MiR-22/GLUT1 Axis Induces Metabolic Reprogramming and Sorafenib Resistance in Hepatocellular Carcinoma

Abstract

The approval of immunotherapy has revolutionized the management of hepatocellular carcinoma (HCC) patients. However, sorafenib remains a first-line therapeutic option for advanced patients and, in particular, for patients not eligible for immune checkpoint inhibitors, but its efficacy is limited by the onset of acquired resistance, highlighting the urgent need for predictive biomarkers. This study investigates the role of miR-22 in metabolic reprogramming and its potential as a biomarker in HCC. The analysis of miR-22 expression was performed in HCC patients and preclinical models by qPCR. Functional analyses in HCC cells evaluated GLUT1 as a direct miR-22 target. Cellular and metabolic assays evaluated the miR-22/GLUT1 axis's role in metabolic changes, tumor aggressiveness, and sorafenib response. Circulating miR-22 was analyzed in sorafenib-treated HCC patients and rats. MiR-22 was downregulated in HCCs and associated with aggressive tumor features. Functionally, miR-22 modulated the HIF1A pathway, enhanced survival in stressful conditions, promoted a glycolytic shift, and enhanced cancer cell plasticity and sorafenib resistance via GLUT1 targeting. In addition, high serum miR-22 levels were associated with sorafenib resistance in HCC patients and rats. GLUT1 inhibition sensitized low miR-22-expressing HCC cells to sorafenib in preclinical models. These findings suggest that circulating miR-22 deserves attention as a predictive biomarker of sorafenib response. GLUT1 inhibition may represent a therapeutic strategy to combine with sorafenib, particularly in patients exhibiting high circulating miR-22 levels.

Keywords: GLUT1; HCC; miR-22; sorafenib.

Figure 1

Deregulated expression of miR-22 in human and rat HCCs and association with clinicopathological features. (A,B) Box plot graphs of miR-22 expression in HCC and surrounding livers from the Bologna cohort (N = 28) and DEN-HCC rats (N = 15). The Y-axes report the 2^−∆∆Ct values corresponding to the miR-22 expression. (C) Correlation graph between the miR-22 and AFP mRNA levels in tumor nodules (N = 24) of DEN-HCC rats. Axes report the 2^−∆∆Ct values corresponding to the miR-22 and AFP levels transformed in a log2 form. (D) Box plot graph of miR-22 expression in HCCs from the Bologna cohort (N = 25) divided according to the presence or absence of microvascular invasion (MVI). The Y-axis reports 2^−∆∆Ct values corresponding to miR-22 expression. (E) Box plot graph of miR-22 expression in HCCs from the Bologna cohort (N = 28) according to Edmondson–Steiner tumor grade. The p-value relative to ANOVA is shown on top of the graph. Statistically significant comparisons between groups are G2 versus G3, p < 0.05; G2 versus G4, p < 0.01 (Tukey’s post hoc test). The Y-axis reports 2^−∆∆Ct values corresponding to miR-22 expression. (F) Box plot graph of miR-22 expression in HCCs from the Bologna cohort (N = 25) divided according to the presence or absence of TP53 mutations. The Y-axis reports 2^−∆∆Ct values corresponding to miR-22 expression. (G) Kaplan–Meier curves of high and low miR-22-expressing HCCs from TCGA cohort. (H) Pathway enrichment analysis of high versus low miR-22-expressing HCCs from TCGA cohort. (A–F) U6RNA and GAPDH or Beta actin were used as housekeeping genes for miRNA and gene quantification, respectively. Real-Time PCR was run in triplicate. Two-tailed unpaired Student’s t-test and Pearson’s correlation were used for comparisons among the two groups.

Figure 2

MiR-22 regulates 2D and 3D cell growth and the HIF-1A pathway in HCC cells. (A) Real-Time PCR analysis of miR-22 expression in stably silenced (MZIP-22) and control (shRNA) HepG2 cells and in stably overexpressing (pMXs-22) and control (shRNA) Huh-7 cells. The Y-axes report the 2^−∆∆Ct values corresponding to miR-22 expression normalized to controls. Mean ± SD values are displayed. U6RNA was used as a housekeeping gene. Real-Time PCR analysis was performed in two independent experiments in triplicate. (B) Growth curves of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related control (shRNA) cells. The growth curves were normalized to T0. Mean ± SD values are reported. Two independent experiments were performed in quadruplicate. WB analysis of apoptotic markers (CASP3 and BAX) in the same cell lines. WB analysis was performed in two independent experiments, and GAPDH was used as a housekeeping gene. (C) Representative images (4× magnification) and related histograms of the miR-22-silenced (MZIP-22) HepG2 spheroids and miR-22-overexpressing (pMXs-22) Huh-7 spheroids and related controls (shRNA) at 96 and 48 h, respectively. The Y-axes report the Feret’s diameter (µm) normalized to controls. Thirty randomly selected spheroids were measured in two independent experiments. Mean ± SD values are displayed. Scale bars: 750 μm. WB analysis of apoptotic genes was performed in the same spheroids. GAPDH was used as a housekeeping gene. WB analysis was performed in two independent experiments. (D) Real-Time PCR and WB analyses of stemness-related genes in the miR-22-silenced (MZIP-22) HepG2 cells (graph above) and miR-22-overexpressing (pMXs-22) Huh-7 cells (graph below) and related controls (shRNA). The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time was performed twice in triplicate, WB was performed in two independent experiments. (E) Real-Time PCR and WB analyses of HIF-1A pathway in miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR and WB analyses were performed in two independent experiments in triplicate and duplicate, respectively. (F) WB analysis of HIF-1A in the subcellular compartments of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). Lamin B and GAPDH were used as housekeeping genes in nuclear and cytoplasmic compartments, respectively. The analysis was performed in two independent experiments. (G) Growth curves of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) treated with CoCl2 (100 nM). The growth curves were normalized to T0. Mean ± SD values are reported. Two independent experiments were performed in quadruplicate. (H) Cell viability and caspase assays in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) treated with CoCl2 (100 nM). The Y-axes report chemiluminescent signals normalized to controls. Mean ± SD values are displayed. Two independent experiments were performed in quadruplicate. WB analysis of apoptotic genes was performed in the same cells. GAPDH was used as a housekeeping gene. The analysis was performed in two independent experiments. *, **, ***, **** mean p < 0.05, p < 0.01, p < 0.001, p < 0.0001, respectively.

Figure 3

MiR-22 regulates migration, tumorigenesis, and angiogenesis in preclinical models. (A) Wound healing assays and related histograms of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). Representative images at initial (T0) and final (T96 or T48 h) times of wound closure are displayed. The Y-axes report the percentage of wound closure at final times with respect to T0. Mean ± SD values are reported. Two independent experiments were performed in duplicate. Real-Time PCR and WB analyses of SNAI1 in the same cell lines. The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR and WB analyses were performed in two independent experiments in triplicate and duplicate, respectively. (B) Box plot graphs representing in vivo tumorigenesis of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) in immunocompromised mice (N = 5 per group) at intermediate timepoints (3 and 6 weeks after cells injection, respectively). The Y-axes report the tumor volume (mm³) measured by a caliper (1/2 D×d²). (C) Real-Time PCR analysis of the HIF1A pathway in xenograft mice obtained following subcutaneous injection of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR analysis was performed in triplicate. (D) Real-Time PCR analysis of SNAI1 in the same xenograft models. The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR analysis was performed in triplicate. (E) WB analysis of apoptotic markers (CASP3, BAX) in xenograft mice obtained following the subcutaneous injection of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). GAPDH was used as a housekeeping gene. The Y-axes report the 2^−∆∆Ct values corresponding to protein expression normalized to controls. Mean ± SD values are displayed. (F) Real-Time PCR analysis of CASP3 expression in the same xenograft models. The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR analysis was performed in triplicate. *, **, ***, mean p < 0.05, p < 0.01, p < 0.001, respectively.

Figure 4

GLUT1 is a target of miR-22 in HCC and regulates cell metabolism. (A) WB analysis of GLUT1 expression in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). GAPDH was used as a housekeeping gene; WB analysis was performed in two independent experiments. GLUT1 molecular weight (MW) is from 45 to 60 kDa, and all the bands in this MW range have been considered for the analysis. (B) Dual-luciferase activity assay of wild-type (WT) and mutant (MUT) GLUT1-3′UTR vectors co-transfected with miR-22 in HepG2 and Huh-7 cells. NC: negative control precursor miRNA. The Y-axes report the Firefly/Renilla ratio normalized to controls (NC). Mean ± SD values are displayed. Analysis was performed in two independent experiments in triplicate. (C) WB analyses of GLUT1 expression in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) grown in acidic conditions (pH 6.8). GAPDH was used as a housekeeping gene; WB analysis was performed in two independent experiments. GLUT1 molecular weight (MW) is from 45 to 60 kDa, and all the bands in this MW range have been considered for the analysis. (D) Cell viability assay of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) grown in acidic conditions (pH 6.8). Rescue experiment was performed in the miR-22-silenced (MZIP-22) HepG2 cells treated with GLUT1-inhibitor BAY-876 (5 µM, 72 h) and grown in acidic conditions (pH 6.8). Vehicle: DMSO. Y-axes report chemiluminescent signals normalized to controls. Mean ± SD values are displayed. Two independent experiments were performed in quadruplicate. (E) Enzymatic assay measuring glucose uptake and HPLC analysis measuring extracellular lactate in the control (shRNA) and miR-22-silenced HepG2 cells (MZIP-22) at different time points (24, 48, 72 h). Data were normalized to protein content obtained at each time point. Mean ± SD values are displayed. Two independent experiments were analyzed in triplicate. (F) Representative images (40× magnification) and relative histograms of PAS staining in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). Rescue experiment was performed in miR-22-silenced (MZIP-22) HepG2 cells treated with GLUT1-inhibitor BAY-876 (5 µM, 48 h). Vehicle: DMSO. The Y-axes report the percentage of PAS-positive cell area normalized to control. Mean ± SD values are displayed. Ten randomly selected fields were analyzed from two independent experiments in triplicate. Scale bars: 50 μm. (G) Lipid droplet (LD) accumulation in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) stained with Nile Red. The Y-axes show the quantification of LD number per cell normalized to control. Mean ± SD values are reported. Two independent experiments were performed in triplicate. (H) Real-Time PCR analysis of FASN and ACLY in the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA). The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR analysis was performed in two independent experiments in triplicate. (I) Growth curves of the miR-22-silenced (MZIP-22) and control HepG2 cells (graph above) or miR-22-overexpressing (pMXs-22) and control Huh-7 cells (graph below) grown in culture medium without glucose. The growth curves were normalized to T0. Mean ± SD values are reported. Two independent experiments were performed in quadruplicate. (J) Correlation graphs between the miR-22 and GLUT1 mRNA levels in HCC patients from Bologna and “LIHC” cohorts and in tumor nodules of DEN-HCC rats. The axes report the 2^−∆∆Ct values corresponding to the miR-22 and GLUT1 levels transformed in a log2 form. U6RNA and GAPDH were used as housekeeping genes for miRNA and gene quantification, respectively. Real-Time PCR analysis was run in triplicate. *, **, ***, **** mean p < 0.05, p < 0.01, p < 0.001, p < 0.0001, respectively.

Figure 5

MiR-22 downregulation is associated with sorafenib resistance in HCC. (A) Growth curves of the miR-22-silenced (MZIP-22) HepG2 cells and miR-22-overexpressing (pMXs-22) Huh-7 cells and related controls (shRNA) under sorafenib treatment. Rescue experiment in the miR-22-silenced (MZIP-22) HepG2 cells treated with the GLUT1-inhibitor BAY-876 or vehicle (DMSO) and sorafenib. The growth curves were normalized to T0. Mean ± SD values are reported. Two independent experiments were performed in quadruplicate. WB analysis of GLUT1 expression in the same setting. WB was performed in two independent experiments, and GAPDH was used as a housekeeping gene. (B,C) Box plot graph of the miR-22 (B) and GLUT1 (C) levels in responder (N = 8) and non-responder (N = 7) HCC nodules from sorafenib-treated rats. The Y-axis reports the 2^−∆∆Ct values corresponding to the miR-22 or GLUT1 levels. Correlation graph between miR-22 (B) or GLUT1 (C) expression and tumor volume of sorafenib-treated DEN-HCC rats. The axes report the 2^−∆∆Ct values corresponding to mRNA levels and tumor size (mm³) of HCC nodules transformed in a log2 form. Beta-actin or U6RNA were used as housekeeping genes. Real-Time PCR analysis was run in triplicate. (D) Correlation graphs between the miR-22 and GLUT1 or CASP3, PUMA, and BMF mRNA levels in HCC nodules from DEN-HCC rats treated with sorafenib (N = 15). The axes report the 2^−∆∆Ct values corresponding to the mRNA levels transformed in a log2 form. Beta-actin or U6RNA were used as housekeeping genes. Real-Time PCR analysis was run in triplicate. (E) Box blot graph of tumor volume in the miR-22-silenced (MZIP-22) or control (shRNA) HepG2 xenograft mice (N = 5 per group) treated with sorafenib (60 mg/kg). (F,G) Real-Time PCR of HIF1A target genes, angiogenic markers (F), and SNAI1 (G) in the same animal model. The Y-axes report the 2^−∆∆Ct values corresponding to gene expression normalized to controls. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. Real-Time PCR analysis was run in triplicate. *, *** mean p < 0.05, p < 0.001, respectively.

Figure 6

MiR-22 circulating levels predict sorafenib resistance in human and rat HCCs. (A) Box plot graph of GLUT1 expression in early HCCs from the Bologna cohort divided based on serum miR-22 levels (high, N = 11; low, N = 10). The Y-axis reports the 2^−∆∆Ct values. Mean ± SD values are displayed. GAPDH was used as a housekeeping gene. (B) Correlation graph between the tissue and serum miR-22 expression levels in early HCCs from the Bologna cohort. The axes report the 2^−∆∆Ct values corresponding to the miR-22 tissue and serum levels. U6RNA or cel-miR-39 were used for data normalization. Real-Time PCR analysis was run in triplicate. (C) Extracellular and intracellular miR-22 levels in HCC cell lines treated with sorafenib (5 µM, 48 h). The Y-axis reports the fold change of the 2^−∆∆Ct values between the treated and untreated cells. Cel-miR-39 and U6RNA were used for data normalization. Real-Time PCR analysis was run in triplicate. The dashed lined indicated the positive of negative fold change of intracellular and extracellular miRNA levels. (D,E) Correlation graph between the serum and tissue miR-22 levels (D) or tumor size (E) in DEN-HCC rats treated with sorafenib (N = 12). The axes report the 2^−∆∆Ct values corresponding to the miR-22 serum and tissue levels, or tumor volume (mm³) transformed in a log2 form. U6RNA and cel-miR-39 were used for data normalization. Real-Time PCR analysis was run in triplicate. (F) Box plot graph of the miR-22 levels in responder (N = 6) and non-responder (N = 6) HCC nodules from sorafenib-treated rats. The Y-axis reports the 2^−∆∆Ct values corresponding to the miR-22 levels. Cel-miR-39 was used as spike-in miRNA for data normalization. Real-Time PCR analysis was run in triplicate. (G) Correlation graphs between the serum miR-22 and CASP3, PUMA, and BMF mRNA levels in HCC nodules from DEN-HCC rats treated with sorafenib (N = 12). The axes report the 2^−∆∆Ct values corresponding to the serum miR-22 and mRNA levels of HCC nodules transformed in a log2 form. Beta-actin or cel-miR-39 was used for data normalization. Real-Time PCR analysis was run in triplicate. (H) Box plot graph of the miR-22 serum levels in responder (N = 31) and non-responder (N = 21) sorafenib-treated patients from the Bologna cohort. The Y-axis reports the 2^−∆∆Ct values corresponding to circulating miR-22 levels. Cel-miR-39 was used as spike-in miRNA for data normalization. Real-Time PCR analysis was run in triplicate.