Background
Circular RNAs (circRNAs) are a class of transcripts characterized by a covalently closed loop structure. They have no 5′ to 3′ polarity or polyA tail [
1]. Most circRNAs are produced by the back splicing of exons, a non-canonical splicing process [
2,
3]. CircRNAs are expressed in a tissue-specific, developmental stage-specific, and disease-specific manner [
4]. They are known to act as microRNA sponges, transcription regulators, and scaffolds for mediating protein interactions or localization [
5]. More recently, circRNAs were found to harbor coding potential [
6‐
8], which is reasonable, as most circRNAs contain exons and are localized in the cytoplasm [
9]. Zhang’s group has previously reported that a panel of circRNAs are translated into proteins that function as tumor suppressors in glioblastomas [
10].
Gastric cancer (GC) has a high mortality rate in China and around the world due to the lack of efficient tools for early diagnosis [
11]. The high stability of circRNA makes it a good candidate as a molecular biomarker for early diagnosis [
1]. The Wnt/β-catenin signaling pathway plays an important role in normal embryo development, tissue differentiation, homeostasis, and oncogenesis [
12]. Mutations in Wnt signaling are observed in the majority of cancers, as it is essential for the viability of cells [
13]. Changes in the stability of cytoplasmic β-catenin is a key switch in the Wnt pathway [
13], which is monitored by the APC/AXIN destruction complex [
14]. The dysfunction of APC [
15] or AXIN [
16] can lead to abnormal β-catenin accumulation. Furthermore, the Wnt/β-catenin signaling pathway is aberrantly active in 30 to 50% of patients with GC [
17,
18], although the mechanism underlying abnormal β-catenin activation in GC is unclear. There has been little research that has addressed the functional role of circRNA in GC development [
19,
20]. Exploring the function and regulatory role of circRNA in the Wnt/β-catenin signaling pathway will facilitate a better understanding of the molecular pathogenesis of GC and pave the way for the development of an early diagnostic biomarker.
In this study, based on the high-throughput sequencing results from five paired GC samples, we identified the elevated expression of circAXIN1 in GC. The circRNA circAXIN1 encodes a novel protein, AXIN1-295aa. Here, we sought to determine how this novel protein interacts with the Wnt/β-catenin signaling pathway and what role it plays in GC.
Materials and methods
Tissue and cell culture
A total of 63 pairs of non-neoplastic gastric and GC tissue samples from patients who attended from Shenzhen Second People’s Hospital, China, were examined in this study. None of the tissues received any radiotherapy or chemotherapy prior to surgery, and they were stored in RNAlater (Thermo Fisher, Shanghai, China) immediately following surgery. All patients provided written informed consent, and the study was approved by the ethics committee of Shenzhen University School of Medicine. Immortalized human normal gastric epithelial cells (HFE-145) were obtained from Dr. Duane T. Smoot of Meharry Medical College, USA. The GC cell lines (AGS, MKN28) were obtained from the American Type Culture Collection (ATCC) and China Infrastructure of Cell Line Resources, respectively, while the GES-1, BGC-823, SGC7901, and NCI-N87(N87) cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). All cells were cultured in DMEM (Hyclone, Logan, Utah) with 10% FBS (Gibco) in an incubator at 37 °C with 5% CO2.
Plasmids and cell transfection
The full length of circAXIN1-3xFLAG was chemically synthesized and cloned into vector pLC5-ciR using the EcoRI and BamHI sites. The pLC5-ciR vector contains artificial side flanking sequences and SA (splice acceptor)/SD (splice donor) sequences. The 3xFLAG was inserted before the stop codon of the putative open reading frame (ORF). The CMV-AXIN1-295aa linear overexpression vector served as a positive control. The mCherry-IRES-GFP was cloned into a psin-EF2 vector. The wild type and mutant internal ribosome entry sites (IRESs) (IRES115–257, IRES115–186, IRES187–257, IRES689–838, IRES689–763, and IRES764–838) were cloned into a P-Luc2-IRES-Report vector using Geneseed. (Guangzhou, China). Plasmids were transfected when cells reached a confluence of 30 to 50%, using Lipofectamine™ 3000 Transfection Reagent (Invitrogen, Shanghai, China). siRNAs for circAXIN1 knockdown were synthesized using Geneseed. (Guangzhou, China); the sequences were: si-hsa_circAXIN1_01 AGAGAGTTCAGGACAGATT; si-hsa_circAXIN1_02: GAGAGTTCAGGACAGATTG; and si-hsa_circAXIN1_03; AGAGTTCAGGACAGATTGA. Cells at 30 to 50% confluence were transfected with 60 nM siRNAs using Lipofectamine RNAiMAX (Invitrogen, Shanghai, China).
RNA-sequencing assay
RNAs from five human GC samples and their adjacent normal tissues were sequenced on an Illumina Hiseq 2500 (Chi Biotech, Shenzhen, China). The reads were aligned to the human reference genome (version GRCh38), using the BWA aligner. Any circRNAs were identified using CIRI software (version 2). The identified circRNAs were then annotated with the gene annotation file corresponding to the reference genome and the full-length circRNA sequences were extracted. The full-length circRNA sequences for all circRNAs were considered to be the reference genome, and the fastq reads were mapped using the bowtie2 aligner. Then, count data were generated using bedtools multiBamCov with the bowtie2 output (converted to bam, sorted, and indexed). The count data were normalized (TPM, transcripts per kilobase million), and the R package limma was used to identify differentially expressed (DE) circRNAs. Fold-change values > 2 and p-values < 0.05 were considered to be the thresholds for defining significantly differentially expressed circRNAs.
RT-PCR and real-time quantitative RT-PCR
Total RNA was extracted using Trizol reagent (251,808, Invitrogen), according to the manufacturer’s protocol. Nuclear and cytoplasmic RNAs were extracted using the Cytoplasmic & Nuclear RNA Purification Kit (Norgen Biotek). For circRNA detection, RNase R (10 U, Geneseed, was used for linear RNA digestion at 37 °C for 30 min. Then, RNA was recovered using the RNeasy MinElute Cleanup Kit (74,204, QIAGEN). Reverse transcription and real-time PCR were performed using GoScript™ Reverse Transcription Mix (A2800 and A6002, Promega). All primers were synthesized by Sangon Biotech; detailed information about the primers is shown in Supplementary Table
1. GAPDH and 18S were used as internal controls.
Western blotting
Protein was extracted with 2x Laemmli sample buffer (Bio-Rad) with a protease inhibitor (Roche). Cytoplasmic and nuclear proteins were isolated using NEPER™ Nuclear and Cytoplasmic Extraction Reagents (Thermofisher Scientific). Western blotting was conducted as previously described [
21]. The antibodies used were: FLAG (F1804, Sigma), Ubiquitin (ab7780, Abcam), HDAC (A0238, ABclonal), AXIN1 (#3323, Cell Signaling), AXIN1(NBP1–31013, Novus Biologicals), AXIN1 (A4747-01A, US biological), β-catenin (#8480, Cell Signaling), GSK3β (#12456, Cell Signaling), Rabbit IgG control (#3900, Cell Signaling), Mouse IgG control (#5415, Cell Signaling), GAPDH (#5174, Cell Signaling), and Wnt/β-Catenin Activated Targets Antibody Sampler Kit (#8655, Cell Signaling).
Co-immunoprecipitation and mass spectrometry
Immunoprecipitation and co-immunoprecipitation were performed using a Pierce Classic Magnetic IP/Co-IP Kit (88,804, Thermo Fisher Scientific). Cells were lysed with cold lysis buffer and supernatant was collected after centrifugation at 13,000 g for 10 min. Approximately 1000 μg protein was incubated with specific IP antibody (1:50) at 4 °C on a rotating platform overnight. Pierce Protein A/G Magnetic Beads (25 μL) were added to the antigen sample/antibody mixture and incubated at room temperature for 1 h. After washing, the target antigen–antibody complex was eluted with 100 μL of Elution Buffer and 10 μL of Neutralization Buffer, followed by Western blotting analysis or mass spectrometry analysis at BGI (BGI, Shenzhen). VeriBlot for IP Detection Reagent (HRP) (ab131366, Abcam) was used to avoid the detection of heavy and light chains.
Chromatin immunoprecipitation
A chromatin immunoprecipitation (ChIP) assay was performed using a Magna ChIP G kit (# 17–409, MAGNA0002, Millipore), according to the manufacturer’s instructions. After crosslinking and sonication, 50 μL of sheared DNA was incubated with 20 μL of protein G magnetic beads and anti-β-catenin (1:50, #8480, Cell Signaling) at 4 °C overnight. Rabbit (DA1E) mAb IgG XP Isotype Control (1:50, #3900, Cell Signaling) was used as a negative control. Then, protein/DNA complexes were eluted and free DNA was purified for the following qRT-PCR assay. Promoter primers for the detection of gene downstream of Wnt signaling (CD44, CMYC, C-Jun) were designed, and the amplified products were confirmed to contain the β-catenin TCF binding site 5′-A/T A/T CAAAG-3′. The primer sequences are shown in Supplementary Table
1. qRT-PCR was performed as previously described [
21].
Dual-luciferase reporter assay
Topflash and Fopflash reporters were obtained from Addgene (Cambridge, MA, USA). Two predicted IRES sequences and truncated sequences were inserted into the P-Luc2-IRES-Report vector (Geneseed, Guangzhou). Firefly and Renilla luciferase activity were measured using a Dual-Glo luciferase assay kit (Promega).
Confocal immunostaining
Cells were inoculated in 35-mm petri dishes (NEST) and transfected with OV-circAXIN1 or circAXIN1 si1 for 48 h. Wnt agonist1 (S8178, Selleckchem) and XAV-939 (S1180, Selleckchem) (10 and 1 μM, respectively) were used as positive controls. After being fixed and permeabilized, cells were incubated with anti-β-catenin antibody (#2677, 1:100, Cell Signaling) at 4 °C overnight. The next day, cells were incubated with Alexa Fluor® 488-conjugated anti-rabbit secondary antibody (#4412, 1:500, Cell Signaling) for 1 h at room temperature in the dark. DAPI II (Abbott Molecular, Abbott Park, Illinois) was used to stain nuclei. Images were taken using a ZEISS confocal microscope.
Immunohistochemistry
Immunohistochemical staining was performed using a DAB kit (ZLI-9017, ZSGB-BIO) and a Mouse Polymer kit (PV6002, ZSGB-BIO). Paraffin sections were first deparaffinized and rehydrated using a series of xylene and ethanol rinses. Slides were heated with EDTA repair solution (10 μM, ZLI-9067, ZSGB-BIO) for 20 min in a microwave oven. The slides were washed with PBS and allowed to cool to room temperature; they were then blocked with 3% hydrogen peroxide for 10 min. After thorough washing, the slides were incubated with a specific antibody (1:100) at 4 °C overnight. Ki-67 (GB111499, Servicebio, 1:1000), TCF-1 (#2203, Cell Signaling, 1:100), β-catenin (#8480, Cell Signaling, 1:100), c-Jun (#9165, Cell Signaling, 1:100), and Met (#8198, Cell Signaling, 1:100) antibodies were used. The next day, the slides were incubated with secondary antibody for 20 min at 37 °C. An equal volume of DAB was added for 10 min, for color rendering. After washing, hematoxylin was used as a counterstain and the slides were sealed using a neutral resin. The intensity scores of stained sections were assessed by two pathologists. The staining intensity was evaluated on a scale from 0 to 3 (0, negative; 1, weakly positive; 2, moderately positive; 3, strongly positive) and the percentage of positive cells was scored from 0 to 4 (0, negative; 1, 1–25% positive; 2, 26–50% positive; 3, 51–75% positive; 4, 76–100% positive). The final scores were calculated as percentage positive multiplied by staining intensity.
Protein structure and prediction of protein–protein interactions
The amino acid sequence of AXIN1-295aa was predicted based on the ORF nucleotide sequence. According to a report in the literature [
22], an SAMP (serine-alanine-methionine-proline) region, comprising 25 aa from APC, interacts directly with the regulators of the G protein signaling (RGS) domain of AXIN1. The AXIN1-295aa and APC SAMP structures were predicted by using the fold recognition method PHYRE [
23] (
http://www.sbg.bio.ic.ac.uk/~phyre2/html/page.cgi?id=index). ZDOCK [
24] (
http://zdock.umassmed.edu/) was used to predict the interaction between AXIN1-295aa and SAMP.
In vitro proliferation, migration, invasion, and colony formation assays
Cell proliferation was determined using an EdU assay (Ribobio, Guangzhou). Following transfection for 48 h, cells were collected to conduct the assays, as previously described [
25]. The cell proliferation rate was calculated by dividing the number of actively dividing cells (red) by the total number of cells (blue). Sterile Transwells® with 8.0-μm pore polycarbonate membrane inserts (3422 and 3428, Corning) were used to determine cell migration and invasion, respectively. After transfection for 48 h, AGS cells were preincubated with mitomycin-C (10 μg/mL, Sigma, St. Louis, MO) for 1 h at 37 °C for migration assay. N87 cells were not preincubated with mitomycin-C. A total of 5 × 10
4 cells were seeded into the upper chamber without FBS. After incubation for 24 h at 37 °C, the cells were fixed and then stained with hematoxylin. The cells in five random fields were counted under a microscope. A wound healing assay was also performed to determine the migratory ability of the cells. Transfected cells were inoculated in 6-well plates and were preincubated with mitomycin-C (10 μg/mL, Sigma, St. Louis, MO) for 1 h at 37 °C. An artificial scratch was made, at time 0 h. The wound width from five random fields was measured at 0, 24, and 48 h. Cell viability was determined by a colony formation assay. Approximately 200 transfected cells were seeded in 6-well plates and their viability was determined after culturing for 2 weeks.. Each experiment was repeated three times.
Female BALB/c nude mice, aged 4 to 6 weeks, were obtained from Charles River Laboratories (Beijing, China) and used for the animal studies. All of the animal experiments were conducted in accordance with the principles of the Institutional Animal Care and Use Committee of Shenzhen University. The xenograft model was successfully established by injecting 5 × 106 AGS cells into the right flank of the nude mice (n = 7 in each group). In the xenograft model group, cholesterol-conjugated circAXIN1-siRNA (10 nmol, Geneseed) was intratumorally injected twice a week for 3 weeks. Tumor volume was measured every 3 days, with the tumors then resected for subsequent experiments. For the in vivo lung metastasis assay, 1 × 106 AGS cells were injected via the tail vein to assess tumor metastasis ability; circAXIN1-siRNA was injected via the tail vein twice a week for 6 weeks. The lung was resected and stained with hematoxylin and eosin (H&E) to count the cancer metastatic lesions.
Statistical analysis
The Student’s t-test or one-way ANOVA was used to evaluate the significance of any differences. The data from three independent experiments are shown as means ±SD. A *p-value < 0.05 was considered statistically significant.
Discussion
Although circular RNA was discovered decades ago, it had been overlooked until the recent advances in and application of parallel sequencing. Research has since revealed that circRNAs are conserved, abundant [
26], and widespread in eukaryotic cells [
27,
28], where they play a vital regulatory role [
29]. In addition to serving as microRNA sponges and interacting with protein complexes, circRNAs have been found to possess coding potential [
30‐
32]. The expression level of most circRNAs is not high and only a few contain multiple perfect microRNA binding sites, which suggests that their role as microRNA sponges may not be their main function [
5]. Recent studies have revealed that functional peptides can be translated from non-coding RNAs, including primary microRNA (pri-miRNAs), long non-coding RNAs (lncRNAs), and circRNAs, thus blurring the distinction between coding and non-coding RNAs. For example, a 34-amino-acid peptide encoded by the lncRNA Dworf localizes to the sarcoplasmic reticulum membrane and enhances muscle contractility [
33]. Regulatory peptides are translated from the plant primaries miR171b and miR165a to delay root development [
34]. In addition, evidence has emerged of the existence of functional peptides encoded by circRNAs, including the novel ZNF609 protein isoform from circ-ZNF609 [
35], FBXW7-185aa from circ-FBXW7 [
36], and β-catenin-370aa from circβ-catenin [
37], to name but a few. In the present study, we discovered that circAXIN1 encodes the novel protein AXIN1-295aa to enhance the progression of GC.
Several criteria must be considered when screening functional circRNAs with coding potential. First, circRNA is differentially expressed at a relatively high level in GC, which makes it easily detectable. Second, circRNA must contain a complete ORF, and this ORF should span the junction site. Third, the parental gene is associated with cancer development. Emerging data have revealed that the function of circRNA is highly associated with its parental gene. Fourth, circRNA that is predicted to encode a protein does not necessarily express that protein. Experimental tools must be applied to verify the actual protein expression from circRNAs.
The number of peptides encoded by circRNAs could be underestimated. In our search for potential coding circRNAs, we considered only junction-spanning ORFs, because only they could be distinguished from those peptides encoded from their linear counterparts. However, peptides other than junction-spanning ORFs might also be translated from circRNAs. Ribosomal profiling, recognized to be the most reliable method for identifying coding circRNAs, has been used to identify translatable circRNAs. We also used ribosomal sequencing to search for coding circRNA; however, we found that the junction site of circAXIN1 was not bound to a ribosome, probably due to the excessively stringent conditions of the specific junction site protected by the ribosome in the ribosome profiling sequencing (data not published). Therefore, although the ribosome profile is the most trusted means for identifying novel coding circRNAs, some translatable circRNAs could be missed.
The functions of these novel proteins remain largely unknown. Most proteins translated from circRNAs share their N-terminus with proteins encoded by their parental genes, which generates decoys or competitors for the parental proteins. For example, the peptide AKT3-174aa shares an N-terminus with AKT3 and functions as a decoy for AKT3, thus playing a negative regulatory role in modulating PI3K/AKT signaling activity [
10]. The translation of circRNA increases upon stress stimulation [
31]. We speculate that circRNA translation may be enhanced during cancer development, taking GC as an example, from early long-term inflammation, intestinal metaplasia, and carcinoma in situ, to distant metastasis. This hypothesis deserves further investigation.
By directly binding to all of the other core components, i.e., β-catenin, APC, CKα, and GSK3β, AXIN1 acts as the central scaffold of the Wnt-pathway destruction complex [
14,
38]. AXIN1 is presumed to be a tumor suppressor [
39,
40] in many types of cancers, especially colorectal cancer (CRC) [
41,
42]. Both somatic and germline mutations in the AXIN1/2 genes have been found in a subset of CRCs and in several other cancer types. AXIN1, because it is the least abundant component, plays a rate-limiting role in the regulation of Wnt activity [
43]. AXIN1-295aa contains the APC binding site RGS domain only; it does not contain the β-catenin and GSK3β binding domains or the oligomerization domain. However, from the MS and co-IP results, we determined that AXIN1-295aa also interacts with β-catenin and GSK3β. We propose that AXIN1-295aa binds indirectly to β-catenin and GSK3β by binding to APC. We assume that AXIN1-295aa saturates the available APC, leaving AXIN1, CKα, and GSK3β unable to form the normal destruction complex. β-catenin subsequently translocates to the nucleus and activates downstream genes.
The high stability of circRNA makes it a good candidate as a biomarker [
44]. In addition, we found that the expression of circAXIN1 was significantly higher in GC tissues than normal tissues and was associated with tumor invasion depth, differentiation, tumor stage, and lymph node metastasis, which implies that circAXIN1 has the potential to serve as a biomarker for GC prognosis. However, this proposal is formulated based on a limited sample size. Further research must be conducted to confirm that circAXIN1 can be used as a prognostic factor. Furthermore, circAXIN1 siRNA exhibited an excellent therapeutic effect in xenografts and a lung metastasis model, with no severe adverse effects. This result has inspired us to search for the fundamental molecular mechanism accounting for this therapeutic effect. Although in this study we only tested circAXIN1 siRNA in mice, we believe that in the future there will be increasing circRNA siRNA-based therapies in clinical trials and subsequent use [
45].
This study is the first to determine that circAXIN1 is translatable and promotes tumorigenesis and aggressiveness in GC. We found circAXIN1 to be highly expressed and associated with lymph node metastasis in GC. We showed that circAXIN1 encodes the functional peptide AXIN1-295aa, which competitively binds to APC, leading to the release and nuclear translocation of β-catenin. Ultimately, β-catenin transactivates the canonical Wnt pathway and induces the expression of Wnt-dependent genes to promote cell proliferation and migration.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.