Background
Human malignant gliomas are the most common and lethal primary brain tumor in adults [
1,
2], and glioma cells are featured as carrying heterogeneous genetic molecular aberrations [
3‐
5]. Despite the application of advanced chemotherapy, radiotherapy, and surgery, patients with this disease suffer a badly median survival [
6‐
8]. Hence, deeper elucidation of the molecular mechanisms underlying glioma malignancy may offer improved treatments of gliomas.
It has been confirmed that a large proportion (>90%) of the human genome is actively transcribed, and most of the transcripts are identified as non-coding RNAs (ncRNAs) [
9]. Circular RNAs (circRNAs), referred to as ncRNAs, are different from the linear RNAs with a circular structure by joining the 3′ end of the RNA to the 5′ end [
10,
11]. Recently, circRNAs have been found to have multiple functions in various mammalian cells. Most circRNAs are derived from exons, and most of them are located in the cytoplasm [
12]. Accumulated evidence showed that circRNAs harbor microRNA (miRNA) binding sites and function as miRNA sponges. For example, miR-138 targets the circular RNA SRY in a sequence-specific manner [
13]. In addition, the circular RNA CiRS-7 possess a binding site of miR-7 and modulate miR-7 expression [
14]. Tau tubulin kinase 2 (TTBK2) was first identified as a kinase which phosphorylated tau and tubulin [
15]. Aberrant expression of linear TTBK2 is related to various diseases. Overexpression of TTBK2 contributes to the progression of amyotrophic lateral sclerosis [
16]. More importantly, increased expression of TTBK2 attenuates the sunitinib-induced apoptosis of kidney carcinoma and melanoma cell lines [
17]. Besides, our preliminary experiment demonstrated that TTBK2 circular RNA (circ-TTBK2, also named has_circ_0000594 according to circBase) was upregulated in glioma tissues. Therefore, we hypothesized that dysregulation of circ-TTBK2 was involved in the regulation of glioma malignancy.
MiRNAs are characterized as a member of small non-coding RNAs and have been confirmed to be involved in both biological and pathological processes [
18]. Expression profiling analysis has depicted a possible tumor-suppressive function of miR-217 in various cancers. As previously reported, miR-217 is robustly downregulated in human epithelial ovarian cancer (EOC) and inhibits cell growth and metastasis [
19]. Additionally, miR-217 expression is downregulated and exerts a tumor-suppressive function in gastric cancer [
20]. Bioinformatics software (Starbase) revealed that circ-TTBK2 harbor a binding site of miR-217. However, the function of miR-217 in glioma and circ-TTBK2/miR-217 functional network in modulating glioma malignant behavior remains unknown.
Hepatocyte nuclear factors (HNFs) were initially identified as a group of transcription factors that were involved in regulating the transcription of liver-specific genes. Hepatocyte nuclear factor-1beta (HNF1β) is the most important member of liver-specific transcription factor and is responsible for sequence-specific DNA binding. It is a homeobox transcription factor functioning as a homodimer or heterodimer with HNF1α [
21]. In addition, HNF1β has been characterized as an oncogene in various tumors. HNF1β is upregulated in hepatocellular carcinoma (HCC), and high level of HNF1β leads to poor overall survival [
22]. Also, HNF1β promotes malignant progression of ovarian clear cell carcinoma via facilitating glucose uptake and glycolytic activity [
23]. More importantly, using bioinformatics softwares (Targetscan, miRanda, and RNAhybrid), a binding site was identified between miR-217 and HNF1β. However, the potential oncogenic function of HNF1β in glioma remains poorly defined.
Derlin-1 participates in the dislocation of misfolded proteins from endoplasmic reticulum (ER) and protects cancer cells from endoplasmic reticulum stress-induced apoptosis. Moreover, Derlin-1 expression is upregulated in various tumors such as human breast carcinoma and colon tumor [
24,
25]. Besides, previous reports unveiled that overexpressed Derlin-1 activated PI3K/AKT and ERK signaling pathways [
26,
27]. Also, by scanning the promoter sequence of Derlin-1, we found a putative binding site of HNF1β. Although the oncogenic role of Derlin-1 is confirmed in many tumors, whether Derlin-1 exerts oncogene function in glioma remains unclear.
In the present study, we investigated the expression and functions of circ-TTBK2, miR-217, HNF1β, and Derlin-1 in glioma tissues and cells. Circ-TTBK2, but not linear TTBK2, exerted oncogenic role in glioma cells. Furthermore, miR-217 targeted circ-TTBK2 in a sequence-specific manner, miR-217 and circ-TTBK2 formed a negative feedback loop possibly mediated by RNA-induced silencing complex (RISC). Moreover, HNF1β was confirmed to harbor a binding site of miR-217 using dual-luciferase assays. These results demonstrated a detailed function of circ-TTBK2 in glioma and provided a novel potential approach for glioma therapy.
Discussion
In this study, we demonstrated that circ-TTBK2 was upregulated in glioma tissues and cell lines. Overexpression of circ-TTBK2 promoted glioma cells malignant progression. In contrary, miR-217 was downregulated in glioma tissues and cell lines. Restoration of miR-217 restrained glioma cells malignant progression. Moreover, miR-217 bound to circ-TTBK2 in a sequence-dependent manner and there was a reciprocal negative feedback between circ-TTBK2 and miR-217. Further, overexpression of circ-TTBK2 increased HNF1β expression via impairing miR-217 expression which negatively regulated HNF1β by targeting its 3′-UTR. HNF1β was upregulated in glioma tissues and cells and promoted cell proliferation, migration, and invasion, while inhibited apoptosis of glioma cells. Meanwhile, HNF1β enhanced the promoter activity and bound to the promoter of Derlin-1. In addition, Derlin-1, identified as an oncogene in glioma tissues and cells, was involved in the HNF1β-mediated promotion of glioma cells malignant progression. Overexpression of Derlin-1 facilitated malignancy of glioma cells. Mechanistically, PI3K/AKT and ERK pathways were involved in circ-TTBK2 regulated malignant progression of glioma cells. Remarkably, the in vivo study demonstrated that the inhibition of circ-TTBK2 and restoration of miR-217 exhibited the lowest tumor volume and the longest survival tumor-bearing nude mice.
Although circRNAs have been found decades ago, their novel functions have remained unclear until recently. Accumulated evidence indicated that dysregulated expression of circRNAs were ubiquitously in heterogeneous tumors and were involved in multiple cellular biological processes in tumor cells [
29,
30]. Notwithstanding, mechanisms of circRNAs’ effect on tumor cells are anfractuous and unclear. As earlier reported, circ-Foxo3 expression was downregulated in several cancer cells than in non-cancer cells, and overexpressed circ-Foxo3 inhibited cell proliferation through binding to CDK2 and p21 [
31]. Meanwhile, circ-Foxo3 expression level is upregulated in aged patient and mice tissues, overexpression of circ-Foxo3 contributes to the progression of senescence in mice Dox-induced cardiomyopathy cells by interacting with ID1, E2F1, FAK, and HIF1A [
32]. Due to our preliminary result, circ-TTBK2 expression was upregulated in glioma tissues. Linear TTBK2 was first characterized as a kinase in brain cells that could phosphorylate Ser 208 and Ser 210 in tau protein [
33]. Also, mutated TTBK2 contributed to spinocerebellar ataxia [
34]. Remarkably, TTBK2 mRNA expression was not changed in glioma tissues and cells. Furthermore, neither circ-TTBK2 nor linear TTBK2 influenced respective expression level. Consistent with several previous findings, we concluded circ-TTBK2 was independent of linear TTBK2 [
31,
35]. Moreover, enhanced circ-TTBK2 facilitated malignant progression of glioma cells. Therefore, circ-TTBK2 might be involved in the modulation of glioma cell biological behavior and exerted critical function in glioma progression.
Accumulated evidence confirmed that circRNAs may act as miRNAs sponges via binding to miRNAs and modulate their function [
36]. For example, circ_005169 exerted oncogenic function via sponging miR-145 and increasing the expression of E2F5, BAG4, and FMNL2 in colorectal cancer cells [
37]. Bioinformatics database (Starbase) showed that a putative binding site exists between circ-TTBK2 and miR-217. We further ascertained that miR-217 targeted circ-TTBK2-Wt. This indicated that circ-TTBK2 might sponge to miR-217 to modulate its function in glioma cell. Interestingly, two putative binding sites were identified in linear TTBK2. Our data showed that miR-217 targeted TTBK2-Wt1, this binding site was the same sequence as circ-TTBK2 harbored. However, the binding site falls within the CDS region of linear TTBK2 (not 3′ UTR). In addition, pre-miR-217 did not change the linear TTBK2 expression. This demonstrated that the binding site between TTBK2 and miR-217 was not functional. Further, we found that there was a reciprocal negative feedback between circ-TTBK2 and miR-217. The RIP assay results showed that circ-TTBK2 and miR-217 were presented in the RISC complex. This might partially explain why the expression of circ-TTBK2 and miR-217 were negatively correlated. We further investigated whether circ-TTBK2 exerted oncogenic function in glioma through regulating miR-217 and found that the restoration of miR-217 robustly reversed the circ-TTBK2-induced promotion of glioma cell malignant progression. These results demonstrated that miR-217 could target circ-TTBK2 in a sequence-specific manner, and there was a reciprocal repression process between circ-TTBK2 and miR-217.
Notoriously, miR-217 was downregulated in various tumors such as EOC and gastric cancer. Meanwhile, overexpressed miR-217 obviously inhibited cell proliferation, colony formation, and invasion, while promoted apoptosis of colorectal cancer cell via targeting AEG-1 3′-UTR [
38]. Similarly, miR-217 was negatively correlated with malignant profiling and exerted tumor-suppressive function by restraining malignant biological behavior of human osteosarcoma cells [
39]. Further, miR-217 expression was significantly lower in lung cancer tissues than in noncancerous tissues, and enhanced miR-217 inhibited cell proliferation, migration, and invasion, while induced apoptosis of SPC-A-1 and A549 cells via targeting KRAS [
40]. Our data demonstrated that miR-217 expression was reduced in glioma tissues and cells. Also, overexpression of miR-217 impeded glioma cells malignancy in vitro and reduced tumor growth in vivo. These findings indicated that miR-217 exerted tumor-suppressive function in glioma cells.
HNF1β was first identified as a liver-specific transcription factor and contributed to the malignant progression of HCC. Recent studies showed that HNF1β functioned as an oncogene in various tumors. HNF1β is upregulated in human prostate cancer and favors cell proliferation and tumor progression [
41,
42]. Also, HNF1β expression is increased in human pancreatic cancer and predicts poor survival [
21]. Due to the oncogenetic role of HNF1β in various tumors, and the putative binding site between miR-217 and HNF1β predicted with bioinformatics databases, we hypothesized that HNF1β might be involved in circ-TTBK2/miR-217 regulation network. Luciferase assay result confirmed that HNF1β was a target of miR-217. Also, our data showed that HNF1β served as an oncogene in glioma cells. We next aimed to investigate whether HNF1β was involved in the circ-TTBK2-mediated regulation of glioma cell progression. As we expected, our results showed that circ-TTBK2 inhibition impaired HNF1β mRNA and protein expressions. Moreover, reintroduction of miR-217 decreased HNF1β mRNA and protein expressions by targeting its 3′-UTR. Additionally, overexpressed miR-217 reversed circ-TTBK2-induced promotion of HNF1β expression. These corroborated the hypothesis that HNF1β was involved in circ-TTBK2/miR-217 regulation network.
In most cases, HNF1β serves as an activator in the transcriptional regulation of targeted genes. Increasing reports showed that Derlin-1 was upregulated in various tumors. By analyzing the promoter sequence of Derlin-1, two putative HNF1β binding sites were identified. ChIP assays corroborate our hypothesis that HNF1β could directly bind to Derlin-1 promoter. Furthermore, we demonstrated that overexpressed HNF1β activated Derlin-1 expression. Similarly, HNF1β played key roles in the regulation of intracellular cholesterol storage and the level of free cholesterol via binding to and activating ACAT promoter [
28]. Overexpressed HNF1β favors tumor progression and inhibits apoptosis of tumor cells via enforcing osteopontin expression, which harbors HNF1β binding sites in its promoter [
43].
Notoriously, Derlin-1 acts as an oncogene in various tumors. Earlier study showed that Derlin-1 was involved in the TCL1-mediated contribution to progression of chronic lymphocytic leukemia in mice [
44]. In addition, Derlin-1 is obviously upregulated in human lung cancer cells, and the inhibition of Derlin-1 attenuates p62 degradation, which leads to the blockage of tumor cell autophagy [
45]. To describe the Derlin-1 profile in glioma tissues and cells, the expression and function of Derlin-1 were determined. Consistent with previously reported, our results demonstrated that Derlin-1 was located in the cytoplasm and had a high expression in glioma tissues and cells. Further, we found an enhanced expression of Derlin-1 that stimulated glioma cell malignant behaviors. More importantly, overexpression of Derlin-1 had been proved to have the ability to activate PI3K/AKT and ERK pathways in tumor cells, which are the key pathways directly related to proliferation, migration, invasion, and apoptosis [
26,
27,
46,
47]. Given that Derlin-1 could be activated by HNF1β, and miR-217 modulated glioma cell progression via targeting 3′-UTR of HNF1β, we next sought to investigate whether Derlin-1 and the downstream pathways were involved in miR-217-induced blunting effect on glioma cells. First, we found that overexpression of HNF1β (without 3′-UTR) reversed inhibition on glioma cell malignant biological behavior induced by miR-217. Then, our results demonstrated that Derlin-1, p-PI3K, p-AKT, p-ERK, and p-MEK1/2 were distinctly restored when glioma cells were co-transfected with miR-217 and HNF1β (without 3′-UTR). These data provided a novel insight into the molecular mechanism of circ-TTBK2 and miR-217. The mechanism underlying tumorgenesis of human glioma cell lines by circ-TTBK2 is schematically presented in Fig.
8c.
Methods
Human tissues specimens
For determination of circ-TTBK2 and miR-217, clinical specimens were divided into five group: NBTs (normal brain tissues) (n = 11), grade I (n = 19), grade II (n = 19), grade III (n = 19), and grade IV (n = 19). For determination of HNF1β and Derlin-1, clinical specimens were divided into three groups: NBTs (n = 8), grade I–II glioma group (low-grade glioma tissues) (n = 16), and grade III–IV glioma group (high-grade glioma tissues) (n = 16) based on the WHO 2007 classification of tumors by two experienced neuropathologists. Normal brain tissues (NBTs) were collected from patients’ fresh autopsy material (donation from individuals who died in traffic accident and confirmed to be free of any prior pathologically detectable conditions) were used as negative control.
Cell culture
Human U87 and U251 glioma cell lines and human embryonic kidney (HEK) 293T cells were purchased from the Shanghai Institutes for Biological Sciences Cell Resource Center. Primary normal human astrocytes (NHA) were purchased from the Sciencell Research Laboratories (Carlsbad, CA, USA). For details, see Additional file
3.
Fluorescence in situ hybridization (FISH)
For identification of circ-TTBK2 and miR-217 rearrangement in glioma tissues, circ-TTBK2 probe (green-labeled, Biosense, Guangzhou, China) (5′ CAATCTTTCTCAATGGTCTGACGTCA 3′) and miR-217 probe (red-labeled, Exiqon, Copenhagen, Denmark) were used. In brief, slides were treated with PCR-grade proteinase-K (Roche Diagnostics, Mannheim, Germany) blocked after with prehybridization buffer (3% BSA in 4 × saline-sodium citrate, SSC). The hybridization mix was prepared with circ-TTBK2 probe or miR-217 probe in hybridization solution. Then the slides was washed with washing buffer; the sections were stained with anti-digoxin rhodamine conjugate (1:100, Exon Biotech Inc, Guangzhou, China) at 37 °C for 1 h away from light. The sections were stained with 4′,6-diamidino-2-phenylindole (DAPI, Beyotime, China) for nuclear staining subsequently. All fluorescence images (100×) were captured using a fluorescence microscope (Leica, Germany).
Reverse transcription and quantitative real-time PCR
Trizol reagent (Life Technologies Corporation, Carlsbad, CA, USA) was used to extract total RNA from the clinical tissues and NHA, U87, and U251 cells. See also Additional file
3.
Western blot
Western blot was performed as previously described [
48]. See Additional file
3 for details and antibodies used.
Cell transfections
Cell transfections were performed as previously described [
48]. See also Additional file
3.
Cell proliferation assay
Cell Counting Kit-8 assay (CCK-8, Dojin, Japan) was used to investigate glioma cell proliferation. Also, see Additional file
3.
Migration and invasion assays
Twenty-four-well chambers with 8-μm pore size (Corning, USA) was used for migration and invasion determination of U87 and U251 cells in vitro. For details, see Additional file
3.
Apoptosis analysis
Cell apoptosis was determined by Annexin V-PE/7AAD staining (Southern Biotech, Birmingham, AL, USA). See also Additional file
3.
Reporter vectors construction and luciferase assays
Dual-luciferase assays were performed as previously described [
48]. See Additional file
3.
RNA immunoprecipitation
RNA immunoprecipitation was performed as previously described [
48]. In brief, glioma cells were lysed by a complete RNA lysis buffer from an EZ-Magna RIP kit (Millipore, Billerica, MA) according to the manufacturer’s protocol. See also Additional file
3.
Immunohistochemistry assays
The slides of human glioma tissue samples (4 μm thick) were dewaxed, rehydrated, and incubated in 0.3% H2O2 for 10 min to inhibit endogenous peroxidase activity before blocking with 10% normal goat serum (MXB, Fuzhou, China) for 30 min and incubating overnight at 4 °C with rabbit polyclonal antibody against HNF1β (1:150, SAB, Chicago, IL). Slides were washed with PBS three times and then incubated with biotinylated rabbit anti-rabbit IgG for 1 h at room temperature. After incubation with avidinbiotin-peroxidase complex for 10 min, samples were stained with 3, 3′-diaminobenzidine. Slides were imaged under a light microscope (Olympus, Japan) at 100× magnification.
Chromatin immunoprecipitation assay
ChIP assay was conducted with Simple ChIP Enzymatic Chromatin IP Kit (Cell signaling Technology, Danvers, Massachusetts, USA) according to the manufacturer’s instruction as previously described [
49]. In brief, glioma cells were fixed with 1% formaldehyde and collected in lysis buffer. Two percent aliquots of lysates were used as an input control and the remaining lysates were immunoprecipitated with normal rabbit IgG or HNF1β antibody. Immunoprecipitated DNA was amplified by PCR using their specific primers (as Additional file
4).
Tumor xenografts in nude mice
The tumor xenograft experiment was performed as previously described [
48]. Stable expression U87 and U251 cells were used for in vivo study. For details, see also Additional file
3.
Statistical analysis
Data are presented as mean + standard deviation (SD). All statistical analyses were evaluated by SPSS 18.0 statistical software with the Student’s t test or one-way analysis of variance ANOVA. Differences were considered to be significant when P < 0.05. Corresponding significance levels were indicated in the figures.