Increased cholesterol synthesis in tumors of HCC patients
Both cholesterol synthesis and excretion and absorption of cholesterol contribute to change in hepatic and blood cholesterol [14]. Hepatic cholesterol accumulation is implicated in HCC patients. However, the effect of cholesterol lowering drugs on HCC is controversial. In HCC patients, compared to adjacent normal livers, cholesterol levels were significantly elevated in tumors (Fig. 1A). Hepatocytes are the major site of cholesterol synthesis. We next analyzed mRNA levels of HMGCR, the rate-limiting enzyme of cholesterol synthesis, in hepatocytes isolated from adjacent normal livers and HCC tumors. As expected, increased HMGCR mRNA was observed in malignant hepatocytes compared to normal hepatocytes (Fig. 1B). In TCGA database, compared to normal individuals (n=50), HCC patients (n=369) exhibited high expression of HMGCR (Fig. 1C), which was associated with poor survival (Fig. 1D). Increased hepatic cholesterol can be caused by activation of cholesterol synthesis, impaired cholesterol excretion and increased cholesterol absorption from the food [14]. In addition, cholesterol-lowering drugs such as statins exhibit no effect on HCC, leading us to speculate cholesterol synthesis as a process rather than cholesterol itself is the major contributor of HCC development. 14C-acetate incorporation into cholesterol was much greater in HCC tumors than in normal livers (Fig. 1E), suggesting activation of de novo cholesterol synthesis in HCC tumors. In sum, cholesterol synthesis was activated in tumors of HCC patients, and high levels of HMGCR in malignant hepatocytes correlated with poor survival of HCC patients.
c-Myc activated hepatic cholesterol synthesis, the pentose phosphate pathway and glycolysis in mice.
Almost 30% of HCC patients show c-MYC gene amplification or overexpression [28]. A positive correlation between HMGCR and c-MYC was observed in tumors of HCC patients from TCGA database (Supplementary Fig. 1A), indicating that c-MYC is a potential driver of cholesterol synthesis. HDI of c-Myc led to c-Myc accumulation and increased expression of Hmgcr in hepatocytes of mice (Supplementary Figure 1B-C) and triggered development of HCC (Fig. 2A). All c-Myc mice died of HCC within eight weeks post injection of c-Myc, while 100% control mice were healthy at that time point (Fig. 2B). As we observed in HCC patients, mRNA levels of Hmgcr, enzyme activity of HMGCR, and hepatic cholesterol were significantly increased in tumors of c-Myc mice (Fig. 2C-E). Cholesterol synthesis, via acetyl-CoA, interfaces with de novo lipogenesis (DNL), glycolysis and the PPP [24], suggesting a possible mechanism of action. In fact, metabolites of glycolysis and the PPP were significantly increased in tumors of c-Myc mice (Fig. 2F); and acetyl-CoA, the major precursor of cholesterol synthesis, was increased in c-Myc mice (Fig. 2F). Consistent with an increase in the glycolytic and the PPP metabolites, expression of the genes controlling glycolysis and the PPP was significantly increased in malignant hepatocytes of c-Myc mice (Fig. 2G). An increase in the glycolytic rate was observed in malignant hepatocytes of c-Myc mice (Fig. 2H). In sum, c-Myc signaling promoted cholesterol synthesis, glycolysis and the PPP in hepatocytes.
Ablation of the PPP reduced cholesterol synthesis and delayed growth of HCC, while ablation of glycolysis did not affect these processes in c-Myc mice.
c-Myc activated the PPP and glycolysis (Fig. 2F-G). Glycolysis produces pyruvate that can be converted to acetyl-CoA, a precursor of cholesterol synthesis. We next determined the effect of glycolysis on cholesterol synthesis. NAD is a driver of glycolysis [24, 29]. We, therefore, treated liver homogenates of pT3 and c-Myc mice with NAD. Although NAD enhanced glycolysis in both pT3 and c-Myc mice (Supplementary Fig. 2A), it did not affect cholesterol synthesis in both pT3 and c-Myc mice (Fig. 3A). We next deleted pyruvate kinase (PKM), the rate-limiting enzyme of glycolysis in c-Myc mice, which ablated glycolysis (Supplementary Fig. 3A-B). Unexpectedly, ablation of Pkm at the time of c-Myc overexpression in murine livers did not affect cholesterol synthesis in c-Myc mice (Fig. 3B). However, cholesterol synthesis is still much higher in c-Myc mice compared to pT3 mice (Fig. 3A), indicating that other pathways activated by c-Myc such as the PPP might be able to drive cholesterol synthesis in mice.
The PPP is a metabolic pathway parallel to glycolysis, which shares a common starting molecule with glycolysis, glucose-6-phosphate (G6P). Two major products of the PPP are R6P and NADPH. As expected, NADPH and R5P as well as expression of G6pd was significantly increased in malignant hepatocytes from c-Myc mice (Fig. 3C-D, Supplementary Fig. 4A). NADPH, as a cofactor of HMGCR, is required for cholesterol synthesis. These established findings led us to speculate that the PPP is potentially involved in enhanced cholesterol synthesis in HCC. To test this speculation, we treated liver homogenates of pT3 and c-Myc mice with NADPH. As expected, NADPH significantly increased enzyme activity of G6PD (Supplementary Fig. 4B), which in turn increased incorporation of 14C-acetate into cholesterol in liver homogenates from pT3 and c-Myc mice (Fig. 3E). Since the PPP is activated in c-Myc mice, cholesterol synthesis, as revealed by 14C-acetate labeling experiment, was much higher in c-Myc mice compared to pT3 mice (Fig. 3E). To confirm this speculation, we ablated the PPP via knocking down G6pd in hepatocytes of c-Myc mice (Supplementary Fig. 5). Knocking down G6pd significantly inhibited the PPP, which was reflected by a decrease in NADPH and R5P (Fig. 3F-G). Consistent with reduced NADPH that is required for cholesterol synthesis, incorporation of 14C-acetate into cholesterol was also significantly reduced in liver homogenates of c-Myc/shG6pd mice (Fig. 4H). Phenotypically, ablation of glycolysis did not affect growth of HCC, while knockdown of G6pd significantly delayed growth of HCC in c-Myc mice (Fig. 3I-J). In sum, the PPP at least in part is required for cholesterol synthesis and hepatocarcinogenesis in c-Myc mice.
A positive feedback between the PPP and cholesterol synthesis drove DNA synthesis and cell proliferation. Activation of the PPP produces more NADPH, which provides a cofactor for cholesterol synthesis. Enhancement of cholesterol synthesis should rapidly deplete NADPH, a major production of PPP. Therefore, we hypothesized that cholesterol synthesis and the PPP formed a positive feedback loop, which amplifies production of R5P, the substrate of DNA synthesis, and NADPH, co-factor for HMGCR. To determine if activation of the PPP drives cholesterol synthesis, DNA synthesis, and proliferation, three groups of hepatocytes were treated empty vector (pT3), c-Myc, or a combination of c-Myc and G6pd shRNA to knock down G6pd (Supplementary Fig. 6A). c-Myc overexpression enhanced the PPP, cholesterol synthesis, DNA synthesis and proliferation of hepatocytes (Fig. 4A-E), while knockdown of G6pd offset the effects of c-Myc overexpression (Fig. 4A-E). These findings indicated that activation of the PPP is required for c-Myc to drive cholesterol synthesis, DNA synthesis and hepatocyte proliferation. To test if enhancement of cholesterol synthesis promotes the PPP, DNA synthesis and proliferation, three groups of hepatocytes were treated with pT3 (control), c-Myc or a combination of c-Myc and Hmgcr shRNA (Supplementary Fig. 6B). c-Myc activated the PPP, cholesterol synthesis, DNA synthesis and hepatocyte proliferation; and knockdown of Hmgcr counteracted the effects of c-Myc (Fig. 4F-J). The significant increase in levels of 14C-acetate-labeled cholesterol and 3H-thymidine incorporation into DNA was observed in c-Myc mice (Fig. 4K-L). In sum, a positive feedback loop between cholesterol synthesis and the PPP enhanced production of cholesterol and R5P, which meets the needs for rapid growth and proliferation of malignant hepatocytes both in vitro and in vivo.
miR-206 repressed expression of HMGCR and G6PD in hepatocytes by binding to their 3’UTRs. HMGCR and G6PD are the rate-limiting enzymes of cholesterol synthesis and the PPP. MicroRNAs (miRNAs) can simultaneously fine tune multiple pathways and exhibit the strong therapeutic potential for cancers and other diseases [30]. We next attempted to identify those miRNAs that can simultaneously target both HMGCR and G6PD. For this purpose, we analyzed murine and human Ago HITS-CLIP databases (high-throughput sequencing of RNAs isolated by crosslinking immunoprecipitation from Argonaute protein complex) of HMGCR and G6PD. Unexpectedly, miR-206 was identified as the only miRNA that can target human and mouse HMGCR and G6PD and 3’UTRs of mouse and human HMGCR and G6PD contains two miR-206 binding site [31] [32] (Supplementary Table 1). To exclude the false positive peaks of Ago-HITS-CLIP, we further used DIANA-microT-CDS to scan the 3’UTRs of murine and human HMGCR and G6PD, confirming the binding sites of miR-206 within the 3’UTRs of murine and human HMGCR and G6PD. 3' UTRs of both human and mouse HMGCR and G6PD mRNAs are 100% complementary to the miR-206 5' seed region exhibiting the highest prediction scores and binding energy (Fig. 5A-B). In addition, levels of miR-206 were significantly reduced in malignant hepatocytes isolate from c-Myc mice (Supplementary Fig. 7). All these findings led us to focus on miR-206. In Fig. 1C-D, high levels of HMGCR predicted poor survival of HCC patients. Similarly, elevated levels of G6PD predicted poor survival of HCC patients TCGA database (Fig. 5C-D). Inclusion of the 3’UTRs of Hmgcr or G6pd into the luciferase reporter constructs reduced luciferase activities upon co-transfection with miR-206 into Hepa1-6 cells (Fig. 5E-F). Mutation of the miR-206 binding sites within the 3’UTRs of Hmgcr and G6pd was necessary to completely offset the inhibitory effects of miR-206 on luciferase activities (Fig. 5E-F), suggesting that miR-206 directly recognized the predicted binding site within the 3' UTR of Hmgcr and G6pd. miR-206 also reduced protein and mRNA levels of Hmgcr and G6pd in Hepa1-6 cells (Fig. 5G-H). MiR-206 was able to inhibit expression of HMGCR and G6PD in human hepatocytes by binding to their 3’UTRs (Supplementary Fig. 8A-C). In sum, HMGCR and G6PD are direct targets of miR-206 in both human and moues hepatocytes.
miR-206 disrupted the positive feedback loop between cholesterol synthesis and the PPP by targeting Hmgcr and G6pd, which impaired DNA synthesis and proliferation of hepatocytes. Since Hmgcr and G6pd are direct targets of miR-206, we hypothesized that miR-206 is able to disrupt the positive feedback loop between cholesterol synthesis and the PPP, thereby inhibiting DNA synthesis, cholesterol synthesis and proliferation of malignant hepatocytes. To test this, CRISPR/Cas9 technique was used to delete the binding sites of miR-206 within the 3’UTRs of both G6pd and Hmgcr in malignant hepatocytes isolated from c-Myc mice [33, 34]. Such a design disrupted the interaction of miR-206 with G6pd and Hmgcr (Supplementary Fig. 9A), allowing us to determine if G6pd and Hmgcr are required for miR-206 to inhibit cholesterol synthesis and the PPP. Overexpression of miR-206 inhibited cholesterol synthesis, the PPP and DNA synthesis, which was reflected by a significant reduction in incorporation of 14C-acetate into cholesterol, the PPP metabolites, 3H-thymidine incorporation into DNA and proliferation of hepatocytes (Fig. 6A-E). In contrast, ablating the miR-206 binding sites within the 3’UTRs of both Hmgcr and G6pd offset the inhibitory effects of miR-206 (Fig. 6A-E).
We assumed that a positive feedback loop between cholesterol synthesis and the PPP amplifies DNA synthesis and proliferation of malignant hepatocytes. Inhibition of either cholesterol synthesis or the PPP should be able to at least in part disrupt this positive feedback loop and thereby prevent DNA synthesis and proliferation. To test this, we ablated the binding sites of miR-206 within the 3’UTR of Hmgcr or G6pd. Ablation of the miR-206 binding sites within the 3’UTR of Hmgcr was able to recover cholesterol synthesis, the PPP, DNA synthesis and cell proliferation that were inhibited by miR-206 (Supplementary Fig. 9B, Fig. 6F-J). Similarly, ablation of the miR-206 binding sites within the 3’UTRs of G6pd also recovered cholesterol synthesis, the PPP, DNA synthesis and proliferation (Supplementary Fig. 9C, Fig. 6K-O). These results indicate that miR-206 is able to disrupt the positive feedback loop between cholesterol synthesis and the PPP by targeting either Hmgcr or G6pd, which subsequently inhibits DNA synthesis and cell proliferation.
miR-206 inhibited cholesterol synthesis and the PPP in c-Myc mice. We next determined if miR-206 was able to simultaneously inhibit cholesterol synthesis and the PPP in vivo. c-Myc mice were injected with pT3-EF1α-miR-206-MM (control) or pT3-EF1α-miR-206. Eight weeks post injection, all miR-206-treated c-Myc mice were healthy, while 100% of c-Myc mice died of HCC (Fig. 7A). Upon dissection, no tumors were observed in c-Myc/miR-206 mice (Fig. 7C). The long-term survival experiment revealed that all c-Myc/miR-206 mice were healthy 24 weeks of post injection of miR-206 (Fig. 7B). Upon dissection, no tumor nodules were observed in livers of this group of c-Myc/miR-206 mice (Fig. 7C). All these findings indicated the long-term effect of miR-206 on preventing HCC. Our hypothesis is that miR-206, by disrupting the positive feedback loop between cholesterol synthesis and the PPP, inhibits HCC. As expected, miR-206 significantly reduced expression of Hmgcr and G6pd in hepatocytes of c-Myc mice (Fig. 7D). Consistent with reduced Hmgcr and G6pd, both cholesterol and the metabolites of the PPP were significantly reduced in miR-206-treated c-Myc mice (Fig. 7E-G). Incorporation of 14C-acetate into cholesterol and 3H-thymidine incorporation into DNA were reduced in miR-206-treated c-Myc mice (Fig. 7H-I). These results established that miR-206 disrupted the positive feedback loop between cholesterol synthesis and the PPP, which subsequently inhibited DNA synthesis and growth of malignant hepatocytes. Unexpectedly, miR-206 treatment significantly induced expression of genes encoding PFK (phosphofructokinase) and PKM (pyruvate kinase), two rate-limiting enzymes of glycolysis, in c-Myc mice (Fig. 7J). The glycolytic rate was also significantly increased in livers of c-Myc/miR-206 mice (Fig. 7K). Although miR-206 induced glycolysis, it still fully prevented c-Myc-induced HCC, further indicating that glycolysis is not required for miR-206 to inhibit HCC. This finding is consistent with our observation that ablation of glycolysis failed to prevent c-Myc-induced HCC. In sum, miR-206 disrupted the positive feedback between cholesterol synthesis and the PPP, which prevented c-Myc-induced HCC.
HMGCR and G6PD are required for miR-206 to prevent c-Myc-induced HCC. We next employed an AAV8-based CRISPR/Cas9 technique to ablate the binding sites of miR-206 within the 3’UTRs of Hmgcr and G6pd in the genome of hepatocytes in c-Myc mice, which disrupted the interaction of miR-206 with Hmgcr and G6pd. Ablation of the miR-206 binding sites impaired the ability of miR-206 to inhibit expression of Hmgcr and G6pd in hepatocytes (Fig. 8A). Phenotypically, 100% c-Myc mice died of HCC within 8 weeks post injection of c-Myc (Fig. 8B). MiR-206 fully prevented c-Myc-induced HCC, while disrupting its interaction with Hmgcr and G6pd resulted in renewed growth of HCC that was fully prevented by miR-206 (Fig. 8B-C). Mechanistically, miR-206 markedly reduced hepatic cholesterol and metabolites of the PPP, while disrupting the interaction of miR-206 with Hmgcr and G6pd recovered levels of hepatic cholesterol and the metabolites of the PPP (Fig. 8D-F). As revealed by 14C-acetate- and 3H-thymidine-labeling experiments, ablating the miR-206 binding sites recovered cholesterol synthesis and DNA synthesis in miR-206-treated c-Myc mice (Fig. 8G-H). In sum, by disrupting the positive feedback loop between the cholesterol synthesis and the PPP, miR-206 inhibited growth of HCC.