HDAC2 knockdown induces cell death in GBM
We first investigated the expression of HDAC2 in human glioma patients via analysis of the GEPIA and TCGA public databases. HDAC1 and HDAC2 were significantly increased in GBM compared to normal tissues (Fig. 1A, B). Especially, HDAC2 was highly increased in GBM as indicated by TCGA Brain statistics (Fig. 1C). In the Shai Brain from TCGA database, HDAC2 was highly expressed in all brain tumors, including astrocytomas, oligodendrogliomas, and glioblastomas (Fig. 1D). Furthermore, we confirmed protein levels of class I HDACs in 8 GBM cells and normal brain cells, and HDAC2 proteins were highly expressed in GBM cells (Additional file 1: Fig. S1A and B). We also analyzed cell viability by using shRNA library (five unique shRNA) targeting 18 types of HDACs in various GBM cells, identifying HDAC2 knockdown was the most effective in promoting GBM cell death (Fig. 1E), and cell viability was also reduced to less than 50% over 6 days (Additional file 1: Fig. S1C). The knockdown of HDACs by lentiviral shRNA was measured and confirmed cell viability (Additional file 1: Fig. S1D and E). As expected, HDAC2 knockdown GBM cells underwent apoptosis more frequently than control cells. Collectively, HDAC2 is might be an essential factor in GBM tumorigenesis and could target a novel epigenetic therapeutic strategy in GBM patients.
To consider HDAC2 is necessary for GBM cell survival, we made DOX-inducible shHDAC2 GBM stable cells that expressed shRNA targeting HDAC2 upon doxycycline treatment (Additional file 1: Fig. S1F). We determined whether HDAC2 knockdown induces GBM cell death using western blot. As a result, the expression of cleaved-PARP, cleaved caspase-3, Bax, Apaf-1, and p21 increased (Fig. 1F), and cell proliferation decreased in DOX-inducible shHDAC2 GBM cells upon doxycycline treatment (Fig. 1G). We also confirmed that HDAC2 is expressed in the nucleus of GBM cells by immunofluorescence (IF) analysis (Additional file 1: Fig. S1G). GBM cell death significantly increased in DOX-inducible shHDAC2 GBM cells upon doxycycline by FACS analysis (Fig. 1H) and TUNEL assay (Fig. 1I), indicating HDAC2 knockdown increased GBM cell death.
Next, we confirmed whether HDAC2 knockdown enhanced apoptotic genes’ transcriptional activity to induce GBM cell death using qPCR. Bax, Puma, and HtrA2 mRNA significantly increased in DOX-inducible shHDAC2 GBM cells upon doxycycline treatment (Fig. 1J). We also validated the transcriptional activity of Puma to determine the functional significance of HDAC2 using luciferase activity assay. The luciferase activity of Puma increased in HDAC2 knockdown GBM cells compared with control GBM cells or Dox-untreated shHDAC2 GBM cells (Additional file 1: Fig. S1H). Additionally, we tested whether Romidepsin, a selective HDAC1/2 inhibitor, affected anti-tumorigenic effects in GBM cells. Interestingly, the cell viability was remarkably decreased in HDAC2 knockdown GBM cells by Romidepsin treatment (Additional file 1: Fig. S1I). Taken together, HDAC2 knockdown regulates apoptosis and anti-proliferation in GBM cells, suggesting that HDAC2 plays an important role in the development and progression of GBM.
HDAC2 knockdown GBM induces cell death by controlling the expression of miR-3189 and GLUT3
Recently, HDAC2 silencing was reported to suppress proliferation and tumorigenesis of GBM [20], but the precise molecular mechanism is unknown. Therefore, to further investigate the physiological relevance of HDAC2 in GBM, we verified the correlation between HDAC2 knockdown and GBM cell death in HDAC2 knockdown A172 cells by RNA-sequencing analysis (Fig. 2A). HDAC2 knockdown efficiently decreased the subset of genes encoding glucose transporter proteins required for glucose metabolism, and GLUT3 was significantly downregulated. Thus, we examined whether HDAC2 knockdown inhibits GLUT3 expression by Immunofluorescence analysis, and GLUT3 expression significantly suppressed in HDAC2 knockdown GBM cells (Additional file 1: Fig. S2A). To confirm the clinical relevance, we generated Kaplan-Meier curves from “Freije”, “Vital” and “Gravendee” datasets. GLUT3 expression poorly affects the survival rate for all datasets (Additional file 1: Fig. S2B). To understand how HDAC2 knockdown suppressed GLUT3 expression and induced GBM cell death, we investigated GLUT3-targeting transcriptional regulators associated with HDAC2 expression from the RNA-sequencing dataset and analyzed miRNA gene expression profiles that regulated GLUT3 expression (Fig. 2B). We found that miR-3189 contained a complementary sequence to the GLUT3–3’UTR, which might inhibit GLUT3 expression (Fig. 2C). To verify this binding potential of miR-3189, we performed miRNA target prediction analysis using the MiRanda, and TargetScan databases. Interestingly, miR-3189-mediated GLUT3 expression has not been reported in GBM and other tumors. We also validated whether GLUT3 mRNA expression could be regulated in miR-3189-mimics transfected GBM cells (Fig. 2D). miR-3189 strongly repressed GLUT3 transcription and siHDAC2 also showed the same results (Additional file 1: Fig. S2C). Thus, HDAC2 knockdown induced GBM cell death via miR-3189-mediated GLUT3 repression. We investigated whether the expression of HDAC2 and GLUT3 was upregulated in GBM tissues (human GBM TMA: US Biomax, Derwood, USA) using IHC. As expected, HDAC2 increased in the nucleus and GLUT3 increased in the cytoplasm in human GBM patient tissues compared to normal brain tissues (Fig. 2E). Bax and Apaf-1 decreased in human GBM patient tissues (Fig. 2F), suggesting that both HDAC2 and GLUT3 positively contribute to GBM progression.
Next, to determine whether HDAC2 knockdown meaningfully inhibit GBM progression in an in vivo preclinical mouse models as in vitro, we orthotopically xenografted DOX-inducible control and DOX-inducible shHDAC2 U87MG cells into immune-deficient BALB/Cnu/nu mice and administrated the drinking water containing doxycycline after 8 days from the experiment starts (Fig. 2G). We observed that while the bodyweight (Additional file 1: Fig. S2D) and survival rate (Fig. 2H) remained steady in DOX-inducible shHDAC2 U87MG-injected mice by doxycycline treatment, and rapidly decreased in DOX-untreated mice and control mice from the fourth week. Also, the tumor growth of Dox-treated shHDAC2 GBM mice was inhibited than Dox-untreated shHDAC2 GBM mice and control mice using H&E histological analysis (Fig. 2I). In addition, MRI images (Fig. 2J) and PET images (Fig. 2K) (SUV: Standardized Uptake Value; mice red image ratio) were compared by GBM tumor scans. Ki67 decreased in HDAC2 knockdown mice GBM brain upon doxycycline treatment compared to control mice brain or DOX-untreated shHDAC2 mice GBM brain using IHC analysis, indicating that HDAC2 knockdown significantly inhibited GBM tumorigenesis (Fig. 2L). Whereas Apaf-1 and Bax increased in mice GBM tissues (Additional file 1: Fig. S2E). As expected, GLUT3 mRNA expression decreased in HDAC2 knockdown mice GBM tissues (Fig. 2M), and HDAC2 expression inhibited in Dox-treated shHDAC2 GBM tissues (Additional file 1: Fig. S2F). Furthermore, we investigated whether miR-3189 expression was increased in DOX-treated HDAC2 knockdown mice GBM tissues upon doxycycline treatment (Fig. 2N). miR-3189 expression was significantly increased by HDAC2 knockdown, suggesting an inverse correlation between miR-3189 and GLUT3 expression in GBM. Taken together, HDAC2 regulates GBM tumorigenesis by controlling miR-3189 and GLUT3 expression.
miR-3189 inhibits tumor growth in orthotopic mouse GBM model
We also investigated whether miR-3189 expression decides the cell fate in GBM cells through analysis of cell viability in miR-3189-overexpressing GBM cells. miR-3189 overexpression inhibited GBM cell growth and activated cell death processes (Fig. 3A and B). Meanwhile, miR-3189-expressing or GLUT3 knockdown GBM cells increased PARP cleavage (Fig. 3C). miR-3189 expectedly suppressed GLUT3 expression and strikingly promoted the expression of Apaf-1, cleaved caspase-3, and Bax (Fig. 3D). Thus, we showed that miR-3189-mediated GBM cell death is dependent on the downregulation of GLUT3 expression.
To determine whether miR-3189 expression effectively inhibits GBM progression in orthotopic mouse models, similar to HDAC2 knockdown, we injected U87MG cells into the brain of immune-deficient BALB/Cnu/nu mice, and after 6 days of GBM cell injection, JetPEI-miR-3189 was directly injected in tumor sites (Fig. 3E). The next day after miR-3189 injection, we scan PET-CT images on the 1st day (1st PET-CT) and 14th day (2nd PET-CT). We observed the bodyweight of U87MG-injected mice by miR-3189 treatment (Fig. 3F) and were steadily maintained by miR-3189 treatment. Whereas miR-3189 untreated mice rapidly decreased from 24 days.
To measure the in vivo efficacy of miR-3189 on tumor growth in GBM mouse models, MRI images and PET images (Fig. 3G) were compared by tumor scans at 0 days and 14 days after miRNA injection. SUV ratio of the red image in mice brain highly increased in miR-control treated mice but not in miR-3189 treated mice (Fig. 3H), indicating that miR-3189 effectively inhibited GBM tumorigenesis. Also, the tumor growth in U87MG injected mice by miR-3189 treatment was inhibited more than miR-control treatment by H&E histological analysis (Fig. 3I). GLUT3 expression was confirmed by western blot (Fig. 3J) and has significantly decreased in U87MG-injected mice brain tissues upon miR-3189 treatment compared to miR-3189 untreated mice using IHC analysis. Bax increased by miR-3189 treatment (Fig. 3K). Additionally, we validated mRNA expression of miR-3189, GLUT3, Bax, and Apaf-1 upon miR-3189 treatment using qPCR. As previous results, GLUT3 mRNA expression decreased in miR-3189-treated U87MG-injected mice brain, whereas the mRNA expression of Bax and Apaf-1 increased (Fig. 3L). Therefore, miR-3189 significantly decreased GBM tumorigenesis by targeting GLUT3 expression in GBM mouse models.
miR-3189 induced GBM cell death via the transcriptional repression of GLUT3
GLUT3 is the essential glucose transporter involved in brain glucose uptake, and its role is well-documented in GBM metabolism. First, we observed GBM survival in GLUT3 knockdown GBM cells by FACS analysis. GLUT3 knockdown induced GBM cell death similar to HDAC2 knockdown or miR-3189 overexpression, and the frequency of apoptosis by FACS analysis was increased in all early (Q2) and late (Q4) stages (Fig. 4A). In addition, we confirmed that the apoptotic cells increased upon GLUT3 knockdown using TUNEL assay (Fig. 4B), and GLUT3 knockdown significantly decreased the colony formation of GBM cells compared to control GBM cells, and miR-3189 also showed the same results (Fig. 4C).
We analyzed whether GLUT3 knockdown or miR-3189 overexpression increased cell death markers’ expression using qPCR. As expected, Pro-apoptosis genes increased in GLUT3 knockdown GBM cells and miR-3189-expressing GBM cells (Additional file 1: Fig. S3A and B). Besides, to validate that miR-3189 directly regulated GLUT3, we performed luciferase reporter assays with miR-3189 mimics and pmirGLO plasmids bearing wild-type or mutant GLUT3 3′-UTR sequences of putative miR-3189 binding sites (Fig. 4D). These results showed that miR-3189 dramatically repressed luciferase activity of pmirGLO-GLUT3wt containing the miR-3189 binding site from the wild type GLUT3 3′-UTR; however, luciferase activity upon pmirGLO-GLUT3mt did not repress by miR-3189 in 293 T and GBM cells (Fig. 4E), indicating that miR-3189 can significantly inhibit GLUT3 expression via binding to GLUT3 3′-UTR. Because HDAC2 knockdown was highly influential in inducing GBM cell death to provide the direct evidence that HDAC2 knockdown induced GLUT3-mediated cell death via miR-3189 upregulation, we investigated the luciferase activity of pmirGLO-GLUT3wt in HDAC2 knockdown GBM cells by doxycycline treatment (Fig. 4F). HDAC2 knockdown effectively decreased the luciferase activity of pmirGLO-GLUT3wt to induce GBM cell death and remained ineffective in repressing luciferase activity in pmirGLO-GLUT3mt-expressing GBM cells. These results suggest that HDAC2 knockdown increased GBM cell death via inhibition of miR-3189-mediated GLUT3 expression.
HDAC2 repression controls the metabolism and proliferation of GBM
In many studies, the reduced glucose uptake and lactate production in glucose metabolism has been known to inhibit tumor cell viability.15 Therefore, to consider whether cell death and proliferation in GBM cells might correlate with glucose metabolism, we examine that the repression of GLUT3 by HDAC2 knockdown induces GBM cell death via inhibition of glucose metabolism. Glucose uptake significantly decreased in DOX-inducible shHDAC2 GBM cells with doxycycline treatment (Fig. 5A) and confirmed the same effect in the HDAC2 siRNA treatment (Additional file 1: Fig. S4A). Indeed, HDAC2 knockdown resulted in reduced glucose uptake and lactate production in GBM cells (Additional file 1: Fig. S4B). Therefore, our results strongly propose that HDAC2 might serve as a master regulator of GBM cell death via regulation of glucose metabolism by miR-3189-mediated GLUT3 expression (Fig. 5B).
We next sought to confirm the functional relevance of reduced glucose metabolism and GBM cell death. Many studies reported that GLUT3 is highly expressed in GBM and contributes to the growth of brain tumors [17]. Because miR-3189 has not been investigated in all cancer-contained brain tumors, to assess the importance of the apoptotic effect of GLUT3 knockdown and miR-3189 expression, we transfected GLUT3 siRNA or miR-3189 mimics into DOX-inducible shHDAC2 GBM cells. Our results definitively show that GLUT3 siRNA-transfected shHDAC2-expressing GBM cells significantly increased PARP cleavage and Bax compared with DOX-inducible control GBM cells or DOX-untreated shHDAC2 GBM cells. Equally important, miR-3189-transfected shHDAC2-expressing GBM cells increased GBM cell death via downregulation of GLUT3 (Fig. 5C and D), suggesting that HDAC2 expression-dependent cell death might occur by miR-3189-mediated GLUT3 inhibition. We also investigated cellular proliferation by expressing GLUT3 siRNA, miR-3189 mimics, and HDAC2 siRNA in GBM cells, confirming that both GLUT3 knockdown and miR-3189 overexpression decreased GBM proliferation, similar to HDAC2 knockdown (Additional file 1: Fig. S4C).
To demonstrate whether GLUT3 downregulation reduced glucose uptake and lactate production to cause GBM cell death, we measured glucose uptake, lactate production, and cell proliferation in DOX-inducible shHDAC2 GBM cells. GLUT3 siRNA or miR-3189 mimics synergistically decreased the glucose uptake (Fig. 5E and F) and lactate production (Fig. 5G and H) in shHDAC2-expressing GBM cells than individually transfected GBM cells, but the change of glucose uptake and lactate production did not observe in GLUT1 or GLUT2 knockdown GBM cells (Additional file 1: Fig. S4D and E). Similarly, the cellular proliferation was decreased in the same condition (Fig. 5I and J), suggesting that the combined treatment of GLUT3 siRNA or miR-3189 remarkably reduced these metabolic changes and cellular proliferation in HDAC2 knockdown GBM cells. We also investigated the Extracellular Acidification Rate (ECAR) and Oxygen Consumption Rate (OCR) in DOX-inducible shHDAC2 GBM cells with doxycycline using a Seahorse extracellular flux analyzer. Both ECAR and OCR significantly decreased in HDAC2 knockdown GBM cells (Fig. 5K) but not in control GBM cells (Fig. 5L). These results strongly support that HDAC2 knockdown was directly associated with metabolite regulation in mitochondrial respiration and glycolysis via GLUT3 inhibition and ultimately induced GBM cell death.
HDAC2 knockdown increases cell death and decreases tumor-sphere formation in GSCs
Most GBM consists of mixed glioma cells and glioma stem cells (GSCs) associated with tumorigenesis and resistance to common therapies in GBM [6]. GSCs also have tumor-initiating, self-renewing properties and the unique ability to grow in microenvironments with limited nutrients [17]. GSCs can promote cancer recurrence and drug resistance by evading cell death [7, 12]. Thus, the discovery of target genes and metabolites characteristic of GSCs is an essential step to enhance apoptosis in designing therapeutic strategies to treat GBM.
To understand these most aggressive and therapeutic-resistant GBM cells, we analyzed the role of HDAC2 in regulating cell death by inhibiting miR-3189-mediated GLUT3 in DOX-inducible shHDAC2 GSCs. HDAC2 knockdown GSCs (GSC20, GSC23, GSC28, and GSC267) significantly decreased glucose uptake levels (Fig. 6A) and increased cleaved PARP (Fig. 6B) upon doxycycline treatment, these results displayed the same results in GBM cells. Importantly, HDAC2 knockdown also inhibited the tumor-sphere formation of GSCs (Fig. 6C and D), indicating that the survival of GSCs was directly regulated by HDAC2 expression level. Thus, we expected that GSCs would be highly sensitive to miR-3189 expression that inhibits GLUT3 expression. We also assessed the cell death effect in miR-3189-expressing GSCs. miR-3189 overexpression dramatically decreased the cell viability of GSCs (Fig. 6E). Indeed in these results, overexpressing miR-3189 repressed GLUT3 transcription in GSCs (Fig. 6F), suggesting that inhibition of miR-3189-mediated GLUT3 reduced tumor growth and cell viability not only in GBM cells but in GSCs.
To further determine whether the direct binding of miR-3189 to the GLUT3 3′-UTR region in GSCs, we observed the transcriptional activity of GLUT3 in pmirGLO-GLUT3wt or pmirGLO-GLUT3mt transfected DOX-inducible shHDAC2 GSCs by using luciferase assay. HDAC2 knockdown GSCs upon doxycycline treatment significantly decreased luciferase activity of GLUT3wt 3′-UTR, but not GLUT3mt 3′-UTR (Fig. 6G), and decreased GSCs proliferation (Fig. 6H). Additionally, we confirmed whether miR-3189 affects the gene expression of HDAC2 or other miRNAs (Additional file 1: Fig. S5A and B). Both expressions of HDAC2 and miRNAs did not change by miR-3189, expecting miR-3189 is involved in the downstream regulatory pathway of HDAC2. These results show that HDAC2 knockdown increased miR-3189 expression which was also recruited to the GLUT3 3′-UTR to inhibit GLUT3 expression, suggesting that HDAC2 knockdown inhibits tumorigenesis and GSC-sphere formation by inducing GSC cell death via miR-3189-mediated GLUT3 downregulation (Fig. 6I). Collectively, HDAC2 is a critical GBM/GSC progression marker and an ideal candidate for targeted therapy.