Glutaminase 2 negatively regulates the PI3K/AKT signaling and shows tumor suppression activity in human hepatocellular carcinoma.

The tumor suppressor p53 and its signaling pathway play a critical role in tumor prevention. As a direct p53 target gene, the role of glutaminase 2 (GLS2) in tumorigenesis is unclear. In this study, we found that GLS2 expression is significantly decreased in majority of human hepatocellular carcinoma (HCC). Restoration of GLS2 expression in HCC cells inhibits the anchorage-independent growth of cells and reduces the growth of HCC xenograft tumors. Interestingly, we found that GLS2 negatively regulates the PI3K/AKT signaling, which is frequently activated in HCC. Blocking the PI3K/AKT signaling in HCC cells largely abolishes the inhibitory effect of GLS2 on the anchorage-independent cell growth and xenograft tumor growth. The GLS2 promoter is hypermethylated in majority of HCC samples. CpG methylation of GLS2 promoter inhibits GLS2 transcription, whereas reducing the methylation of GLS2 promoter induces GLS2 expression. Taken together, our results demonstrate that GLS2 plays an important role in tumor suppression in HCC, and the negative regulation of PI3K/AKT signaling contributes greatly to this function of GLS2. Furthermore, hypermethylation of GLS2 promoter is an important mechanism contributing to the decreased GLS2 expression in HCC.


INTRODUCTION
Tumor suppressor p53 plays a crucial role in tumor prevention, including hepatocellular carcinoma (HCC) [1][2][3]. p53 is the most frequently-mutated gene in human tumors; over 50% of human tumors harbor mutations in the p53 gene. In HCC, around 40-60% of HCCs contain DNA mutations in the p53 gene [4,5]. Disruption of normal p53 function is in many circumstances a prerequisite for the development or progression of tumors [6,7]. As a transcription factor, p53 mainly exerts its tumor suppression function through its transcriptional regulation of its downstream target genes. Through the regulation of the expression of many downstream target genes, p53 regulates cell cycle arrest, apoptosis, senescence, cellular energy metabolism and anti-oxidant defense, all of which contribute to the role of p53 in tumor suppression [1][2][3]8].
Glutaminase 2 (GLS2) is a liver-type mitochondrial glutaminase, which can catalyze the hydrolysis of glutamine to glutamate in mitochondria in cells [9,10]. GLS2 is specifically expressed in very few tissues, including the liver. To date, little is known about the biological functions of GLS2 in cells except for its function as a glutaminase. Recently, GLS2 was identified as a novel p53 downstream target gene by our group and another group [11,12]. p53 induces the GLS2 expression under both stressed and non-stressed conditions. Importantly, GLS2 mediates the functions of p53 in www.impactjournals.com/oncotarget regulation of energy metabolism and anti-oxidant defense [11,12]. Considering the critical role of p53 and its pathway in tumor suppression, as a novel p53 target gene, GLS2 might play an important role in tumor suppression. However, the role of GLS2 in tumorigenesis is not wellunderstood.
HCC is the fifth most frequent cancer worldwide [13][14][15]. HCC is a highly malignant tumor type with average survival rates less than 1 year following diagnosis. One main reason for the high HCC mortality is because most patients are diagnosed when the disease is already at an advanced stage, and the cancerous tissue cannot be surgically removed [13][14][15]. Therefore, further understanding the molecular mechanism of liver tumorignesis will provide potential molecular biomarkers for early diagnosis and novel therapeutic strategies for HCC.
In this study, we investigated the role of GLS2 in liver tumorigenesis. Our results demonstrated that GLS2 protein levels were significantly decreased in majority of HCCs that we examined. GLS2 inhibited the anchorageindependent growth of HCC cells and the growth of HCC xenograft tumors. Furthermore, GLS2 negatively regulated the PI3K/AKT signaling, which is frequently activated in various tumors, including HCC, and plays a pivotal role in tumorigenesis [16][17][18]. Blocking the PI3K/AKT signaling largely abolished the inhibitory effect of GLS2 liver tumorigenesis. CpG hypermethylation in gene promoters is an important epigenetic mechanism that contributes to decreased expression of tumor suppressor genes in cancer, including HCC [19][20][21]. Our results strongly suggested that hypermethylation of GLS2 promoter is an important mechanism that contributes to the down-regulation of GLS2 expression in HCC. Taken together, results from this study demonstrated an important role of GLS2 in tumor suppression in HCC through its negative regulation of the PI3K/AKT signaling.

RESULTS
The GLS2 protein expression is frequently decreased in human primary HCC Liver is one of few tissues that specifically express GLS2. Previously, we examined the levels of GLS2 mRNA in a set of primary HCCs at different stages, and found that GLS2 mRNA levels were greatly decreased in majority of primary HCCs compared with normal liver  Supplementary Tables S1 & S2. www.impactjournals.com/oncotarget tissues or tumor adjacent liver tissues [11]. These results are consistent with the results from another study [12]. However, it was unclear whether the change of GLS2 protein levels is consistent with the change of GLS2 mRNA levels in HCCs.
To investigate whether GLS2 protein expression is decreased in HCCs, we analyzed the levels of GLS2 protein in two different sets of primary HCC samples by immunohistochemistry (IHC) staining assays. One set of samples was provided by US Biomax (Rockville, MD), which includes totally 110 primary HCCs and 125 nontumor liver tissues in three tissue microarrays (TMAs). Another set of samples was collected at University of Texas MD Anderson Cancer Center, which includes 21 pairs of primary HCCs and their matched adjacent nontumor liver tissues. As shown in Figure 1, IHC staining results showed that in 125 non-tumor liver samples in the TMA, 118 samples showed positive staining for GLS2 (+; ≥10% positive staining cells) and 7 samples showed weak staining for GLS2 (±; <10% positive staining cells). In contrast, in 110 HCC samples in the TMA, 96 HCC samples showed negative for GLS2 (-; 0% positive staining cells), and 10 samples showed weak staining for GLS2 (±), whereas only 4 HCC samples showed positive staining for GLS2 (+) (p<0.0001; HCCs vs. nontumor liver tissues). The decreased expression of GLS2 protein appeared to be a common event in HCCs, which was observed in majority of HCCs at different stages and different histological grades (Supplementary Tables S1  and S2).
Similar results were observed in another set of HCC samples. IHC staining results showed that in 21 primary HCC samples, 19 samples showed negative staining for GLS2 (-), 1 showed weak staining (±) and 1 showed positive staining (+). In contrast, in the matched 21 adjacent non-tumor liver tissues, 20 samples showed positive staining for GLS2 (+) and 1 showed weak staining for GLS2 (±) (Supplementary Figure S1A; p<0.0001; HCCs vs. adjacent non-tumor liver tissues). Again, the decreased expression of GLS2 protein was observed in majority of HCCs at different stages and different histological grades (Supplementary Table S3). Consistent with the results from IHC staining, results from Taqman real-time PCR assays showed that GLS2 mRNA levels were clearly decreased in majority of HCCs; 18 of 21 HCCs showed decreased mRNA levels compared with their matched adjacent non-tumor liver tissues (the cut-off is 2-fold difference). The average reduction fold was ~8-fold (Supplementary Figure S1B). Taken together, these results clearly demonstrated that GLS2 expression is significantly decreased at both mRNA and protein levels in majority of primary HCCs, which strongly suggests a potential important role of GLS2 in tumor suppression of HCC.

GLS2 inhibits anchorage-independent growth of HCC cells and the growth of HCC xenograft tumors
We further investigated whether GLS2 plays a role in tumor suppression in HCC. As shown in Figure 2A, compared with non-tumor liver tissues (3 tissues from Origene), GLS2 mRNA levels were clearly decreased in different HCC cell lines, including Huh1, Huh7, PLC/ PRF/5, and Hep3B cells. To investigate the role of GLS2 in HCC, GLS2 was ectopically expressed in Huh1 and Huh7 cells by stable transduction of a retroviral vector, which expresses GLS2 ( Figure 2B, left panel). Control cells were transduced with empty vectors. These cell lines were then employed for both anchorage-independent growth assays in soft agar and xenograft tumor assays in nude mice. As shown in Figure 2B (middle and right panels), ectopic expression of GLS2 in Huh1 and Huh7 cells significantly inhibited the anchorage-independent growth of cells in soft agar compared with their control cells. Similarly, ectopic expression of GLS2 greatly reduced the growth of xenograft tumors formed by Huh1 and Huh7 cells compared with tumors formed by their control cells ( Figure 2C).
To investigate whether loss of GLS2 expression promotes tumorigenesis in HCC, endogenous GLS2 in PLC/PRF/5 cells was knocked down by stable transduction of shRNA vectors against GLS2 ( Figure 3A). PLC/PRF/5 cell line was used for shRNA knockdown since it shows relatively higher mRNA levels of GLS2 compared with other HCC cell lines that we tested ( Figure 2A). To avoid off-target effects, two different shRNA vectors against GLS2 were employed ( Figure 3A). As shown in Figure  3B, knockdown of GLS2 in PLC/PRF/5 cells significantly promoted the anchorage-independent cell growth in soft agar compared with their control cells transduced with control shRNA. Furthermore, knockdown of GLS2 also clearly promoted the growth of xenograft tumors formed by PLC/PRF/5 cells ( Figure 3C). These results together strongly suggest an important role of GLS2 in tumor suppression in HCC.

GLS2 negatively regulates the PI3K/AKT signaling
The PI3K/AKT signaling is frequently activated in various tumors, including HCC, which plays a critical role in promoting tumor cell growth, proliferation and survival [16,17]. The PI3K/AKT signaling was reported to be activated in over 50% of HCCs [18]. Interestingly, we found that GLS2 negatively regulated the PI3K/ AKT signaling in HCC cells. It is well-established that phosphorylation of AKT at Ser473 and Thr308 leads to the activation of AKT [16,17]. Ectopic GLS2 expression in Huh1 and Huh7 cells clearly reduced the phosphorylation  of AKT at Ser473 and Thr308 ( Figure 4A). The downregulation of AKT activities by GLS2 was also observed in HCC xenograft tumors; results from both Western-blot and IHC staining assays showed that AKT phosphorylation at Ser473 was clearly reduced in xenograft tumors formed by Huh1 cells with ectopic expression of GLS2 compared with tumors formed by control cells ( Figure  4B). Furthermore, knockdown of the endogenous GLS2 in PLC/PRF/5 cells clearly enhanced the phsphorylation of AKT at Ser473 and Thr308 ( Figure 4C). Consistently, AKT phosphorylation at Ser473 was clearly increased in xenograft tumors formed by PLC/PRF/5 cells with GLS2 knockdown compared with tumors formed by control cells ( Figure 4D). Considering the critical role of the oncogenic PI3K/AKT signaling in HCC, the negative regulation of the PI3K/AKT signaling by GLS2 could contribute greatly to the tumor suppression activity of GLS2 in HCC.

The negative regulation of the PI3K/AKT signaling by GLS2 contributes to tumor suppression activity of GLS2 in HCC
We further investigated whether negative regulation of the PI3K/AKT signaling by GLS2 contributes to GLS2's role in tumor suppression in HCC. To this end, Huh1 cells with ectopic GLS2 expression and control cells were stably transduced with a vector expressing a dominant negative AKT (DN-AKT; K179M), which inhibits AKT activities in cells [22,23] ( Figure 5A). The DN-AKT expression clearly inhibited the anchorageindependent growth of Huh1 cells ( Figure 5B) and the growth of Huh1 xenograft tumors ( Figure 5C). Notably, DN-AKT expression largely abolished the inhibitory effect of GLS2 on the anchorage-independent growth and the growth of xenograft tumors. The differences in colony number and growth rates of tumors between cells co-expressing GLS2 and DN-AKT (GLS2+DN-AKT) and cells expressing DN-AKT (DN-AKT) were much less pronounced compared with the differences between cells expressing GLS2 (GLS2) and control cells (Con) ( Figure  5B&C).
To investigate whether blocking the PI3K/AKT signaling can largely abolish the promoting effect of GLS2 knockdown on the anchorage-independent growth of HCC cells and the growth of HCC xenograft tumors, the DN-AKT (K179M) was stably expressed in PLC/ PRF/5 cells with stable GLS2 knockdown ( Figure 5D). Whereas GLS2 knockdown in PLC/PRF/5 cells clearly increased the colony number in soft agar and the growth rates of xenograft tumors, DN-AKT expression largely abolished the promoting effect of GLS2 knockdown on the anchorage-independent growth ( Figure 5E) and the growth of xenograft tumors ( Figure 5F). These results strongly suggest that the negative regulation of PI3K/ AKT signaling by GLS2 contributes greatly to the tumor suppressive activity of GLS2 in HCC.

Hypermethylation of the CpG sites in the promoter region of GLS2 in human HCC cells and primary HCCs.
Hypermethylation of the promoter is an important epigenetic mechanism that contributes to the decreased expression of tumor suppressor genes in human cancer, including HCC [19][20][21]. The promoter region of GLS2 gene is a CpG rich domain ( Figure 6A). To investigate whether promoter hypermethylation is an important mechanism contributing to the reduced expression of GLS2 in HCC, the methylation status of GLS2 promoter was analyzed in human HCC cell lines and primary HCCs. First, methylation-specific PCR (MSP) assays were employed to analyze the methylation status of the GLS2 promoter in multiple HCC cell lines, including Huh1, Huh7, Hep3B and PLC/PRF/5. Compared with the nontumor liver tissues (3 tissues from Origene), all of these cell lines showed clearly decreased mRNA levels of GLS2 (Figure 2A). MSP analysis revealed that methylation was detected in the GLS2 promoter region in all of these HCC cell lines. In contrast, no clear methylation or much lower methylation in GLS2 promoter region was detected in these 3 non-tumor liver tissues ( Figure 6B). To examine the methylation status of the GLS2 promoter in more detail, bisulfite genomic sequencing (BGS) assays were used to analyze the 2 non-tumor liver samples (Origene), Huh1 and Huh7 cells ( Figure 6C). Consistent with MSP results, BGS analysis showed that CpG sites in the GLS2 promoter region were extensively methylated on majority of CpG sites that we analyzed in these 2 HCC cell lines. In contrast, for these 2 non-tumor liver tissues, no methylation or much lower levels of methylation were observed in these CpG sites that showed hypermethylation in those HCC cells ( Figure 6C).
We further investigated whether the GLS2 promoter is hypermethylated in primary HCCs. MSP assays were employed to detect the methylation levels in GLS2 promoter in the above-mentioned 21 pairs of HCCs and their adjacent non-tumor liver tissues. MSP assays showed that the GLS2 promoter was hypermethylated in majority of HCCs. Compared with their matched non-tumor tissues, much higher levels of methylation (>2 fold higher) were detected in 16 of 21 tumors, similar levels of methylation were detected in 2 tumors (e.g. #10) and no clear methylation (e.g. # 7) was detected in 3 tumors ( Figure 6D). The hypermethylation of these 2 non-tumor liver tissues could be caused by the contamination of tumor cells in non-tumor tissues. Furthermore, GLS2 promoter hypermethylation is significantly associated with decreased GLS2 expression in HCCs (p<0.001; χ 2 test). These results strongly suggest that hypermethylation of GLS2 promoter is an important mechanism contributing to the decreased expression of GLS2 in HCC.

CpG methylation inhibits the transcriptional activity of the GLS2 promoter.
To directly investigate the effect of CpG hypermethylation on transcriptional activity of GLS2 promoter, GLS2 promoter region (-458 to +86 relative to the transcription start site) was amplified by PCR and subcloned into the promoter region of pGL2 luciferase reporter vectors. Luciferase reporter assays showed that the insertion of GLS2 promoter clearly activated the luciferase activities of the reporter vectors by over 100fold in Huh1 and Huh7 cells compared with control reporter vectors without insertion of GLS2 promoter ( Figure 7A). The reporter vectors were treated with CpG methyltransferase M. SssI to methylate cytosine residues of CpG in the vectors [24,25]. As a control, the vectors were mock-methylated. The completion of CpG methylation was confirmed by digestion of HpaII, a methylation-sensitive restriction enzyme ( Figure 7B). The GLS2 promoter DNA fragments cut from methylated and mock-methylated PGL2 vectors were then re-inserted into unmethylated pGL2 vectors by ligation reactions. After ligation reactions, the unmethylated pGL2 vectors containing methylated or mock-methylated GLS2 promoter were transfected into HCC cells for luciferase reporter assays. Reporter vectors containing methylated GLS2 promoter showed much weaker luciferase activities compared with vectors containing mock-methylated GLS2 promoter; GLS2 promoter methylation reduced the luciferase activities by over 30-40-fold in Huh1 and Huh7 cells ( Figure 7C). These results indicate that the hypermethylation of GLS2 promoter greatly reduces the transcriptional activities of the GLS2 promoter.
Methytransferase inhibitor 5-Aza-2'-deoxycytidine (5-Aza-dC) is a widely used DNA-demethylating agent, which induces expression of many genes with hypermethylated promoters [26,27]. To further investigate whether hypermethylation of GLS2 promoter is an important mechanism leading to the down-regulation of GLS2 expression in HCCs, multiple HCC cell lines, including Huh1, Huh7, Hep3B and PLC/PRF/5 cells, were treated with 5-Aza-dC. MSP analysis showed that 5-Aza-dC greatly reduced the methylation of CpG sites in the GLS2 promoter in these 4 HCC cell lines ( Figure 7D). Notably, 5-Aza-dC greatly induced the mRNA levels of GLS2 in these 4 HCC cell lines (by 3-7 folds) as measured by Taqman real-time PCR assays ( Figure 7E), indicating that demethylation of the GLS2 promoter increased GLS2 expression in HCC cells. Taken together, results in Figures 6 &7 suggest that GLS2 promoter hypermethylation is an important mechanism contributing to the decreased

Loss of heterozygosity of GLS2 in primary HCCs
Loss of heterozygosity (LOH) is frequently observed in various tumors, including HCC, which leads to the decreased expression and loss of function of many tumor suppressor genes [28,29]. To determine whether LOH contributes to the decreased expression of GLS2 in HCC, Taqman real-time PCR copy number assays were employed to determine the copy number of the gene in aforementioned 21 primary HCCs and their adjacent nontumor liver tissues. Three different copy number assays, which measured the 5', 3' and the central region of the GLS2 gene, respectively, were employed and the same results were obtained. Only 1 HCC lost a copy of GLS2 gene whereas the rest 20 HCCs retained both copies of GLS2 (Table 1), indicating that although LOH occurs in GLS2 gene in HCC, LOH is not a major mechanism accounting for the frequently observed down-regulation of GLS2 in HCC.

DISCUSSION
GLS2 was recently identified as a novel p53 target gene [11,12]. However, the role of GLS2 in tumorigenesis is unclear. In this study, we found that the GLS2 protein expression was significantly decreased in majority of primary HCCs that we examined compared with nontumor liver tissues, including liver tissues with cirrhosis and hepatitis (Supplementary Table S2). This result is consistent with previous reports from our group and another group showing that GLS2 mRNA expression was greatly decreased in majority of HCCs [11,12]. These results clearly demonstrated that the down-regulation of GLS2 is a common event in primary HCCs, which has the potential to be developed as a biomarker for early detection and diagnosis of HCCs. Our results further showed that ectopic GLS2 expression in HCC cells greatly reduced the anchorage-independent growth of cells and the growth of xenograft tumors. Consistently, knockdown of GLS2 in HCC cells clearly promoted the anchorage-independent growth of cells and growth of xenograft tumors. Our results demonstrated an important role of GLS2 in tumor suppression in HCC. In addition to the liver, GLS2 is also specifically expressed in the brain tissues. Interestingly, it was reported that the expression of GLS2 is also reduced in brain tumors [30]. It has been well-established that p53 plays a critical role in prevention of HCC [4,5]. As a direct p53 target, GLS2 could contribute greatly to the function of p53 in tumor suppression in HCC.
The aberrant activation of the PI3K/AKT signaling is frequently observed in HCC, which plays a critical role in liver tumorigenesis [16][17][18]. Results from this study showed that GLS2 is an important negative regulator of the PI3K/AKT signaling in HCC; ectopic GLS2 expression clearly reduced AKT activities in HCC cells, whereas GLS2 knockdown enhanced AKT activities. Blocking the PI3K/AKT signaling in HCC cells largely abolished the inhibitory effect of GLS2 on the anchorageindependent cell growth and xenograft tumor growth. These results strongly suggest that the negative regulation of PI3K/AKT signaling contributes greatly to GLS2's role in suppression of HCC. It remains unclear how GLS2 negatively regulates the PI3K/AKT signaling. GLS2 was reported to interact with other proteins through the PDZ domain in its C-terminus [31], which could contribute to its function in negative regulation of the PI3K/AKT signaling. Recently, GLS2 was also reported to regulate the expression of genes, although its mechanism is unclear [30]. Therefore, it is possible that GLS2 may change the expression of genes in the PI3K/AKT signaling pathway, contributing to the down-regulation of the PI3K/AKT signaling in cells. Further studies should shed light on the underlying mechanisms.
Our results further showed that the promoter region of GLS2 is hypermethylated in a high percentage of HCCs but not in their matched adjacent non-tumor liver tissues. The hypermethylation of GLS2 promoter was also observed in different HCC cell lines. Demethylation of GLS2 promoter by 5-Aza-dC greatly induced GLS2 expression in these cells. Luciferase reporter assays further showed that in vitro methylation of CpG sites in GLS2 promoter region greatly reduced the transcriptional activities of the GLS2 promoter. These results strongly Total 21 21 LOH of GLS2 in HCCs was detected by Taqman copy number assays. Three different copy number assays, which measured the 5', 3' and the central region of the GLS2 gene, respectively, were employed and the same results were obtained. Primary HCCs and their matched adjacent non-tumor liver samples (n=21; MD Anderson Cancer Center) were used for assays.
suggest that GLS2 promoter hypermethylation is an important mechanism contributing to the decreased GLS2 expression in HCC.
In summary, the results in this study clearly demonstrated that GLS2 protein expression is significantly decreased in majority of HCC, and hypermethylation of GLS2 promoter is an important mechanism that contributes to the decreased expression of GLS2 in HCC. Our results further showed that GLS2 plays an important role in tumor suppression in HCC, and the negative regulation of the PI3K/AKT signaling by GLS2 contributes greatly to the function of GLS2 in tumor suppression in HCC.

Tissue samples
Three tissue microarrays (TMAs) which include totally 110 HCC samples and 125 non-tumor liver tissues was provided from US Biomax (Rockville, MD). Twentyone cases of formalin-fixed and paraffin-embedded (FFPE) primary HCC samples and their matched adjacent non-tumor liver tissues were collected at the University of Texas MD Anderson Cancer Center with approved IRB. Three fresh frozen non-tumor liver samples were provided by Origene (Rockville, MD).

Immunohistochemistry (IHC) assays
The expression levels of GLS2 in tissues were determined by IHC assays as previously described [32]. GLS2 antibody was prepared as previously described [11].
Tyramide Signal Amplification plus dinitrphenyl (DNP)-HRP system (Perkin Elmer) was employed to increase the IHC signal. Rabbit IgG, replacing the primary antibody, was used as a negative control. Two known normal liver tissues were used as positive controls to normalize the staining efficiency from batch to batch. The IHC results were scored according to the percentage of tumor cells showing positive GLS2 staining: -: 0%; ±: < 10%; +: ≥10%.

Taqman real-time PCR and semi-quantitative RT-PCR assays
Total RNA from cells and fresh frozen tissues was prepared by using an RNeasy kit (Qiagen). Total RNA from FFPE tissues was prepared by using an RNeasy FFPE kit (Qiagen). cDNA was prepared using a TaqMan reverse transcription kit. Taqman real-time PCR was performed using TaqMan PCR mixture (Applied Biosystems) as previously described [11]. The expression levels of the genes were normalized with Actin gene. The primers used for semi-quantitative RT-PCR were as follows: For GLS2, 5'-AGGCGAGAGTGTGCTGAGTGCTG-3' and 5'-GCTGGTCCCCCTATGGCTGTTC-3'; For Actin, 5'-GGACTTCGAGCA AGAGATGG-3' and 5'-AGCACTGTGTTGGCGTACAG-3'.

Construction of luciferase reporter vectors, in vitro methylation and luciferase activity assays
The GLS2 promoter region (-458 to +86 relative to the transcription start site) was PCR amplified and subcloned into pGL2 luciferase reporter vectors (Promega) at BglII and HindIII sites. In vitro methylation of GLS2 promoter was performed as described [24,25]. In brief, the luciferase reporter vectors containing GLS2 promoter (10 µg) were methylated using CpG methyltransferase M.SssI, which methylates all cytosine residues of CpG. The completion of CpG methylation was confirmed by digestion of HpaII, a methylation-sensitive restriction enzyme which can not cut CCGG sites when CpG is methylated. In a parallel control reaction, the reporter vectors containing GLS2 promoter (10 µg) were mockmethylated. Methylated and mock-methylated reporter vectors were then digested with BglII and HindIII, and the GLS2 promoter DNA fragments were purified and then re-ligated into unmethylated pGL2 vectors, which were digested with BglII and HindIII. After ligation, the unmethylated pGL2 vectors containing methylated or mock-methylated GLS2 promoter were used directly for luciferase reporter assays. Luciferase reporter assays were performed as described [35]. Briefly, the reporter vectors containing methylated or mock-methylated GLS2 promoter were transfected into cells along with pRL-SV40 vectors expressing renilla luciferase as an internal control to normalize the transfection efficiency. The luciferase activity was measured 24 h after transfection.

Taqman real-time gene copy number analysis
Gene copy number was determined by duplex Taqman copy number assays using Taqman genotype Mix [32,36]. FAM-labeled primers for GLS2 and VIC-labeled primer for TERT as an internal control were obtained from Applied Biosystems. Three primers (Hs01976936_cn, Hs02840568_cn and Hs03075375_cn) which measure the 5' (intron 2-exon 3), 3' (intron 15-exon 16) and the central region of the GLS2 gene (intron 10-exon 11), respectively, were employed. The ΔΔCt method was used for data analysis.

Anchorage-independent growth assays
Anchorage-independent growth assays were performed in dishes coated with media containing 0.6% agarose. Cells were seeded on top of this layer in media containing 0.3% agarose. Colonies were stained and counted after 2-3 weeks [37].

Xenograft tumorigencity analysis
Xenograft tumorigenicity assays were performed as previously described [38]. Seven-week-old BALB/c nu/ nu male athymic nude mice (Taconic) were used. Cells (6x 10 6 for Huh1 cells, and 1x 10 7 for Huh7 and PLC/PRF/5 in 0.2 mL PBS) were injected (s.c.) into nude mice (n = 10/group). After injection, mice were examined and tumor volumes were measured 3 times/week for 3-5 weeks. Tumor volume = ½ (length × width 2 ). All experiments with animals were performed with the approval of the Institutional Animal Care and Use Committee.

Statistical analysis
The differences in tumor growth among groups were analyzed for statistical significance by ANOVA, followed by Student's t-tests. All other P values were obtained using Student t-tests or χ2 tests. **: p<0.001; *: p<0.01; #: p<0.05.