Circular RNA FOXP1 promotes tumor progression and Warburg effect in gallbladder cancer by regulating PKLR expression

A detailed description of RNA-sequencing analysis was provided in Additional file 1: Supplementary Materials and Methods section. The GEO Accession number is GSE100363.

Forty human GBC and adjacent normal tissue specimens were collected from patients who underwent radical resection at Xinhua Hospital between March 2009 and January 2013. The eleven patients were male and twenty-nine were female. Ages ranged from 35 to 82 years (the mean value = 54.85 years). Each tissue sample was snap-frozen in liquid nitrogen for further analysis. All of patients in this study belonged to the same ethnic group. The patients were selected according to the criteria: (1) All clinicopathological diagnoses were confirmed by two pathologists. (2) None of the patients received any treatments before surgery. (3) None of the patients received radiotherapy or chemotherapy during follow-up period. (4) availability of complete follow-up data and not lost follow-up. (5) no death in the perioperative period. (6) no history of other synchronous malignancies. (7) For patients undergoing radical resection of gallbladder cancer. The study procedure was approved by the Human Ethics Committee of Xinhua Hospital. All patients signed consent forms. Follow-ups after surgery were performed according to patient survival time until March 8, 2016. The clinicopathological data were shown in Table 1.

Human GBC cell lines NOZ, GBC-SD, EHGB-1, SGC-996 and OCUG-1 and the human intrahepatic biliary epithelial cell line H69 and another normal biliary epithelia cell line HIBEC were used in the present study. GBC-SD and OCUG-1 cell lines were purchased from Cell Bank of the Chinese Academy of Science (Shanghai, China). The NOZ cell line was purchased from the Health Science Research Resources Bank (Osaka, Japan). EHGB-1 and SGC-996 cells were a generous gift from Eastern Hepatobiliary Surgical Hospital and Institute, The Second Military University, Shanghai, China. The cells were cultured in Dulbecco’s modified Eagle’s medium (Gibco BRL, Grand Island, NY, USA); the cell media contained 10% fetal bovine serum (FBS, HyClone, Invitrogen). Cells were maintained in a humidified incubator at 37 °C in the presence of 5% CO₂.

Cells were transfected using lipofectamine 2000 according to the manufacturer’s instructions. Additional information about the performed experiments can be found in the Additional file 1: Supplementary Materials and Methods. The sequences of all oligonucleotide used in the study were shown in Additional file 2: Table S1.

Cell proliferation was assessed using Cell Counting Kit 8 (Dojindo, Japan); cell migration and invasion was assessed with transwell assays. Cell cycle distribution and cell apoptosis rate analysis was analyzed using flow cytometry. Detailed descriptions of experiments can be found in Additional file 1: Supplementary Materials and Methods.

RNA isolation, qRT-PCR and western blot analysis were performed as described previously [19]. Detailed descriptions of experiments can be found in Additional file 1: Supplementary Materials and Methods. The sequences of all primers used in the study were shown in Additional file 2: Table S1.

NOZ cells (1 × 10⁶) stably expressing sh-NC, sh-circFOXP1–1, sh-circFOXP1–2 or GBC-SD cells (1 × 10⁶) stably expressing pLCDH-vector or pLCDH-circFOXP1 were subcutaneously injected into either side of the flank area of 3-week-old male nude mice (n = 5 mice per group). Tumor volumes were measured (0.5 × length × width²) and tumor weights were evaluated in mice on a weekly basis. After 4 weeks, mice were sacrificed, and the tumors were excised. The study protocol was approved by the Animal Care and Use committee of Xinhua Hospital (approval ID: 2014041). All animal experiments were performed in the animal laboratory center at Xinhua Hospital and conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH publication number 85–23, revised 1996).

RNA pull-down assays were performed as described previously [20, 21]. Briefly, cells extract (2 μg) was mixed with biotinylated RNA (100 pmol). Washed streptavidin agarose beads (100 ml) were added to each binding reaction and further incubated at room temperature for 1 h. Beads were washed briefly three times and boiled in SDS buffer, and the retrieved protein was determined by western blot analysis.

RIP assays were performed using an EZ-Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, Billerica, MA, USA) according to the manufacturer’s instructions. Cells at approximately 90% confluence was lysed using complete RIP lysis buffer containing RNase Inhibitor (Millipore) and protease inhibitor and then 100 μl of whole cell extract was incubated with RIP buffer containing magnetic beads conjugated to specific antibodies. The negative control was normal mouse anti-IgG antibody (Cell Signaling Technology, USA), and the positive control was anti-SNRNP70 antibody (Millipore, USA).

To explore circRNAs expression profile in GBC, we performed RNA sequencing analyses of ribosomal RNA-depleted total RNA from four pairs of GBC tissues and adjacent normal tissues. The cluster heat map demonstrated the differentially expressed circRNAs over 3.5-fold change (Fig. 1a, left). The scatter plots showed that upregulated 44 circRNAs and downregulated 81 circRNAs in GBC tissues compared with adjacent normal tissues (Fig. 1a, right). Compared with previously reported database obtained from circBase [18], 34 of the top 44 upregulated circRNAs were overlapped circRNAs and were listed in Fig. 1a, left. Several circRNAs including circXPO1 (chr2_61710091_61717911_-, hsa_circ_0002607), circFOXP1 (chr3_71090478_71102924_-,hsa_circ_0008234),circMAPK1(chr22_22153300_22162135_-, hsa_circ_0004872), circSENP1 (chr12_48477373_48491907_-, hsa_circ_0006222) and circSMAD2 (chr18_45391429_45423180_-, hsa_circ_0000847) were chosen for further investigation because their parental genes are involved in cell proliferation, invasion and cell cycle regulation in some tumors [22‐26]. We examined these investigated circRNA candidates in GBC tissues and paired normal tissue samples from 40 patients. The results showed that expression of circFOXP1, circMAPK1 and circSAMD2 were upregulated in GBC tissues (Fig. 1b, top and Additional file 3: Figure S1A). GBC patients were classified into two groups: high-circFOXP1 group (n = 20, circFOXP1 expression ratio ≥ median ratio) and low-circFOXP1 group (n = 20, circFOXP1 expression ratio < median ratio). Kaplan-Meier analysis and log rank tests indicated that higher circFOXP1 expression levels were positively correlated with a shorter OS in patients (Fig. 1b, low), while circSMAD2 and circMAPK1 expression showed no statistical significance with OS in patients (Additional file 3: Figure S1B). CircFOXP1 expression was upregulated in several tumors compared with normal tissues through RNA sequencing data in a previous study [12]. However, the clinical significance and functional role of circFOXP1 remains unknown.

As seen in Table 1, the statistical analysis revealed that high circFOXP1 expression levels were strongly correlated with lymph node metastasis (P = 0.011) and advanced TNM stage (III-IV) (P = 0.027) in GBC patients, but no correlation was found with the other factors (Table 1). Univariate and multivariate Cox analysis revealed that lymph node metastasis (HR = 3.749, 95% CI: 1.578–8.905, P < 0.05) and high circFOXP1 expression (HR = 3.658, 95% CI: 1.587–8.431, P < 0.05) were independent risk factors for the OS of GBC patients (Table 2). As shown in Additional file 4: Figure S2A, the transcript levels of circFOXP1 were also significantly higher in several GBC cells than in the bile duct epithelial cell line H69 or another normal biliary epithelia cell line HIBEC. RNA was isolated and RT-PCR with Sanger sequencing confirmed the circular form in GBC cells (Additional file 4: Figure S2B). RNA isolated from the above cell lines was subjected to northern blot using a probe specific for circFOXP1, confirming that circFOXP1 was present in these cells (Fig. 1c). In addition, we confirmed that circFOXP1 was resistant to RNase R after digestion of RNA using RNase R exonuclease (Fig. 1d and Additional file 4: Figure S2C), which was consistent with a previous study [27], the circHIPK3 was used as the control, which was reported as a circRNA in previous study [12]. Using fluorescence in situ hybridization (FISH) and qRT-PCR analysis, we also noted that circFOXP1 localized in the cytoplasm and nucleus, but was predominately enriched in the cytoplasm in GBC cells (Fig. 1e and Additional file 4: Figure S2D). Thus, the above results indicated that circFOXP1 may serve as a prognostic biomarker for GBC.

To further explore the biological significance of circFOXP1 in GBC progression, gain-and loss-of-function studies were performed. Based on endogenous expression of circFOXP1 in several GBC cell lines, we developed NOZ and SGC-996 cells with circFOXP1 stably silenced by shRNA-circFOXP1 and EHGB1 or GBC-SD cells with circFOXP1 stably overexpressed using pLCDH-circFOXP1. Meanwhile, to confirm the specificity of circFOXP1 silencing or overexpression, we showed that the cells transfection had no effects on mRNA expression of FOXP1 in GBC cells. In addition, circFOXP1 was also overexpressed after digestion of RNA using RNase R by using pLCDH-circFOXP1 (Additional file 4: Figure S2E-2F). By performing CCK8 assays and flow cytometry analysis in vitro, we observed that knockdown of endogenous circFOXP1 significantly suppressed cell proliferation capacity, G1-S arrest, and increased the cell apoptosis rate in NOZ and SGC-996 cells, but upregulated expression of circFOXP1 dramatically promoted the cell proliferation capacity, beyond the G1-S transition, and reduced the cell apoptosis rate in GBC-SD cells (Fig. 2a-c). Consistent with the decreased cell proliferation capacity, NOZ and SGC-996 cells exhibited reduced expression levels of proliferating cell nuclear antigen (PCNA), MMP9 and AKT, but increased Caspase-3 expression after knockdown of endogenous circFOXP1, and conversely, upregulated expression of circFOXP1 had the opposite effects in GBC cells (Additional file 5: Figure S3A). As shown in Additional file 5: Figure S3B-3C, migration and invasion capacities were impaired in NOZ and SGC-996 cells after knockdown of endogenous circFOXP1, but were dramatically promoted in GBC-SD cells after circFOXP1 overexpression.

To evaluate the biological function of circFOXP1 in vivo, a xenograft tumor model was constructed by inoculating different clones of NOZ and GBC-SD cells subcutaneously into nude mice. The results confirmed that the mean tumor volumes and weights were larger and tumor growth was rapid in the upregulated expression of circFOXP1 group, compared with the control group. However, the tumor growth was inhibited, as shown by decreasing mean volumes and weights, in the knockdown of endogenous circFOXP1 group compared with the control group (Fig. 2d). Immunohistochemical staining of tumor tissues revealed an increased proportion of proliferating cells (Ki67+) in the upregulated expression of circFOXP1 group compared with the control group. In contrast, knockdown of endogenous circFOXP1 led to reduced expression of Ki67 compared with the control group (Fig. 2e). These results indicated that circFOXP1 promoted GBC growth in vitro and in vivo.

The Warburg effect, characterized by abnormal metabolic phenomena that enhance glycolysis and reduce oxidative phosphorylation, induces significant differences between cancer cells and normal cells and affects tumor progression [28]. Based on the above findings, we investigated whether the expression of circFOXP1 affected the Warburg effect by measuring the levels of pyruvate production and lactate production and the glycolytic rate in GBC cells. The statistical analysis revealed a dramatically lower level of lactate production and pyruvate production after circFOXP1 knockdown in NOZ and SGC-996 cells, conversely, were significantly higher after circFOXP1 overexpression in GBC-SD cells (Fig. 3a). Furthermore, glycolysis and extracellular acidification rates (ECAR) were analyzed using an XF24 Extracellular Flux analyzer (Seahorse), The results confirmed that circFOXP1 knockdown reduced the ECAR in NOZ and SGC-996 cells, while upregulated circFOXP1 expression increased ECAR in GBC-SD cells (Fig. 3b-c). we also measured oxygen consumption rate (OCR, a marker of OXPHOS) and found that knockdown of circFOXP1 enhanced OCR ability in NOZ and SGC-996 cells, in contrast, upregulated circFOXP1 expression reduced OCR ability in GBC-SD cells (Fig. 3d-e). The results confirmed that circFOXP1 knockdown impaired the glycolysis rate and glycolytic capacity in NOZ and SGC-996 cells. On the other hand, upregulated circFOXP1 expression significantly enhanced the rate of glycolysis and glycolytic capacity in GBC-SD cells (Fig. 3f-g). In addition, knockdown of circFOXP1 in NOZ and SGC-996 cells also elevated the formation of ATP produced by OXPHOS, but upregulated circFOXP1 has an decreased formation of ATP in GBC-SD cells (Fig. 3g). These results indicated that upregulation of circFOXP1 promoted the Warburg effect in GBC cells.

We further investigated the mechanisms by which circFOXP1 promoted the Warburg effect in GBC cells. Recent evidence has indicated that circRNAs participate in molecular regulation by interacting with several proteins [13]. Based on this hypothesis, we identified circFOXP1-interacting proteins by performing an RNA pull-down assay combined with Liquid Chromatography-Mass Spectrometry (LC-MS) in NOZ cells. The RNA-related proteins were determined using SDS-polyacrylamide gel electrophoresis (SDS/PAGE) and silver staining (Fig. 4a). By using LC-MS and comparing the results with those from antisense circFOXP1 experiments, proteins were identified by LC-MS that specifically associated with circFOXP1 (Additional file 6: Table S2). PTBP1 was further detected due to higher expression in the circFOXP1-sense probe compared with the circFOXP1-antisense sample or no RNA sample, and previous studies have verified that PTBP1 is associated with the Warburg effect in cancer cells through regulation of the PKM1/PKM2 ratio [29]. Western blot analysis with anti-PTBP1 antibody indicated the existence of PTBP1 within the circFOXP1 sense RNA probe pull-down samples in NOZ and SGC-996 cells (Fig. 4b). Meanwhile, a RIP assay with PTBP1 antibody showed that endogenous PTBP1 directly bound to circFOXP1 in NOZ and SGC-996 cells (Fig. 4c). Moreover, we observed that knockdown of endogenous circFOXP1 decreased the protein expression of PTBP1 in NOZ and SGC-996 cells, but upregulated expression of circFOXP1 upregulated the expression of PTBP1 in GBC-SD cells (Fig. 4d). By Immunofluorescence and western blot analysis, we further demonstrated that exogenous expression of circFOXP1 enhanced expression of PTBP1 by increased transportation levels of PTBP1 from the nucleus to the cytoplasm, but knockdown of circFOXP1 also revert PTBP1 back to the nucleus (Fig. 5a-b).

In mammalian cells, pyruvate kinase is encoded by 4 isozymes: M1, M2, liver (PKL), and red blood cell (PKR). While the M1, PKL, and PKR isozymes are described to exhibit tissue-specific expression, the pyruvate kinase M2 isoform is highly expressed across cancer types. Glycolytic deregulation, including activation of upregulated PKM2 and liver and RBC (PKLR), is a driver that promotes cancer progression [15, 30, 31]. Previous reports confirmed that PTBP1 regulates AS of the pyruvate kinase gene (PKM) in the acquisition of oncogenic features by increasing recruitment to the PKM pre-mRNA to promote PKM2 splicing, which contributes to the Warburg effect [32]. First, we speculated whether decreased circFOXP1 led to exertion of the function of PTBP1 to regulate AS. The results confirmed that knockdown of endogenous circFOXP1 resulted in no significant change in the mRNA and protein levels of PKM1 in NOZ and SGC-996 cells but had a partial effect on the PKM2 protein level (Fig. 5c-d). PKLR is also known as a promoter in modulation of the Warburg effect [15]. Interestingly, dramatically decreased PKLR expression levels were observed after downregulation of circFOXP1 in NOZ and SGC-996 cells or in NOZ cells in vivo, but increased expression of PKLR was observed after upregulation of circFOXP1 in GBC-SD cells or in vivo (Fig. 5e-f and Additional file 7: Figure S4A). In addition, we showed that PKLR silencing impaired the glycolysis rate and glycolytic capacity in NOZ and SGC-996 cells, which inhibited the Warburg effect in GBC (Additional file 7: Figure S4B-4F).

PKLR acts as a driver of tumor growth and metastasis [15, 33]. We sought to elucidate the underlying mechanism by which circFOXP1 affect PKLR expression. PTBP1, an RNA-binding protein, exerts various molecular functions, including RNA metabolism, for example, control of mRNA stability or degradation [34, 35], determination of mRNA localization [36], and protection of mRNAs from decay [37]. Compared with normal tissues, PTBP1 was upregulated in gallbladder cancer tissues and the protein and mRNA of PKLR was also upregulated in gallbladder cancer tissues (Fig. 6a). Furthermore, we showed that the proteasome inhibitor MG-132 could increase PKLR expression and abolished the reduction in PKLR protein levels in circFOXP1-knockdown NOZ cells (Additional file 8: Figure S5A). The protein synthesis inhibitor cycloheximide (CHX) decreased the expression of PKLR proteins by inhibiting protein synthesis, but upregulation of PKLR in circFOXP1-overexpressing SGC-996 and GBC-SD cells was not abolished by treatment with CHX (Additional file 8: Figure S5B). These results showed that circFOXP1 may affect PKLR degradation in GBC.

Based on these findings, we hypothesized that circFOXP1 may enhance the capacity of PTBP1 to bind to PKLR mRNA and protect it from mRNA decay and consequently increasing the PKLR protein levels. To test this hypothesis, we utilized a RIP assay to detect the association between PKLR mRNA and PTBP1. The results demonstrated that PTBP1 could bind to PKLR mRNA in NOZ and SGC-996 cells (Fig. 6b). Using the online prediction software RBP map (http://​rbpmap.​technion.​ac.​il/​), we found that PTBP1 could bind to the 3’UTR region and coding region (CDS) of PKLR mRNA (UCUU binding bites) (Additional file 9: Figure S6A-6B). Furthermore, an RNA pull-down assay with the 3’UTR region or CDS region of PKLR biotin-labeled RNA probe confirmed that PTBP1 directly bound to PKLR mRNA in NOZ and SGC-996 cells (Fig. 6c). A RIP assay with PTBP1 antibody demonstrated that downregulation of circFOXP1 reduced the enrichment of PKLR mRNA in NOZ and SGC-996 cells, but enhanced the enrichment of PKLR mRNA after overexpression of circFOXP1 in GBC-SD cells (Fig. 6d-e). We further detected the effects of PTBP1 knockdown on PKLR expression. Downregulation of PTBP1 reduced both PKLR mRNA and protein in NOZ and SGC-996 cells, and knockdown of PTBP1 abrogated the effect of overexpressed circFOXP1 on PKLR expression in GBC-SD cells (Additional file 10: Figure S7A and Fig. 6f). These results revealed that circFOXP1 promoted PKLR mRNA expression by interacting with PTBP1 in GBC cells.

Endogenous circRNAs have been found to act as microRNA (miRNA) sponges in human cancers [38]. As circFOXP1 was predominantly localized in the cytoplasm, we hypothesized that circFOXP1 could also regulate PKLR expression by binding to specific miRNAs. To confirm this hypothesis, we performed a search for miRNAs that have complementary base pairing with circFOXP1 using the online software tools circinteractome (http://​circinteractome.​nia.​nih.​gov) and circRNAs from RNA sequencing targeted miRNAs predicted analysis by miRanda (www.​microrna.​org) (Fig. 7a, left, Additional file 11: Table S3). The results from RNA sequencing showed that miR-370 could form complementary base pairing with circFOXP1 and PKLR, respectively (Fig. 7a, right, and Fig. 7c). To verify that circFOXP1 could bind to miR-370, the wild-type (WT) and two mutant-type (MUT) circFOXP1 reporter vectors were constructed. We observed that a miR-370 mimic significantly reduced the luciferase activity of the circFOXP1-WT-1 (37%) and circFOXP1-WT-2 (33%) reporter vectors but not that of circFOXP1-MUT-1 or circFOXP1-MUT-2, which suggested that miR-370 was a target of circFOXP1 in a sequence-specific manner (Fig. 7b). To further clarify the regulatory association between circFOXP1 and the 3’UTR of PKLR, we constructed wild-type (WT) and mutant-type (MUT) PKLR 3’UTR reporter vectors (Fig. 7c). The results revealed that the miR-370 mimic significantly reduced the luciferase activity of the WT PKLR 3’UTR reporter vector but not that of the mutant PKLR 3’UTR reporter. Co-transfection with the WT PKLR 3’UTR and miR-370 mimic and pLCDH-circFOXP1 antagonized the effects (Fig. 7d). Furthermore, we confirmed that miR-370 expression was significantly downregulated in GBC tissues and cells, compared with adjacent normal tissues and H69 cell, respectively (Additional file 10: Figure S7B-7C). Moreover, an inverse correlation was observed between circFOXP1 and miR-370 expression in GBC tissues (r = − 0.45, P < 0.05, Additional file 10: Figure S7D). The PKLR mRNA and protein expression was downregulated after transfection of NOZ and SGC-996 cells with sh-circFOXP1, but was rescued by co-transfection with miR-370 inhibitor and sh-circFOXP1 (Fig. 7e). Conversely, the PKLR mRNA and protein expression was increased after transfection of GBC-SD cells with a pLCDH-circFOXP1, but was rescued by co-transfection with pLCDH-circFOXP1 and miR-370 mimic (Fig. 7f). These results showed that circFOXP1 promoted PKLR expression by sponging miR-370 in GBC cells.