochondrial NDUFA4L2 protein by HIF-1alpha prognostic significance of peroxiredoxin 1 expression in human hepato- decreases oxygen consumption by inhibiting Complex I activity. Cell cellular carcinoma. Med Oncol 2014;31:786. Metab 2011;14:768–79. 44. Guo W, Zhao Y, Zhang Z, Tan N, Zhao F, Ge C, et al. Disruption of xCT 24. Castelblanco E, Gallel P, Ros S, Gatius S, Valls J, De-Cubas AA, et al. Thyroid inhibits cell growth via the ROS/autophagy pathway in hepatocellular paraganglioma. Report of 3 cases and description of an immunohisto- carcinoma. Cancer Lett 2011;312:55–61. chemical profile useful in the differential diagnosis with medullary thyroid 45. Harris IS, Treloar AE, Inoue S, Sasaki M, Gorrini C, Lee KC, et al. Gluta- carcinoma, based on complementary DNA array results. Hum Pathol thione and thioredoxin antioxidant pathways synergize to drive cancer 2012;43:1103–12. initiation and progression. Cancer Cell 2015;27:211–22.

www.aacrjournals.org Clin Cancer Res; 22(12) June 15, 2016 3117

NDUFA4L2 Fine-tunes Oxidative Stress in Hepatocellular Carcinoma

Robin Kit-Ho Lai, Iris Ming-Jing Xu, David Kung-Chun Chiu, et al.

Clin Cancer Res 2016;22:3105-3117. Published OnlineFirst January 27, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-15-1987

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2016/01/27/1078-0432.CCR-15-1987.DC1

Cited articles This article cites 45 articles, 11 of which you can access for free at: http://clincancerres.aacrjournals.org/content/22/12/3105.full#ref-list-1

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://clincancerres.aacrjournals.org/content/22/12/3105. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.