Patient characteristics

From January 2013 to December 2014, 49 pairs of LAC samples were collected. Among these patients (Table S1), 19 were female and 30 were male. The mean age was 62.0 (range: 46–77) years, with 22 patients aged younger than 60 years and 27 older than 60 years. Among the 49 patients, 12 (24.49%), 16 (32.65%), 20 (40.82%) and 1 (2.04%) had LACs classified as stage I, II, III and IV, respectively.

General profiles of the circRNA microarray

Using a microarray technique, we evaluated the circRNA profiles of three paired LAC samples. The results of hierarchical clustering show distinct circRNA expression profiling among the samples (Fig. 1A). A Volcano plot was constructed to visualize the differential expression between tumour and nontumour tissues (Fig. 1B). A fold change > 2.0 and P < 0.05 were considered to indicate a significant difference. Thus, the transcription of a total of 59 circRNAs was altered, including 39 up-regulated and 20 down-regulated circRNAs, in patients with LAC compared to adjacent normal controls (Tables S2 and S3).

Subsequently, we examined the most up-regulated circRNA, hsa_circ_0013958, which is located at chromosome 1q21 and is composed of three exons (Fig. 1C). A set of specific divergent primers were designed for hsa_circ_0013958 PCR (Table S4). PCR was performed using the circRNA as the template, and the melting curve of the yielded products was a single peak (Fig. 1D), indicating the specificity of the primers. In addition, PCR with genomic deoxyribonucleic acid (gDNA) as the template yielded no products (Fig. 1E) and the sequences were completely consistent with those from circBase (http://www.circbase.org; Fig. 1F). These results show that these circRNAs could be specifically amplified by RT-PCR. To minimize the effects of degradation of the formalin-fixed paraffin-embedded (FFPE) tissues, we used short amplicon primers (~ 60–70 bp) [13] and our results show that the total RNAs were of high quality for detection. The primers were applied in the subsequent experiments.

Hsa_circ_0013958 level was up-regulated in LAC

To identify hsa_circ_0013958, we first examined its expression in LAC cells and tissues using real-time PCR. It was confirmed that hsa_circ_0013958 expression was increased in seven LAC cell lines compared to the normal human bronchus epithelium cell line BEAS-2B (control; Fig. 2A). We next assessed hsa_circ_0013958 expression in LAC tissues. The hsa_circ_0013958 levels were high in 87.76% of LAC tissue samples (43 of 49; Fig. 2B–E). In addition, we measured another 30 plasma samples from LAC patients and found high levels of hsa_circ_0013958 in 90.00% of them (with U6 as the reference gene; P < 0.001; Fig. 3F,G). These data indicate that abnormal hsa_circ_0013958 expression may be related to LAC progression.

Next, we aimed to determine whether a high level of hsa_circ_0013958 in patients was associated with clinicopathological parameters. As shown in Table 1, 49 patients were grouped into high-level (n = 39) and low-level (n = 10) groups using an hsa_circ_0013958 cut-off value of 0.00101. hsa_circ_0013958 expression showed no significant differences between groups of different sexes (P = 0.931), ages (P = 0.723), smoking/nonsmoking (P = 0.588), tumour sizes (P = 0.076) or tumour differentiation stages (P = 0.474). TNM stage was correlated with the hsa_circ_0013958 expression level using a t-test (P = 0.009). Based on the TNM stage, hsa_circ_0013958 expression was significantly associated with patient pathogenesis (P < 0.001; Fig. 2D). There were 58.33% (seven of 12) of patients with a high level of hsa_circ_0013958 at stage I, and this rate increased to 75.00% (12 of 16) and 95.23% (20 of 21) at stages II and III–V, respectively. Additionally, the hsa_circ_0013958 level was closely related to lymphatic metastasis (P < 0.001; Fig. 2E).

Diagnostic accuracy of hsa_circ_0013958 for LAC

A receiver operating characteristic (ROC) curve analysis was used to analyze the diagnostic accuracy of hsa_circ_0013958. When a comparison was made between the cancer group and the normal controls, the area under the ROC curve (AUC) was 0.815 [95% confidence interval (CI) = 0.727–0.903; P < 0.001]. Using 0.00101 as a cut-off value for the hsa_circ_0013958 relative expression level, the sensitivity and specificity of hsa_circ_0013958 for the diagnosis of LAC were 0.755 and 0.796, respectively (Fig. 3A). Furthermore, the TNM subgroups were analyzed, and the AUC of stages III–IV was the highest in all subgroups. Overall, these results indicate that the performance of hsa_circ_0013958 was better in the late stage of LAC compared to that in the early stage (Fig. 3B–D). In addition, plasma hsa_circ_0013958 distinguished LAC cases from healthy controls with an AUC of 0.794 (95% CI = 0.703–0.912; P < 0.001; Fig. 3E).

The biological functions of hsa_circ_0013958 in LAC cells

We then treated A549 and H1299 cells with hsa_circ_0013958 small interfering RNA (siRNA) oligos and the efficiency of the three oligonucleotides (siRNA1, siRNA2 and siRNA3) was examined using real-time PCR 24 h after transfection. The results suggest that the siRNA oligos could knockdown hsa_circ_0013958 expression (Fig. 4A,B). In addition, siRNA1 was the most effective siRNA, and thus it was selected for the subsequent experiments.

To explore whether hsa_circ_0013958 was involved in cellular proliferation, we performed CCK8 and 5-ethynyl-2′-deoxyuridine (EDU) assays. As shown in Fig. 4C,D, siRNA-1 significantly inhibited LAC cell proliferation, whereas the control siRNA without hsa_circ_0013958-binding activity had no effect on cell proliferation. Similar results were obtained with the EDU assay (Fig. 4E). These data reveal that knockdown of hsa_circ_0013958 suppressed proliferation in LAC cells. Next, we determined whether hsa_circ_0013958 suppressed cell apoptosis in LAC. Two different LAC cell lines, A549 and H1299 cells, were transfected with hsa_circ_0013958 siRNA or negative control, and apoptosis of LAC cells was determined using flow cytometry. Induction of cell apoptosis was observed after silencing hsa_circ_0013958 in LAC cells (Fig. 4F), as expected.

Considering that hsa_circ_0013958 was closely related to lymphatic metastasis, transwell assays were performed with LAC cells transfected with either hsa_circ_0013958 siRNA or negative control siRNA. We further determined whether hsa_circ_0013958 promoted cell migration and invasion in LAC cells. Inhibition of cell migration and invasion was observed after silencing hsa_circ_0013958 in A549 and H1299 cells, as expected (Fig. 4G,H). These results demonstrate that hsa_circ_0013958 promoted cell migration and invasion in LAC cells. As the representative micrographs clearly demonstrate, silencing hsa_circ_0013958 led to potent inhibition of cell migration and invasion.

The mechanism of hsa_circ_0013958

To determine the subcellular localization of hsa_circ_0013958, we performed quantitative RT-PCR and fluorescence in situ hybridization (FISH). The results demonstrate that the circular form of hsa_circ_0013958 preferentially localized in the cytoplasm (Fig. 5). It has been proposed that circRNAs in the cytoplasm may act as competing endogenous RNAs to bind miRNAs [14, 15]. Thus, we speculated that hsa_circ_0013958 could target miRNAs to inhibit their expression.

We utilized two public databases [circbase (http://www.circbase.org) and starbase, version 2.0 (http://starbase.sysu.edu.cn/)] to screen for targeted miRNAs and the results obtained show that miR-545-3p, miR-134-5p, miR-660-3p, miR-509-3p and miR-629-3p had a binding site for hsa_circ_0013958 (Fig. 6). To identify the miRNAs that bind to hsa_circ_0013958, we performed a luciferase screening assay. Each miRNA mimic was co-transfected with the luciferase reporters into HEK-293T cells. Compared to the control RNA, only miR-134-5p reduced the luciferase reporter activity by at least 50% (Fig. 7A). Interestingly, hsa_circ_0013958 did not affect the expression of miR-134 (Fig. 7B). In addition, miR-134 expression was negatively correlated with hsa_circ_0013958 expression in LAC tissues (r = −0.771, P < 0.001; Fig. 7C). Knockdown of hsa_circ_0013958 inhibited the proliferation of A549 and H1299 cells. However, silencing miR-134 reversed the effect of hsa_circ_0013958 (P < 0.05; Fig. 7D). These results support the idea that hsa_circ_0013958 promotes the proliferation of LAC cells by inhibiting miR-134. These results also suggest that hsa_circ_0013958 may function as a sponge for miR134.

It has been reported that miR-134 suppresses NSCLC development through down-regulation of cyclin D1 (CCND1) [16] and, thus, we predicted that hsa_circ_0013958 exerts its functions via miR-134-CCND1. Western blotting showed that ectopic miR-134 reduced the levels of CCND1 and also that hsa_circ_0013958 increased the protein levels of CCND1 in LAC cells (Fig. 7E). These results confirm that miR-134 was a sponge for hsa_circ_0013958 and subsequently up-regulates CCND1 protein.

Patients and specimens

The methodologies used in the present study were approved by the Research Ethics Board at Zhongda Hospital. The experiments were undertaken with the understanding and written consent of each subject. To screen candidate circRNAs, fresh tumour samples of primary LAC and noncancer tissues were collected in Zhongda Hospital. Furthermore, in total, 49 paraffin-embedded LAC samples collected between June 2013 and December 2014 were examined to validate the candidate. Another 30 plasma samples from LAC patients and 30 from healthy volunteers were collected and examined to validate hsa_circ_0013958. Patients who had received preoperative radiotherapy or chemotherapy were excluded from the study. The patients were classified according to the National Comprehensive Cancer Network (NCCN) tumour classification by two pathologists and independently reviewed by an expert LAC pathologist. The study methodologies conformed with the standards set by the Declaration of Helsinki.

Total RNA isolation

Total RNA was isolated from tumour and adjacent normal tissues using a TRIzol kit (Invitrogen, Carlsbad, CA, USA) in accordance with the manufacturer's instructions. Total RNA inewas isolated using an RNAprep Pure FFPE Kit in accordance with the manufacturer's instructions (TIANGEN, Beijing, China). The purity and concentration of the total RNA samples were measured with a BioMate 3S spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

Microarray processing

An Arraystar Human circRNA Microarray, version 2.0 (Arraystar Inc., Rockville, MD, USA) was used for the global profiling of human circRNAs. Sample preparation and array hybridization were performed in accordance with the manufacturer's instructions. Briefly, total RNA was digested with Rnase R (Epicentre, Inc., Madison, WI, USA) to remove linear RNAs. Then, the enriched circular RNAs were amplified and hybridized. circRNAs differentially expressed between two samples were identified via fold change filtering. Hierarchical clustering was performed to show the distinguishable circRNA expression pattern among samples.

Quantitative RT-PCR

Quantitative RT-PCR of circRNAs and the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene was performed on a CFX96 Detection System (Bio-Rad, Hercules, CA, USA) using an RT Kit (TakaRa, Dalian, China) and PCR Master Mix (TakaRa) in accordance with the manufacturer's instructions. Briefly, 500 ng of total RNA was reverse transcribed into cDNA with random primers in a total volume of 20 μL. The reactions were initiated in a 96-well optical plate at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s and 60 °C for 20 s. The primers used for real-time PCR are shown in Table S4.

Cell culture

The cells used in the present study were from Cobioer Biosciences Co., Ltd. (Nanjing, China). All of the cells were cultured in RPMI1640 medium (HyClone; GE Healthcare Life Sciences, Milwaukee, WI, USA) with 10% fetal bovine plasma. Culture plates were incubated at 37 °C with 5% CO₂. The siRNA specifically targeting hsa_circ_0013958 was constructed by GenePharma (Shanghai, China) and the sequence is given in Table S5. A549 and H1299 cells with a high level of hsa_circ_0013958 were transfected with siRNA targeting hsa_circ_0013958 using siRNA-Mate (GenePharma) in accordance with the manufacturer's instructions. The over-expression vector of hsa_circ_0013958 (pLCDH-ciR-hsa_circ_0013958) was constructed by Geneseed Co., Ltd. (Guangzhou, China). The miR-134 mimic and inhibitor were obtained from RiboBio Co., Ltd. (Guangzhou, China).

Cell proliferation assay

To measure whether hsa_circ_0013958 is involved in LAC cell proliferation, we performed CCK8 and EDU assays. A549 and H1299 cells were transfected with siRNA for 48 h and, after transfection, cells were seeded into 96-well plates at a density of 2 × 10³ cells per well and cultured for 24, 48, 72 or 96 h. Cell proliferation was determined using Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan). A₄₅₀ was monitored. All of the experiments were performed in triplicate.

For the EDU assays, 48 h after seeding, EDU was added to a final concentration of 10 μm. After the indicated incubation time with EDU, cells were fixed for 15 min with 4% formaldehyde solution in PBS. Labelling of EDU with an azide derivative of Apollo 643 was performed using an C10310-2Cell-Light™ EDU Apollo®643 In Vitro Imaging Kit (RiboBio Co., Ltd). For excitation of Apollo 643, a 652 nm laser was used. Confocal laser-scanning microscopy images were obtained with an Fluoview FV1000 confocal laser-scanning microscope (Olympus, Tokyo, Japan) using a ×100 quartz objective. Composite images were prepared using imagej, version 1.41 (National Institutes of Health, Betheswda, MD, USA).

Cell migration and invasion assays

To examine the possible effect on metastasis of LAC cells, migration and invasion transwell assays were used. For the migration test, 1 × 10⁴ cells in 200 μL of plasma-free medium were placed in the upper chamber of the transwell (pore size, 8 μm; Corning Inc., Tewksbury, MA, USA). For the invasion test, 2 × 10⁴ cells in 200 μL of plasma-free medium were placed in the upper chamber, which was coated with Matrigel (BD Biosciences, San Joses, CA, USA). RPMI medium containing 20% fetal bovine serum was added to the lower chamber. After the cells had been incubated for 24 h at 37 °C, the cells remaining in the upper membrane were discarded and those on the lower surface of the membrane were fixed and stained with crystal violet. Ten random fields were counted in the high-power field under a microscope. Three wells were measured to determine cell viability in each group.

Flow cytometry analysis

Cells were harvested at a density of 1 × 10⁶ cells·mL⁻¹, fixed with 70% ice-cold ethanol, stained with 400 μL of propidium iodide solution (Keygen, Nanjing, China) for 30 min, and then subjected to cycle analysis using a flow cytometer (BD Biosciences).

Western blotting analysis

Western blotting was performed as described previously. In brief, samples were collected and electrophoresed on a 12% Tris-HCl gel and transferred to nitrocellulose membranes (Millipore, Billerica, MA, USA). After treatment with blocking buffer, the membranes were incubated with a primary antibody overnight at 4 °C and then with a secondary antibody for 1 h each at room temperature. The antibodies used were: anti-CCND1 (dilution 1 : 1000; Cell Signaling Technology, Beverly, MA, USA) and goat anti-rabbit secondary antibody (dilution 1 : 5000; Cell Signaling Technology). The results were normalized to the internal control β-actin, and the results were quantified and the images processed using imagej, version 1.41 (National Institutes of Health).

Luciferase reporter assay

HEK293T cells transfected with hsa_circ_0013958 or hsa_circ_0013958-mutant were seeded in 96-well plates and co-transfected with the reporter vector and 50 nm miR-NC, miR-545-3p, miR-134-5p, miR-660-3p, miR-509-3p or miR-629-3p mimic using Lipofectamine 2000 (Life Technologies, Grand Island, NY, USA). The five miRNA mimics were obtained from RiboBio Co., Ltd. Firefly and Renilla luciferase activities were measured with a Dual-Luciferase Reporter System (Promega, Madison, WI, USA) 48 h after transfection. The effect of each miRNA on the luciferase reporter with hsa_circ_0013958 was then normalized to that of the luciferase reporter with hsa_circ_0013958-mutant. Finally, the fold change between each miRNA compared to NC was calculated.

RNA-FISH

The RNA-FISH procedure was performed as described previously [39]. In brief, A549 cells were grown to 80–95% confluence at the time of fixation. After pre-hybridization, cells were hybridized in hybridization buffer with digoxigenin (DIG)-labelled probes specific to hsa_circ_0013958 at 60 °C overnight. A double FISH assay was performed in A549 cells after co-transfection with hsa_circ_0013958 and miR-134 expression vectors. Biotin-labelled probes specific to hsa_circ_0013958 and Dig-labelled locked nucleic acid miR-134 probes were used in the hybridization. The signals of biotin-labelled probes were detected using Cy5-Streptavidin (Life Technologies). The signals of Dig-labelled locked nucleic acid miR-134 probes were detected using a tyramide-conjugated Alexa 488 fluorochrome TSA kit. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). The images were obtained with a confocal microscope (Olympus).

Statistical analysis

The circRNA expression levels were normalized to the reference gene GAPDH. The relative circRNA expression between tumour and normal samples was calculated using the equation ▵Cq = Cq (target) – Cq (GAPDH); relative expression was calculated using: relative expression = 2^−▵Cq or relative expression = 2^−▵▵Cq. Data are expressed as the mean ± SD and statistical analysis was performed using anova followed by a t-test. Overall survival was defined as the period from the date of surgery to the date of death or last follow-up, and disease-free survival was calculated from the date of surgery to the date of recurrence or the date of last follow-up. The distribution of the time-to-event variables was estimated using the Kaplan–Meier method with log-rank testing. Both univariate and multivariate analyses were performed to calculate the association between clinical features and overall survival and disease-free survival using Cox proportional hazard models. P < 0.05 (two-sided) was considered statistically significant for all of the statistical calculations. All statistical analyses were performed with prism (GraphPad Software Inc., San Diego, CA, USA) and spss (IBM Corp., Armonk, NY, USA).