Abstract
Deregulation of microRNAs (miRNAs) and c-Myc (Myc) contributes to hepatocellular carcinoma (HCC) progression, but how miRNAs and Myc regulate each other in hepatocarcinogenesis is still poorly understood. Using a functional screen, we identified miR-129-5p as a miRNA that inhibits HCC cell growth. miR-129-5p targets the mitochondrial matrix protein pyruvate dehydrogenase kinase 4 (PDK4), which leads to decreased phosphorylation of the E1α subunit of pyruvate dehyrogenase (PDH) complex, inhibition of glycolysis, retarded tumor growth, and impaired lung colonization. Enforced expression of PDK4 refractory to inhibition by miR-129-5p rescued all of these phenotypes. Targeting PDK4 by shRNA recapitulated the effects caused by miR-129-5p. miR-129-5p is transcriptionally repressed by a complex comprised of Myc, histone deacetylase 3 (HDAC3), and enhancer of zeste 2 polycomb repressive complex 2 (EZH2). Levels of miR-129-5p negatively correlated with clinical stages in human HCC. Restoring miR-129-5p expression suppressed the diethylnitrosamine (DEN)-induced hepatocarcinogenesis in mice. Thus, we concluded that miR-129-5p, which is a negative target of Myc, blocks glycolysis to retard hepatocarcinogenesis via targeting PDK4. The critical link between miR-129-5p and PDK4 in the progression of HCC suggests potential points of therapeutic intervention for this disease.
Introduction
c-Myc (Myc) acts as a transcriptional activator or repressor to regulate target genes, including those encoding miRNAs (Dang, 2012). Through a myriad of downstream targets, Myc plays crucial roles in cell growth, metabolism, and tumorigenesis. Myc is frequently altered in many human cancers, including hepatocellular carcinoma (HCC) (Nesbit et al., 1999; Laurenti et al., 2009; Dang, 2012). HCC is the third most common cause of cancer-related death and the incidence of HCC is expected to increase worldwide (Forner et al., 2012). Although innate immunity, cytokine-mediated inflammation, and RNA editing play critical roles during the early stages of HCC pathogenesis in mice (Coulouarn et al., 2011; Hatziapostolou et al., 2011; Forner et al., 2012; Chen et al., 2013; He et al., 2013), multistage hepatocarcinogenesis is also influenced by microenvironmental factors and changes in both genetic and epigenetic landscape (Thorgeirsson and Grisham, 2002). For example, Myc activation coupled with prolonged stress in hepatocytes leads to liver hyperplasia that eventually degenerates into HCC (Chan et al., 2004; Shachaf et al., 2004). Myc upregulation and its transcription signature reprogramming are critical steps (Kaposi-Novak et al., 2009; Thorgeirsson, 2011), which are also required to maintain cancer stem cell properties in HCC progression in mice (Holczbauer et al., 2013; Akita et al., 2014). However, the precise mechanisms through which Myc operates to promote HCC are poorly understood.
miRNAs are small noncoding RNAs that induce mRNA degradation or translational inhibition and thus control many physiological and disease processes, including hepatocarcinogenesis (Bartel, 2004; Thorgeirsson, 2011; Mendell and Olson, 2012). miRNA deregulation is linked to cancer progression and clinical outcomes (Calin and Croce, 2006; Datta et al., 2008; Lujambio et al., 2008; Iorio and Croce, 2009). miRNAs and Myc can regulate each other (Bui and Mendell, 2010). How miRNAs and Myc regulate each other in hepatocarcinogenesis is still poorly understood.
The Warburg effect, whereby cellular energy production is driven by glycolysis even in the face of oxygen levels that are sufficient to support oxidative phosphorylation (Oxphos), is a core metabolic hallmark of cancer (Koppenol et al., 2011). In addition to generating ATP, aerobic glycolysis provides glycolytic intermediates needed to address the biosynthetic needs of fast-growing tumors. Oncogenes, tumor suppressor genes, miRNAs, and their reciprocal interplay are critical regulators of the Warburg effect. Precisely how the Warburg effect is regulated by these various factors and how it contributes to hepatocarcinogenesis are currently unclear (Dang et al., 2009; Vander Heiden et al., 2009; Koppenol et al., 2011; Masui et al., 2013).
We report here the results of a high-content screen to identify miRNAs that inhibit HCC cell growth. miR-129-5p was identified as one of the miRNAs that exerted a potent anti-proliferative effect on HCC growth. And we focused on the functional significance and regulatory mechanism of miR-129-5p in the regulation of the Warburg effect and tumor growth in HCC.
Results
Functional library screening identifies miRNAs inhibiting HCC cell growth
To identify miRNAs involved in HCC progression, we created a human miRNA expression library by cloning nearly all annotated human miRNAs [1254 miRNA loci, 1746 mature miRNAs, miRBase release 18.0 (2012); Supplementary Table S1] into the miRNA expression vector. We subsequently performed a high-content, function-based miRNA screen in HepG2 cells with ∼1254 individual miRNA vectors and used an MTT assay to identify miRNAs that inhibited cell growth. In this first screening, 32 miRNAs, including known tumor suppressor-miRNAs such as let-7 and miR-1 (Calin and Croce, 2006; Datta et al., 2008; Lujambio et al., 2008; Iorio and Croce, 2009), showed marked inhibitory effects on cell growth (Log2-relative growth ratio <−0.7, Figure 1A and Supplementary Table S2). Then for the 32 miRNAs that most potently inhibited HepG2 growth, we expanded the screen to include two additional HCC lines, BEL-7402 and FHCC98. miR-129-2 and miR-148a significantly inhibited all three HCC lines by >2-fold compared with control vector (Figure 1A). The miR-129-2 locus encodes two miRNAs, miR-129-5p/3p. miR-129-5p is one member of a family of evolutionarily conserved miRNAs and is expressed in most tissues (Supplementary Figure S1A and B). The miR-129-1 locus also encodes miR-129-5p. Therefore, Pri/Pre-miR-129-1 and Pri/Pre-miR-129-2 expression were detected in normal hepatocytes HL-7702 and mouse liver tissues, and Pri/Pre-miR-129-2 expression was significantly higher compared with Pri/Pre-miR-129-1 (Supplementary Figure S1C and D). Based on these observations, we selected miR-129-5p and miR-148a for further analyses in this and another study, respectively.
miR-129-5p inhibits growth of HCC cells and is downregulated in human HCC. (A) Left: high-content functional library screening results of miRNAs inhibiting HepG2 cell growth. miR-129-5p is indicated in white. Right: relative growth rate in three HCC lines treated with indicated miRNAs. (B) Expression of Pri/Pre-miR-129-2 and miR-129-5p in normal human hepatocytes HL-7702 and four human liver tumor-derived cell lines HepG2, BEL-7402, FHCC98, and Huh7. Expression levels of Pri/Pre-miR-129-2 and miR-129-5p were measured by stem-loop real-time RT-PCR in RNA purified from the above log-phase cells. Data present mean ± SD in A and B. (C) Expression of Pri/Pre-miR-129-2 and miR-129-5p normalized for U6 was measured with real-time RT-PCR in total RNA extracted from 88 HCC specimens and their adjacent noncancerous tissues. Significance was performed using Wilcoxon signed rank test. The horizontal lines in the box plots represent the median, the boxes represent the interquartile range, and the whiskers represent the minimal and maximal values.
miR-129-5p expression is downregulated in human HCC
Using qRT-PCR analysis, Pri/Pre-miR-129-2 and miR-129-5p were detected in normal hepatocytes HL-7702 and four HCC cells. As shown in Figure 1B, expression of Pri/Pre-miR-129-2 and miR-129-5p was much lower in BEL-7402, HepG2, and Huh7 compared with HL7702 and FHCC98. To test whether this is also clinically relevant, we quantified the RNA levels of Pri/Pre-miR-129-2 and miR-129-5p in primary human HCC specimens derived from 88 HCC patients (Supplementary Table S3). Pri/Pre-miR-129-2 and miR-129-5p were significantly downregulated in HCC samples compared with their adjacent noncancerous tissues (Figure 1C). These results indicate that the expression of miR-129-5p is downregulated in primary HCC.
PDK4 is a miR-129-5p direct target involved in the Warburg effect
Revealing miR-129-5p targets is essential for understanding its biological functions. Therefore we performed prediction analyses to identify targets of miR-129-5p. Across five database analyses, we found that the 3′UTR of PDK4 contains one miR-129-5p-binding site (Supplementary Figure S2A and B). PDK4 is a member of the pyruvate dehydrogenase kinase family comprised of four proteins (PDK1, PDK2, PDK3, and PDK4) that are critical components of the Warburg effect (Vander Heiden et al., 2009; Koppenol et al., 2011). Then we quantified the RNA levels of four PDK members in 32 primary HCC specimens (Supplementary Table S3). Among the four members, PDK4 was most significantly upregulated in HCC samples compared with their adjacent noncancerous tissues (Supplementary Figure S2C). Therefore, we focused on PDK4 for further studies. We first transfected Pri-miR-129-2 in HEK-293 and HepG2 cells and miR-129-5p sponge in FHCC98 cells. As shown in Supplementary Figure S2D and E, the expression of miR-129-5p sponge decreased miR-129-5p level, while the expression of Pri-miR-129-2 significantly increased miR-129-5p level. To explore whether PDK4 represents a direct target of miR-129-5p, we conducted luciferase reporter assays to determine whether the putative miR-129-5p-binding site in the 3′UTR of PDK4 is important for miR-129-5p-mediated suppression. Indeed, ectopic miR-129-2 expression repressed the activity of human PDK4 3′UTR reporter vector in dual luciferase reporter assays, while mutation in miR-129-5p-binding site abrogated this repression; inhibition of miR-129-5p using its sponge increased the activity of human PDK4 3′UTR reporter vector in the same assays, while mutation in miR-129-5p-binding site abrogated this upregulation (Figure 2A and Supplementary Figure S2F). Thus, the human PDK4 mRNA is directly regulated by miR-129-5p via seed-matching sequences. PDK negatively regulates the activity of PDH by catalyzing the phosphorylation of three serine residues on the E1α subunit of PDH (Dang et al., 2009; Vander Heiden et al., 2009; Koppenol et al., 2011). As expected, ectopic expression of miR-129-5p in HEK-293 and HepG2 cells decreased both mRNA and protein levels of endogenous PDK4, leading to a reduction in the phosphorylation of PDH-E1α. Opposing results were obtained with expression of miR-129-5p sponge in FHCC98 cells (Figure 2B and C).
miR-129-5p directly targets PDK4. (A) Luciferase activity of the reporter vector containing the wild-type or miR-129-5p binding mutant 3′UTR of PDK4 was determined after co-transfection with control or miR-129-2 and control miR sponge or miR-129-2 sponge expression vectors in HEK293 or HepG2 cells and HEK293 or FHCC98 cells, respectively. (B) Effects of overexpression or inhibition of miR-129-5p on endogenous mRNA expression of PDK4 in HCC cells. (C) Effects of overexpression or inhibition of miR-129-5p on endogenous protein level of PDK4. (D) Real-time RT-PCR analysis was performed to detect the miR-129-5p and PDK4 RNA levels incorporated into RISC derived from miR-129-2 overexpression or control vector in HepG2 cells. (E) Real-time RT-PCR and western blot assays show PDK4 mRNA and protein levels in normal hepatocytes HL-7702 and four HCC cells, respectively. (F) Left: expression of PDK4 normalized for GAPDH was measured with real-time RT-PCR in total RNA extracted from 88 HCC specimens and their adjacent noncancerous tissues. Significance was performed using Wilcoxon signed rank test. The horizontal lines in the box plots represent the median, the boxes represent the interquartile range and the whiskers represent the minimal and maximal values. Right: western blot assays show PDK4 protein levels in the indicated HCC samples. (G) Correlative analysis of PDK4 mRNA level and miR-129-5p expression from F and Figure 1C. Data present mean ± SD.
To further study whether PDK4 is a direct target of miR-129-5p, RNAs associated with RNA-induced silencing complex (RISC) using AGO-2 antibody were pulled down to quantify mRNAs enriched in RISC after miR-129-5p overexpression by real-time RT-PCR. As a positive control, miR-129-5p incorporation into RISC was significantly increased in miR-129-5p-overexpressing cells. Results showed that PDK4 was also significantly elevated enrichment in miR-129-5p-overexpressing cells compared with the control group (Figure 2D). Taken together, these data indicate that miR-129-5p can directly target PDK4.
PDK4 is upregulated in human HCC
Using qRT-PCR and western blot analyses, PDK4 expression was detected in normal hepatocytes HL-7702 and four HCC cells, which was upregulated in HCC cells compared with normal hepatocytes HL7702 (Figure 2E). To test the clinical relevance, we quantified PDK4 expression in primary HCC specimens derived from 88 HCC patients (Supplementary Table S3). PDK4 was upregulated in HCC samples compared with their adjacent noncancerous tissues at both mRNA and protein levels (Figure 2F). Furthermore, a negative correlation was detected between miR-129-5p and PDK4 mRNA levels (Figure 2G) in primary HCC specimens. These results suggest that the expression of PDK4 is upregulated in primary HCC.
miR-129-5p affects the Warburg effect, tumor growth, and lung colonization
The Warburg effect is a metabolic hallmark of cancer (Dang et al., 2009; Vander Heiden et al., 2009; Koppenol et al., 2011; Masui et al., 2013). As miR-129-5p targets a critical component of the Warburg effect, we speculated that miR-129-5p affects the Warburg effect and tumor growth. Among HCC cell lines used in this study, HepG2 and FHCC98 show low and high levels of miR-129-5p expression, respectively. To examine the consequences of miR-129-5p activation or inhibition on tumor growth and metabolism, HepG2 cells were transfected with Pri-miR-129-2 in the presence or absence of PDK4 without the complete 3′UTR (so that miR-129-5p cannot lower PDK4 expression from this vector). Stable FHCC98 cells expressing forced miR-129-5p sponge were established. HepG2 cells stably expressing Pri-miR-129-2 with mutated seed region were also established. As shown in Supplementary Figure S3A and B, Pri-miR-129-2 overexpression decreased PDK4 protein and PDH E1α phosphorylation levels, while ectopic expression of PDK4 without the complete 3′UTR blocked Pri-miR-129-2-mediated reduction in PDK4 protein and PDH E1α phosphorylation levels. We next measured lactate and ATP levels, as well as pH, in these cells. Forced Pri-miR-129-2 expression caused decreases in lactate and ATP and increased the pH in HepG2 cells whereas the opposite results were obtained with expression of miR-129-5p sponge in FHCC98 cells (Figure 3A and Supplementary Figure S3C). However, ectopic expression of PDK4 lacking the complete 3′UTR prevented Pri-miR-129-2 from inhibiting the Warburg effect. These results suggest that Pri-miR-129-2 modulates the Warburg effect by regulating the levels of PDK4 via an interaction with the 3′UTR of PDK4 mRNA.
miR-129-5p affects glycolysis, tumor growth, and lung colonization. (A) miR-129-5p inhibits glycolysis. Lactate and ATP levels from stable HepG2 and FHCC98 cell lines established as indicated. (B) Tumor growth in nude mice. Left: growth curves of tumors derived from the cells as indicated. Indicated cells were injected subcutaneously into the flanks of nude mice (n = 3 mice per group). Right: western blot analysis of representative excised tumor xenograft using indicated specific antibodies. (C) Tumor volumes in mouse livers after 4 weeks of intrahepatic inoculation with the indicated cells. (D) Representative images show colonization foci in lungs of mice. Indicated cells (1 × 105) were injected via tail vein into the mice. Colonization nodules in the lungs of mice were analyzed. (E) Quantitation of tumor nodules in lungs of mice during Days 30–45 after tail vein injection with 1 × 105 indicated cells (n = 3 mice per group). Data present mean ± SD.
We next asked whether miR-129-5p affected the biology of liver cells. As shown in Supplementary Figure S3D, obvious inhibition in soft agar formation was observed in HepG2 cells stably expressing Pri-miR-129-2. In contrast, enhanced colony formation was seen in FHCC98 cells expressing miR-129-5p sponge. Following subcutaneous inoculation into the flanks of nude mice, FHCC98 cells stably expressing miR-129-5p sponge reproducibly gave rise to more rapidly growing tumors, whereas HepG2 cells stably expressing Pri-miR-129-2 showed significantly retarded tumor growth (Figure 3B, Supplementary Figure S3E and F). Consistent with their different rates of growth, Pri-miR-129-2-expressing and miR-129-5p sponge-expressing tumors showed low and high levels of proliferating cell nuclear antigen (PCNA) staining, respectively (Supplementary Figure S3G). Moreover, Pri-miR-129-2-expressing tumors showed reduced PDK4 protein and PDH E1α phosphorylation levels whereas miR-129-5p sponge-expressing tumors showed an opposite pattern (Figure 3B). Ectopic expression of PDK4 lacking its complete 3′UTR, which is not suppressed by miR-129-5p, blocked Pri-miR-129-2-mediated tumor growth inhibition, as well as increased PCNA staining, PDK4 protein level, and PDH E1α phosphorylation in tumors (Figure 3B and Supplementary Figure S3E–G). However, ectopic expression of Pri-miR-129-2 with mutated seed region failed to show any effects on the Warburg effect and anchorage-independent growth of HCC cells (Supplementary Figure S3B–D).
To further explore the role of miR-129-5p on tumor growth, we used an orthotopic liver cancer model. As shown in Figure 3C and Supplementary Figure S3H–J, the tumors formed by miR-129-5p sponge-expressing FHCC98 cells were larger than those formed by control cells and expressed higher levels of PCNA. In contrast, Pri-miR-129-2-expressing HepG2 cells formed smaller tumors and demonstrated less intense PCNA staining. Ectopic expression of PDK4 lacking its complete 3′UTR blocked Pri-miR-129-2-mediated tumor growth retardation. These findings suggest that miR-129-5p retards tumor growth through targeting PDK4.
HepG2 and FHCC98 cells have high and low lung colonization potential, respectively. To test whether miR-129-5p levels could affect lung colonization, we used a mouse model whereby tumor cells are directly injected via tail vein into nude mice and quantified lung colonization 8 weeks later. We found that Pri-miR-129-2 overexpression significantly reduced the incidence of lung colonization arising from HepG2 cells whereas reducing miR-129-5p expression in FHCC98 cells increased lung colonization (Figure 3D and E). Histologic analysis and PCNA staining confirmed the presence of highly proliferative lung colonization foci (Supplementary Figure S3K and L). Moreover, ectopic expression of PDK4 blocked miR-129-5p-mediated lung colonization suppression. These findings strongly supported the role of miR-129-5p as a suppressor of lung colonization via inhibition of PDK4.
Targeting PDK4 affects the Warburg effect, tumor growth, and lung colonization
Given that PDK4 is an important mediator of miR-129-5p function, we hypothesized that PDK4 would be an important determinant of HCC behavior. To evaluate the role of PDK4 in the Warburg effect, tumor growth, and lung colonization of HCC cells, we targeted PDK4 in HepG2 cells using specific shRNAs, which significantly decreased PDK4 protein and PDH E1α phosphorylation levels (Supplementary Figure S4A and B). Metabolic assays performed on these cells showed that PDK4 inhibition significantly decreased ATP and lactate levels and increased the pH (Figure 4A and Supplementary Figure S4C). This result strongly supported the notion that inhibiting PDK4 decreased the Warburg effect.
Inhibition of PDK4 affects glycolysis, tumor growth, and lung colonization. (A) PDK4 inhibition decreases glycolysis. Lactate and ATP levels from stable HepG2 cells expressing sh-PDK4 or control sh-RNA expression vector. (B) Top: tumor growth in nude mice. The indicated cells were injected subcutaneously into the flanks of nude mice (n= 3 mice per group). Bottom: western blot analysis of representative excised tumor xenografts using indicated specific antibodies. (C) Tumor volumes in mouse livers after 4 weeks of intrahepatic inoculation with stable HepG2 cells expressing sh-PDK4 or control sh-RNA expression vector (n= 3 mice per group). (D) Stable HepG2 cells (1 × 105) expressing sh-PDK4 or control sh-RNA expression vector were injected via tail vein into nude mice. Colonization nodules in the lungs of mice were analyzed. Top: representative H&E image show colonization nodules in the lungs. Middle: PCNA expression was assessed by immunostaining. Bottom: apoptosis was assessed by caspase-3 immunostaining. Scale bar, 25 µm. (E) Quantitation of tumor nodules in lungs of mice (n= 3 mice per group). Data present mean ± SD.
Furthermore, injections of HepG2 cells expressing PDK4 shRNAs into the flanks, left hepatic lobe, or tail vein of nude mice all resulted in significantly smaller tumors or reduced lung colonization (Figure 4B–E and Supplementary Figure S4D–K). All of these tumors had reduced levels of PDK4 protein and PDH E1α phosphorylation (Figure 4B). Supporting experiments indicated that tumors arising from HepG2 cells expressing PDK4 shRNAs showed lower levels of PCNA and higher levels of caspase-3 staining, indicating the induction of apoptosis (Figure 4D and Supplementary Figure S4G, H, J, K). These results suggest that inhibition of PDK4 may affect cancer growth and lung colonization by inducing apoptosis.
To test whether miR-129-2 affects anchorage-independent growth, tumor growth, and lung colonization in the absence of PDK4, we conducted PDK4 inhibition by specific shRNAs in FHCC98 cells expressing miR-129-5p sponge. As shown in Supplementary Figure S5A–C, inhibition of PDK4 at least partially blocked cancer growth and lung colonization caused by reducing miR-129-5p in HCC cells.
miR-129-5p expression is directly regulated by Myc
To study why miR-129-5p expression is downregulated in human HCC, we examined the Pri-miR-129-2 promoter for transcription factor-binding sites and identified highly conserved Myc-binding E-boxes, located within CpG islands 5 kb upstream and 2 kb downstream of the transcription start site (TSS) (Figure 5A and Supplementary Figure S6A). Previously, we reported that Myc is activated in and essential for HCC progression (Han et al., 2013). Thus, we tested whether Myc regulates Pri-miR-129-2 and miR-129-5p expression. As shown in Supplementary Figure S6B and C, inhibition of Myc in HCC cells resulted in upregulation of Pri-miR-129-2 and miR-129-5p. To further address whether Myc directly binds to the promoter of Pri-miR-129-2, we conducted a chromatin IP (ChIP) assay in HCC cells. This revealed that Myc binds to the Pri-miR-129-2 promoter containing the highly conserved E-boxes in the CpG island (Figure 5A). These results provide evidence that Pri-miR-129-2 is directly regulated by Myc.
miR-129-5p expression is regulated by Myc, HDAC3, and PRC2. (A) qPCR and chromatin IP analyses of Myc, HDAC3, and PRC2 binding to the differential promoter regions of Pri-miR-129-2 transfected with si-GFP or si-Myc. (B) Effects of inhibition of Myc, HDAC3, and PRC2 on endogenous expression of Pri-miR-129-2 and miR-129-5p. (C) Co-IP of Myc with HDAC3, EZH2, and SUZ12 in HepG2 cells. (D) Expression of Myc, HDAC3, EZH2, and SUZ12 normalized to GAPDH was measured with real-time RT-PCR in total RNA extracted from 88 HCC specimens and their adjacent noncancerous tissues. Significance was performed using Wilcoxon signed rank test. The horizontal lines in the box plots represent the median, the boxes represent the interquartile range, and the whiskers represent the minimal and maximal values. (E) Correlative analysis of the Pri/Pre-miR-129-2 or miR-129-5p RNA level and Myc, HDAC3, EZH2, or SUZ12 mRNA level from D and Figure 1C. Data present mean ± SD.
HDAC3 and PRC2 affect miR-129-5p expression at the transcription level
Other researchers have reported that Myc represses the transcription of its targets by recruiting the histone deacetylase HDAC3 or/and polycomb repressive complex 2 (PRC2) to their promoters (Kurland and Tansey, 2008; Au et al., 2012; Zhang et al., 2012; Cole, 2014; Wang et al., 2014). Therefore, we studied whether Myc collaborates with HDAC3 and PRC2 in silencing miR-129-5p in HCC cells. First we tested the potential roles of Myc, HDAC3, and PRC2 in regulation of Pri-miR-129-2 and miR-129-5p expression by using siRNAs to inhibit Myc, HDAC3, and two of the main subunits of PRC2, EZH2 and suppressor of zeste 12 homolog (SUZ12). Inhibition of any of these molecules significantly decreased Myc expression and enhanced Pri-miR-129-2 and miR-129-5p expression (Figure 5B and Supplementary Figure S6C). Inhibition of HDAC3, EZH2, or SUZ12 also significantly decreased growth and anchorage-independent soft agar growth of HCC cells, as indicated by cell cycle and soft agar experiments (Supplementary Figure S6D–F).
We next performed Co-IP experiments to determine whether HDAC3 and PRC2 form a co-repressor complex with Myc to inhibit miR-129-5p expression. We tested the endogenous interaction among Myc, HDAC3, and PRC2 in HepG2 cells. As shown in Figure 5C and Supplementary Figure S6G, cell lysates immunoprecipitated with a Myc-specific antibody contained HDAC3 and EZH2 but no SUZ12. The reverse endogenous co-immunoprecipitates of Myc, HDAC3, EZH2, and SUZ12 by HDAC3 or EZH2 antibody were also demonstrated in HepG2 cells. However, cell lysates immunoprecipitated with a SUZ12-specific antibody contained HDAC3 and EZH2 but no endogenous Myc. Overall, these results suggest that Myc interacts with HDAC3 and EZH2 to form a co-repressor complex.
EZH2, the core subunit of PRC2, is the sole histone methyltransferase that catalyzes trimethylation of histone H3 at lysine 27 (H3K27me3), thereby mediating epigenetic silencing (Zhang et al., 2012; Cole, 2014). We therefore again performed ChIP experiments to determine whether HDAC3, EZH2, and SUZ12 could be recruited to the Pri-miR-129-2 promoter by Myc and whether HDAC3, EZH2, SUZ12, and EZH2-catalyzing H3K27me3 mediated Myc-induced Pri-miR-129-2 repression. The results showed that antibodies against HDAC3, EZH2, SUZ12, and H3K27me3 efficiently immunoprecipitated the Pri-miR-129-2 promoter in HepG2 cells treated with si-GFP but not si-Myc (Figure 5A). Of note, the binding sites for HDAC3, EZH2, SUZ12, and H3K27me3 correspond to Myc-binding sites. These results indicate that Myc plays a central role in recruiting HDAC3 and PRC2 to the Pri-miR-129-2 promoter and that EZH2-mediated H3K27me3 and HDAC3-catalyzing histone deacetylation might contribute to Myc-induced Pri-miR-129-2 repression.
Next, we examined RNA levels of Myc, HDAC3, and PRC2 in primary HCC specimens derived from 88 HCC patients. As shown in Figure 5D, they were significantly overexpressed in HCC samples compared with adjacent normal liver tissues. Furthermore, we noted a negative correlation between the mRNA level of Myc, HDAC3, EZH2, or SUZ12 and the expression of Pri-miR-129-2 or miR-129-5p (Figure 5E).
In other experiments, we showed that siRNA-mediated Myc, HDAC3, and PRC2 inhibition decreased PDK4 mRNA level (Supplementary Figure S6H). In contrast, Myc overexpression increased PDK4 protein level, which was blocked by enforced miR-129-5p expression (Supplementary Figure S6I). A positive correlation between Myc, HDAC3, or PRC2 level and the expression of PDK4 was also determined (Supplementary Figure S6J). Taken together, these data indicate that Myc, HDAC3, and PRC2 complex can affect the expression of miR-129-5p and PDK4.
The expression of miR-129-5p, Myc, or PDK4 correlates with tumor stage and differentiation in human HCC
Interestingly, we found that HCC patients with Stage III/IV tumors displayed significantly decreased levels of miR-129-5p and elevated expression of Myc and PDK4 compared with patients with Stage I/II disease (Figure 6A–C). Moreover, histologically well-differentiated HCC showed a significant association with increased expression of miR-129-5p and decreased expression of Myc and PDK4 relative to poorly differentiated tumors (Figure 6A–C). However, there is no any correlation between the expression of miR-129-5p, Myc, or PDK4 with tumor size, encapsulation, and liver cirrhosis (Supplementary Figure S7). These results show that the expression of miR-129-5p, Myc, or PDK4 correlates with HCC stage and tumor differentiation.
Myc, miR-129-5p, and PDK4 are correlated with clinical stage and differentiation in human HCC, and miR-129-5p network molecules are dysregulated in DEN-induced HCC in mice. (A–C) Association of miR-129-5p, Myc, and PDK4 expressions with clinical stages and tumor differentiation in sporadic HCC. The expression of miR-129-5p negatively and positively correlates with disease stage and differentiation, respectively. In contrast, the expression of Myc and PDK4 positively and negatively correlates with disease stage and differentiation, respectively. Significance was performed using Mann–Whitney U-test. The horizontal lines in the box plots represent the median, the boxes represent the interquartile range, and the whiskers represent the minimal and maximal values. (D) H&E staining of liver tissues from differential phases during DEN-induced hepatocarcinogenesis (n= 5 mice per group). Scale bar, 25 µm. (E) qRT-PCR detection for RNA levels of miR-129-5p network molecules in liver tissues during DEN-induced hepatocarcinogenesis. Data present mean ± SD.
miR-129-5p network molecules are dysregulated in DEN-induced HCC in mice
Based on the above findings, we speculated that miR-129-5p network molecules including miR-129-5p, PDK4, and Myc may be dysregulated during the development of DEN-induced mouse hepatocarcinoma in vivo. We determined the expression levels of miR-129-5p network molecules in liver tissues derived from DEN-treated mice and found that they were indeed dysregulated in liver tissues during HCC development in mice. PDK4 and Myc were not changed between Weeks 1 and 8 and both upregulated in Week 44, whereas miR-129-5p was downregulated although this was not seen until Week 44 (Figure 6D and E). These data are consistent with those derived from human HCC clinical samples.
Restoring miR-129-5p expression shows therapeutic effects on DEN-induced HCC and xenograft mouse models
To further test in vivo therapeutic potential of manipulating miR-129-5p, we asked whether restoring its expression would inhibit tumor growth in HCC mouse models. HepG2 cells were injected subcutaneously in the flanks of nude mice. One week post-injection, tumor xenografts in the control group were locally treated with concentrated lentivirus containing a pHAGE-GFP expression vector. Tumor xenografts from the other two groups were treated with concentrated lentivirus containing a pHAGE-Pri-miR-129-2 expression vector, twice per week for 4 weeks. Tumor xenografts in Pri-miR-129-2-E group were treated from Day 10, while tumors in Pri-miR-129-2-L group were treated from Day 20. The results showed that lentivirus expressing pHAGE-miR-129-2 retarded HepG2 xenograft tumor growth and these tumors expressed significantly less PDK4 (Figure 7A).
Restoring miR-129-5p expression suppresses xenograft tumors and DEN-induced hepatocarcinogenesis in mice. (A) Left: restoring miR-129-5p expression retards growth of xenograft tumors in nude mice. HepG2 cells (1 × 107) were injected into the flanks of nude mice (n= 3 mice per group). One week post-injection, tumor xenografts from the control group were locally treated with lentivirus containing a pHAGE-GFP vector. Tumor xenografts from the other two groups were treated with lentivirus containing a pHAGE-Pri-miR-129-2 expression vector from Day 10 (Pri-miR-129-2-E group) or Day 20 (Pri-miR-129-2-L group) after injection of HepG2 cells. Right: endogenous RNA expression of miR-129-5p and PDK4 in tumor xenografts. (B) Left: treatment schedule with lentivirus containing a pHAGE-Pri-miR-129-2 expression vector during DEN-induced hepatocarcinogenesis in mice. Right: tumor size and quantitation of liver nodules in tumors (n= 5 or 6 mice for control or Pri-miR-129-2 group, respectively). (C) Tumor nodules in the livers of mice were analyzed for H&E, PCNA staining, and apoptosis. Scale bar, 25 µm. (D) qRT-PCR detection of miR-129-5p and PDK4 RNA levels in tumors derived from B. (E) Lactate and ATP levels in tumors derived from B. (F) Schematic representation of the miR-129-5p network in hepatocarcinogenesis. Data present mean ± SD.
To determine whether restoring miR-129-5p expression could inhibit autologous HCC tumor growth in DEN-treated mice, concentrated lentivirus containing a pHAGE-Pri-miR-129-2 expression vector was administered to DEN-treated mice for several weeks (Weeks 38–41). The tumor burden was assessed after the mice were sacrificed in Week 44. We found that the lentiviral vector encoding miR-129-2 but not the control vector inhibited HCC growth and induced robust apoptosis (Figure 7B and C). Tumors treated with lentivirus expressing pHAGE-miR-129-2 also showed increased miR-129-5p expression and reduced PKD4 expression (Figure 7D). Such treatment inhibited PDK activity, as evidenced by decreased lactate and ATP levels (Figure 7E) as well as serum AFP level (Supplementary Figure S8A and B).
Discussion
Our functional miRNA library screening revealed that miR-129-2 inhibited HCC cell proliferation. miR-129-5p was first reported to exert significant growth inhibition and to promote cell death in bladder carcinoma and liver cancer (Dyrskjøt et al., 2009; Liu et al., 2012). Most notably, miR-129-5p was drastically suppressed in many primary HCC cell lines (Figure 1C), suggesting that loss of miR-129-5p is a prerequisite for HCC development. Restoration of miR-129-5p repressed the Warburg effect in HCC cells, resulting in impairment of tumor growth and lung colonization (Figure 3).
miRNA expression is tightly regulated at the transcriptional level (Kurland and Tansey, 2008; Au et al., 2012; Zhang et al., 2012; Cole, 2014; Wang et al., 2014). Our results indicate that miR-129-5p was suppressed by a complex comprised of Myc, EZH2, and HDAC3 through binding to the Pri-miR-129-2 promoter (Figure 7F). Others have also reported that Myc and EZH2 inhibit miRNA transcription (Kurland and Tansey, 2008; Au et al., 2012; Zhang et al., 2012; Cole, 2014; Wang et al., 2014). The resultant downregulation of miR-129-5p was notably correlated with high clinical stages and a poorly differentiated phenotype.
It is generally believed that the primary purpose of the Warburg effect in tumor cells is to ensure the supply of macromolecular precursors and energy necessary to sustain rapid tumor growth (Dang et al., 2009; Vander Heiden et al., 2009; Koppenol et al., 2011; Masui et al., 2013). Previous evidence indicates that c-Myc can regulate the Warburg effect by directly upregulating the majority of glycolytic genes (Dang, 2012). In addition, c-Myc represses many miRNAs that function to suppress glycolytic genes (Dang, 2012). One such example is seen in the case of miR-34A, whose suppression by c-Myc increases LDHA expression (Dang, 2012). We have shown here that the suppression of miR-129-5p by the c-Myc–HDAC3–EZH2 repressor complex markedly increases PDK4 expression. Consistent with this, inhibition of PDK4 blocked glycolysis, tumor growth, and metastasis of HCC cells. PDK4 plays distinct and critical roles in promoting the Warburg effect. PDK phosphorylates PDH and blocks the conversion of pyruvate to acetyl CoA, thus depriving the mitochondria of its most critical entry-level substrate (Bonnet et al., 2007; Vander Heiden et al., 2009; Koppenol et al., 2011). Recently, studies suggest that dysregulation of miRNAs facilitates the Warburg effect to contribute to the development of HCC. It has been reported that Lin28/let-7 axis facilitates the Warburg effect to promote HCC progression (Ma et al., 2014). Another report revealed a novel role for miR-199a-5p/HK2 in reprogramming the metabolic process in liver cancer cells (Guo et al., 2015). Mineralocorticoid receptor has been reported to modulate miR-338-3p–PKLR axis and suppress HCC progression by its ability to suppress the Warburg effect (Nie et al., 2015).
The restoration of miR-129-5p expression affected DEN-induced hepatocarcinogenesis in mice. A few studies have studied the therapeutic delivery of miRNAs in vivo. Restoration of miR-26a expression has been reported to suppress liver tumorigenesis in liver-specific Myc transgenic mice without any cytotoxic effects (Kota et al., 2009). miR-124 administration prevented DEN-induced hepatocellular carcinogenesis in mice (Hatziapostolou et al., 2011). Our findings show that PDK4 inhibition decreased PDH E1α phosphorylation level and inhibited DEN-induced hepatocarcinogenesis. These findings suggest that restoring miR-129-5p expression might be viable and perhaps synergistic clinical anticancer therapeutic approaches for HCC (Figure 7F). The dysregulation of the miR-129-5p network molecules was strongly correlated with HCC stage and differentiation. Normalizing this pathway at any one of several points could potentially provide novel therapeutic points of therapeutic intervention for HCC.
Materials and methods
Human liver tissues
Primary human HCC samples were obtained from patients undergoing tumor resection. Informed consent was obtained at the Union Hospital in Wuhan, China. The diagnosis of HCC was confirmed in each case by histological review and none of them received chemotherapy prior to hepatectomy (Han et al., 2013).
Mouse experiments
All animal studies were approved by the Animal Care Committee of Wuhan University. Mouse experiments including DEN-induced mouse HCC model and therapeutic protocols were performed according to the principles previously described (Hatziapostolou et al., 2011). Male BALB/c nude mice were purchased from Shanghai and Changsha (China) SLAC Laboratory Animal Corporation and maintained in microisolator cages (Han et al., 2013). For DEN-induced mouse HCC model, mice were intraperitoneally administrated once with 25 mg/kg of diethylnitrosamine (DEN) (Sigma) on Day 15 and then monitored for development of HCC (Hatziapostolou et al., 2011).
Statistical analyses
The SPSS statistical package for Windows (SPSS 16), the GraphPad Prism Software (version 5.01), and Microsoft Excel (Microsoft Office 2007 for Windows) were used for data analysis. Data are expressed as mean ± SD or SEM. The correlation was analyzed using a Spearman rank correlation test. Other statistical analyses were performed using the unpaired, two-tailed Student's t-test, Wilcoxon signed rank test, or Mann–Whitney U-test, and P< 0.05 was considered statistically significant (Han et al., 2013).
Extended materials and methods are described in the Supplementary Materials and methods.
Supplementary material
Supplementary material is available at Journal of Molecular Cell Biology online.
Funding
This work was supported by grants from the 973 Program of China (2014CB910600 and 2012CB967000), the National Natural Science Foundation of China (81472549), and the China 111 Project (B06018).
Conflict of interest: none declared.
Acknowledgements
The authors thank Drs Hongbing Shu and Baoliang Song (Wuhan University, Wuhan, China) for encouraging suggestions, Edward Prochownik for revising the manuscript, Ying Liu for cell sorting, and all members in our lab for technical help.
