Although circular RNAs (circRNA) are known to modulate tumor initiation and progression, their role in hepatocellular carcinoma (HCC) metastasis remains poorly understood. Here, three metastasis-associated circRNAs identified in a previous circRNA-sequencing study were screened and validated in two HCC cohorts. CircRPN2 was downregulated in highly metastatic HCC cell lines and HCC tissues with metastasis. Patients with HCC with lower circRPN2 levels displayed shorter overall survival and higher rates of cumulative recurrence. Mechanistic studies in vitro and in vivo revealed that circRPN2 binds to enolase 1 (ENO1) and accelerates its degradation to promote glycolytic reprogramming through the AKT/mTOR pathway, thereby inhibiting HCC metastasis. CircRPN2 also acted as a competing endogenous RNA for miR-183–5p, which increases forkhead box protein O1 (FOXO1) expression to suppress glucose metabolism and tumor progression. In clinical samples, circRPN2 expression negatively correlated with ENO1 and positively correlated with FOXO1, and expression of circRPN2, either alone or in combination with ENO1 and FOXO1, was a novel indicator of HCC prognosis. These data support a model wherein circRPN2 inhibits HCC aerobic glycolysis and metastasis via acceleration of ENO1 degradation and regulation of the miR-183–5p/FOXO1 axis, suggesting that circRPN2 represents a possible therapeutic target in HCC.

Significance:

The circRNA circRPN2 is a potential prognostic biomarker and therapeutic target in hepatocellular carcinoma that suppresses aerobic glycolysis and metastasis.

The most frequent primary liver cancer, hepatocellular carcinoma (HCC), is the sixth most common neoplasm and the third leading cause of cancer-related deaths in the world (1). During HCC progression, genetic and epigenetic alterations accumulate as it changes from an abnormal outgrowth to a life-threatening metastatic tumor (2–4). Distinguishing key alterations that promote metastasis from the thousands of less significant changes is crucial for inhibiting this process but remains a major challenge. Critically, cancers must reprogram metabolic pathways to meet redox, bioenergetic, and biosynthetic demands. Thus, modifications associated with sustaining metabolic homeostasis may contribute to tumor metastasis (5, 6). However, the precise mechanisms related to HCC metastasis and glycolysis reprogramming remain mostly unknown.

Endogenous noncoding RNA known as circular RNA (circRNA) has a closed-loop structure and is mainly produced from precursor RNAs through variable-shear processing. CircRNAs can participate in many stages of tumor progression (7), including tumor apoptosis, differentiation, metastasis, and proliferation. Nevertheless, the effects of circRNA on glycolysis are not well studied. In particular, the potential role of circRNA in HCC metastasis and/or recurrence via its effect on glycolysis, and the mechanisms underlying a possible circRNA-mediated metabolic switch, require further investigation.

In our previous study, we used circRNA sequencing (circRNA-seq) to identify HCC metastasis-associated circRNAs, focusing on those that are upregulated in patients with metastasis (8). Here, we investigated three circRNAs identified in this prior study that are downregulated in metastasis, using samples from two HCC patient cohorts and both in vivo and in vitro assays, to find their roles and mechanisms in the intervention of tumor progression and metastasis and then validated the results in clinical samples.

Patients and follow-up

A total of 600 patients with HCC from three independent cohorts were included in this study. Cohort 1 was previously used for circRNA sequencing and included 30 patients (15 patients with pulmonary metastasis and 15 patients without pulmonary metastasis). Cohort 2, which contained snap-frozen tumor tissues used for qRT-PCR analysis, included 190 patients who underwent curative resection during the period from January 2010 to December 2010 (Liver Cancer Institute, Zhongshan Hospital, Fudan University, Shanghai, China). Cohort 3 contained paraffin-embedded tissue samples used for RNAscope assays and IHC, and included 380 patients who underwent curative resection in 2006 and were monitored until March 15, 2013. Curative resections were defined as complete resection of tumor nodules, with the tumor margins confirmed to be free of cancer on histologic examination, and resection of regional lymph nodes, including the hilar, hepatoduodenal ligament, and caval lymph nodes, with no cancerous thrombus in the portal vein (main trunk or two major branches), hepatic veins, or bile duct. Patients who underwent palliative surgery only, had prior interventions (e.g., transhepatic artery embolization, chemotherapy, or radiotherapy), or were diagnosed with other primary malignancies or inflammatory diseases during the follow-up were excluded from the study.

All diagnoses of HCC were based on histopathology and followed World Health Organization criteria. Tumor grades were assigned using the system Edmondson–Steiner system (9), and Child–Pugh scores (10) were used for liver function assessment. Tumor stage was determined using the Union for International Cancer Control tumor, node, metastasis classification system (11). All patients were monitored after surgery every 3 to 6 months, as previously described (12). Tumor recurrence was diagnosed on the basis of CT scans, MRI, digital subtraction angiography, and elevated serum alpha-fetoprotein levels, with or without histological confirmation. Overall survival (OS) time was defined as the time from surgery to death or from surgery to the last follow-up visit. Time to recurrence (TTR) was defined as the time from surgery to detection of any intrahepatic recurrence or to detection of extrahepatic metastasis. The research ethics committee of Zhongshan Hospital (Shanghai, China) approved the ethical use for this study, and written informed consent was obtained from each patient.

Other materials and methods

Additional methodological details are provided in the Supporting Materials and Methods.

CircRPN2 is downregulated in HCC with metastasis or recurrence

In our previous study, we used circRNA-seq, in which RNA samples are enriched for circRNA by RNase R digestion, to identify 76 downregulated circRNAs and 144 upregulated circRNAs (fold-change >2, P < 0.05, average RPM >0.5) in patients with HCC with metastasis, relative to those without metastasis (8). With the same 30 primary HCC tissues, we tested typical metastasis makers' expression using PCR assay to evaluate the validity of these samples. The coexpression analysis showed that the identified circRNAs were closely correlated with these makers (Fig. 1A). Of all the identified downregulated circRNAs, we chose the three showing downregulation with the greatest significance (i.e., hsa_ circ_0004851/circCAPRIN1, hsa_circ_0080960/circABCB4, and hsa_circ_0007746/circRPN2; Supplementary Fig. S1A) for further analysis. We first confirmed amplification of these three circRNAs using PCR and Sanger sequencing of the products and ensured there were no changes in circRNA levels after treating with ribonuclease R, indicating that all three circRNAs are circular instead of linear (Fig. 1B and C; Supplementary Fig. S1B and S1C). We then examined their expression in human HCC cells with diverse metastatic potential and found that of these circRNAs, only expression of circRPN2 was lower in HCC cell lines than in the nontumorous L0–2 cell line (Fig. 1D; Supplementary Fig. S1D). CircRPN2 expression was also found to be significantly lower in the metastatic lines, MHCC97L, MHCC97H, and HCCLM3, than in the nonmetastatic cell lines, HepG2 and PLC/PRF/5 (Fig. 1D).

Figure 1.

CircRPN2 is associated with HCC metastasis and prognosis. A, Heatmap showing coexpression analysis between results of previous circRNA-seq and qRT-PCR analysis of metastasis makers in the same samples. The sub-heatmap on the left indicates the correlation between circRNAs and metastasis makers via the Spearman method. Blue, negative correlation; red, positive correlation; R2 value, correlation coefficient. The sub-heatmap below indicates differential expression of metastasis makers. Blue, low expression correlation; red, high expression correlation. Each column represents individual patients, and the stacked bar graph on the right shows the names of groups and markers. B, Sanger sequencing of circRPN2 and schematic illustrating circRNA production. C, Levels of circRPN2, and its associated mRNAs by qRT-PCR, before and after RNase R treatment. D, Relative expression of circRPN2 in seven human HCC cell lines and one human noncancerous cell line (L0-2), as detected by qRT-PCR. E, Expression of circRPN2 in HCC tumors with or without metastasis/recurrence, as detected by qRT-PCR. F, Expression of circRPN2 in tumors (T) and paratumors (P) from metastasis or recurrence group as detected by qRT-PCR. G, Scatter plots depict the relationships between circRPN2 expression and levels of E-cadherin, N-cadherin, vimentin, snail, c-myc, CD44, and TNFα. H and I, The prognostic values of circRPN2, as assessed by Kaplan–Meier analysis and Cox regression analysis. H, High, mean overall survival unreached; low, mean overall survival 51.2 months. I, High, mean recurrence-free survival unreached; low, mean recurrence-free survival 38.8 months. Length of the horizontal line represents the 95% confidence interval (CI) for each group. Data are shown as the mean ± SD of three independent experiments. *, P < 0.05, **, P < 0.01.

Figure 1.

CircRPN2 is associated with HCC metastasis and prognosis. A, Heatmap showing coexpression analysis between results of previous circRNA-seq and qRT-PCR analysis of metastasis makers in the same samples. The sub-heatmap on the left indicates the correlation between circRNAs and metastasis makers via the Spearman method. Blue, negative correlation; red, positive correlation; R2 value, correlation coefficient. The sub-heatmap below indicates differential expression of metastasis makers. Blue, low expression correlation; red, high expression correlation. Each column represents individual patients, and the stacked bar graph on the right shows the names of groups and markers. B, Sanger sequencing of circRPN2 and schematic illustrating circRNA production. C, Levels of circRPN2, and its associated mRNAs by qRT-PCR, before and after RNase R treatment. D, Relative expression of circRPN2 in seven human HCC cell lines and one human noncancerous cell line (L0-2), as detected by qRT-PCR. E, Expression of circRPN2 in HCC tumors with or without metastasis/recurrence, as detected by qRT-PCR. F, Expression of circRPN2 in tumors (T) and paratumors (P) from metastasis or recurrence group as detected by qRT-PCR. G, Scatter plots depict the relationships between circRPN2 expression and levels of E-cadherin, N-cadherin, vimentin, snail, c-myc, CD44, and TNFα. H and I, The prognostic values of circRPN2, as assessed by Kaplan–Meier analysis and Cox regression analysis. H, High, mean overall survival unreached; low, mean overall survival 51.2 months. I, High, mean recurrence-free survival unreached; low, mean recurrence-free survival 38.8 months. Length of the horizontal line represents the 95% confidence interval (CI) for each group. Data are shown as the mean ± SD of three independent experiments. *, P < 0.05, **, P < 0.01.

Close modal

We next measured expression of the candidate circRNAs in 190 human HCC tumors and observed significantly lower levels of circRPN2 in tumors from patients with postoperative metastasis or recurrence (Fig. 1E). Conversely, lower levels of circCAPRIN1 and circABCB4 were not observed in these samples, as compared with those from patients with no postoperative metastasis or recurrence (Supplementary Fig. S1E). We also detected decreased circRPN2 in HCC tissues relative to normal tissues, particularly in samples from the metastasis or recurrence subgroup (Fig. 1F; Supplementary Fig. S1F). Consistent with this observation, circRPN2 expression was found to be positively correlated with ECAD (E-cadherin) and negatively correlated with NCAD (N-cadherin), VIM (vimentin), SNAIL (snail), MYC (c-Myc), CD44, and TNFα (Fig. 1G), indicating that it shows a good tumor-suppressor index. Notably, we observed lower OS and recurrence-free survival (RFS) in patients with tumors expressing lower circRPN2 levels [mean OS, 51.2 months vs. unreached; mean RFS (mRFS): 38.8 months vs. unreached], but not lower levels of circCAPRIN1 or circABCB4 (Fig. 1H and I; Supplementary Fig. S1G–S1I). Furthermore, multivariate Cox regression analyses showed that circRPN2 expression is an independent predictor for OS and RFS rates in patients with HCC. Taken together, these results indicate that circRPN2 is frequently downregulated in patients with metastatic or recurrent HCC.

CircRPN2 inhibits HCC progression and metastasis

To investigate the functional role of circRPN2 in HCC, we established stable circRPN2-knockdown and overexpression cell lines via lentiviral infection, which had no effect on liner mRPN2. Specifically, circRPN2 was knocked down in HepG2 and PLC/PRF/5 cells using short hairpin (sh)RNA (HepG2-sh-circRPN2 and PLC/PRF/5-sh-circRPN2, respectively) and overexpressed in MHCC97H and HCCLM3 cells (MHCC97H-circRPN2 and HCCLM3-circRPN2, respectively). We then used qRT-PCR to measure levels of circRPN2 and RPN2 mRNA in these cell lines and confirm circRPN2 knockdown and overexpression (Supplementary Fig. S2A and S2B). Biological functions of knockdown and overexpression lines were further measured using the Cell Counting Kit-8 (CCK-8), as well as colony formation, wound-healing, Transwell migration, and Transwell invasion assays. These revealed that in HepG2 and PLC/PRF/5 cells, circRPN2 knockdown promotes cell proliferation, colony formation, migration, and invasion, whereas circRPN2 overexpression significantly suppresses these activities in MHCC97H and HCCLM3 cells (Fig. 2AG; Supplementary Fig. S2C).

Figure 2.

CircRPN2 suppresses HCC malignancy in vitro and in vivo. A, Proliferation of knockdown and overexpression cells was measured using the Cell Counting Kit-8 (CCK-8) assay. B, Colony formation assays with circRPN2 knockdown and overexpression HCC cells. C, Quantification of colony formation assays. D and E, Monolayers of circRPN2 knockdown and overexpression cells were wounded and monitored at 0 and 48 hours for wound channel closure; the cleaned area was measured and plotted as the percentage of the original time point (0 hours). F and G, Migration and invasive potential of circRPN2-knockdown and overexpression HCC cells measured using Transwell migration and Matrigel invasion assays, respectively. Scale bar, 100 μm. H–K, Mouse liver orthotopic tumor models established using circRPN2 knockdown and overexpression lines. H and J, Representative bioluminescence images of liver tumors (top), pulmonary metastasis (middle), and hematoxylin and eosin staining of metastatic nodules (bottom; scale bar, 200 μm) in lungs from experimental and control mice at day 35 after inoculation with HepG2 tumors (H) and HCCLM3 tumors (J). I and K, Relative circRPN2 expression of xenografts, liver photon flux growth curve and lung photon flux, and number of pulmonary metastasis nodules in experimental and control mice with HepG2 tumors (I) and HCCLM3 tumors (K). Data are shown as the mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Figure 2.

CircRPN2 suppresses HCC malignancy in vitro and in vivo. A, Proliferation of knockdown and overexpression cells was measured using the Cell Counting Kit-8 (CCK-8) assay. B, Colony formation assays with circRPN2 knockdown and overexpression HCC cells. C, Quantification of colony formation assays. D and E, Monolayers of circRPN2 knockdown and overexpression cells were wounded and monitored at 0 and 48 hours for wound channel closure; the cleaned area was measured and plotted as the percentage of the original time point (0 hours). F and G, Migration and invasive potential of circRPN2-knockdown and overexpression HCC cells measured using Transwell migration and Matrigel invasion assays, respectively. Scale bar, 100 μm. H–K, Mouse liver orthotopic tumor models established using circRPN2 knockdown and overexpression lines. H and J, Representative bioluminescence images of liver tumors (top), pulmonary metastasis (middle), and hematoxylin and eosin staining of metastatic nodules (bottom; scale bar, 200 μm) in lungs from experimental and control mice at day 35 after inoculation with HepG2 tumors (H) and HCCLM3 tumors (J). I and K, Relative circRPN2 expression of xenografts, liver photon flux growth curve and lung photon flux, and number of pulmonary metastasis nodules in experimental and control mice with HepG2 tumors (I) and HCCLM3 tumors (K). Data are shown as the mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Close modal

To further investigate the effects of circRPN2 on progression and metastasis of HCC in vivo, we established mouse liver orthotopic tumor models using knockdown and overexpression lines. Results of bioluminescence imaging reveal that HepG2-sh-circRPN2–derived xenografts produce tumors that are significantly larger than those of HepG2 cells (Fig. 2H and I). In contrast, HCCLM3-circRPN2 xenografts form significantly smaller tumors relative to HCCLM3 xenografts (Fig. 2J and K). Analysis of tumor weights yields similar results (Supplementary Fig. S2D and S2E). Notably, HepG2-sh-circRPN2 mice display higher pulmonary metastasis rates (3/6, 50%) than animals with HepG2 xenografts (0/6, 0%), whereas mice with HCCLM3-circRPN2 (4/6, 66.7%) xenografts show lower pulmonary metastasis rates than those with HCCLM3 xenografts (6/6, 100%). Similar results were obtained when lungs were examined for metastatic tumor nodules (Fig. 2I and K). Thus, these results suggest that circRPN2 inhibits HCC progression and metastasis.

CircRPN2 regulates aerobic glycolysis reprogramming in HCC cells

To further investigate the mechanism underlying the role of circRPN2 in HCC, we used RNA-seq to measure gene expression in HepG2 and HCCLM3 cells with circRPN2 knockdown or overexpression, respectively, and in the parental control cells. We then performed DESeq analysis (Supplementary Fig. S3A) to identify differentially expressed genes (DEG) relative to controls as follows: |log2FoldChange|>1, with P < 0.05. A total of 661 DEGs were identified, of which, 167 were upregulated and 216 were downregulated in circRPN2-knockdown cells, and 74 were upregulated and 219 were downregulated in circRPN2-overexpressing cells (Fig. 3A; Supplementary Fig. S3B). ClusterProfiler was then used to perform Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, in which DEGs were annotated by the KEGG pathway, and the number of DEGs associated with each pathway were determined. P values were calculated by the hypergeometric distribution method, with P < 0.05 used as the cutoff value for significance, to identify pathways that are significantly enriched among DEGs compared with the whole-genome background. Interestingly, our results indicate that circRPN2 induces distinct metabolic changes, including changes in many glucose-related genes, such as ADH4, PCK1, PKLR, ADH6, ALDOC, ADH1C, FBP1, and ATOH8 (Fig. 3A and B; Supplementary Fig. S3B; Supplementary Table S1), which are upregulated in circRPN2-knockdown cells and downregulated in circRPN2-overexpressing cells (Fig. 3C).

Figure 3.

CircRPN2 regulates glycolysis reprogramming in HCC. A, Volcano plot of mRNA transcript abundance and heatmap of DEGs related to glucose metabolism in circRPN2-knockdown cells analyzed by RNA-seq. Horizontal dashed line corresponds to P = 0.05. Vertical dashed lines indicate upregulation (fold change >2) and downregulation (fold change < −2). B, KEGG pathway enrichment analysis of DEGs. The stacked bar graph below shows the significant pathways identified. C, Validation of DEGs related to glucose metabolism by qRT-PCR. D, Heatmap of differentially expressed glycolytic metabolites in circRPN2-knockdown and circRPN2-overexpressing cells identified by metabolic mass spectrometry. E, Changes in relative glucose uptake, ATP levels, and lactate production in human HCC cells with circRPN2 knockdown or overexpression compared with the respective negative control cells. F–I, Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HepG2 cells with or without circRPN2 knockdown (F and G) and in HCCLM3 cells with or without circRPN2 overexpression (H and I). Data are shown as the mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Figure 3.

CircRPN2 regulates glycolysis reprogramming in HCC. A, Volcano plot of mRNA transcript abundance and heatmap of DEGs related to glucose metabolism in circRPN2-knockdown cells analyzed by RNA-seq. Horizontal dashed line corresponds to P = 0.05. Vertical dashed lines indicate upregulation (fold change >2) and downregulation (fold change < −2). B, KEGG pathway enrichment analysis of DEGs. The stacked bar graph below shows the significant pathways identified. C, Validation of DEGs related to glucose metabolism by qRT-PCR. D, Heatmap of differentially expressed glycolytic metabolites in circRPN2-knockdown and circRPN2-overexpressing cells identified by metabolic mass spectrometry. E, Changes in relative glucose uptake, ATP levels, and lactate production in human HCC cells with circRPN2 knockdown or overexpression compared with the respective negative control cells. F–I, Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HepG2 cells with or without circRPN2 knockdown (F and G) and in HCCLM3 cells with or without circRPN2 overexpression (H and I). Data are shown as the mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Close modal

We next performed targeted metabolomics and mass spectrometry (MS) analysis on circRPN2-knockdown and circRPN2-overexpressing cells, to validate our findings that circRPN2 modulates glucose metabolism in HCC. These experiments revealed that circRPN2 knockdown promotes production of products related to glucose metabolism, whereas circRPN2 overexpression inhibits production of these products (Fig. 3D; Supplementary Fig. S3C). KEGG pathway analysis of liquid chromatography (LC)/MS data further confirmed the significant enrichment of glucose-related processes (Supplementary Fig. S3D). In addition, we found that knockdown of circRPN2 expression in HepG2 and PLC/PRF/5 cells leads to increases in glucose uptake, ATP levels, and lactate production (Fig. 3E). In contrast, MHCC97H and HCCLM3 cell lines with overexpression of circRPN2 display proportionately opposite changes (Fig. 3E). We then performed the Seahorse experiment to detect the specific effect of circRPN2 on HCC glucose metabolism. We found that knockdown of circRPN2 reduces levels of cellular respiration and increases glycolysis levels in HepG2 cells (Fig. 3F and G), whereas overexpression of circRPN2 enhances respiration and decreases glycolysis levels in HCCLM3 cells (Fig. 3H and I). Thus, our data suggest that circRPN2 promotes HCC glycolytic reprogramming and inhibits conversion of oxidative phosphorylation to glycolysis in HCC cells.

CircRPN2 binds to enolase 1 and accelerates ubiquitin/proteasome-dependent ENO1 degradation

CircRNA has been shown to modulate tumor progression and metastasis by binding to key proteins involved in these processes. Therefore, we used RNA antisense purification technology with a circRPN2-specific probe to capture circRPN2-binding proteins. PCR and DNA electrophoresis were used to confirm probe efficacy (Fig. 4A; Supplementary Fig. S4A), and precipitated proteins were separated by 10% SDS-PAGE and detected by silver-staining (Supplementary Fig. S4B). Proteins pulled down using the circRPN2 probe were identified by LC-MS/MS, and the results are presented in Supplementary Table S2. On the basis of the Gene Ontology analysis, 39.2% of these proteins were predicted to be involved in metabolic processes. Of these, glycolysis was the most common process, suggesting that some of these binding proteins are involved in circRPN2-mediated aerobic glycolysis in HCC cells (Fig. 4B). Among the 10 proteins that were preliminarily screened using score criteria (i.e., score >350), enolase 1 (ENO1) had the highest binding score, with multiple meaningful combined fragments (Supplementary Fig. S4C). In addition, CatRAPID software analysis (http://service.tartaglialab.com/page/catrapid_group) predicts that circRPN2 binds to ENO1 (Supplementary Fig. S4D), and this was confirmed using Western blot analysis (Fig. 4C). RNA immunoprecipitation assays with anti-ENO1 antibody consistently revealed a significantly higher enrichment of circRPN2 with ENO1 antibody than with IgG control (Fig. 4D and E), and FISH assays show that circRPN2 colocalizes with ENO1 in the cytoplasm (Fig. 4F). Taken together, these results indicate that circRPN2 binds to ENO1.

Figure 4.

CircRPN2 binds to ENO1 and accelerates ENO1 degradation by ubiquitin-mediated proteolysis. A–C, Identification of circRPN2-binding proteins using RNA antisense purification (RAP) technology and LC-MS/MS analysis. A, PCR assay to confirm efficacy of the circRPN2 probe. RAP, circRPN2 probe; NC, control probe. B, Gene Ontology and KEGG analysis of predicted circRPN2-binding proteins. C, Western blot analysis of ENO1 derived from RNA pulldown using the circRPN2 probe and Lac Z (negative control) probe. The input is the total protein used for RNA pulldown. D and E, qRT-PCR analysis of RNAs derived from RNA immunoprecipitation (RIP) assays with anti-ENO1 antibody (D) and agarose electrophoresis of PCR products from the RIP assay (E). F, Localization of circRPN2 and ENO1 by FISH. Scale bar, 50 μm. G, Relative mRNA expression, protein levels, and enzymatic activity of ENO1 in different human HCC cell lines with circRPN2 knockdown or overexpression. H, CircRPN2-knockdown and circRPN2-overexpressing cells and their respective controls were treated with actinomycin D (ActD, 4 μg/mL) for the indicated periods of time, and ENO1 mRNA levels were analyzed by qRT-PCR. I, CirRPN2 knockdown and overexpressing cells and respective controls were treated with cycloheximide (CHX) for the indicated periods of time, and ENO1 protein levels were measured by Western blot. J, Ubiquitin levels of ENO1 in circRPN2-knockdown HepG2 cells with HA-ubiquitin overexpression. K, Ubiquitin levels of ENO1 in circRPN2-overexpressing HCCLM3 cells with HA-ubiquitin overexpression. For J and K, the plasmid transfection time was 24 hours, and the treatment time with the proteosome inhibitor MG132 (10 μmol/L) was 4 hours. Data are shown as the mean ± SD of three independent experiments; ns, not significant; *, P < 0.05; **, P < 0.01.

Figure 4.

CircRPN2 binds to ENO1 and accelerates ENO1 degradation by ubiquitin-mediated proteolysis. A–C, Identification of circRPN2-binding proteins using RNA antisense purification (RAP) technology and LC-MS/MS analysis. A, PCR assay to confirm efficacy of the circRPN2 probe. RAP, circRPN2 probe; NC, control probe. B, Gene Ontology and KEGG analysis of predicted circRPN2-binding proteins. C, Western blot analysis of ENO1 derived from RNA pulldown using the circRPN2 probe and Lac Z (negative control) probe. The input is the total protein used for RNA pulldown. D and E, qRT-PCR analysis of RNAs derived from RNA immunoprecipitation (RIP) assays with anti-ENO1 antibody (D) and agarose electrophoresis of PCR products from the RIP assay (E). F, Localization of circRPN2 and ENO1 by FISH. Scale bar, 50 μm. G, Relative mRNA expression, protein levels, and enzymatic activity of ENO1 in different human HCC cell lines with circRPN2 knockdown or overexpression. H, CircRPN2-knockdown and circRPN2-overexpressing cells and their respective controls were treated with actinomycin D (ActD, 4 μg/mL) for the indicated periods of time, and ENO1 mRNA levels were analyzed by qRT-PCR. I, CirRPN2 knockdown and overexpressing cells and respective controls were treated with cycloheximide (CHX) for the indicated periods of time, and ENO1 protein levels were measured by Western blot. J, Ubiquitin levels of ENO1 in circRPN2-knockdown HepG2 cells with HA-ubiquitin overexpression. K, Ubiquitin levels of ENO1 in circRPN2-overexpressing HCCLM3 cells with HA-ubiquitin overexpression. For J and K, the plasmid transfection time was 24 hours, and the treatment time with the proteosome inhibitor MG132 (10 μmol/L) was 4 hours. Data are shown as the mean ± SD of three independent experiments; ns, not significant; *, P < 0.05; **, P < 0.01.

Close modal

We then performed PCR and Western blot analyses in HCC cells to further examine the relationship between circRPN2 and ENO1. Results indicate that circRPN2 does not alter ENO1 mRNA levels but can decrease ENO1 protein levels and enzymatic activity (Fig. 4G). We hypothesized that this occurs due to protein degradation mediated by posttranslational modification. To confirm that circRPN2 does not affect ENO1 mRNA stability, we performed RNA polymerase II Inhibitor actinomycin chase assays. Results show that ENO1 mRNA levels are not significantly altered in circRPN2-knockdown or -overexpressing cells (Fig. 4H). In contrast, following treatment with the protein synthesis inhibitor cycloheximide, we found that circRPN2 knockdown increases half-life of ENO1 protein, whereas circRPN2 overexpression decreases ENO1 half-life (Fig. 4I).

A previous study found that >80% of protein degradation occurs via the ubiquitin–proteasome pathway (13). We therefore measured ubiquitination in HCC cells to determine whether circRPN2 is involved in ubiquitin-mediated ENO1 degradation. Our results show that ubiquitin-mediated ENO1 degradation is inhibited in circRPN2-knockdown cells and enhanced in circRPN2-overexpressing cells, compared with the respective control cells (Fig. 4J and K), suggesting that circRPN2 decreases ENO1 protein levels via ubiquitin-mediated degradation. Collectively, these results indicate that circRPN2 binds to ENO1 and downregulates this protein via ubiquitin/proteasome-dependent degradation.

CircRPN2 inhibits aerobic glycolysis in HCC cells via intervention in the ENO1-mediated Akt/mTOR pathway

We next examined whether ENO1 acts as the key factor in circRPN2-mediated inhibition of aerobic glycolysis. To this end, we performed Seahorse assays to test the effect of circRPN2 and ENO1 on metabolic activity. We found that treatment with ENOblock (AP-III-a4), a novel small molecule that directly binds to enolase and inhibits its activity but does not affect protein levels (Supplementary Fig. S5A and S5B), reverses the circRPN2 knockdown-mediated decrease in respiration levels and increased glycolysis rates observed in HepG2 cells (Fig. 5AC). Furthermore, similar to circRPN2 overexpression, ENOblock enhances respiration levels and reduces glycolytic activity in HCCLM3 cells (Fig. 5DF). These results suggest that ENO1 is critical for circRPN2-mediated inhibition of aerobic glycolysis in HCC cells.

Figure 5.

CircRPN2 suppresses HCC aerobic glycolysis through the ENO1/AKT/mTOR pathway. A–C, OCR and ECAR levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HepG2 circRPN2-knockdown cells, with or without ENOblock treatment. D–F, OCR and ECAR levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HCCLM3 cells, with or without ENOblock treatment. G and H, HepG2 circRPN2 knockdown cells or HCCLM3 circRPN2 overexpression cells were treated with ENOblock (G) or MK2206 (H), and levels of AKT, pAKT, mTOR, and p-mTOR were measured by Western blot. I and J, Changes in ECAR and maximal glycolytic levels in the HepG2 (I) and HCCLM3 (J) groups treated with MK2206. Cells were treated with 10-μmol/L ENOblock or 5-nmol/L MK2206 for 24 hours before the Seahorse assay. Data are shown as the mean ± SD of three independent experiments; ns, not significant; *, P < 0.05; **, P < 0.01.

Figure 5.

CircRPN2 suppresses HCC aerobic glycolysis through the ENO1/AKT/mTOR pathway. A–C, OCR and ECAR levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HepG2 circRPN2-knockdown cells, with or without ENOblock treatment. D–F, OCR and ECAR levels were measured using the Seahorse assay, and the basal/maximal respiration and glycolysis levels were calculated accordingly in HCCLM3 cells, with or without ENOblock treatment. G and H, HepG2 circRPN2 knockdown cells or HCCLM3 circRPN2 overexpression cells were treated with ENOblock (G) or MK2206 (H), and levels of AKT, pAKT, mTOR, and p-mTOR were measured by Western blot. I and J, Changes in ECAR and maximal glycolytic levels in the HepG2 (I) and HCCLM3 (J) groups treated with MK2206. Cells were treated with 10-μmol/L ENOblock or 5-nmol/L MK2206 for 24 hours before the Seahorse assay. Data are shown as the mean ± SD of three independent experiments; ns, not significant; *, P < 0.05; **, P < 0.01.

Close modal

ENO1 promotes AKT activation to exert its metabolic effects, and the AKT/mTOR signaling pathway is essential for glucose metabolism (14, 15). We found that changes in circRPN2 expression do affect AKT phosphorylation in HCC cell lines (Fig. 5GH). Therefore, we investigated whether circRNP2 inhibits glycolysis in HCC cells via effects on the ENO1-mediated AKT/mTOR pathway. Notably, we detected increased levels of phosphorylated AKT (pAKT) and mTOR (p-mTOR) in circRPN2-knockdown HepG2 cells and decreased levels of these proteins in HCCLM3 cells overexpressing circRPN2. Furthermore, ENOblock treatment ameliorates the increases in pAKT and p-mTOR induced by cirRPN2 knockdown in HepG2 cells and reduces pAKT and p-mTOR in HCCLM3 cells. These results indicate that cirRPN2 modulates phosphorylation of AKT and mTOR via its effects on ENO1 (Fig. 5G). We also used the AKT inhibitor, MK2206, to confirm the inhibitory role of circRPN2 on the AKT/mTOR signaling pathway and found that it markedly attenuates the circRPN2 knockdown-mediated increase of pAKT and p-mTOR in HepG2 cells. In addition, similar to the effect of circRPN2 overexpression, treatment with MK2206 inhibits phosphorylation of AKT and mTOR in HCCLM3 cells (Fig. 5H). Consistent with these observations, MK2206 treatment attenuates the circRPN2 knockdown-mediated increased in glycolysis levels observed in HepG2 cells, and in HCCLM3 cells, it promotes similar reduction in glycolytic levels as circRPN2 overexpression (Fig. 5I and J). Taken together, these results indicate that circRPN2 promotes metabolic changes in aerobic glycolysis via its effects on the ENO1-mediated AKT/mTOR pathway.

CircRPN2 inhibits HCC progression and metastasis through ENO1

To further investigate whether circRPN2 can inhibit HCC progression and metastasis through ENO1, we used ENOblock to probe the biological role of ENO1 in vitro. Using CCK-8, Transwell migration, and Transwell invasion assays, we found that ENOblock ameliorates the increases in tumor proliferation and metastatic ability observed in circRPN2-knockdown Hep2G cells (Fig. 6A). It also decreases proliferation and metastasis in HCCLM3 cells, similar to the effect of circRPN2 overexpression (Fig. 6B).

Figure 6.

CircRPN2 inhibits HCC progression and metastasis in vitro and in vivo through ENO1. A and B, CCK-8 and Transwell migration and invasion assays to measure proliferation, migration, and invasion abilities of the HepG2 (A) and HCCLM3 (B) groups, with and without ENOblock treatment. Scale bar, 100 μm. C and D, Representative bioluminescence images of liver tumors (top), pulmonary metastasis (middle), and hematoxylin and eosin staining of metastatic nodules (bottom; scale bar, 200 μm) in lungs from different groups of mice at day 35 after inoculation with HepG2 tumors (C) and HCCLM3 tumors (D). The tumor growth curve and histogram depict the photon flux emitted from mouse livers and pulmonary metastasis in the HepG2 mouse model (C) and HCCLM3 mouse model (D). E and F, Tumor weights of xenografts, efficiency of pulmonary metastasis, and number of pulmonary metastasis nodules in the HepG2 mouse model (E) and the HCCLM3 mouse model (F). Data are shown as mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Figure 6.

CircRPN2 inhibits HCC progression and metastasis in vitro and in vivo through ENO1. A and B, CCK-8 and Transwell migration and invasion assays to measure proliferation, migration, and invasion abilities of the HepG2 (A) and HCCLM3 (B) groups, with and without ENOblock treatment. Scale bar, 100 μm. C and D, Representative bioluminescence images of liver tumors (top), pulmonary metastasis (middle), and hematoxylin and eosin staining of metastatic nodules (bottom; scale bar, 200 μm) in lungs from different groups of mice at day 35 after inoculation with HepG2 tumors (C) and HCCLM3 tumors (D). The tumor growth curve and histogram depict the photon flux emitted from mouse livers and pulmonary metastasis in the HepG2 mouse model (C) and HCCLM3 mouse model (D). E and F, Tumor weights of xenografts, efficiency of pulmonary metastasis, and number of pulmonary metastasis nodules in the HepG2 mouse model (E) and the HCCLM3 mouse model (F). Data are shown as mean ± SD of three independent experiments. *, P < 0.05; **, P < 0.01.

Close modal

We then used mouse orthotopic tumor models to evaluate whether circRPN2 inhibits HCC progression and metastasis through ENO1 in vivo. Bioluminescence imaging results revealed that ENOblock reverses the effect of circRPN2 knockdown in HepG2-sh-circRPN2 xenograft mice (Fig. 6C), and similar to HCCLM3-circRPN2 xenografts, HCCLM3 + ENOblock xenografts produce significantly smaller tumors than HCCLM3 xenografts (Fig. 6D). Analysis of tumor weights yields similar results (Fig. 6E and F). In addition, ENOblock suppresses circRPN2 knockdown-mediated induction of metastasis in HepG2-sh-circRPN2 mice (1/6; 16.7%; Fig. 6E), and HCCLM3 + ENOblock xenografts show a lower rate of pulmonary metastasis (2/6; 33.3%) than HCCLM3 xenografts (6/6; 100%; Fig. 6F). Thus, in summary, our results indicate that circRPN2 suppresses HCC progression and metastasis in vitro and in vivo via ENO1 mediation.

CircRPN2 enhances FOXO1 expression by sponging miR-183–5p

CircRNAs often exert their biological effects by acting as competing endogenous RNAs (ceRNA) for miRNAs in the cytoplasm. Here, using FISH and nuclear mass separation assays, we found that circRPN2 is primarily localized in the cytoplasm of HCC cells (Supplementary Fig. S6A and S6B). Therefore, we examined whether circRPN2 can influence HCC progression by binding to miRNAs. To this end, we used the miRanda miRNA Target Prediction Tool to predict 111 potential circRPN2-binding miRNAs (Supplementary Table S3). On the basis of our previous study (16), six miRNAs were selected with rich abundance (TPM ≥10) in human HCC cells (Supplementary Table S4; Fig. 7A). We then performed luciferase reporter assays in 293T cells with all six miRNAs to determine whether any of these species could interact with circRPN2 (Supplementary Fig. S6C and S6D; Fig. 7B). Our results showed that miR-183–5p and miR-148a-5p can reduce activity from a wild-type (WT) LUC-circRPN2 reporter gene construct. However, The Cancer Genome Atlas (TCGA) database indicates that only miR-183–5p is associated with tumor prognosis (Supplementary Fig. S6E; Fig. 7C). Furthermore, using RNA FISH, we found that circRPN2 colocalizes with miR-183–5p in the cytoplasm (Fig. 7D), and miR-183-5p is strongly enriched among purified RNAs pulled down by the circRPN2-specific probe (Fig. 7E). Notably, however, circRPN2 does not affect miR-183–5p expression, and miR-183–5p has no effect on circRPN2 expression (Supplementary Fig. S6F–S6H), suggesting that circRPN2 might function as a ceRNA sponge for miR-183–5p.

Figure 7.

CircRPN2 enhances FOXO1 expression by sponging miR-183–5p and represses HCC glycolysis by regulating the miR-183–5p/FOXO1 axis. A, Putative circRPN2 target microRNAs predicted using the miRanda miRNA Target Prediction Tool and miR-seq data from HCC cell lines. B, Luciferase reporter assay to measure activity from a WT LUC-circRPN2 or LUC-circRPN2-mutant construct in 293T cells overexpressing putative target miRNAs. C, Expression and prognosis values of miR-183–5p in HCC from TCGA database. D, Colocalization of circRPN2 and miR-183–5p in the cytoplasm detected by FISH. Scale bar, 50 μm. E, Binding of circRPN2 to miR-183–5p demonstrated by a circRPN2-antisense pulldown assay.F, Luciferase reporter assay to measure activity from a WT LUC-FOXO1 or LUC-FOXO1-mutant construct in 293T cells overexpressing miR-183–5p. G, FOXO1 protein levels in HepG2 and HCCLM3 cells with circRPN2 knockdown or overexpression, respectively, with or without miR-183–5p mimic/inhibitor treatment, as detected by Western blot. H, ECAR levels were measured using the Seahorse assay, and basal/maximal glycolysis levels were calculated accordingly in HepG2 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 knockdown. I, Relative glucose uptake, ATP levels, and lactate production in HepG2 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 knockdown. J, ECAR levels were measured using the Seahorse assay, and the basal/maximal glycolysis levels were calculated accordingly in HCCLM3 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 overexpression. K, Relative glucose uptake, ATP levels, and lactate production in HCCLM3 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 overexpression. Data are shown as the mean ± SD of three independent experiments. ns, not significant; *, P < 0.05; **, P < 0.01.

Figure 7.

CircRPN2 enhances FOXO1 expression by sponging miR-183–5p and represses HCC glycolysis by regulating the miR-183–5p/FOXO1 axis. A, Putative circRPN2 target microRNAs predicted using the miRanda miRNA Target Prediction Tool and miR-seq data from HCC cell lines. B, Luciferase reporter assay to measure activity from a WT LUC-circRPN2 or LUC-circRPN2-mutant construct in 293T cells overexpressing putative target miRNAs. C, Expression and prognosis values of miR-183–5p in HCC from TCGA database. D, Colocalization of circRPN2 and miR-183–5p in the cytoplasm detected by FISH. Scale bar, 50 μm. E, Binding of circRPN2 to miR-183–5p demonstrated by a circRPN2-antisense pulldown assay.F, Luciferase reporter assay to measure activity from a WT LUC-FOXO1 or LUC-FOXO1-mutant construct in 293T cells overexpressing miR-183–5p. G, FOXO1 protein levels in HepG2 and HCCLM3 cells with circRPN2 knockdown or overexpression, respectively, with or without miR-183–5p mimic/inhibitor treatment, as detected by Western blot. H, ECAR levels were measured using the Seahorse assay, and basal/maximal glycolysis levels were calculated accordingly in HepG2 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 knockdown. I, Relative glucose uptake, ATP levels, and lactate production in HepG2 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 knockdown. J, ECAR levels were measured using the Seahorse assay, and the basal/maximal glycolysis levels were calculated accordingly in HCCLM3 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 overexpression. K, Relative glucose uptake, ATP levels, and lactate production in HCCLM3 cells subjected to miR-183–5p mimic/inhibitor treatment and circRPN2 overexpression. Data are shown as the mean ± SD of three independent experiments. ns, not significant; *, P < 0.05; **, P < 0.01.

Close modal

To test this possibility, we first predicted the target genes for miR-183–5p by analyzing two databases (TargetScan and miRTarBase) and common tumor-suppressor genes (17, 18). Three putative target genes were identified (Supplementary Fig. S7A)—SMAD family member 4 (SMAD4), BAF chromatin remodeling complex subunit BCL11B (BCL11B), and forkhead box protein O1 (FOXO1), and these were verified by PCR and Western blot analyses (Supplementary Fig. S7B–S7E). Notably, we found that upregulation of miR-183–5p decreases FOXO1 mRNA and protein expression in HepG2 and PLC/PRF/5 cells, whereas anti–miR-183–5p has the opposite effect in MHCC97H and HCCLM3 cells. However, no significant changes in response to altered miR-183–5p expression were detected for the two other candidate target genes (SMAD4 and BCL11B). We note that ENO1 is not a predicted target of miR-183–5p, and its expression levels are also unaffected by changes in miR-183–5p expression (Supplementary Fig. S7F and S7G). Using luciferase assays, we obtained further evidence that miR-183–5p may interact with FOXO1 (Fig. 7F). In addition, TCGA database analysis revealed that FOXO1, which is decreased in HCC tumor tissues and can affect patient prognosis, is negatively correlated with miR-183–5p expression (Supplementary Fig. S7H–S7J). Thus, these results suggest that FOXO1 is a target of miR-183–5p.

We then investigated whether circRPN2 can influence FOXO1 expression by sponging miR-183–5p. We found that circRPN2 knockdown promotes decreased FOXO1 expression in HepG2 cells, whereas overexpression of circRPN2 has the opposite effect in HCCLM3 cells (Fig. 7G). Furthermore, miR-183-5p inhibitor abolishes the circRPN2 knockdown-mediated decrease in FOXO1 observed in HepG2 cells, and miR-183–5p mimic reverses the increase FOXO1 expression induced by circRPN2 overexpression in HCCLM3 cells (Fig. 7G). Thus, in total, our results suggest that circRPN2 may upregulate FOXO1 expression by sponging miR-183–5p.

CircRPN2 suppresses HCC glycolysis and progression via the miR-183–5p/FOXO1 axis

FOXO1 negatively regulates glycolysis in many tissues or cancers (19–22). To investigate whether circRPN2 can suppress HCC glycolysis and progression via the miR-183–5p/FOXO1 axis, we performed the Seahorse assay and other glucose metabolism-related functional tests to detect the effect of miR-183–5p/FOXO1 on HCC glycolysis. We found that miR-183–5p mimic enhances glycolysis, glucose uptake, ATP levels, and lactate production in HepG2 cells, whereas miR-183–5p inhibitor decreases glycolysis, glucose uptake, ATP levels, and lactate production in HCCLM3 cells (Fig. 7HK). In addition, miR-183–5p inhibitor abolishes the circRPN2 knockdown-induced increases in glycolysis, glucose uptake, ATP levels, and lactate production in HepG2 cells, whereas miR-183–5p mimic reverses the decreases in glycolysis, glucose uptake, ATP levels, and lactate production observed in HCCLM3 cells in response to circRPN2 overexpression (Fig. 7HK). We then tested whether circRPN2 can inhibit HCC progression and metastasis through the miR-183–5p/FOXO1 axis using CCK-8 and Transwell migration and invasion assays. We found that, consistent with the effects on glycolysis, miR-183–5p mimic increases proliferation and metastatic ability in HepG2 cells, whereas miR-183–5p inhibitor has the opposite effect in HCCLM3 cells (Supplementary Fig. S8A and S8B). And miR-183-5p inhibitor abolishes the circRPN2 knockdown-induced increases of proliferation and metastatic ability in HepG2 cells, whereas miR-183–5p mimic reverses the decreases of proliferation and metastatic ability in HCCLM3 cells with circRPN2 overexpression (Supplementary Fig. S8A and S8B). Furthermore, FOXO1 inhibitor (AS1842856) reverses the decreases in glycolysis, proliferation, and metastasis observed in HCCLM3 cells in response to circRPN2 overexpression (Supplementary Fig. S8C–S8E). Overall, these results suggest that circRPN2 inhibits HCC glycolysis and progression via the miR-183–5p/FOXO1 axis.

Levels of circRPN2, ENO1, and FOXO1 correlate in HCC tissues and predict the prognosis of patients with HCC

To investigate the clinical prognostic value of circRPN2, ENO1, and FOXO1 and uncover possible associations among these molecules in HCC, we examined their expression using the matched-tissue microarrays for RNAscope and immunohistochemical staining of primary tumor tissues from 380 patients with HCC. Representative results are presented in Fig. 8A. We found that HCC tumors with low circRPN2 levels show increased ENO1 expression and decreased FOXO1 expression, whereas tumors with high circRPN2 levels often display decreased ENO1 expression and increased FOXO1 expression. These relationships were further shown through correlation analysis (Fig. 8B). Notably, 1-, 3-, and 5-year OS rates for patients with lower circRPN2 levels were found to be lower than for those with high circRPN2 levels (84.7% vs. 91.6%, 62.4% vs. 75.8%, and 50.5% vs. 64.5%, respectively). In addition, groups of patients with low circRPN2 levels show higher cumulative recurrence rates (34.4% vs. 20.5%, 53.5% vs. 36.8%, and 62.5% vs. 43.3%, respectively; Fig. 8C). OS rates among patients with high levels of ENO1 or low levels of FOXO1 were also found to be lower, with higher cumulative recurrence rates observed in these individuals (Fig. 8D and E).

Figure 8.

Expression and prognostic values of circRPN2, ENO1, and FOXO1 in HCC. A, Expression levels of circRPN2, ENO1, and FOXO1 in representative patients with HCC are shown. Scale bar, 50 μm. B, Scatter plots depicting the relationships between circRPN2 and ENO1 or FOXO1 in HCC tissues. C–F, Kaplan–Meier analysis to determine the prognostic values of circRPN2, ENO1, and FOXO1. C, circRPN2: high, mean overall survival (mOS) unreached; low, mOS 61.8 months; high, mean cumulative recurrence (mCR) 73.2 months; low, mCR 29.5 months. D, ENO1: high, mOS 58.5 months; low, mOS unreached; high, mCR 29.5 months; low, mCR 66.3 months. E, FOXO1: high, mOS unreached; low, mOS 62.4 months; high, mCR 70.3 months; low, mCR 32.4 months. F, Group I, circRPN2 high/FOXO1 high/ENO1 low; Group III, circRPN2 low/FOXO1 low/ENO1 high; Group II, all other patients. Group I, mOS unreached; Group II, mOS 78.2 months; Group III, mOS 45.3 months; Group I, mCR 74.3 months; Group II, mCR 58.9 months; Group III, mCR 22 months. Data are the representative of three independent experiments. G, The model illustrates the proposed role of circRPN2–ENO1–AKT/mTOR and circRPN2/miR-183–5p/FOXO1 signaling in regulating HCC glycolysis and metastasis.

Figure 8.

Expression and prognostic values of circRPN2, ENO1, and FOXO1 in HCC. A, Expression levels of circRPN2, ENO1, and FOXO1 in representative patients with HCC are shown. Scale bar, 50 μm. B, Scatter plots depicting the relationships between circRPN2 and ENO1 or FOXO1 in HCC tissues. C–F, Kaplan–Meier analysis to determine the prognostic values of circRPN2, ENO1, and FOXO1. C, circRPN2: high, mean overall survival (mOS) unreached; low, mOS 61.8 months; high, mean cumulative recurrence (mCR) 73.2 months; low, mCR 29.5 months. D, ENO1: high, mOS 58.5 months; low, mOS unreached; high, mCR 29.5 months; low, mCR 66.3 months. E, FOXO1: high, mOS unreached; low, mOS 62.4 months; high, mCR 70.3 months; low, mCR 32.4 months. F, Group I, circRPN2 high/FOXO1 high/ENO1 low; Group III, circRPN2 low/FOXO1 low/ENO1 high; Group II, all other patients. Group I, mOS unreached; Group II, mOS 78.2 months; Group III, mOS 45.3 months; Group I, mCR 74.3 months; Group II, mCR 58.9 months; Group III, mCR 22 months. Data are the representative of three independent experiments. G, The model illustrates the proposed role of circRPN2–ENO1–AKT/mTOR and circRPN2/miR-183–5p/FOXO1 signaling in regulating HCC glycolysis and metastasis.

Close modal

We then assigned patients to one of three groups (Group I, high levels of circRPN2 and FOXO1, low levels of ENO1; Group III:, low levels of circRPN2 and FOXO1, high levels of ENO1; Group II, remaining patients) to evaluate their combined effects on HCC prognosis. We detected significantly higher OS rates for Group I than for Group III (90.8% vs. 81.1%, 72.3% vs. 58.6%, and 67.3% vs. 44.8%, respectively; Fig. 8F); cumulative recurrence rates in Group I were also found to be significantly lower than rates for Group III (23.5% vs. 41.4%, 35.7% vs. 57.8%, and 41.8% vs. 65.8%, respectively; Fig. 8F). Finally, we performed univariate and multivariate Cox regression analyses and found that expression levels of circRPN2, ENO1, FOXO1 alone, as well as the circRPN2/ENO1/FOXO1 coindex, are independent predictors for OS and TTR rates in patients with HCC (Supplementary Table S5).

CircRNAs were previously thought to be splicing by-products that are present only at low levels in a cell. Subsequently, however, thousands of circRNAs have been identified using high-throughput sequencing techniques and bioinformatic analyses (23). Many of these display development-specific and tissue-specific expression patterns, indicating involvement in key physiological and pathophysiological processes. Critically, circRNAs are important potential biomarkers in a variety of malignant cancers (24). However, the roles of circRNAs in tumor metastasis or recurrence, which is the main cause of treatment failure in patients with malignant tumors (25, 26), remain to be thoroughly studied. In our previous study, we used circRNA-seq to identify circRNAs that are differentially expressed in primary HCCs from patients with metastasis and those without metastasis, and found circASAP1 as a key regulator in HCC metastasis by means of miR-326/miR-532–5p-MAPK1/CSF-1 signaling (8). Here, we examined the role of circRPN2, which was found to be downregulated in metastatic HCC in our circRNA-seq data. Notably, we show that circRPN2 is significantly downregulated in HCC tissues, particularly in HCC with metastasis or recurrence, and for the first time, we provide evidence that circRPN2 can inhibit HCC progression and metastasis by promoting the degradation of ENO1 and regulating the miR-183–5p/FOXO1 Axis.

An increasing ability to metastasize is usually accompanied by enhancement of tumor proliferation. Instead of mitochondrial oxidative phosphorylation, glycolysis is preferentially used by cancer cells to produce glucose-dependent ATP and glycolytic intermediates for meeting the needs of rapid growth, even when an abundant oxygen supply is available (27). Thus, the important relationship between tumor metastasis and aberrant tumor metabolism has become increasingly recognized. However, few studies have examined the role of circRNA in this critical link between tumor metastasis and glycolysis reprogramming. Notably, in this study, we found that altered levels of circRPN2 lead to glycolytic reprogramming in HCC cells.

Most studies on circRNAs have focused on their ability to act as miRNA sponges, and thus, their other biological roles, such as direct protein binding to regulate biological processes, have rarely been examined (28). Here, we used RNA pulldown assays and MS analysis and found that circRPN2 binds to the glycolytic enzyme ENO1. We then examined the relationship between circRPN2 and ENO1 and found that cirRPN2 accelerates ENO1 degradation via ubiquitin-mediated proteolysis, thereby decreasing its activity. However, it remains unknown whether circRPN2 promotes ubiquitination of ENO1 directly by inducing a shift in its conformational structure that exposes a ubiquitin ligase recognition domain or if it acts as a scaffold to provide a platform for interaction between ENO1 and specific proteins. Thus, this key question requires further investigation.

ENO1 is a metalloenzyme that catalyzes dehydration of 2-phospho-D-glycerate to phosphoenolpyruvate in the glycolytic pathway. Previous studies have shown that ENO1-mediated metabolic reprogramming occurs in cancer cells (29), and critically, ENO1 inhibition successfully restores mitochondrial respiration from glycolysis and inhibits cancer cell proliferation and metastasis (30, 31). Silencing of ENO1 further promotes adaptation to catabolic pathways and restoration of bulk acetyl-CoA via enhanced β-oxidation, and it also fuels the tricarboxylic acid cycle via cataplerotic reactions associated with tyrosine and glutamine catabolism (29, 32). Consistent with previous studies showing that ENO1 regulates progression, metastasis, metabolic shift, and glycolytic reprogramming in cancer (14, 31), we found that circRPN2-mediated ENO1 degradation suppresses HCC progression and metastasis, enhances respiration levels, and significantly reduces glycolytic levels in HCC cells. Critically, pharmacological inhibition of ENO1 with ENOblock attenuates the effects of circRPN2 on glucose metabolism, progression, and metastasis. Thus, our data suggest that by binding with ENO1 and targeting this protein for degradation, circRPN2 may play a key role in HCC metastasis and glycolytic reprogramming.

Cancer cells increase glucose consumption via a shift to glycolysis, which provides a sufficient carbon source for cell proliferation and allows cells to engage in autocrine signaling and local metabolite-based paracrine communication to facilitate progression (33). Consequently, glycolysis pathway suppression is a rational strategy for cancer therapy. Notably, recent studies have found that the ENO1-catalyzed glycolytic step is reversible (14). In addition, although enolase includes three isoforms, only ENO1 is relatively stabilized, resulting in less toxicity and enhanced specificity for ENO1-targeting compounds, and suggesting that ENO1 may be an ideal target for glycolysis-inhibition cancer therapy. In this study, we inhibited ENO1 by modulating the levels of circRPN2 and observed similar effects on HCC metastasis and glycolysis, as previously reported. Thus, our results support the hypothesis that ENO1 is a promising therapeutic target in cancer. Notably, we found that ENOblock effectively inhibits HCC progression and metastasis in vitro and in vivo, indicating that it may show a beneficial effect in patients with circRPN2-associated HCC (i.e., those with reduced levels of circRPN2 expression). Critically, however, although ENOblock-mediated suppression of metabolic activity in HCC is consistent with findings from previous studies (34), the specific mechanism of action may be complex and requires further study (35).

The AKT/mTOR signaling pathway affects cellular biosynthesis and aerobic glycolytic processes in malignant tumors (36, 37), and AKT downstream targets, such as mTOR, have been implicated in metabolic changes (38). Consistent with these observations, we found that in HCC cells, cirRPN2 inhibits the AKT/mTOR signaling pathway via ENO1 degradation to increase the inhibitory effect on glycolysis. This leads to decreased lactate production and amelioration of malignant cancer cell phenotypes. Through use of the AKT inhibitor, MK2206, we validated circRPN2-mediatated inhibition of the AKT/mTOR pathway and found that, consistent with our hypothesis, glycolytic reprogramming and increased levels of pAKT and p-mTOR in circRPN2 knockdown cells are attenuated by MK2206 treatment. Collectively, these results indicate that the effects of circRPN2 on glycolytic reprogramming and progression in HCC occur via the ENO1-activated AKT/mTOR pathway.

Many circRNAs promote their biological activities by acting as a “sponge” bind specific miRNAs and block their interactions with miRNA response elements (39). Furthermore, numerous miRNAs are known to play important roles in cancer cells by regulating key target genes (40), thus suggesting the possibility of a circRNA—miRNA–mRNA regulatory axis in HCC. We therefore investigated whether circRPN2 can function as a miRNA sponge in HCC and found that circRPN2 can bind to miR-183–5p, which acts as a tumor enhancer in many tumor types by directly targeting FOXO1 (41, 42). A number of previous studies have reported that FOXO1 plays an important glucose metabolism and functions as a tumor suppressor. As a critical downstream effector of STAT3 signaling, FOXO1 participates in glucose homeostasis, cell proliferation, and apoptosis (22). Notably, FOXO1/3/4 knockout mice show altered glucose metabolism compared with WT mice, including decreased gluconeogenesis and increased glycolysis (43). In epithelial cells, FOXO1 acts as a gatekeeper of endothelial quiescence, decreasing metabolic activity by reducing glycolysis through c-MYC signaling (20). In addition, FOXO1 upregulates PDK4 expression and to further inhibit glucose metabolism (44). In this study, we used the FOXO1 inhibitor (AS1842856) to confirm that the effects of circRPN2 on HCC glycolysis and progression are mediated by the circRPN2/miR-183–5p/FOXO1 axis, a finding that is consistent with the previously reported role of FOXO1.

Finally, to further investigate the relevance of circRPN2, ENO1, and FOXO1 in vivo, we examined expression of these molecules in clinical HCC samples and determined whether they could be used for prediction of patient outcomes and tumor recurrence. We found that expression of circRPN2 was negatively correlated with ENO1 and positively correlated with FOXO1 in these samples. Notably, patients with tumors expressing low levels of circRPN2, high levels of ENO1, or low levels of FOXO1 had a poorer prognosis and higher cumulative recurrence rates. Thus, our results indicate that circRPN2, ENO1, and FOXO1 may be used as significant indicators of HCC prognosis.

In conclusion, our cumulative results show that circRPN2 is downregulated in patients with HCC with metastasis or recurrence and acts as a key regulator of HCC aerobic glycolysis, progression, and metastasis via acceleration of ENO1 degradation and regulation of miR-183–5p/FOXO1 signaling. Thus, we propose that circRPN2 expression, either alone or in combination with FOXO1 and ENO1, may be an important a novel prognostic marker for HCC.

No disclosures were reported.

J. Li: Data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. Z.-Q. Hu: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. S.-Y. Yu: Data curation, software, formal analysis, validation, investigation. L. Mao: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. Z.-J. Zhou: Resources, software, supervision, methodology. P.-C. Wang: Resources, data curation, formal analysis, visualization. Y. Gong: Resources, software, validation, investigation, visualization, methodology. S. Su: Software, validation, investigation, visualization. J. Zhou: Resources, supervision, methodology, project administration. J. Fan: Resources, supervision, methodology, writing–review and editing. S.-L. Zhou: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, project administration, writing–review and editing. X.-W. Huang: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, project administration, writing–review and editing.

This study was jointly supported by the Major Special Projects of the Ministry of Science and Technology (2018ZX10302207) and National Natural Science Foundation of China (No. 91942313, 81773069, 81972708, 82072681, and 82003082).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Forner
A
,
Reig
M
,
Bruix
J
.
Hepatocellular carcinoma
.
Lancet North Am Ed
2018
;
391
:
1301
14
.
2.
Zhou
SL
,
Yin
D
,
Hu
ZQ
,
Luo
CB
,
Zhou
ZJ
,
Xin
HY
, et al
.
A positive feedback loop between cancer stem-like cells and tumor-associated neutrophils controls hepatocellular carcinoma progression
.
Hepatology
2019
;
70
:
1214
30
.
3.
Zhou
SL
,
Zhou
ZJ
,
Hu
ZQ
,
Huang
XW
,
Wang
Z
,
Chen
EB
, et al
.
Tumor-associated neutrophils recruit macrophages and T-regulatory cells to promote progression of hepatocellular carcinoma and resistance to sorafenib
.
Gastroenterology
2016
;
150
:
1646
58
.
4.
Chin
L
,
Gray
JW
.
Translating insights from the cancer genome into clinical practice
.
Nature
2008
;
452
:
553
63
.
5.
Li
Q
,
Pan
X
,
Zhu
D
,
Deng
Z
,
Jiang
R
,
Wang
X
.
Circular RNA MAT2B promotes glycolysis and malignancy of hepatocellular carcinoma through the miR-338–3p/PKM2 axis under hypoxic stress
.
Hepatology
2019
;
70
:
1298
316
.
6.
Hay
N
.
Reprogramming glucose metabolism in cancer: can it be exploited for cancer therapy?
Nat Rev Cancer
2016
;
16
:
635
49
.
7.
Yu
J
,
Xu
QG
,
Wang
ZG
,
Yang
Y
,
Zhang
L
,
Ma
JZ
, et al
.
Circular RNA cSMARCA5 inhibits growth and metastasis in hepatocellular carcinoma
.
J Hepatol
2018
;
68
:
1214
27
.
8.
Hu
ZQ
,
Zhou
SL
,
Li
J
,
Zhou
ZJ
,
Wang
PC
,
Xin
HY
, et al
.
Circular RNA sequencing identifies CircASAP1 as a key regulator in hepatocellular carcinoma metastasis
.
Hepatology
2020
;
72
:
906
22
.
9.
Wittekind
C
.
[Pitfalls in the classification of liver tumors]
.
Pathologe
2006
;
27
:
289
93
.
10.
Wan
SZ
,
Nie
Y
,
Zhang
Y
,
Liu
C
,
Zhu
X
.
Assessing the Prognostic performance of the child-pugh, model for end-stage liver disease, and albumin-bilirubin scores in patients with decompensated cirrhosis: a large Asian cohort from gastroenterology department
.
Dis Markers
2020
;
2020
:
5193028
.
11.
Chun
YS
,
Pawlik
TM
,
Vauthey
JN
.
8th edition of the AJCC cancer staging manual: pancreas and hepatobiliary cancers
.
Ann Surg Oncol
2018
;
25
:
845
7
.
12.
Zhou
SL
,
Dai
Z
,
Zhou
ZJ
,
Wang
XY
,
Yang
GH
,
Wang
Z
, et al
.
Overexpression of CXCL5 mediates neutrophil infiltration and indicates poor prognosis for hepatocellular carcinoma
.
Hepatology
2012
;
56
:
2242
54
.
13.
Wang
J
,
Maldonado
MA
.
The ubiquitin-proteasome system and its role in inflammatory and autoimmune diseases
.
Cell Mol Immunol
2006
;
3
:
255
61
.
14.
Dai
J
,
Zhou
Q
,
Chen
J
,
Rexius-Hall
ML
,
Rehman
J
,
Zhou
G
.
Alpha-enolase regulates the malignant phenotype of pulmonary artery smooth muscle cells via the AMPK–Akt pathway
.
Nat Commun
2018
;
9
:
3850
.
15.
Wu
Y
,
Zhang
Y
,
Qin
X
,
Geng
H
,
Zuo
D
,
Zhao
Q
.
PI3K/AKT/mTOR pathway-related long non-coding RNAs: roles and mechanisms in hepatocellular carcinoma
.
Pharmacol Res
2020
;
160
:
105195
.
16.
Zhou
SL
,
Hu
ZQ
,
Zhou
ZJ
,
Dai
Z
,
Wang
Z
,
Cao
Y
, et al
.
miR-28–5p-IL-34-macrophage feedback loop modulates hepatocellular carcinoma metastasis
.
Hepatology
2016
;
63
:
1560
75
.
17.
Futreal
PA
,
Coin
L
,
Marshall
M
,
Down
T
,
Hubbard
T
,
Wooster
R
, et al
.
A census of human cancer genes
.
Nat Rev Cancer
2004
;
4
:
177
83
.
18.
Malhotra
S
,
Alsulami
AF
,
Heiyun
Y
,
Ochoa
BM
,
Jubb
H
,
Forbes
S
, et al
.
Understanding the impacts of missense mutations on structures and functions of human cancer-related genes: a preliminary computational analysis of the COSMIC Cancer Gene Census
.
PLoS ONE
2019
;
14
:
e0219935
.
19.
Zhang
W
,
Patil
S
,
Chauhan
B
,
Guo
S
,
Powell
DR
,
Le
J
, et al
.
FoxO1 regulates multiple metabolic pathways in the liver: effects on gluconeogenic, glycolytic, and lipogenic gene expression
.
J Biol Chem
2006
;
281
:
10105
17
.
20.
Wilhelm
K
,
Happel
K
,
Eelen
G
,
Schoors
S
,
Oellerich
MF
,
Lim
R
, et al
.
FOXO1 couples metabolic activity and growth state in the vascular endothelium
.
Nature
2016
;
529
:
216
20
.
21.
Zheng
M
,
Cao
MX
,
Luo
XJ
,
Li
L
,
Wang
K
,
Wang
SS
, et al
.
EZH2 promotes invasion and tumour glycolysis by regulating STAT3 and FoxO1 signalling in human OSCC cells
.
J Cell Mol Med
2019
;
23
:
6942
54
.
22.
Zheng
M
,
Cao
MX
,
Yu
XH
,
Li
L
,
Wang
K
,
Wang
SS
, et al
.
STAT3 promotes invasion and aerobic glycolysis of human oral squamous cell carcinoma via inhibiting FoxO1
.
Front Oncol
2019
;
9
:
1175
.
23.
Nicolet
BP
,
Engels
S
,
Aglialoro
F
,
van den Akker
E
,
von Lindern
M
,
Wolkers
MC
.
Circular RNA expression in human hematopoietic cells is widespread and cell-type specific
.
Nucleic Acids Res
2018
;
46
:
8168
80
.
24.
Fan
L
,
Cao
Q
,
Liu
J
,
Zhang
J
,
Li
B
.
Correction to: circular RNA profiling and its potential for esophageal squamous cell cancer diagnosis and prognosis
.
Mol Cancer
2020
;
19
:
113
.
25.
Chaffer
CL
,
Weinberg
RA
.
A perspective on cancer cell metastasis
.
Science
2011
;
331
:
1559
64
.
26.
Zhou
SL
,
Zhou
ZJ
,
Hu
ZQ
,
Song
CL
,
Luo
YJ
,
Luo
CB
, et al
.
Genomic sequencing identifies WNK2 as a driver in hepatocellular carcinoma and a risk factor for early recurrence
.
J Hepatol
2019
;
71
:
1152
63
.
27.
Koppenol
WH
,
Bounds
PL
,
Dang
CV
.
Otto Warburg's contributions to current concepts of cancer metabolism
.
Nat Rev Cancer
2011
;
11
:
325
37
.
28.
Du
WW
,
Yang
W
,
Liu
E
,
Yang
Z
,
Dhaliwal
P
,
Yang
BB
.
Foxo3 circular RNA retards cell-cycle progression via forming ternary complexes with p21 and CDK2
.
Nucleic Acids Res
2016
;
44
:
2846
58
.
29.
Principe
M
,
Ceruti
P
,
Shih
NY
,
Chattaragada
MS
,
Rolla
S
,
Conti
L
, et al
.
Targeting of surface alpha-enolase inhibits the invasiveness of pancreatic cancer cells
.
Oncotarget
2015
;
6
:
11098
113
.
30.
Zhu
X
,
Yu
H
,
Li
B
,
Quan
J
,
Zeng
Z
,
Li
G
, et al
.
Targetting an LncRNA P5848–ENO1 axis inhibits tumor growth in hepatocellular carcinoma
.
Biosci Rep
2019
;
39
:
BSR20180896
.
31.
Yu
S
,
Li
N
,
Huang
Z
,
Chen
R
,
Yi
P
,
Kang
R
, et al
.
A novel lncRNA, TCONS_00006195, represses hepatocellular carcinoma progression by inhibiting enzymatic activity of ENO1
.
Cell Death Dis
2018
;
9
:
1184
.
32.
Capello
M
,
Ferri-Borgogno
S
,
Riganti
C
,
Chattaragada
MS
,
Principe
M
,
Roux
C
, et al
.
Targeting the Warburg effect in cancer cells through ENO1 knockdown rescues oxidative phosphorylation and induces growth arrest
.
Oncotarget
2016
;
7
:
5598
612
.
33.
Hsu
PP
,
Sabatini
DM
.
Cancer cell metabolism: Warburg and beyond
.
Cell
2008
;
134
:
703
7
.
34.
Yang
T
,
Shu
X
,
Zhang
HW
,
Sun
LX
,
Yu
L
,
Liu
J
, et al
.
Enolase 1 regulates stem cell-like properties in gastric cancer cells by stimulating glycolysis
.
Cell Death Dis
2020
;
11
:
870
.
35.
Jung
DW
,
Kim
WH
,
Park
SH
,
Lee
J
,
Kim
J
,
Su
D
, et al
.
A unique small-molecule inhibitor of enolase clarifies its role in fundamental biological processes
.
ACS Chem Biol
2013
;
8
:
1271
82
.
36.
Ma
Y
,
Zhang
D
,
Wu
H
,
Li
P
,
Zhao
W
,
Yang
X
, et al
.
Circular RNA PRKCI silencing represses esophageal cancer progression and elevates cell radiosensitivity through regulating the miR-186–5p/PARP9 axis
.
Life Sci
2020
;
259
:
118168
.
37.
DeBerardinis
RJ
,
Lum
JJ
,
Hatzivassiliou
G
,
Thompson
CB
.
The biology of cancer: metabolic reprogramming fuels cell growth and proliferation
.
Cell Metab
2008
;
7
:
11
20
.
38.
Pena
A
,
Kobir
A
,
Goncharov
D
,
Goda
A
,
Kudryashova
TV
,
Ray
A
, et al
.
Pharmacological inhibition of mTOR kinase reverses right ventricle remodeling and improves right ventricle structure and function in rats
.
Am J Respir Cell Mol Biol
2017
;
57
:
615
25
.
39.
Tay
Y
,
Rinn
J
,
Pandolfi
PP
.
The multilayered complexity of ceRNA crosstalk and competition
.
Nature
2014
;
505
:
344
52
.
40.
Croce
CM
.
Causes and consequences of microRNA dysregulation in cancer
.
Nat Rev Genet
2009
;
10
:
704
14
.
41.
Yan
R
,
Li
K
,
Yuan
D
,
Wang
H
,
Chen
W
,
Zhu
K
, et al
.
miR-183–5p promotes proliferation and migration in hepatocellular carcinoma by targeting IRS1 and its association with patient survival
.
Int J Biol Markers
2020
;
35
:
83
9
.
42.
Shang
Y
,
Sun
Y
,
Xu
J
,
Ge
X
,
Hu
Z
,
Xiao
J
, et al
.
Exosomes from mmu_ circ_0001359-modified ADSCs attenuate airway remodeling by enhancing FoxO1 signaling-mediated M2-like macrophage activation
.
Mol Ther Nucleic Acids
2020
;
19
:
951
60
.
43.
Xiong
X
,
Tao
R
,
DePinho
RA
,
Dong
XC
.
Deletion of hepatic FoxO1/3/4 genes in mice significantly impacts on glucose metabolism through downregulation of gluconeogenesis and upregulation of glycolysis
.
PLoS ONE
2013
;
8
:
e74340
.
44.
Piao
L
,
Sidhu
VK
,
Fang
YH
,
Ryan
JJ
,
Parikh
KS
,
Hong
Z
, et al
.
FOXO1-mediated upregulation of pyruvate dehydrogenase kinase-4 (PDK4) decreases glucose oxidation and impairs right ventricular function in pulmonary hypertension: therapeutic benefits of dichloroacetate
.
J Mol Med
2013
;
91
:
333
46
.

Supplementary data