Background
Lung cancer is the leading cause of cancer-related deaths worldwide. Non-small cell lung cancer (NSCLC) is primary histological subtype and accounts for ~ 85% of lung cancers, including two major histological types, lung adenocarcinoma and lung squamous cell carcinoma [
1,
2]. Recently, although the treatment of advanced NSCLC has made significant progress by targeted therapies and immunotherapies, five-year survival rate of NSCLC patients is still less than 15% [
3]. One of the major challenges is to find genes essential for NSCLC progression and understand the mechanisms of them [
4,
5]. In the past few decades, the investigation into the NSCLC has typically focused on protein-coding genes. However, the limitation of NSCLC therapy is still not fully overcome. Considering that the protein-coding gene accounts for only about 2% of the whole transcriptome and the remaining 98% is transcribed into non-coding RNAs (ncRNAs) [
6], it is critical to examine the potential of ncRNAs and comprehend the mechanisms underlying the progression of NSCLC.
Circular RNAs (circRNAs) constitute a novel class of ncRNAs. Pervasive expression of circRNAs is a recently discovered feature in various kinds of eukaryotes, including human cells [
7,
8]. circRNAs are usually considered as a master regulator of cellular processes due to their unique properties, such as abundant, stable, cell- and tissue specific expressions [
9]. An increasing number of studies have revealed that circRNAs are correlated with clinical features of NSCLC patients and play as regulators in NSCLC. For example, circTP63 is positively correlated with tumor size and TNM stage, exerting oncogenic potential by sponging to miR-873 in lung squamous cell carcinoma [
10]. Another circRNA, circPRKCI, also plays as an oncogene, and its therapeutic potential is confirmed in the nude mouse xenograft model and patient-derived tumor xenografts model. circPRKCI acts as a sponge for both miR-545 and miR-589 and abolishes their suppression on the oncogene E2F7 [
11]. However, only a small fraction of circRNAs in the genome has been investigated until now, thus further insight into molecular mechanisms and clinical relevance of more functional circRNAs may facilitate the identification of potential biomarkers and therapeutic targets for NSCLC.
Family with sequence similarity 83-member D (FAM83D), a microtubule-associated protein, has been implicated as an oncogene in the various kinds of malignancy. Wang et al. have shown that both of the mRNA and protein levels of FAM83D are upregulated in gastric cancer, and high expression of FAM83D predicts worse overall survival and disease-free survival [
12]. Liao et al. report that FAM83D is elevated in hepatocellular carcinoma, and the upregulation of FAM83D significantly promotes cell proliferation and invasion [
13]. Furthermore, Yan et al. and He et al. demonstrate that FAM83D is regulated by miR-495 and miR-210 [
14,
15]. Cyclin D1 (CCND1) and cyclin E1 (CCNE1) are two essential downstream mediators of FAM83D [
16].
In this study, we explored the role of circFOXM1(has_circ_0025039), a circular RNA obviously upregulated in our previous microarray results (GSE126533). We first identified the circular form of circFOXM1 and then verified the upregulated expression in NSCLC tissues compared to adjacent normal tissues. circFOXM1 was correlated with advanced clinical stage and worse overall survival in NSCLC patients. In vitro and in vivo assays showed that circFOXM1 promoted cell proliferation and cell cycle progression. RNA-sequencing results suggested that cell-cycle related genes were modulated by circFOXM1. circFOXM1 functioned as a ceRNA to sponge miR-614 and upregulated FAM83D level. circFOXM1/miR-614/FAM83D regulatory network may play essential roles in cell growth of NSCLC.
Materials and methods
Tissues and cell lines
48 paired samples of tumor tissues(T) and their corresponding nontumor tissues(N) from patients with NSCLC were obtained from the Shanghai Chest Hospital, Shanghai Jiao Tong University (Shanghai, China). The NSCLC tissues were immediately stored in liquid nitrogen when they were removed from patients in surgery. The tissues were transported to sample bank of Shanghai Chest hospital and stored in liquid nitrogen. Subsequently, RNAs were extracted from tissues and only qualified RNAs were retained for study. An apart of qualified RNAs was used as template for cDNA reverse transcription. The RNAs and cDNAs were divided into different tubes (RNase-free and DNase-free) to avoid freeze and thaw repeatedly. The qualified RNAs and cDNAs were stored at − 80 °C until use. These samples were identified by two pathologists independently. The detailed clinicopathological features were described in Table
1. All tissues were collected from July 2013 to September 2014. Written consent approving the usage of tissues in our study was obtained from each patient. This study was approved by the Ethics Committees of the Shanghai Chest Hospital, Shanghai Jiao Tong University. Human NSCLC cell lines, including H1299, H2170, A549, and H1703, as well as BEAS-2B were purchased from the American Type Culture Collection (ATCC) and were tested negative for mycoplasma contamination. H1299, H2170, A549, and H1703 cells were cultured in RPMI 1640 medium (Life Technologies). BEAS-2B cells were cultured in MEM medium (Life Technologies). All medium was supplemented with 10% fetal bovine serum and 100 units/mL penicillin-streptomycin (Life Technologies). All cells were maintained in humidified incubator at 37 °C in a CO
2 incubator.
Table 1Correlations between circFOXM1 and clinicopathological parameters of NSCLC tissues
Gender |
male | 41 | 22 | 19 | 0.219 |
female | 7 | 2 | 5 |
Age |
≥ 60 | 33 | 17 | 16 | 0.755 |
<60 | 15 | 7 | 8 |
Tumor size |
≥ 4 | 25 | 18 | 7 | 0.001** |
<4 | 23 | 6 | 17 |
Lymphatic metastasis |
positive | 27 | 15 | 12 | 0.382 |
negative | 21 | 9 | 12 |
History type |
adenocarcinoma | 15 | 6 | 9 | 0.459 |
squamous | 33 | 16 | 17 |
TNM stage |
I/II | 25 | 9 | 16 | 0.043* |
III/IV | 23 | 15 | 8 |
RNA and DNA extraction
Total RNAs of tissues and cells were extracted by using Trizol reagent (Invitrogen). All experiment operations were followed the manufacturer’s instruction of Trizol reagent. The procedure of RNAs extracted from nuclear fractions or cytoplasmic fractions were according to PARIS Kit (Life Technologies) manufacturer’s protocol. For DNA extraction, cells were rinsed with PBS twice and then extracted by Genomic DNA Isolation Kit (Sangon Biotech, China).
RNase R treatment
RNase R (Epicentre Technologies) was used to treat with total RNAs. Briefly, extracted RNAs aliquots from H1299 and H2170 cells were split into two parts: one for RNase R digestion and another for control with digestion buffer only. For RNase R digestion, 2 μg of total RNA was mixed with 2 μl 10 × RNase R Reaction Buffer and 2 μl RNase R (20 U/μl); for control, RNase R was replaced with DEPC-treated water. Then, the RNA samples were incubated at 37 °C water bath heater for 30 min. The detection of circFOXM1 and FOXM1 mRNA was analyzed by PCR, RT-PCR or qRT-PCR. RNase R treated RNA was used only for detecting resistance of circFOXM1 to RNase R exonuclease digestion. All primers were listed in Additional file
1: Table S1.
Reverse transcription PCR(RT-PCR) and quantitative real-time PCR (qRT-PCR)
For RT-PCR, 500 ng RNA was treated with gDNA wiper for 2 min at 42 °C and then was used to synthesize cDNA by using Hiscript Revert 1st First Strand cDNA Synthesis Kit (Vazyme, China). cDNA was used as templates to amplify by DNA Polymerase (Life Technologies), and products were further verified by using 1.5% agarose gel electrophoresis.
For qRT-PCR, only the cDNA was used as template and qRT-PCR assays were investigated by AceQ qPCR SYBR Green Master Mix (Vazyme, China) kits on ABI 7500 qPCR system. The circRNA and mRNA levels were normalized by β-actin. miRNA level was normalized by U6. The relative expression levels were determined by the 2−ΔCt or 2−ΔΔCt method. To determine the absolute quantity of RNA, the purified PCR product amplified from cDNA corresponding to the circFOXM1 and FAM83D sequence was serially diluted to generate a standard curve, respectively. Briefly, circFOXM1 and FAM83D form cDNAs were amplified, purified and measured. Then they were serially diluted to be as templates for qRT-PCR. The standard curves were drawn according to the Ct values at different concentrations. According to the standard curves, copy numbers of circFOXM1 and FAM83D in NSCLC cell lines were calculated.
Plasmid construction and transfection
To construct circFOXM1 ectopic overexpression plasmid, the sequences of exon 4 and exon 5 in FOXM1(amplified from cDNAs of H1299 cells) were cloned into pZW-circRNA vector (a gift from Ling-Ling Chen Lab) [
17]. To construct circFOXM1 knockdown plasmids (sh-circFOXM1), fragments targeting the circFOXM1 junction sites were cloned into pGreenPuro vector (System Biosciences). Likely, fragments targeting FAM83D mRNAs were constructed into pGreenPuro vector to generate FAM83D knockdown plasmids (sh-FAM83D). For dual-luciferase assay, wild-type and mutant fragments of circFOXM1 as well as FAM83D 3′ UTR were cloned into pmirGLO vector (Promega) to form luciferase reporter vector. The sequences of primers were listed in Additional file
1: Table S1.
For pZW-circFOXM1, sh-circFOXM1 or sh-FAM83D transfection, 2 × 105 cells were seed in 60 mm dishes for 24 h before transfection. For shRNA-circFOXM1 or shRNA-FAM83D stable cell line construction, 1 × 105 cells were seed in 60 mm dishes for 24 h before virus infection. Lentivirus was added into culture medium with polybrene, followed by selection with puromycin (2 μg/ml) for 2 weeks. For dual-luciferase assay, plasmids with wild-type or mutant fragments were co-transfected with miR-614, respectively.
Transcriptome sequencing
Total RNAs were isolated from H1299 cells with circFOXM1silencingorcontrol by using TRIzol reagent (Invitrogen) and purified by RNeasy Mini Kit (Qiagen). Transcriptome sequencing was conducted using Illumina HiSeq™ 2000 by Sangon (Sangon Biotech, China). Differently expressed transcripts were selected by |fold change| > 2 and
FDR < 0.1. Results were collected in Additional file
2: Table S2.
Gene set enrichment analysis
Gene set enrichment analysis (GSEA) software (version 3.0,
www.broadinstitute.org/gsea/) was employed to identify gene sets that were significantly overrepresented among genes up- or down-regulated in circFOXM1 silencing cells. Briefly, gene expression profiles were generated from circFOXM1knockdown and control cells by RNA-sequencing. GSEA v3.0 software was used to explore the distribution of members of the gene sets from the MSigDB database. In this bioinformatics analysis, if the most members in a gene set were positively or negatively correlated with the circFOXM1 expression, the set was termed associated with circFOXM1.
CCK-8 assay
1 × 103 cells were seeded in 96-well plates and were cultured for 1 ~ 5 day, followed by incubation with 10 μl of CCK-8 assay solution per well for 2 h (Dojindo Laboratories, Japan). The absorbance values at 450 nm were then assessed.
A density of 1 × 103 cells per well were incubated in 6-well plates. After 2 weeks’ incubation, a total of 1% of crystal violet (Beyotime Biotechnology, China) was applied to stain cell clones which were fixed with methanol.
Flow cytometry analysis
After harvesting by trypsinization, cells were washed with pre-cold PBS buffer, and then fixed in 75% ice-cold ethanol. Before staining, cells were resuspended in cold PBS and subjected to digestion with 2 μg/ml RNase A at 37 °C for 30 min, followed by labeling with 15 μg/ml propidium iodide (Beyotime Biotechnology, China) for 15 min at room temperature. Cell cycle profiles of labeled-cells were analyzed using a FACS Calibur flow cytometer (BD Biosciences).
Fluorescence in situ hybridization (FISH)
Cells were rinsed by PBS and then fixed in 4% formaldehyde for 10 min room temperature. The cells were further permeabilized in PBST (0.5% Triton X-100) on ice for 10 min, and then prehybridized with prehybridization buffer. After rinsing once in 2 × SSC, hybridization was performed using Cy3-labelled probe at 42 °C overnight. After co-staining with DAPI, the signals of the probe were captured by confocal microscopy (Zess 7100).
RNA pull-down assay
For miR-614 pull-down assay, the 3’end biotinylatedmiR-614 mimics or miRNA control (Sangon, China) were transfected into cells at a final concentration of 100 nM for 48 h before harvest. Then IP Cell lysis Buffer (Sangon Biotech, China) and complete protease inhibitor cocktail (Roche Applied Science) were added into the cell pellets, and incubated on ice for 10 min. Biotin-coupled RNA complex was pulled down by incubating the cell lysates with streptavidin-coated magnetic beads (Life Technologies) by centrifugation at 10,000×g for 10 min. The abundance of targets was evaluated by qRT-PCR analysis.
For circFOXM1 pull-down assay, 1 × 107cells that expressed circFOXM1 were harvested, lysed, and sonicated. The biotin-coupled probe of circFOXM1 or probe control was incubated with magnetic beads, respectively. After 2 h incubation, cell lysates were incubated with the probes overnight at 4 °C. After the incubation, the bound RNAs were washed for six times with wash buffer and purified for the qRT-PCR analysis. The biotinylated-probe was designed and synthesized by Sangon (Shanghai, China).
Immunoblotting
Extracted protein samples were quantified and boiled in SDS/β-mercaptoethanol buffer, then loaded into polyacrylamide gels. After separation by electrophoresis, the proteins were transferred onto PVDF membranes (Millipore). The membrane was incubated with rabbit anti-FAM83D antibody (ab236882, Abcam) or anti-CCND1 antibody (26939–1-AP, proteintech) or anti-CCNE1(11554–1-AP, proteintech) or anti-FOXM1 antibody (13147–1-AP, proteintech) or anti-β-actin antibody (20536–1-AP, proteintech) at 4 °C for 12 h, followed by an incubation with secondary antibody (proteintech) for 1 h. Bands were detected by a Bio-Rad ChemiDoc XRS system.
RNA immunoprecipitation (RIP)
RIP assays were performed by using a Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the protocol. Ago2 antibody was used for RIP (Cell Signaling Technology). Co-precipitated RNA was detected by qRT-PCR.
Luciferase assay
Cells were plated onto 96-well plates and grown to 70% confluence. For circFOXM1 and miR-614 luciferase assay, the cells were co-transfected with 50 nM miR-614 mimics and 200 ng pmirGLO-circFOXM1 vector, or corresponding mutant type (mut). For FAM83D and miR-614 luciferase assay, the cells were co-transfected with 50 nM miR-614 mimics and 200 ng pmirGLO-FAM83D vector, or corresponding mutant type (mut). Dual luciferase assay kit (promega) was used for dual luciferase assay. At 48 h post-transfection, cells were collected, and Renilla luciferase activity was assessed. Results are assessed as the ratio of Firefly luciferase activity to Renilla luciferase activity.
Xenografts experiments
Five-weeks’ old male BALB/c nude mice were chosen and randomly divided into four groups for our experiment. H2170 cells transfected with vector control, pZW-circFOXM1, sh-circFOXM1 or negative control (sh-NC) were subcutaneously injected into the back of the nude mice (1 × 106, 100 μl). The volume and weight of tumors were measured for 35 days. All animal experiments were performed under approval by the Shanghai Medical Experimental Animal Care Commission.
Immunohistochemistry
Tumor sections were subjected to deparaffinize and rehydrate with gradient ethanol solutions. After being immersed in antigen retrieval solution (AR0023, Boster, China) and heated for 15 min, sections were then incubated with anti-human Ki-67 antibody (M7240, Dako, Denmark) for 1 h and counterstained with hematoxylin (BA-4226, BASO, China) for 2 min.
Statistical analysis
Data were expressed as the mean ± standard deviation from at least three independent experiments. Survival analysis was performed with the Kaplan-Meier method, and the log-rank test was used for comparisons. Statistical results were analyzed using Prism software (GraphPad Software). Student’s t-test was used to compare two experimental groups. A probability of 0.05 or less was considered as statistical significance.
Discussion
circRNAs are first observed in the 1970s [
7]. In the initial studies, the evidence of circRNAs is provided in the viral genetic materials [
22‐
24]. Then, many genes are reported to produce individual circRNAs, such as ETS-1, SRY, and P450 in different cells. However, circRNAs are still considered as rare events with unclear biological functions [
25‐
28]. Until 2012, with the development of high-throughput RNA sequencing and bioinformatics analyses, the landscape of human circRNAs has been disclosed in different cells and tissues [
6,
29]. Since then, the focus of circRNA research has shifted to elucidate the association with human diseases, including cancers. Serval studies have been reported that the abnormally expressed circRNAs are correlated with clinical features and have the potential to be biomarkers in NSCLC [
9,
30,
31]. In this study, we found a circular RNA, has_circ_0025039 derived from the exons 4 and 5 of FOXM1 (named circFOXM1), was obviously upregulated in NSCLC tissues. High expression of circFOXM1 positively correlated with tumor size and advanced stage, and predicted poor prognosis for NSCLC patients.
To date, A large number of studies report that many genes, previously considered as protein-coding genes, can produce circRNAs by back-splicing [
32,
33]. Interestingly, it has been described that these circRNAs can be similar or completely different from their maternal gene functions [
34‐
36]. In this study, functional assays showed that circFOXM1 played as oncogenic roles in NSCLC, which is similar to function of its parent gene FOXM1. Liu et al. report that hsa_circ_0025033, another circular RNA derived from the exons 2–10 of FOXM1, also promotes the NSCLC progression [
37]. Nowadays, a total of 14 FOXM1-derived circRNAs have been recorded in the circbase database (
http://www.circbase.org/) including hsa_circ_0025033, hsa_circ_0025031, hsa_circ_0025035, hsa_circ_0025032, hsa_circ_0025041, hsa_circ_0025039 (circFOXM1), hsa_circ_0025038, hsa_circ_0025037, hsa_circ_0025042, hsa_circ_0025040, hsa_circ_0025030, hsa_circ_009831, hsa_circ_0025036, and hsa_circ_0025034. In our microarray data (GSE126533), 13 circRNAs derived from FOXM1 gene were detected except for hsa_circ_0098318. Among the 13 circRNAs, 5 circRNAs including hsa_circ_0025032, hsa_circ_0025038,hsa_circ_0025042,hsa_circ_0025031, and hsa_circ_0025039 (circFOXM1) were differential expressed (|Fold Change| > 2,
P < 0.01). Only hsa_circ_0025039 (circFOXM1) was abundant in NSCLC cells (Fig.S
6).
The ceRNA hypothesis proposes that RNA transcripts share the same miRNA response elements, resulting in competing for binding to miRNAs, then modulating the expression of each other [
38]. In our study, we firstly combined the bioinformatic analyses and circFOXM1 pull-down assays to screen miRNAs, which bind with circFOXM1. Simultaneously, we designed the circFOXM1 luciferase reporter to identify the direct interaction between circFOXM1 and miRNA. We found that miR-614 was the most enriched miRNA in circFOXM1 pull-down assay, and miR-614 could reduce the luciferase activity of circFOXM1 luciferase reporter by about 50% (Fig.
4h). However, some miRNAs, such as miR-198, miR-516, and miR-6812, could also be enriched in H1299 cells, but not be enriched in H2170 cells. This observation suggests that the mechanism of circFOXM1 may be diverse in different cell lines.
We next analyzed the data of RNA-sequencing and results of miR-614 target prediction, and found FAM83D, a cell-cycle related gene, shared the same miR-614 response elements with circFOXM1. Based on the ceRNA hypothesis, FAM83D should be modulated by circFOXM1/miR-614 ceRNA regulatory network. After a series of experiments, such as RNA pull-down assay, luciferase assay, western blot assay, and qRT-PCR assay, we confirmed that FAM83D was indeed regulated by circFOXM1. Recently, similar mechanisms have been discovered, such as circMOT1/miR-9/p21 regulatory axis [
39] and circNRIP1/miR-149/AKT regulatory axis [
40].
FAM83D, a microtube associated protein, is reported frequently in various kinds of cancers as an oncogene. So far, FAM83D is considered to be associated with proliferation, migration, and invasion in cancer cells [
41]. In NSCLC, it has been reported that FAM83D is a master cell-cycle regulator, resulting in CCND1 and CCNE1 alteration [
16]. Although FAM83D function has been studied comprehensively, the regulators of FAM83D are largely unknown. Recently, several studies indicate that FAM83D can be regulated by miRNAs. For example, Yan et al. report that FAM83D is a direct target of miR-495 in colorectal cancer cells [
14]. He et al. show that FAM83D expression is repressed by miR-210 during cell mitosis [
15]. In our study, we first report that FAM83D is directly repressed by miR-614 and it is an essential partner of the circFOXM1 ceRNA regulatory network.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.