Hsa_circ_0006948 enhances cancer progression and epithelial-mesenchymal transition through the miR-490-3p/HMGA2 axis in esophageal squamous cell carcinoma

Expression profiles of circRNAs and characterization of hsa_circ_0006948 in ESCC

Briefly, the circRNA expression profiles of three paired ESCC tissue samples were analyzed using a microarray we deposited at Gene Expression Omnibus previously. Distinct circRNA expression profiles were shown in the hierarchical clustering (Figure 1A). A volcano plot indicated differential expression between tumor and normal tissues (Figure 1B). A significant difference was defined as a fold change > 2.0 and P < 0.05. We examined the 10 most upregulated circRNAs using qRT-PCR and found that hsa_circ_0006948 expression was significantly high in five ESCC cell lines normalized to HEEC cells (Figure 1C). Therefore, hsa_circ_0006948 was further studied. Next, we tried to investigate the relative abundance of hsa_circ_0006948 compared to its cognate linear RNA. The results showed that hsa_circ_0006948 was significantly lower than housekeeping mRNA GAPDH, however, it was abundant as the linear FNDC3B (linFNDC3B) (Figure 1D). It has been reported that the ratio between most circRNAs and their linear counterpart is about 1% [16]. Given the expression of hsa_circ_0006948 in comparison with linFNDC3B, we speculated that hsa_circ_0006948 is functional in ESCC cell lines.

Figure 1. The identification and characteristics of hsa_circ_0006948 in ESCC cells. (A) Heat map showing the differential expression and hierarchical clustering of circRNAs between ESCC and adjacent normal tissues. (B) Volcano plot, x-axis: log2 (fold change); y-axis: -log10 (P-value). The vertical lines correspond to 2.0-fold up and down, and the horizontal line represents a P-value of 0.05. The red points in the plot represent differentially expressed circRNAs with statistical significance. (C) The relative hsa_circ_0006948 was significantly high in ESCC cells. (D) q RT-PCR analyses of expression of hsa_circ_0006948, linFNDC3B and GAPDH in various ESCC cell lines. Y-axis is the raw CT value. (E) Above: Divergent primers detected circular RNAs in cDNA but not gDNA. Below: Three exons form hsa_circ_0006948 by back splicing from chromosomal region and Sanger sequencing of hsa_circ_0006948 showed the back-splice junction (∇). (F) Fluorescence in situ hybridization assay was conducted to determine the subcellular localization of hsa_circ_0006948.

Two sets of primers were designed. Divergent primers were used to amplify only the circular form, hsa_circ_0006948, while convergent primers were used to amplify only the linear form, FNDC3B mRNA. The PCR products were validated by electrophoresis using cDNA and genomic DNA (gDNA) as templates. The results revealed that the single product of the expected size was amplified distinctly, with the divergent primers for cDNA but not gDNA (Figure 1E). The Sanger sequencing results further confirmed head-to-tail splicing. These results indicated that hsa_circ_0006948 was derived from exon 2, exon 3 and exon 4 of the FNDC3B gene, which was consistent with the hsa_circ_0006948 data from circBase (Figure 1E). Further fluorescence in situ hybridization (FISH) indicated that hsa_circ_0006948 was localized preferentially in the cytoplasm (Figure 1F).

Hsa_circ_0006948 is upregulated in ESCC and related to poor survival

We examined hsa_circ_0006948 expression in 153 pairs of ESCC tissues and adjacent normal tissues using qRT-PCR analysis and found that hsa_circ_0006948 was significantly higher in cancer tissues than in normal tissues (Figure 2A). Additionally, hsa_circ_0006948 expression was high in 83.6% of ESCC patients (128 of 153) (Figure 2B). We further analyzed ROC of hsa_circ_0006948 expression to investigate its diagnostic value and found that hsa_circ_0006948 level could differentiate ESCC from normal control, with AUC of 0.85 (cutoff = 0.415, sensitivity = 0.74, specificity = 0.88) (Figure 2C). Then, the relationship between hsa_circ_0006948 and clinicopathological features was analyzed after patients were classified into high and low level groups via the median hsa_circ_0006948 expression (Table 1). Briefly, high hsa_circ_0006948 was positively correlated with lymphatic metastasis (Figure 2D), and no significant correlation was observed for other clinicopathological features, including age, T stage, sex, blood type, tumor location and differentiation. Kaplan–Meier survival curves showed that patients with higher hsa_circ_0006948 expression had a significantly shorter OS than patients with low expression (P=0.0009) (Figure 2E). To assess the prognosis value of hsa_circ_0006948, survival ROC was performed. As shown in Figure 2F, the AUC for hsa_circ_0006948 in prediction of OS was 0.71 (cutoff = 3.71, sensitivity = 0.39, specificity = 0.85). Furthermore, univariate and multivariate COX analyses were performed (Table 2). We found that the expression of hsa_circ_0006948 was an independent prognostic factor for ESCC prognosis (HR = 1.94; 95% CI 1.11-3.39; P = 0.021).

Figure 2. The hsa_circ_0006948 was up-regulated in ESCC tissues and correlates with poor patients prognosis. (A) hsa_circ_0006948 expression was assessed in cancer tissues and normal tissues. (B) The relative hsa_circ_0006948 expression was assessed in 153 ESCC patients. Blue: low level: Red: high level. (C) Present the ROC curve analysis of hsa_circ_0006948 for the diagnosis of ESCC. (D) The expression of hsa_circ_0006948 was significantly higher in patients with lymph nodes at N1-N3 stage than in those with lymph nodes at NO stage. (E) Kaplan-Meier’s analyses of correlations between the hsa_circ_0006948 expression levels and OS (overall survival) of 153 ESCC patients. (F) Present the ROC curve analysis of hsa_circ_0006948 for the prognosis of ESCC.

Hsa_circ_0006948 promotes the proliferation, migration and invasion of ESCC cells

After transfection with siRNAs (siRNA1 and siRNA2) targeting the back-splice region, hsa_circ_0006948 expression in TE-1 and KYSE30 cells was knocked down, and there was no significant change in its linear counterpart, FNDC3B mRNA (Figure 3A, Additional file 1: Supplementary Figure 1A). However, siRNA1 was more effective than siRNA2 in the knockdown of hsa_circ_0006948 expression. Thus, siRNA1 was selected for subsequent studies. Additionally, the hsa_circ_0006948 overexpression plasmid was constructed successfully, and did not affect FNDC3B mRNA (Figure 3B, Additional file 1: Supplementary Figure 1B). Colony formation assays proved that the clonogenic ability of ESCC cells was inhibited by hsa_circ_0006948 downregulation (Figure 3C, Additional file 1: Supplementary Figure 1C). To investigate the functions of hsa_circ_0006948 in cellular proliferation, CCK8 assays were performed. Compared to the negative control conditions, knockdown of hsa_circ_0006948 significantly inhibited ESCC cell proliferation (Figure 3D, Additional file 1: Supplementary Figure 1D). Then, transwell assays were performed using ESCC cells transfected with siRNA.

Figure 3. The function of hsa_circ_0006948 in ESCC cells. (A) Expression of hsa_circ_0006948 and FNDC3B mRNA in TE-1 cells transfected with siRNAs. and (B) KYSE30 cells overexpressing hsa_circ_0006948. (C and D) The effect of hsa_circ_0006948 on cell proliferation in vitro using colony formation assay and CCK8 assay after knocking down hsa_circ_0006948 in TE-1. (E) Cell migration and invasion abilities were assessed by transwell assay after knocking down hsa_circ_0006948 in TE-1 cells. (F and G) The effect of hsa_circ_0006948 on cell proliferation in vitro using colony formation assay and CCK8 assay after overexpressing hsa_circ_0006948 in KYSE30 cells. (H) Cell migration and invasion abilities were assessed by transwell assay after overexpressing hsa_circ_0006948 in KYSE30 cells. (I) Western blot analysis comparing downregulated and upregulated-hsa_circ_0006948 ESCC cells with control cells were shown for vimentin, E-cadherin and N-cadherin* P<0.05,**P<0.01, ***P<0.001.

The transwell assay results showed that hsa_circ_0006948 downregulation inhibited ESCC cell migratory and invasion abilities. (Figure 3E, Additional file 1: Supplementary Figure 1E). In contrast, hsa_circ_0006948 upregulation yielded the reverse effects. Overexpression of hsa_circ_0006948 in ESCC cells promoted the proliferation and increased the migratory and invasion abilities in ESCC cells (Figure 3F–H, Additional file 1: Supplementary Figure 1F–1H).

Hsa_circ_0006948 induces ESCC cell EMT

Epithelial-mesenchymal transition (EMT) is an important process in tumor aggressiveness. Therefore, we assessed whether hsa_circ_0006948 could induce EMT in ESCC cells, and EMT markers were assessed by Western blot in the present study. Western blot analysis demonstrated that knockdown of hsa_circ_0006948 led to the upregulated expression of the epithelial marker E-cadherin and downregulated expression of the mesenchymal markers vimentin and N-cadherin. On the other hand, the expression of E-cadherin was decreased, whereas that of vimentin and N-cadherin was increased in ESCC cells overexpressing hsa_circ_0006948 (Figure 3I). These results implied that hsa_circ_0006948 promotes ESCC cell EMT.

Hsa_circ_0006948 functions as a miR-490-3p sponge in ESCC cells

For circRNAs, acting as miRNA sponges is the most commonly reported function pattern, and circRNAs in the cytoplasm may act as completing endogenous RNAs to bind miRNAs [17, 18]. Given that FISH showed that hsa_circ_0006948 was localized mainly in the cytoplasm, we speculated that hsa_circ_0006948 could sponge miRNAs to inhibit their function. Therefore, we tried to identify candidate miRNAs that may sponge hsa_circ_0006948. The top five miRNA targets are displayed according to bioinformatics analysis (Figure 4A). One of these five miRNAs, miR-490-3p was reported to inhibit EMT in ESCC cells by targeting HMGA2 in a previous ESCC study[19], and the molecular interaction between hsa_circ_0006948 and miR-490-3p is shown in Figure 4B. Therefore, we hypothesized that hsa_circ_0006948 could induce EMT by enhancing HMGA2 by sponging miR-490-3p. To identify whether miR-490-3p can bind to hsa_circ_0006948, a luciferase reporter assay was performed. We found that the luciferase reporter activity was reduced by nearly 50% in only cells co-transfected with miR-490-3p mimics and luc-circ_0006948-wild type containing the complete hsa_circ_0006948 sequence, and overexpression of miR-490-3p did not affect the luciferase activity of the vector with the mutant miR-490-3p binding site in cells (Figure 4C). In addition, hsa_circ_0006948 expression was negatively correlated with miR-490-3p expression in 153 ESCC patients according to q RT-PCR (r=-0.6212, p<0.001) (Figure 4D). Furthermore, we performed RIP assay for AGO2 in TE-1 and KYSE30 cells. The results of qRT-PCR showed that hsa_circ_0006948 was significantly enriched in AGO2-containing beads samples compared to control IgG immunoprecipitates, suggesting miR-490-3p targets directly hsa_circ_0006948 in an AGO2-dependent manner (Figure 4E, 4F). These results indicated that hsa_circ_0006948 may function as a competitive endogenous RNA by sponging miR-490-3p.

Figure 4. Hsa_circ_0006948 could serve as a miR-490-3p sponge in ESCC cells. (A) Top five miRNA targets are displayed. (B) A schematic drawing showing the putative binding sites of the miR-490-3p associated with hsa_circ_0006948. (C) Luciferase reporter assay for the luciferase activity of Luc-circ_0006948 WT or Luc-circ_0006948 mutant in cells cotransfected with miRNA mimics. Data are the mean±SD of three experiments. (D) Hsa_ciic_0006948 expression was negatively correlated with miR-490-3p expression in 153 patients with ESCC. (E and F) RNA immunoprecipitation (RIP) assays were performed using an anti-AGO2 antibody with the transfection of miR-490-3p mimics (miR-490-3p) or miR-NC in TE-1 and KYSE30 cells to detect hsa_circ_0006948 expression according to qRT-PCR. ***P<0.001.

miR-490-3p functions as a tumor suppressor and targets HMGA2 in ESCC

A previous study has verified that miR-490-3p inhibited the proliferation and metastasis of ESCC through experiments using miR-490-3p upregulation in ESCC cells [19]. To further confirm its role as a tumor suppressor, loss-of-function and gain-of-function assays were performed. Briefly, proliferative, migratory and invasive abilities were significantly inhibited after ESCC cells were transfected with miR-490-3p mimics (Figure 5A–5C). When the cells were transfected with a miR-490-3p inhibitor, the opposite results were observed (Figure 5D–5F). Additionally, previous study illustrated that HMGA2 was shown to be a direct target of miR-490-3p and induces EMT [19]. Therefore, 3’UTR of HMGA2 were constructed and inserted into vector. The result showed that luciferase activity decreased after cotransfection of miR-490-3p mimics with HMGA2-3′UTR compared with mutant vector, (Figure 5G) suggesting that miR-490-3p could bind to the 3′UTR of HMGA2.

Figure 5. The function of miR-490-3p in ESCC cells. (A) and (B) The effect of miR-490-3p on cell proliferation in vitro using CCK8 assay and colony formation assay after overexpressing miR-490-3p in TE-1. (C) Cell migration and invasion abilities were assessed by transwell assay after overexpressing miR-490-3p in TE-1 cells. (D and E) The effect of miR-490-3p on cell proliferation in vitro using CCK8 assay and colony formation assay after knocking down miR-490-3p in KYSE30 cells. (F) Cell migration and invasion abilities were assessed by transwell assay after knocking down miR-490-3p in KYSE30 cells. (G) Luciferase reporter assay for the luciferase activity of HMGA2-3 ′UTR WT or HMGA2-3 ’UTR mutant in cells cotransfected with miRNA mimics. (H) The invasion ability was evaluated by transwell Matrigel invasion assays (I) Knockdown of HMGA2 inhibits EMT (J) Western blot analysis comparing upregulated and downregulated- miR-490-3p ESCC cells with control cells were shown for vimentin, E-cadherin, N-cadherin and HMGA2. (K) The expression of HMGA2 and EMT markers was detected by Western blot after transfection with mimics or HMGA2 overexpression plasmids.* P<0.05,**P<0.01, **P<0.001.

HMGA2 have been reported to promote EMT, which was further verified in this study (Figure 5I). Additionally, miR-490-3p could regulate the expression of HMGA2 and EMT biomarkers (Figure 5J). Our further results indicated that the invasion ability inhibited by miR-490-3p was abolished by HMGA2 overexpression (Figure 5H). Similarly, upregulated miR-490-3p led to decreased expression levels of N-cadherin and vimentin, and this effect was reversed by HMGA2 (Figure 5K). Together, these results indicated that miR-490-3p inhibits EMT of ESCC by targeting HMGA2.

Hsa_circ_0006948 upregulates HMGA2 and induces EMT by sponging miR-490-3p

Next, we tried to analyze whether hsa_circ_0006948 exerts its cancerous effect on ESCC by sponging miR-490-3p, and a “rescue” experiment was performed to assess the functional interaction of “hsa_circ_0006948/miR-490-3p”. Compared with the migratory and invasive abilities of ESCC cells transfected with empty vector and miR-490-3p mimics, those of upregulated hsa_circ_0006948 cells transfected with miR-490-3p mimics significantly increased, suggesting that hsa_circ0006948 promotes ESCC progression and partly reverses the tumor suppressive effect of miR-490-3p (Figure 6A 6B). Further Western blot analysis indicated that knockdown of hsa_circ_0006948 could induce downregulation of HMGA2, vimentin and N-cadherin while increasing E-cadherin. Additionally, the above phenomenon was partially restored after treatment with a miR-490-3p inhibitor in downregulated hsa_circ_0006948 ECSS cells (Figure 6C). In contrast, the opposite effect on the expression of HMGA2, E-cadherin, vimentin, and N-cadherin was observed when miR-490-3p mimics were coinfected in hsa_circ_0006948 overexpressing cells (Figure 6D). Accordingly, the result revealed that hsa_circ_0006948 might act as a miR-490-3p sponge to regulate HMGA2 expression and induces EMT.

Figure 6. Overexpression of hsa_circ_0006948 reverses the suppressive roles of miR-490-3p in ESCC. The migration (A) and invasion (B) abilities inhibited by miR-490-3p were reversed after co-transfection with hsa_circ_0006948 using transwell assay. (C) The expression of levels of vimentin, N-cadherin and HMGA2 were inhibited by down-regulated hsa_circ_0006948, but this effect was reversed by miR-490-3p inhibitor presented in Western blot analysis. (D) The expression of levels of vimentin, N-cadherin and HMGA2 were enhanced by up-regulated hsa_circ_0006948, but this effect was reversed by miR-490-3p mimics presented in Western blot analysis. * P<0.05, **P<0.01, ***P<0.001.

Silencing of HMGA2 reverses the oncogenic effect induced by upregulation of hsa_circ_0006948

To investigate whether hsa_circ_0006948 could promote cancer progression and induce EMT by enhancing HMGA2, we transfected ESCC cells with HMGA2 siRNA and hsa_circ_0006948 overexpression plasmid. Cell proliferation assay indicated that hsa_circ_0006948 promoted cell viability of ESCC cells, while downregulation of HMGA2 reversed this effect (Figure 7A, 7B). Additionally, transwell invasion assays showed that cell invasion ability was enhanced by hsa_circ_0006948, but was restored by transfection of si-HMGA2 (Figure 7C). Moreover, knockdown of HMGA2 abrogated hsa_circ_0006948-induced increase on HMGA2, vimentin and N-cadherin, and decrease on E-cadherin (Figure 7D). Collectively, these results illustrated that HMGA2 plays a vital role in the hsa_circ_0006948/HMGA2/EMT aixs (Figure 7E).

Figure 7. Knockdown of HMGA2 abolishes the oncogenic effect induced by hsa_circ_0006948 in ESCC. (A and B) The cell proliferation was measured by CCK8 assays. (C) The invasion ability was evaluated by transwell Matrigel invasion assays. (D) The upregulation of vimentin, N-cadherin. HMGA2 and the downregulation of E-cadherin in TE-1 and KYSE30 cells transfected with hsa_ciic_0006948 overexpression plasmid were abolished by knockdown of HMGA2 as detected by Western blot analysis. (E) A mechanistic model: hsa_circ_0006948 functions as a miR-490-3p sponge and regulates HMGA2 through inhibiting miR-490-3p activity in ESCC cells' EMT. *P<0.05, **P<0.01, ***P<0.001.

Hsa_circ_0006948-miR-490-3p affects ESCC growth and proliferation in vivo

To further assess whether hsa_circ_0006948 exerts a tumor-promoting effect in vivo, a xenograft mouse model was established by subcutaneously injecting TE-1 cells (n=5 for each group). After 12 days, negative control, si-hsa_circ_0006948 and combined si-hsa_circ_0006948 and agomiR-490-3p were injected intratumorally every two days for two weeks. The results indicated that the tumor weight and growth rates were significantly lower in the si-hsa_circ_0006948 group than in the control group. Importantly, the tumor volume and weight were lower in the combined si-hsa_circ_0006948 and agomiR-490-3p group than in the si-hsa_circ_0006948 group alone. We further used IHC to evaluate the tumor tissues. The IHC results showed that the expression levels of HMGA2, N-cadherin and vimentin were significantly inhibited in the combined si-hsa_circ_0006948 and agomiR-490-3p group compared with those in control group or si- hsa_circ_0006948 group alone, which implied that silencing hsa_circ_0006948 combined with miR-490-3p overexpression exhibits an additive inhibitory effect on ESCC growth in xenograft animal models (Figure 8).

Figure 8. The images of tumor-bearing nude mice from three treatment groups (n = 5 for each group) on the 30^th day. (A) Tumor volumes were monitored with a caliper during the time course of 30 days, and tumor weights were also measured at the end of this study.(B) IHC analysis of E-cadherin, N-cadherin, vimentin and HMGA2. **P<0.01 vs control group, ## P<0.01 vs si-hsa_circ_0006948 group.

Ethical statement and tissue collection

This study was approved by the Ethics Committee of Sun Yat-sen Memorial Hospital. All clinical data were collected after each surgery. The esophageal cancer tissues and adjacent normal samples were obtained from patients who had undergone surgery at the Sun Yat-sen Memorial Hospital between January 1, 2014 and December 31, 2016. These 153 pairs of tumor and adjacent tissue specimens were frozen immediately and stored at −80 °C. After being confirmed by experienced clinical pathologists, the tumor and adjacent normal tissues were subjected to further analysis. We obtained written informed consent from all patients before their participation in this research.

Microarray data processing

circRNA Microarray (GSE131969), which have been deposited at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) previously, was used for the global profiling of human circRNAs. Differentially expressed circRNAs with statistical significance between the two groups were identified through volcano plot filtering. Differentially expressed circRNAs between the two sample types were identified through fold Change filtering. Distinguishable circRNA expression patterns among the samples were shown through hierarchical clustering.

Total RNA extraction and qRT-PCR

Total RNA was extracted with RNAiso Plus (TaKaRa Japan) in accordance with the manufacturer’s instructions. The purity and concentration of the total RNA samples were measured with a NanoDrop 2000 (Thermo Scientific, Wilmington, DE, USA). All cDNA was generated with PrimeScript RT Master Mix (TaKaRa, Japan).qRT-PCR for circRNAs was performed on a LightCycler® 96 System (Roche, Switzerland). The relative circRNA expression was calculated using:relative expression=2^-ΔCt or relative expression=2^-ΔΔCt.

Cell culture

ESCC cells (TE-1, KYSE30, TE-7, TE-13, KYSE150 cell lines) and normal epithelial esophageal cells (HEEC cell line) were purchased from the Shanghai Institutes for Biological Science, China. Briefly, esophageal cells were cultured in DMEM (Gibco; Thermo Fishier Scientific, Suzhou, China) with 10% fetal bovine plasma at 37 °C in a humidified atmosphere with 5% CO₂.

Oligonucleotide transfection

Before transfection, TE-1 and KYSE30 cells were seeded in 6-well plates to approximately 60% confluence. siRNA, miRNA mimics and inhibitors were transfected using a lipofectamine RNAiMAX transfection kit (Invitrogen, USA), and overexpression plasmids were transfected using a lipofectamine3000 transfection kit (Invitrogen, USA) following the manufacturer’s instructions.

Cell proliferation assay

To determine whether hsa_circ_0006948 is involved in ESCC cell proliferation, CCK8 assays were performed. TE-1 and KYSE30 cells transfected with si_circ_0006948 and overexpression vectors were seeded into 96-well plates at 5×10³ /well and cultured for 24, 48, 72, 96, 120 and 144 h. Cell proliferation was determined using a Cell Counting Kit-8 (CCK-8; Immuno Way Biotechnology Company Plano, TX, USA). OD450 values were then determined. Three biological repeats were performed for the statistical analysis. We used colony-forming assays to evaluate the clonogenic ability of transfected TE-1 and KYSE30 cells. Cells were seeded into 6-well plates (1000/well) and incubated for approximately 2 weeks. The visible colonies were counted after staining with crystal violet.

Transwell migration and invasion assays

Transwell assays were used to evaluate the migration and invasion of ESCC cells using transwell chambers (Costar, USA) precoated with or without Matrigel. For the assay, 2×10⁵ cells in serum-free medium were added to the upper chambers (pore size, 8 μm; Corning Inc., Tewksbury, MA, USA). DMEM with 10% fetal bovine serum was added to the lower chambers. After incubation for 24 h, the esophageal cancer cells migrated into the lower chambers were fixed in 4% paraformaldehyde and stained with a crystal violet staining solution. Random fields were digitally imaged and counted.

RNA fluorescence in situ hybridization

To study the location of hsa_circ_0006948, RNA fluorescence in situ hybridization (RNA-FISH) was performed using a Fluorescent in Situ Hybridization Kit (RiboBio, Guangzhou, China) according to the manufacturer’s guidelines. Cy3-labeled hsa_circ_0006948 probes were measured by the Fluorescent in Situ Hybridization Kit, followed by visualization with confocal microscopy.

RNA immunoprecipitation

RNA immunoprecipitation (RIP) assays were performed using an anti-AGO2 antibody with the transfection of miR-490-3p mimics (miR-490-3p) or miR-NC in TE-1 and KYSE30 cells to detect hsa_circ_0006948 expression according to qRT-PCR.

Luciferase reporter assay

Hsa_circ_0006948 sequences (or HMGA2-3’UTR sequences) containing wild-type or mutated miR-490-3p binding sites were synthesized and inserted into luciferase vectors respectively. HEK-293T cells were seeded into 24-well plates at a density of 3 × 10⁴ cells per well. After co-transfection with miR-490-3p mimics and constructed luciferase plasmids for 48 h, we measured luciferase activity using a dual-luciferase reporter assay system (Promega, USA) according to the manufacturer′s protocol. Relative luciferase activity was normalized to the Renilla luciferase internal control. Three independent experiments were performed in triplicate.

Western blot analysis

Proteins were extracted using RIPA buffer (CWBIO, China). After being electrophoresed by SDS-PAGE, the samples were transferred to nitrocellulose membranes and incubated with primary antibodies specific for E-cadherin (diluted 1:1000, Proteintech, USA), vimentin (diluted 1:1000, ABclonal, China), N-cadherin (diluted 1:1000, ABclonal, China), HMGA2 (diluted 1:1000, ABclonal, China) and GAPDH (diluted 1:1000, ABclonal, China) at 4 °C overnight. Then the membranes were incubated with secondary antibodies at room temperature for 1 h. Signals were detected with images acquisition using Immobilon ECL substrate (Millipore, Germany) and Optimax X-ray Film Processor (Protec, Germany).

Mouse xenograft model

Ethical approval was obtained from the Institutional Research Ethics Committee of Sun Yat-sen University. All animal care and procedures were performed in accordance with institutional guidelines. To assess the effect of hsa_circ_0006948 on tumor growth in xenograft models, TE-1 cells (5×10⁶/0.2 ml PBS) were injected subcutaneously into the right backs of 4-week-old BALB/c nude mice. After 12 days, the mice were treated with intertumoral injection of negative control, si-hsa_circ_0006948 and combined si-hsa_circ_0006948 and agomiR-490-3p every two days, respectively. The tumors were measured every three days and their volumes were calculated according to the following formula: tumor volume = (length × width²)/2. Thirty days later, the mice were sacrificed, and the tumors were excised for further immunohistochemistry. Primary antibodies against E-cadherin (Proteintech, USA), vimentin (ABclonal, China), N-cadherin (ABclonal, China) and HMGA2 (ABclonal, China) were used. We captured images using a Nikon Eclipse 80i system with NIS-Elements software (Nikon, Japan).

Statistical analysis

The relationship between hsa_circ_0006948 expression and clinicopathological features was analyzed by Chi-square test. Overall survival (OS) was assessed by Kaplan-Meier analysis and compared by log-rank test. Univariate and multivariate Cox proportional hazards regression analyses were used to analyze the independent prognosis factors. ROC curve were drawn to assess the diagnosis value and prognosis value of hsa_circ_0006948. P <0.05 was considered statistically significant. Data analyses were performed using PRISM (GraphPad Software Inc., San Diego, CA, USA), and Stata version 13.1 (StataCorp, College Station, TX).