Stat3-mediated activation of miR-23a suppresses gluconeogenesis in hepatocellular carcinoma by downregulating G6PC and PGC-1α

Abstract

Considerable effort has been made in elucidating the mechanism and functional significance of high levels of aerobic glycolysis in cancer cells, commonly referred to as Warburg effect. Here, we investigated whether gluconeogenic pathway is significantly modulated in hepatocarcinogenesis, resulting in altered levels of glucose homeostasis. To test this possibility, we used a mouse model (mice fed choline deficient diet) that develops nonalcoholic steatohepatitis (NASH), preneoplastic nodules and hepatocellular carcinoma (HCC), along with human primary HCCs and HCC cells. This study demonstrated marked reduction in the expressions of G6pc, Pepck, and Fbp1 encoding the key gluconeogenic enzymes glucose-6-phosphatase, phosphoenolpyruvate carboxykinase, fructose-1,6-phosphatase and the transcription factor Pgc-1α in HCCs developed in the mouse model that correlated with reduction in serum glucose in tumor bearing mice. The mRNA levels of these genes were also reduced by ~80% in the majority of primary human HCCs compared to matching peritumoral livers. The expression of miR-23a, a candidate microRNA targeting PGC-1α and G6PC, was upregulated in the mouse liver tumors as well as in primary human HCC. We confirmed PGC-1α and G6PC as direct targets of miR-23a and their expressions negatively correlated with miR-23a expression in human HCCs. G6PC expression also correlated with tumor grade in human primary HCCs. Finally, this study showed that the activation of IL-6-Stat3 signaling caused the upregulation of miR-23a expression in HCC.

Conclusion

Based on these data, we conclude that gluconeogenesis is severely compromised in HCC by IL6-Stat3-mediated activation of miR-23a which directly targets PGC-1α and G6PC, leading to decreased glucose production.

Introduction

Until recently, little effort has been made to understand the significance of Otto Warburg’s pioneering demonstration more than 8 decades ago that proliferating cancer cells metabolize 8–10-fold more glucose to lactic acid than corresponding normal tissues under aerobic conditions (1). The relatively heavy dependence of the tumors on glycolysis even in the presence of abundant oxygen occurs with concomitant reduction in oxidative phosphorylation. The predominance of aerobic glycolysis characteristic of most cancer cells facilitates conversion of pyruvate produced during glycolysis to lactate, which is secreted into the blood instead of oxidation to completion (2–4). Increased glucose uptake and metabolism correlate with poor prognosis of many tumor types (2, 5–6), supporting the notion that metabolic alterations may contribute to the malignant phenotype (7).

As articulated by Levine and Puzio-Kuter (3), several compelling issues contributed to the lack of interest in the altered carbohydrate metabolism of cancer cells. First, while the conversion of pyruvate to lactate results in the recovery of NAD+ required for the maintenance of glycolysis and for continued cell proliferation in vivo (8), this process is energetically inefficient, producing just two out of 36 ATP molecules for each mole of glucose expended. Second, it is not known how such a dramatic shift in energy metabolism occurs and how this altered metabolism results in cancer phenotype. Third, the relationship between metabolism deregulation and oncogene activation/tumor suppressor gene inactivation has not been established. After more than eight decades since Warburg proposed his hypothesis for the elicitation of cancer phenotype, the long abandoned idea for the dependence of cancer cells on increased utilization of glycolysis has been resurrected, and has also provided an impetus for the development of novel anticancer drugs (9).

While considerable effort has been made in understanding the functional significance of higher rates of glycolysis and production of lactate in malignant cells, the possibility of altered gluconeogenesis that could potentially facilitate glycolytic pathway has not been explored. A mouse model for hepatocellular carcinoma is an ideal system to determine potential modifications in gluconeogenesis pathway for two reasons. First, only the liver contains a full complement of all the enzymes essential for gluconeogenesis, particularly glucose-6-phosphatase (G6PC) that converts glucose-6-phosphate to glucose. Glucose is then released to circulation and subsequently utilized by other tissues including muscle. Second, a suitable mouse model for HCC is now available, which can induce liver tumors by feeding choline deficient and amino acid defined (CDAA) diet in the absence of any exogenous chemicals or virus (10). Further, tumorigenesis in this model involves steatosis, inflammation, fibrosis and insulin resistance that are the hallmarks of human HCC (11–12). We have used this model to study altered expression of microRNAs (miRs) at different stages of hepatocarcinogenesis (13–14). Recently, these mice were used to demonstrate resistance of jnk1−/− (c-Jun N-terminal kinase) mice to CDAA-diet induced steatohepatitis and liver fibrosis (11). In another study, steatohepatitis and liver fibrosis in mice fed CDAA diet were significantly reduced in TLR (Toll-like receptor) null mice (12).

In the present study, we used this diet-induced model to study alterations in gluconeogenesis during the process of liver tumorigenesis. Surprisingly, we found dramatic inhibition of key gluconeogenic enzymes and of a transcription factor involved in this metabolic pathway in the liver tumors. We also explored the molecular mechanisms underlying the suppression of these enzymes and probable functional significance of this observation.

Materials and methods

Mice and Diet

The animal model and the dietary regimen to induce HCC are described earlier (13). All animal experiments were carried out under protocols approved by the Ohio State University Institutional Laboratory Animal Care and Use Committee.

Human HCC Specimens

Primary human hepatocellular tumor and adjacent peritumoral tissue specimens were obtained from the Cooperative Human Tissue Network at The Ohio State University James Cancer Hospital. Tissue specimens were procured in accordance with The Ohio State University Cancer Internal Review Board guidelines. The detailed information of patients was listed in Supplementary Table 1.

Statistical Analysis

Statistical analysis of differences between groups was performed by unpaired Student’s t test, and P≤0.05 was considered to be statistically significant. Paired Student’s t test was used to analyze differences in expression of microRNAs and mRNAs levels among tumors and paired peritumoral tissues in real-time RT-PCR (qPCR) analysis. The correlation between miR-23a level, and G6PC, PEPCK and PGC-1α mRNA levels was analyzed by two-tailed Pearson Correlation Test. Correlation between G6PC expression and clinico-pathological characteristics in human HCCs was analyzed by Wilcoxon Mann Whitney test or One-Way ANOVA test.

All other materials and methods are described in the supplementary information.

Result

Development of HCC in mice fed CDAA diet for 71 to 84 weeks

Previously, we have shown that C57BL/6 mice fed choline deficient and amino acide defined (CDAA) diet progressively developed NASH (~32 week) and preneoplastic nodules (~65 week) (13). In this study, mice were fed with CDAA and control CSAA diet for longer time (71 to 84 weeks) and examined the development of HCC. As previously reported (11), mice fed CDAA diet showed significant weight increase up to preneoplastic stage (~65 week) (Fig. 1A). Although the body weight started to decrease after 65 week due to tumor development, all mice still gained weight when comparing the final to initial body weight (Fig. 1A). After 71 week, almost all mice fed CDAA diet (20 out of 21) developed NASH, and the majority of these mice (16 out of 20) developed hepatic tumors (Fig. 1B). In contrast, although 15 out of 21 mice fed CSAA diet developed NASH, because of higher calorie and fat content in CSAA diet compared to standard chow diet, only 6 of them developed tumors. Moreover, the liver to body weight ratio, tumor number and largest tumor weight were much higher in mice fed CDAA than those fed CSAA diet (Fig. 1C). Histopathological analysis confirmed that most of these tumors were grade II HCC (Fig. 1B and 1D). H&E staining showed massive inflammation in the livers of mice fed CDAA diet compared to those fed CSAA diet (Fig. 1E). Masson’s trichrome staining revealed severe fibrosis in CDAA diet livers, but not in CSAA diet livers (Fig. 1E). A two fold increase in hydroxyproline level, which reflects the amount of collagen, corroborated severe fibrosis in CDAA diet livers compared to CSAA diet livers (Fig. 1E). Ki-67 staining indicated that cells in CDAA diet livers were highly proliferative (Fig. 1F). Furthermore, increased ALT level in sera suggested liver damage in tumor bearing mice fed CDAA diet compared to CSAA diet (Supplementary Table 2). We also observed significant decrease in serum glucose level in tumor bearing mice after overnight fasting (167.55±14.42 vs 256.36±21.89 mg/dL, P=0.0029) (Supplementary Table 2).

Figure 1. Mice fed CDAA diet for 71 to 84 weeks developed fibrosis and HCC.

(A) Weight gain of mice fed CDAA and CSAA diet. Asterisks denote P<0.05. (B) Representative photographs and H&E staining of mouse liver from mice fed CDAA and CSAA diet. (C) Liver to body weight ratio, tumor number and tumor weight in mice fed CSAA and CDAA diet. (D) Histopathological analysis of liver sections from mice fed CSAA and CDAA diet. (E) Representative H&E and Masson’s trichrome staining sections of CSAA and CDAA diet livers, as well as quantification of hydroxyproline in the livers. (F) Representative Ki-67 staining sections from CSAA and CDAA diet livers.

The genes encoding glucose 6-phosphatase (G6PC), phosphoenolpyruvate carboxykinase (PEPCK), fructose-1,6-bisphosphatase (FBP1), pyruvate carboxylase (PC) and peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PGC-1α) are down-regulated in CDAA diet-induced mouse HCC model

We postulated that higher glycolysis in hepatocellular carcinoma is accompanied by suppression of gluconeogenesis preventing conversion of glucose 6-phosphate to glucose and its release into blood as well as other tissues. To test this possibility, we used CDAA diet induced mouse HCC model described above. To determine the gluconeogenesis level in CDAA diet induced mouse HCC, we first examined the expression of genes involved in gluconeogeensis at different stages of tumorigenesis. qPCR analysis showed that the levels of mRNAs corresponding to glucose 6-phosphatase (G6pc), phosphoenolpyruvate carboxykinase (Pepck), fructose-1,6-bisphosphatase (Fbp1) and pyruvate carboxylase (Pc), key enzymes involved in gluconeogenesis, dramatically decreased in the liver tumors developed in mice at 84 week. The tumor formed in one mouse at 65 week also exhibited markedly reduced levels of these enzymes (Fig. 2A). Similarly, the level of PPARGC1A (Pgc-1α), a transcription factor that regulates G6pc and Pepck expression (15), also decreased in tumors (Fig. 2A). Western blot analysis showed that G6PC and PGC-1α protein levels were significantly diminished in tumors (Fig. 2B). The G6pc phosphatase activity in liver tissues also declined gradually during tumorigenesis (Fig. 2C).

Figure 2. The expression of gluconeogenic genes is reduced in CDAA diet-induced mouse HCC model.

(A) G6pc, Pepck, Fbp1, Pc and Pgc-1α mRNA level decreased during hepatocarcinogenesis. The mRNA levels were measured by qPCR in the liver of mice fed CSAA and CDAA diet for 32, 65 and 84 weeks (n = 4–5 per group with the exception of no tumor in 32 weeks and 1 tumor in 65 weeks). Data are represented as mean ± SD. (B) Western blot analysis of G6PC and PGC-1α protein level in liver extracts from mice fed CSAA and CDAA diet as in (A). (C) G6PC phosphatase activity in liver tissues was measured by colorimetric methods using glucose-6-phosphate as substrate (n = 4–5 mice per group). Data are represented as mean ± SD.

G6PC, PEPCK, FBP1 and PGC-1α expression are significantly reduced in primary human HCC

Next, we assessed the expression of these gluconeogenic enzymes and PGC-1α expression in human primary HCC. qPCR analysis showed that the expression of most of these genes was significantly reduced in human primary HCCs compared to matching livers (Fig. 3A). The mRNA levels of G6PC, PEPCK and PGC-1α decreased by 73–80% in 24, 26 and 24 pairs, respectively out of 30 pairs of HCC examined. The mRNA level of FBP1 was also reduced by 48–95% in 15 out of 17 HCC specimens (Fig. 3A), whereas the mRNA level of PC was not significantly altered in human HCC. Western blot analysis showed dramatic reduction (50%) in G6PC protein level in primary HCC (Fig. 3B). The curtailed G6PC enzyme activity corroborated with its low level of expression in the human HCC specimens (Fig. 3C).

Figure 3. G6PC, PEPCK, FBP1 and PGC-1α expression are significantly inhibited in primary human HCC.

(A) G6PC, PEPCK, FBP1, PC and PGC-1α mRNA levels were measured by qPCR in HCC and pair-matched peritumoral liver tissues. Each data point represents one patient, and horizontal bars denote the mean ± SEM. (B) Western blot analysis of G6PC protein level in 6 pairs of HCC and matching livers. The signal was quantified using Kodak Imaging software and the data was normalized to Ku-70. T: tumor, N: non-tumor matching liver. (C) G6PC phosphatase activity in 6 pairs of HCC and matching livers.

G6PC expression correlates with tumor grade in human primary HCCs

To address the clinical relevance of gluconeogenesis inhibition, we analyzed the correlation between G6PC mRNA level and clinical features in 29 human primary HCC patients by comparing the mRNA level of G6PC in different subgroups of patients based on clinical parameters. One way ANOVA analysis revealed significant correlation between G6PC expression and tumor grade (P=0.0066) (Table 1). The mRNA level of G6PC was ~32% and ~19% in moderately and poorly differentiated tumors compared to that in well differentiated tumors. However, there was no significant association between G6PC expression and tumor invasion status and tumor stage. No significant difference was observed in G6PC expression among HCC subgroups with regard to age, gender, and tumor size.

Table 1.

Correlation between G6PC mRNA levels and clinico-pathological characteristics in human primary HCC specimens. T and N denote tumor and peritumoral liver tissues, respectively

Characteristics	no	G6PC Expression (T/N)		P value
		Mean±SEM	Median
Age				P=0.62*
>60	14	0.60±0.21	0.29
≤60	15	0.53±0.23	0.23
Gender				P=0.73*
Male	17	0.60±0.21	0.27
female	12	0.48±0.19	0.23
Tumor grade (histological differentiation)				P=0.0066**
well	7	1.24±0.38	0.95
moderate	16	0.39±0.18	0.15
poor	6	0.23±0.10	0.17
Tumor size				P=0.89*
>5cm	23	0.49±0.15	0.27
≤5cm	6	0.84±0.49	0.14
Vascular invasion				P=0.35*
Yes	14	0.49±0.23	0.23
No	15	0.63±0.21	0.31
Tumor stage				P=0.49**
Stage I	14	0.67±0.22	0.35
Stage II	10	0.62±0.32	0.19
Stage III	5	0.16±0.05	0.23

^*Wilcoxon Mann Whitney test.

^**One-Way ANOVA test.

Expression of miR-23a, a candidate microRNA targeting PGC-1α and G6PC, is elevated in CDAA diet induced liver tumors and in human primary HCC

Next, we sought to identify possible mechanisms underlying the downregulation of genes encoding the key proteins involved in gluconeogenesis in diet-induced HCC model and primary HCC. MicroRNAs (miRs) regulate various biological processes including metabolism by inhibiting target gene expression (16). We explored the potential involvement of specific miRs in regulating gluconeogenesis. Analysis of the 3′UTR of PGC-1α mRNA revealed two conserved miR-23a sites predicted by three different databases (TargetScan, Pictar and miRanda) (Fig. 4A). Interestingly, G6PC 3′UTR also harbors one miR-23a site predicted by TargetScan. Previously, miR-23a has been shown to regulate glutamine metabolism by targeting glutaminase in lymphoma and prostate cancer cells (17). Another study demonstrated its upregulation in HCC (18). To determine whether miR-23a can regulate PGC-1α and G6PC expression in liver tumors, we first measured its expression in diet-induced HCC model by qPCR. The results displayed gradual increase in miR-23a expression during tumorigenesis in the livers of mice fed CDAA diet compared to those fed CSAA (control) diet (Fig. 4B). Its expression increased from 1.65-fold (P=0.08) in the livers from mice fed CDAA diet at 65 week to 2.5-fold (P=0.019) in the tumors from mice fed CDAA diet at 84 week compared to those fed CSAA diet. It should be noted that 2.5-fold difference in microRNA levels is considered highly significant, as this level of increase in a single miR can decrease expression of multiple genes (19–20).

Figure 4. Expression of miR-23a is elevated in CDAA diet induced tumors and human primary HCC.

(A) miR-23a sites in PGC-1α and G6PC 3′UTR. (B) miR-23a level was measured by qPCR in the livers of mice fed CSAA and CDAA diet for 32, 65 and 84 weeks (n = 4–5 mice per group). The data was normalized to RNU6B. Data are represented as mean ± SD. (C) miR-23a level was measured by qPCR in 30 HCC and pair-matched peritumoral liver tissues. Each data point represents one patient, and horizontal bars denote the mean ± SEM. (D) Inverse correlation between G6PC (r=−0.55, P=0.002), PEPCK (r=−0.35, P=0.059), PGC-1α (r=−0.37, P=0.043) mRNA level and miR-23a expression in HCCs determined by qPCR analysis in 30 HCC samples.

To address the functional relevance of enhanced miR-23a levels in human HCC, miR-23a expression was analyzed in human primary HCCs and pair-matched peritumoral liver tissues. Among the 30 HCC samples analyzed, miR-23a expression was significantly elevated (~2-fold) in 21 HCC samples (P=0.0046) (Fig. 4C). Comparison of miR-23a level and mRNA levels corresponding to PGC-1α, G6PC and PEPCK in HCCs exhibited inverse correlation between miR-23a and PGC-1α (r=−0.37, P=0.047), G6PC (r=−0.55, P=0.002) and PEPCK (r=−0.35, P=0.059) (Fig. 4D). In contrast, there is no correlation between miR-23a and FBP1 or PC levels (data not shown).

PGC-1α and G6PC are direct targets of miR-23a

To determine whether PGC-1α and G6PC are direct targets of miR-23a, wild-type and mutant 3′UTRs lacking miR-23a binding sites (3′UTRΔ) were cloned into psiCHECK2 luciferase reporter vector downstream of renilla luciferase coding region. The constructs were then co-transfected with miR-23a precursor or negative control (NC) RNA into H293FT cells. The overexpression of miR-23a (~15 fold) was confirmed by real-time PCR (Supplementary Fig. 1). It should be noted that ectopic expression of specific miR in cell lines after transfecting miR mimics results in higher expression of the miR than in vivo level (21). The relative luciferase activity was reduced by 40% (P=0.0001) and 25% (P=0.002) in PGC-1α and G6PC wild type 3′UTR, respectively, but not in respective mutant 3′UTRs or psiCHECK2 vector alone (Fig. 5A), suggesting that miR-23a cognate sites are essential for negative regulation of luciferase expression driven by PGC-1α-3′-UTR and G6PC-3′UTR.

Figure 5. PGC-1α and G6PC are direct targets of miR-23a.

(A) Luciferase assay. H293FT cells were co-transfected with the reporters and miR-23a or NC RNA (50 nM). After 48 h, RLU2/RLU1 activity was measured. (B and C) Overexpression of miR-23a reduces PGC-1α and G6PC expression in Hep3B cells and mouse hepatocytes. Hep3B cells and mouse hepatocytes were transfected with miR-23a or NC RNA (50 nM) followed by measurements of PGC-1α and G6PC mRNA, protein levels and G6PC phosphotase activity. (D and E) Depletion of miR-23a increases PGC-1α and G6PC expression in Hep3B cells and mouse hepatocytes. Hep3B cells and mouse hepatocytes were transfected with anti-miR-23a or anti-NC RNA (100 nM) followed by measurement of PGC-1α and G6PC mRNA, protein levels and G6PC phosphotase activity.

To confirm that miR-23a can indeed suppress expression of endogenous PGC-1α and G6PC, Hep3B cells and mouse hepatocytes were transfected with miR-23a or negative control RNA, followed by measurements of their mRNA and protein levels. PGC-1α and G6PC mRNA levels decreased by ~40–50% (P=0.01 and 0.02, respectively) in both cells overexpressing miR-23a (Fig. 5B and 5C). More than 30% reduction in PGC-1α protein level was observed in both Hep3B and hepatocytes (Fig. 5B). G6PC protein level was reduced by 28% upon expression of miR-23a in hepatocytes compared to negative control RNA (Fig. 5C). Next, we measured G6PC enzyme activity since it was not detectable by western blot analysis in Hep3B cells. The glucose-6-phosphatase activity in Hep3B cells transfected with miR-23a precursor indeed decreased by ~50% compared to cells transfected with control RNA (Fig. 5C).

We also performed the reverse experiment by transfecting anti-miR-23a in Hep3B cells and mouse hepatocytes that resulted in ~90% depletion of miR-23a (data not shown). PGC-1α (Fig. 5D) and G6PC (Fig. 5E) mRNA levels increased by 20%~70% in both cells depleted of miR-23a. The respective protein levels and glucose-6-phosphatase activity also increased by 30~49% in these cells (Fig. 5D and 5E). These results indicate that miR-23a interferes with PGC-1α and G6PC expression at both mRNA and protein levels. Taken together, these data demonstrate that PGC-1α and G6PC are direct targets of miR-23a. This is a functionally significant observation, as nearly 90% of glucose-6 phosphatase is present in the liver. Consequently, suppression of G6PC will block the last critical reaction essential for the production of glucose and its release to blood for uptake by glucose-requiring tissues.

miR-23a inhibits glucose production in hepatocytes

As indicated earlier, we observed significant reduction in the glucose levels in the sera of tumor-bearing mice fed CDAA diet (Supplementary Table 2) consistent with inhibition of gluconeogenesis in HCC. Because PGC-1α and G6PC are involved in gluconeogenesis in liver, we investigated potential regulation of hepatic gluconeogenesis by miR-23a, resulting in decreased glucose production. For this purpose, miR-23a precursor or anti-miR inhibitor was transfected into hepatocytes isolated from mouse liver followed by replacement of the culture medium with glucose production medium 48 hours later. After 6 hours, glucose synthesis was analyzed by measuring glucose concentration in the culture supernatant. The amount of glucose released from hepatocytes revealed a 40% reduction in glucose production after ectopic expression of miR-23a (Fig. 6A) whereas suppressing miR-23a expression increased its formation by 30%. Overexpression or depletion of miR-23a was confirmed by real-time PCR analysis (Fig. 6B). Taken together, these data indicate that miR-23a negatively regulates glucose production.

Figure 6. miR-23a regulates glucose production in vitro.

Mouse hepatocytes were transfected with miR-23a mimic (50 nM) or anti-miR-23a inhibitor (100 nM) followed by measurement of glucose levels (A) and miR-23a levels (B).

IL-6 and Stat3 signaling pathway modulate the expression of miR-23a

Next, we sought to delineate the potential mechanisms underlying upregulation of hepatic miR-23a during hepatocarcinogenesis in mice fed CDAA diet. Analysis of miR-23a promoter region using rVISTA program (22) identified a potential Stat3 binding site (23) that is conserved in human and mouse (Supplementary Fig. 2). Stat3 has been shown to inhibit gluconeogenesis by suppressing PGC-1α, G6PC and PEPCK expression (24). This observation coupled with the present demonstration of miR-23a mediated inhibition of gluconeogenesis prompted us to hypothesize that Stat3 may be involved in the regulation of miR-23a expression in liver. To test this hypothesis, we first measured Stat3 levels in the extracts of livers from CDAA diet-fed mice by western blot analysis. The phospho-Stat3 level indeed was higher in the tumor tissues compared to the matching liver tissues (Supplementary Fig. 3). To assess further the potential regulation of miR-23a expression by Stat3, Hep3B cells were treated with cryptotanshinone (25), a specific inhibitor of Stat3 phosphorylation, for 12 hours followed by measurements of miR-23a level. Inhibition of Stat3 mitigated miR-23a expression by ~30% compared to the vehicle (DMSO) treated cells (Fig. 7A). Similarly, knocking down Stat3 expression in Hepa cells by siRNA reduced miR-23a expression by more than 50% (Fig. 7A), suggesting that Stat3 positively regulates miR-23a expression.

Figure 7. IL-6-Stat3 signaling pathway activates miR-23a expression.

(A) Inhibition of Stat3 activation decreases miR-23a expression. Hep3B and Hepa cells were treated with Stat3 inhibitor (cryptotanshinone) (10 μM) or transfected with siRNA (40 nM) followed by western blot analysis of p-Stat3 and Stat3 protein levels, and qPCR analysis of miR-23a expression. (B) Measurements of serum IL-6 level from tumor bearing and age-matching control mice by ELISA. Each data point represents one mouse, and horizontal bars denote the mean ± SEM. (C) IL-6 increased miR-23a expression. Hep3B and Hepa cells were treated with IL-6 (10ng/ml) for 12h and subjected to qPCR analysis of miR-23a levels. (D) miR-23a promoter luciferase assay. Hepa cells were co-transfected with the reporters along with control (siNC) or Stat3 (siStat3) siRNA (50 nM). After 48 h, RLU2/RLU1 activity was measured. For IL-6 treatment, Hepa cells transfected with the reporters were treated with/without IL-6 (25ng/ml) for 12h before analysis of luciferase activity. psiCHECK2 vector without promoter (del-promoter) was used as a control plasmid. (E) Stat3 directly binds to miR-23a promoter region. Hepa cells were transfected with Stat3 siRNA or treated with/without IL-6 followed by chromatin immunoprecipitation (ChIP) analysis.

It is well known that IL-6 activates Stat3 to suppress gluconeogenesis (26). Serum IL-6 level was indeed ~2 fold higher in tumor bearing mice fed CDAA diet compared to control mice fed CSAA diet (Fig. 7B). This observation suggested the possibility that IL-6 may regulate miR-23a expression. To confirm this notion Hep3B and Hepa cells were treated with IL-6 for 12 hours, and miR-23a levels were analyzed by real-time PCR that showed ~2-fold increased expression of miR-23a in both cell lines (Fig. 7C). To further test if Stat3 signaling pathway regulates miR-23a expression by activating its promoter, we cloned the mouse miR-23a promoter region into psiCHECK2 vector upstream of renilla luciferase coding region and performed luciferase assay after knocking down Stat3 or with/without IL-6 treatment. Knockdown of Stat3 expression in Hepa cells with siRNA reduced the luciferase activity by ~30% compared to control siRNA transfected cells (Fig. 7D). In contrast, IL-6 treatment increased luciferase activity by ~50% compared to untreated cells. As expected, luciferase activity driven by psiCHECK2 alone did not respond to siRNA or IL-6 treatment (Fig. 7D).

To determine if Stat3 directly binds to miR-23a promoter region, chromatin immunoprecipitation (ChIP) analysis was performed in Hepa cells transfected with Stat3 siRNA (siStat3) or control siRNA (siNC), and Hepa cells treated with or without IL-6. The results showed ~25-fold enrichment of Stat3 at the miR-23a promoter region compared to no antibody control samples (Fig. 7E). Similarly, we observed ~2.6-fold increase in Stat3 enrichment at miR-23a promoter in Hepa cells treated with IL-6 compared to untreated cells. In contrast, recruitment of Stat3 to the same region was reduced by 80% in Stat3 depleted cells (Fig. 7E). These data, taken together, indicate that Stat3 can directly bind to the promoter region of miR-23a and increase its expression.

DISCUSSION

Overall significance of this study

While there has been considerable progress in understanding the functional significance and probable mechanisms of Warburg effect, the potential role of gluconeogenesis during transformation of normal cells to cancer cells, particularly in liver cancer, has not been explored. We took advantage of a mouse model system that mimics many of the biochemical and pathological changes characteristic of nonalcoholic steatohepatitis and produces hepatocellular carcinoma. It should be noted that this model has now been successfully utilized by many laboratories to address important genetic, epigenetic and signaling mechanisms involved in hepatocarcinogenesis (11–14). We have now shown that this model can be effectively used to study alterations in gluconeogenesis at different stages of hepatocarcinogenesis and the potential mechanisms. A key observation in the present study was almost complete suppression of glucose-6-phosphatase gene expression and its enzymatic activity in tumor. Because it is the last enzyme in the gluconeogenesis pathway as well as one of the four irreversible enzymes in this pathway, and the majority of this enzyme (>80%) is present in the liver, its inhibition will have major functional consequences. Accordingly, the glucose production via gluconeogenesis will be severely blocked, resulting in markedly reduced release of glucose into circulation and consequently curtailing the supply of glucose for energy requirements in other tissues such as exercising muscles. Suppression of the genes encoding the rate-limiting enzyme FBP1, and PEPCK, an enzyme acting upstream of FBP1 as well as a key transcription factor involved in the expression of G6PC and PEPCK further leads to drastic reduction of gluconeogenesis.

Role of miR-23a in suppression of gluconeogenesis

A key observation is that enhanced phospho-Stat3 levels (Supplementary Fig. 3) caused marked increase in miR-23a expression in HCC produced in the mouse model (Fig. 4B), causing suppression of two key miR-23a targeted genes involved in gluconeogenesis. The difference in phosphorylation between the tumors and peritumoral liver tissues in 84w CDAA group was not as dramatic as 65w, because the peritumoral regions were full of microscopic tumors at 84w. It was technically difficult to dissect the tumor-free matching liver tissues by micro-dissection at this stage of tumor development. To our knowledge, this is the first report that demonstrates the role of a specific microRNA in regulating gluconeogenesis in HCC. The direct role of Stat3 in miR-23a expression was confirmed by (a) significant reduction in miR-23a expression following treatment of HCC cells with Stat3 specific inhibitor or by Stat3-specific siRNA, (b) enhanced miR-23a expression upon treatment of the two HCC cell lines Hep3B and Hepa cells with the cytokine IL-6 that is known to cause inflammation and activation of the oncogenic transcription factor Stat3 in the liver (27) and (c) demonstration of direct binding of Stat3 to miR-23a promoter and its enrichment at the miR promoter following treatment of cells with IL-6. IL-6-Stat3 signaling pathway has been shown to inhibit gluconeogenesis by suppressing G6pc and Pepck in stat3 null mice (24). Previously, Stat3 has been shown to suppress G6PC and PEPCK expression by directly binding to the promoter region (28). Our study has revealed another layer of complexity in the regulation of these enzymes involving miR-23a. The present study has shown that IL-6-Stat3 signaling pathway inhibits gluconeogenesis, in part, by upregulating miR-23a and compromising glucose production by directly targeting G6PC and PGC-1α (See Supplementary Fig. 4 for a proposed model).

Recently, SIRT1 has been shown to regulate acetylation and phosphorylation of Stat3 and thus suppresses its inhibitory effect on gluconeogenesis (29). Our results showed that activation of Stat3 upregulates miR-23a expression. It will be of interest to determine if SIRT1 regulates miR-23a expression. miR-23a does not appear to be involved in the regulation of FBP1 expression, as there is no miR-23a binding site in its 3′UTR.

How does reduced gluconeogenesis facilitate tumorigenesis?

The drastic reduction in gluconeogenesis is reflected in the significant decrease in the level of glucose released from hepatocytes following increased expression of miR-23a, or in augmented glucose release in response to miR-23a suppression (Fig. 6A). Further, the glucose level in sera from the tumor bearing mice was markedly diminished compared to normal sera (Supplementary Table 2). Although the level of upstream enzyme PEPCK was also curtailed, drastic reduction in G6PC is likely to result in accumulation of glucose-6-phosphate (G6P). Because G6P is utilized in the hexose monophosphate (HMP) shunt pathway for glucose metabolism, the tumors could metabolize excess G6P via this pathway and produce ribose-5-phosphate used in nucleotide synthesis. The enhanced production of this key building block of nucleic acids is probably an important means of meeting the basic requirement for rapid cell division and growth of tumors. The increased utilization of glycolytic and HMP shunt pathways in HCC depending on the extent of accumulation of G6P due to block in gluconeogenesis may contribute to the survival of the tumor cells under hypoxic conditions.

Role of glutaminase in HCC proliferation and survival

The role of other factors relevant to cellular energy metabolism in cancer merits discussion. One such factor is glutamine that has been recently rediscovered as an important growth stimulating amino acid after nearly four decades since the initial observations of its potential role in energy metabolism (30–31). Recent study has shown that glutaminase 1 (GLS), which converts glutamine to glutamate, is upregulated in lymphoma and prostate cancer cell lines by Myc oncogene through its repression of miR-23a/b, two microRNAs targeting GLS (17). Interestingly, we observed marked stimulation of GLS expression in the diet induced mouse HCC (data not shown) irrespective of the increased expression of miR-23a, suggesting that GLS is probably regulated by some other mechanisms in liver cancer. The increased expression of GLS results in elevated levels of reduced glutathione (GSH) that scavenges reactive oxygen species. It is conceivable that NADPH required for GSH synthesis could be efficiently generated by augmented HMP shunt pathway resulting from inhibition of G6PC and accumulation of glucose 6-phosphate in HCC. Further study is required to address this issue which is beyond the scope of this work.

Abbreviations

NASH

nonalcoholic steatohepatitis

HCC

hepatocellular carcinoma

G6PC

glucose-6-phosphatase, catalytic subunit

PEPCK

phosphoenolpyruvate carboxykinase

PGC-1α

peroxisome proliferator-activated receptor gamma, coactivator 1 alpha

miR

microRNA

CDAA

choline deficient and amino acid defined

CSAA

choline sufficient and amino acid defined

G6P

glucose-6-phosphate

GLS

glutaminase

HMP

hexose monophosphate

qPCR

real-time RT-PCR