Fresh NSCLC tissue samples and corresponding para-carcinoma nonmalignant tissue samples were collected at Sun Yat-sen University Cancer Center and immediately stored in liquid nitrogen. Total RNA was isolated and subjected to circRNA microarray analysis. The research protocols involving human samples and animals were approved by the Ethics Committee of Sun Yat-sen University Cancer Center and were performed according to the Declaration of Helsinki. All patients were informed and signed consent forms.

Normal lung cells (Beas2b cells) and NSCLC cells (PC9, H1299, H1975 and A549 cells) were purchased from ATCC (USA). Cell authenticity was confirmed by DNA fingerprinting. Assays were routinely performed to detect mycoplasma infection.

The Lipofectamine 3000 system (Invitrogen, USA) was used to perform transfection. circUCP2 siRNAs, miR-149 mimics, and inhibitors were supplied by Ruibo (China). The siRNA sequences were as follows: si-NC, UUCUCCGAACGUGUCACGUTT; si-circUCP2 #1, ACCTTGGGGACCTCTCCCAAT; si-circUCP2 #2, AGCTCAACCTTGGGGACCTCT; si-circUCP2 #3, AACCTTGGGGACCTCTCCCAA.

Total RNA isolation was completed using TRIzol (Sigma, USA). Nuclear and cytoplasmic RNA isolation was completed by Beyotime (China). Takara PrimeScript^TM RT reagent Kit and TB Green Premix Ex Taq^TM (Japan) were used for RT‒qPCR assays. circRNA levels were quantified via a method that was similar to the method for quantifying of mRNA levels. RNase R removal eliminated the effects of linear RNA. The 2^−ΔΔCt method was applied to calculate the expression levels of circRNAs and mRNAs. U6 and GAPDH were used as the controls for circRNA and mRNA expression, respectively. The following oligonucleotide primers were supplied by Ruibo (China): GAPDH, Forward (5′-3′), TGTTCGTCATGGGTGTGAAC, Reverse (5′-3′), ATGGCATGGACTGTGGTCAT; 18s, Forward (5′-3′), TTAATTCCGATAACGAACGAGA, Reverse (5′-3′), CGCTGAGCCAGTCAGTGTAG; UCP2, Forward (5′-3′), GGAGGTGGTCGGAGATACCAA, Reverse (5′-3′), ACAATGGCATTACGAGCAACAT; circUCP2, Forward (5′-3′), TCAGCCAGAATCTTCGTCCT, Reverse (5′-3′), CATAGGTCACCAGCTCAGCA.

RNA (2 μg) was isolated from PC9 and H1975 cells and incubated with or without 3 U/μg RNase R at 37°C for 20 min. Then, the RNAs were purified with an RNeasy MinElute Cleanup Kit (Qiagen, USA) and subjected to qRT‒PCR.

To block transcription, PC9 and H1975 cells were treated with 2 μg/mL actinomycin D (Sigma, USA) for 8, 16, and 24 h. Then, the cells were harvested, and the stability of circUCP2 and UCP2 mRNA was measured by qRT‒PCR.

Ninety-six-well plates were seeded with 10³ cells after transfection, and then, the cells were incubated under proper conditions for 48 h. Then, 10 μL CCK-8 solution (Dojindo, Japan) was added to the culture medium and incubated for 2 h. The OD values were measured at 450 nM on a Bio-Tek EPOCH2 microtiter plate reader (USA), and the data were recorded.

Six-well plates were seeded with 10³ cells after transfection. After incubation for 14 days at 37°C, when the clones were visible to the naked eye, the clones were fixed with paraformaldehyde, stained with crystal violet, and counted under a microscope.

For the invasion assay, 10⁴ cells were seeded into upper chambers that were precoated with Matrigel (BD Bioscience, USA) and incubated in serum-free medium. Medium supplemented with 10% fetal bovine serum was added to the lower chambers. Twenty-four hours later, the cells that had invaded were fixed in cold methanol and stained with 1% crystal violet, and then, the number of stained cells were counted under a microscope.

H1975 cells (2 × 10⁶) were subcutaneously administered into male nude mice (n = 5, five weeks old). Then, si-NC and si-circUCP2 were injected into the tumors every four days (40 μL). Twenty-eight days later, the tumors were excised, weighed, and subsequently subjected to pathological studies. A total of 10⁵ H1975 cells were injected into each mouse through the tail vein (n = 5, five weeks old). After 8 weeks, the lungs were excised for subsequent pathological studies. The number of metastatic sites in the lungs were counted under a microscope.

A 96-well plate was seeded with 3 × 10⁴ cells per well. Luciferase reporter plasmids carrying wild-type or mutant sequences of circUCP2 or UCP2 3′-UTR, miR-149 mimics, miR-149 inhibitors and their controls were purchased and cotransfected into cells. The activities of Renilla and firefly luciferase were measured 48 h later (Promega, USA).

MS2bs-circUCP2, MS2bs-circUCP2-mt and MS2bs-Rluc were transfected into NSCLC cells. Forty-eight hours later, the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA) was used for the RIP test. The beads were purified using the RNA-quick Purification Kit (Yishan, China) to obtain the eluted RNAs. qRT‒PCR assays were used to measure the levels of circUCP2, UCP2 and miR-149.

Statistical analyses were performed with SPSS 20.0 software. t-tests were used to compare the differences between groups. All the data are presented as the mean ± SD. Statistical significance was set at p < 0.05.

A high-throughput circRNA microarray from four matched NSCLC tissue samples and neighboring nonmalignant tissue samples was used to analyze the expression of circRNAs. The 20 most significantly upregulated and 20 most downregulated circRNAs are shown in Fig. 1A. We established that the hsa_circ_0023525 content was higher in NSCLC tissues than in neighboring nonmalignant tissues. According to the human reference genome (GRCh37/hg19), hsa_circ_0023525 is located at chr11:73685715-73687787 and originates from the uncoupling protein 2 (UCP2) gene. Thus, we named it “circUCP2”. Further experiments showed that circUCP2 is expressed at high levels in NSCLC cells (Fig. 1B). Then, we assessed circUCP2 expression in 40 pairs of NSCLC tissues and their controls. The data showed that circUCP2 was overexpressed in NSCLC tissues (Fig. 1C). To further confirm that circUCP2 is a circRNA, its resistance to digestion by the exonuclease RNase R was determined. As shown in Fig. 1D, the expression levels of the circRNA isoform transcript circUCP2 were not affected by RNase R, while the expression levels of the linear transcript UCP2 were significantly decreased. Moreover, actinomycin D assays further confirmed that the half-life of circUCP2 was more than 24 h, showing that it was far more stable than the linear transcript UCP2 in NSCLC cells (Fig. 1E).

We used siRNAs to knockdown the expression of circUCP2, and si-circUCP2#1 was selected for subsequent experiments (Fig. 2A). We carried out a CCK-8 assay and established that circUCP2 silencing suppressed cell proliferation (Fig. 2B). Additionally, knockdown of circUCP2 inhibited NSCLC cell colony formation (Fig. 2C and 2D). Moreover, the invasive ability of NSCLC cells was repressed when circUCP2 was knocked down (Figs. 2E and 2F). Additionally, we measured the expression of MMP2 and MMP9, which are indicators of cancer metastasis. Knockdown of circUCP2 suppressed the expression levels of both MMP2 and MMP9 in NSCLC cells (Fig. 2G).

Silencing of circUCP2 repressed growth ability in mouse xenograft model experiments (Figs. 3A and 3B). Further immunohistochemistry (IHC) staining for Ki67 showed that Ki67 expression was decreased in circUCP2-knockdown tumor tissues, indicating that circUCP2 inhibition reduced cell proliferation in xenograft tumors (Fig. 3C). Moreover, knockdown of circUCP2 repressed metastasis in mice (Figs. 3D and 3E). In addition, we measured the expression of MMP2 and MMP9 and found that the expression levels of both MMP2 and MMP9 were decreased in circUCP2-knockdown tumor tissues (Fig. 3F).

Recent studies have shown that circRNAs repress miRNA expression by sequestering miRNAs [13]. Some circRNAs in the cytoplasm act as competing endogenous RNAs (ceRNAs) and sponge miRNAs to prevent them from binding to and repressing their target mRNAs [14]. Thus, we evaluated the cellular location of circUCP2 in NSCLC cells. The data showed that circUCP2 was primarily localized in the cytoplasm (Fig. 4A), which could allow it to function as an miRNA sponge. Thus, we searched CircInteractome to find possible circRNA/miRNA interactions, and miR-149 was predicted. The docking sites of miR-149 in the circUCP2 sequence are shown in Fig. 4B. Next, we explored the level of miR-149 in PC9, H1299, H1975 and A549 NSCLC cells, and the miR-149 content was found to be decreased (Fig. 4C). The luciferase enzyme reporter assay revealed a reduced luciferase intensity in cells cotransfected with plasmid containing wild-type sequences, as well as miR-149 mimics (Fig. 4D). To further investigate the direct docking between circUCP2 and miR-149, we performed an RIP assay. The results revealed that miR-149 was most abundant in the MS2bs-circUCP2 group (Fig. 4E), showing that circUCP2 could directly cross talk with miR-149 and sponge miR-149. Finally, we inhibited the expression of miR-149 with LNA in NSCLC cells to confirm the functions of miR-149 (Fig. 4F). The CCK-8 assay revealed that circUCP2 silencing suppressed cell proliferation, while inhibition of miR-149 reversed these effects of circUCP2 (Fig. 4G), indicating that circUCP2 functioned partly by regulating miR-149 in NSCLC.

To investigate whether circUCP2 sponges miR-149 to allow the expression of downstream targets, a possible binding site of miR-149 was searched using TargetScan. The docking site of miR-149 in the UCP2 3′UTR was predicted (Fig. 5A). Therefore, we assessed UCP2 expression in NSCLC cells and tissues and found that it was overexpressed (Figs. 5B and 5C). A luciferase enzyme reporter assay revealed a reduced luciferase intensity in cells that were cotransfected with wild-type luciferase reporter and miR-149 mimics (Fig. 5D). Furthermore, the luciferase intensity was elevated in cells cotransfected with the wild-type luciferase reporter and miR-149 inhibitor (Fig. 5D). Next, we examined UCP2 expression in NSCLC cells and found that miR-149 could inhibit UCP2 expression and that the miR-149 suppressor promoted UCP2 expression levels (Fig. 5E). The results verified that miR-149 regulated UCP2 expression levels.

Ago2 was predominantly enriched by circUCP2, UCP2 and miR-149 in the RIP assay (Fig. 5F), which indicated that circUCP2 as well as UCP2 can be mobilized to Ago2-linked RISC followed by binding to miR-149. In addition, circUCP2 silencing decreased circUCP2 binding to Ago2 while promoting UCP2 binding (Fig. 5G), which illustrated that circUCP2 bound to miRNAs by competing with UCP2. Silencing of circUCP2 reduced UCP2 expression, but miR-149 inhibitor cotransfection reversed the effect (Fig. 5H); this indicated that circUCP2 sponged miR-149 to regulate UCP2 expression in NSCLC. sssMoreover, the CCK-8 assay revealed that circUCP2 silencing suppressed cell proliferation, while overexpression of UCP2 reversed these effects of circUCP2 (Fig. 5I), indicating that circUCP2 functioned partly by regulating UCP2 in NSCLC.