A total of 76 patients diagnosed with primary PA in The First Affiliated Hospital (Xijing Hospital) of the Air Force Medical University between June 2006 and May 2010 were recruited in this study. The inclusion criteria were as following: (1) the primary tumor, (2) both radiological and pathological characteristics diagnosed with PA, and (3) no treatment of chemotherapy or radiotherapy. Frozen tissue specimens of PA were subjected to RNA examining after operation. This research was approved by the ethics committee of The First Affiliated Hospital (Xijing Hospital) of the Air Force Medical University. Informed consent was obtained from all the enrolled patients prior to recruitment. All tissue specimens were conducted and made anonymous by complying with the ethical and legal standards. The collected clinical samples were first flash-frozen in liquid nitrogen and then stored at −80°C refrigerator before examination.

HP75 cell line was obtained from the Institute of Biochemistry and Cell Biology, Chinese Academic of Sciences. Isolation of cells for primary culture of PA tissue was performed as described previously (10). Cells were cultured in high-glucose Dulbecco’s modified Eagle medium (DMEM; 11995040, Gibco, USA) supplemented with 10% fetal bovine serum (10099141C, Gibco, USA) in humidified incubator containing 5% CO₂ at 37°C. si-circDDX17 (5′-CCACAAATTTGGAGCAAGA-3′, knock-down circDDX17), si-circNC (5′-TTCTCCGAACGTGTCACGT-3′, negative control), si-CADM2 (5′-TACGTCCAAGGTCGGGCAGGAAGA-3′, knock-down CADM2), and si-NC (5′-CGTACGCGGAATACTTCGA-3′, negative control) were synthesized from GenePharma, Shanghai, China. To construct a circDDX17-overexpressing plasmid (circDDX17), we synthesized circDDX17 cDNA and cloned it into the pcD-ciR vector (Geneseed Biotech Co., Guangzhou, China). An empty plasmid served as the negative control (vector). All synthetic compounds were transfected by Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA).

Cytoplasmic and nuclear fractions from PA cells were isolated using the Cytoplasmic & Nuclear RNA Purification Kit (Norgen Biotek, Thorold, Canada).

Reverse transcription–quantitative PCR (RT-qPCR) was performed by SYBR Green PCR mixture (Invitrogen) with a gene-specific primer/probe mix and DNA template. Reaction mixtures were incubated at 95°C for 5 min, followed by 40 cycles of 95°C for 10 s, 60°C for 30 s, and 72°C for 1 s. Signals were detected at the end of each cycle. SDS 2.4 software (Applied Biosystems) was performed to calculate the cycle threshold (Ct) values. The 2^−ΔΔCT method was used to calculate relative expression. Additional File 1 , Table S1 lists the specific primers.

The genomics sequences including the wild-type (WT) binding site or the mutated (MUT) binding site were all cloned into the firefly luciferase reporter vector. The associated and regulated relationship of circDDX17 and miR-1279 was verified by dual-luciferase reporter assay (Promega, Madison, WI, USA). Another luciferase reporter assay was applied to investigate the targeting relationship of miR-1279 and CADM2. PA cells were co-transfected with WT or MUT luciferase reporter vector. After co-transfection, a dual-luciferase reporter system was applied to measure the luciferase activity following 48 h of incubation.

Total protein lysates were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA). Primary antibodies detecting targeted proteins were added in the blot and incubated for 12 h. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as control. Finally, the combined signals were detected by Chemistar™ High-sig ECL Western Blot Substrate (Tanon, Shanghai, China). The antibodies CADM2 (1:1,000, AB_2854219), Matrix metalloproteinase 2 (MMP2) (1:1,000, AB_2792342), MMP9 (1:1,000, AB_2809982), GAPDH (1:1,000, AB_2107310), and Horseradish Peroxidase (HRP)-conjugated Immunoglobulin G (IgG) anti-rabbit (1:10,000, AB_228341) were commercially acquired from Thermo Fisher Scientific (Waltham, MA, USA).

Lipofectamine™ 2000 was applied to transfect biotinylated control or circRNA probe into PA cells. RNA pull-down assay was performed by using the Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. The recruited mRNA was following examined by RT-qPCR analysis.

RNA immunoprecipitation (RIP) was performed by an RNA Binding Protein Immunoprecipitation kit purchased from Millipore. PA cells were lysed using RIP lysis buffer. Cell lysates were centrifuged at 15,000 × g for 20 min at 4°C, and the supernatant was incubated with magnetic beads coated with 10 μg of control IgG or antibody overnight at 4°C, followed by digestion with proteinase K buffer. The magnetic beads were precipitated by a magnetic bar and then washed three times with the washing buffer. At the last, TRIzol reagent was applied to extract the immunoprecipitation RNA examined by RT-qPCR analysis.

Approximately 4 × 10⁵ PA cells were seeded in six-well plates for wound healing assays. Then, a 200-μl pipette tip was applied to make a straight scratch. Every plate was washed with PBS more than three times and replaced with serum-free medium. Images of wounds were captured at the appropriate time to estimate the area occupied by migratory cells.

Transwell assay (Costar, New York, NY, USA) was performed to examine the invasion and migration capacities of PA cells in vitro according to the manufacturer’s protocol. Cells (4 × 10⁵) were collected and re-suspended in serum-free DMEM and then inoculated in the upper chamber, and 10% serum-containing medium was added to the lower chamber. In the migration assay, un-coated upper chambers were directly utilized for further analysis. The transwell was precoated with Matrigel (BD Biosciences) for cell invasion determination. After 24 h, the invaded or migrated cells were fixed with 4% paraformaldehyde for 10 min and 0.5% stained crystal violet for 20 min at room temperature. Finally, the filter membrane was photographed and counted in three random fields.

All animal experiment procedures were carried out according to the regulations and internal biosafety and bioethics guidelines of The First Affiliated Hospital (Xijing Hospital) of the Air Force Medical University. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. Six- to eight-week-old female BALB/c nude mice (Vital River, Beijing, China) were used in vivo detection. Thirty mice were randomly divided into two groups: vector and circDDX17. Transfected HP75 cells (5 × 10⁶ per mice) were injected into the right flanks of mice. After 28 days observation, the mice were performed euthanasia by CO₂, and subcutaneous tumors were extracted for further experiments.

SPSS version 13.0 (SPSS, Chicago, IL, USA) was applied to assess and analyze statistical data. Paired two-tailed Student’s t-test or Chi-square test was performed to analyze the significance between groups. The Pearson’s correlation coefficient analysis was performed to analyze the correlationships. Log-rank test was used to discover the significance, and Kaplan–Meier curves were for survival analysis. Statistical significance was determined as P < 0.05 (*) or 0.01 (**).

To identify the circRNAs involved in the occurrence and progression of PA, RT-qPCR was initially applied to detect the relative expression levels of circDDX17 as demonstrated in Figure 1 . Analysis of the expression circDDX17 in patients with PA (n = 76) revealed a significant difference in contrast to healthy controls (n = 15) after normalization (p < 0.01; Figure 1A ). Further experiments confirmed a decreasing tendency of circDDX17 in the invasive group compared with the non-invasive group (p < 0.01; Figure 1B ). In the subsequent analysis, we conducted and analyzed the associations between clinic-pathological characteristics and levels of circDDX17 expression. The significantly statistical differences of circDDX17 expression were observed among different Knop stages, tumor size, and recurrent status; however, no obvious differences were found in age and gender ( Table 1 ). To evaluate the accuracy of circDDX17 levels for the diagnosis of patients with PA, Receiver Operating Characteristic Curve (ROC) curve analysis and the corresponding area under curve (AUC) values were calculated to discriminate PA from the general population. ROC analyses exhibited that circDDX17 levels could act as a valuable biomarker to distinguish patients with PA from healthy controls, with AUC value of 0.9632 (95% CI, 0.9251–1.000) ( Figure 1C ). Subsequently, to analyze the effect of circDDX17 expression to the prognosis of patient with PA, all recruited patients were divided into two groups: low circDDX17 group (n = 38) and high circDDX17 group (n = 38) according to the median value of circDDX17 expression ( Figure 1D ). Survival analysis demonstrated a low level expression of circDDX17 correlated with poor progression-free survival (PFS) in contrast to the high expression of circDDX17 ( Figure 1E ).

The above results have suggested that circDDX17 was downregulated in PA tissues and correlated with a poor clinical prognosis; next, we performed a series of experimental assays to explore the function of circDDX17 in PA progression. We transiently transfected circDDX17 vector to increase its expression and explore the changes of biological functions in HP75 cells and primary PA cells ( Figure 2A ). As depicted in Figure 2B , upregulation of circDDX17 significantly suppressed cell migration analyzed by wound healing assay. Increasing circDDX17 effectively restrained the migrated ability of PA cells compared with control vector group ( Figure 2C ). Consistently, Matrigel invasion assays revealed that the invasive abilities of PA cells were also inhibited by overexpression of circDDX17 ( Figure 2D ). Moreover, si-circDDX17 was transfected into HP75 cells and primary PA cells to knockdown circDDX17 expression ( Figure 2E ). Knockdown of circDDX17 expression obviously increased the migration and invasion capabilities in contrast to si-circNC control group in PA cells as consistent with the circDDX17 overexpression tendency ( Figures 2F–H ). Together, these results indicated that circDDX17 could significantly inhibit PA cells invasive and migrating capacity.

As demonstrated in Figure 3A , circDDX17 was predominantly localized in the cytoplasm, occasionally in the nuclei of HP75 cells and primary PA cells. As circRNAs were reported as molecular sponges of miRNAs to regulate expression of targeted miRNAs, the publicly available “circBank” and “circInteractome” databases tools were employed to predict potentially possible binding miRNA candidates of circDDX17. miR-1279 was predicted by both circRNA database tools and chosen to further verification. Next, the expression levels of miR-1279 were examined in PA specimens by RT-qPCR, and the association of circDDX17 and miR-1279 was assessed. As depicted in Figure 3B , the expression of miR-1279 in PA tissues was obviously increased compared with the normal tissues. Pearson’s correlation coefficient analysis confirmed a significant negative correlation between the expression of circDDX17 and miR-1279 in PA tissues (r = −0.745, p < 0.01; Figure 3C ). The biotin-labeled probe pull-down assay was implemented to investigate the directly binding interactions between circDDX17 and miR-1279, and the results revealed abundant enrichment of miR-1279 when treated with circDDX17 probe in contrast to the control oligo probe ( Figures 3D, E ). Furthermore, AGO2 RIP assay also demonstrated that AGO2-containing complex could accumulate both circDDX17 and miR-1279 in large amount compared with IgG-control ( Figure 3F ). Subsequently, the putative binding sites between circDDX17 and miR-1279 were predicted to further confirm the sponging function ( Figure 3G ). Dual-luciferase reporter assay demonstrated that the presence of miR-1279 mimic significantly inhibited the luciferase activity of the WT circDDX17 reporter, whereas no obvious change in the MUT reporter ( Figure 3H ). Collectively, the series findings proved that circDDX17 as a specific miRNA sponge could negatively regulate and directly bind to miR-1279 in PA.

As shown in Figure 4A , upregulating the expressions of circDDX17 significantly inhibited the proliferation of PA cells, whereas decreasing circDDX17 levels promoted the cell proliferation compared with the corresponding control group. Overexpressed miR-1279 enhanced the proliferative ability when transfected with miR-1279 mimics ( Figure 4B ). In addition, rescued experiments were designed to confirm the inhibiting effects of circDDX17 through sponging miR-1279 and examined the aggressive biology by co-transfecting circDDX17 vector and miR-1279 mimics in HP75 cells and primary PA cells. Remarkably, upregulated miR-1279 levels could partially restrain the tumor-suppressing effects of circDDX17 in the migratory and invasive abilities ( Figures 4C–E ). Moreover, knockdown of circDDX17 expression obviously increased the migration and invasion capabilities, and the promoting effects were also alleviated by miR-1279 inhibitors ( Figures 4F–H ). Consistently, these results suggested that circDDX17 regulated PA cells’ migratory and invasive abilities through sponging miR-1279.

To explore the underlying mechanism of miR-1279 and predict the downstream targets, web-based software program, TargetScan, was applied and CADM2 searched as a candidate gene. In the beginning, the expression levels of CADM2 were examined in PA specimens by RT-qPCR, and the association of circDDX17 and CADM2 was evaluated. As demonstrated in Figures 5A, B , the expression of CADM2 in PA tissues was obviously reduced compared with the normal tissues and decreased even much more in the invasive group. Survival analysis demonstrated a low level expression of CADM2 correlated with poor survival, and, next, Pearson’s correlation coefficient analysis was applied to confirm a significant positive correlation between the expression of circDDX17 and CADM2 in PA tissues (r = 0.711, p < 0.01; Figures 5C, D ). Upregulation of circDDX17 levels significantly increases the mRNA and protein expression ( Figures 5E, F ). Next, we investigated whether circDDX17 interfered the expression of CADM2 through regulating intermediate factor miR-1279. The correlation analysis displayed inverse association between miR-1279 and CADM2 (r = −0.775, p < 0.01; Figure 5G ). As TargetScan calculated, the 3′ Untranslated Region (3′UTR) of CADM2 contained the putative binding sites of miR-1279 ( Figure 5H ). In addition, dual-luciferase reporter assay and Western blot confirmed the negative regulation and directly binding correlation between miR-1279 and CADM2 ( Figures 5I, J ). Moreover, upregulated circDDX17 also caused a significantly increasing on the expression level of CADM2 protein, whereas miR-1279 partly reversed the enrich containing of CADM2 ( Figure 5K ). Together, these data demonstrated that circDDX17 suppressed the invasive progression by eliminating miR-1279 and then interfering CADM2 expression.

In this section, we were aimed to explore whether circDDX17 participated in the biologically invasive prognosis by forming circDDX17/miR-1279/CADM2 regulating axis and ultimately affecting the activity of CADM2. We co-transfected circDDX17 vector and CADM2 siRNA to explore the changes of biological functions in HP75 cells and primary PA cells. As expected, declined CADM2 containing significantly attenuated the migration suppressing effects of circDDX17 enrichment in PA cells analyzed by wound healing assay ( Figure 6A ). Analogously, increasing circDDX17 effectively inhibited the migrated and invasive abilities of PA cells in the previous results, whereas CADM2 downregulation also overrode this suppression effects ( Figures 6B, C ). Furthermore, increasing circDDX17 levels inhibited the expression of MMP2 and MMP9 levels, which was also reversed by CADM2 knockdown ( Figure 6D ). Collectively, these data indicated that the significant effects of circDDX17 in invasive ability were mediated by the activity of CADM2 expression via sponging miR-1279.

HP75 cells were used to establish subcutaneous tumors for the further experiment in vivo. Xenograft growth assay indicated that overexpression circDDX17 could significantly inhibit tumor growth compared with the control groups from the second week ( Figure 7A ). RT-qPCR was used to examine solid tumors, and the levels of circDDX17 were upregulated, whereas the levels of miR-1279 were downregulated in the group of transfected with circDDX17 ( Figure 7B ). In addition, immunohistochemical analysis was performed to examine the downstream proteins. CircDDX17 overexpression significantly increased the downstream target of CADM2 expression, whereas migration-related proteins MMP2/9 were inhibited ( Figure 7C ). In the previous study, we have demonstrated that the levels of MMP2/9 were correlated with high invasive ability to cavernous sinus in PA (11). In addition, hematoxylin and eosin staining indicated that overexpression of circDDX17 could inhibit invasion of PA cells in vivo ( Figure 7D ). In conclusion, these data supported that overexpression of circDDX17 efficiently inhibits the growth and invasion of PA bearing subcutaneous tumors in vivo.