circRAE1 promotes colorectal cancer cell migration and invasion through modulating miR-338-3p/TYRO3 axis

Background: Growing evidences have revealed that long non-coding RNAs (lncRNAs) including circular RNAs (circRNAs) involve in numerous carcinogenesis. However, the roles of circRNAs in the cancer biology of colorectal cancer (CRC) remain vague. Methods: qRT-PCR and western-blot were used to detecte the circRAE1 levels in CRC tissues and CRC cell lines.,Cell proliferation, migration and invasion were detected using wound healing assays, and transwell assays. The interaction between circRAE1 and miR-338-3p and TRYO3 was conrmed by dual-luciferase reporter assays. Results: We uncovered that a novel circRNA Hsa_circ_0060967 (also known as circRAE1) was remarkably increased in CRC tissues, and high circRAE1 level was positively associated with advanced tumor stage, lymph node metastasis, and tumor size. Loss-of-function assay indicated that circRAE1 accelerated cell proliferation, migration and invasion. Besides, miR-338-3p , lowly expressed in CRC tissues and CRC cell lines. dual-luciferase reporter assays showed that circRAE1 could sponge miR-338-3p, which targeted TRYO3 in CRC cells. Furthermore, overexpression of circRAE1 could recue the impaired migration and invasion triggered by miR-338-3p mimics or si-TYRO3 in CRC cells and vice versa. Conclusion : we gured out the network of circRAE1, miR-338-3p, and TYRO3 in CRC cells and revealed that increased circRAE1 served as an oncogene through sponging miR-338-3p, resulting in upregulated TYRO3 expression, which suggested that circRAE1 would be a potential therapeutic target and diagnostic marker for CRC treatment.


TYRO3 in CRC cells and vice versa.
Conclusion : we gured out the network of circRAE1, miR-338-3p, and TYRO3 in CRC cells and revealed that increased circRAE1 served as an oncogene through sponging miR-338-3p, resulting in upregulated TYRO3 expression, which suggested that circRAE1 would be a potential therapeutic target and diagnostic marker for CRC treatment.

Background
Colorectal cancer (CRC) is one of the most commonly diagnosed cancers worldwide, leading to more than 600,000 deaths each year [1,2]. Approximately 25% of patients with CRC were reported to have distant metastasis at the time of diagnosis. In addition, 25% of the patients experienced metastasis during treatment, and the prognosis of advanced patients was poor, with a ve-year survival rate for less than 15% of them [1]. With the advances in the diagnosis and treatment of CRCs, the ve-year relative survival rates of patients with stage I or II disease have increased to 91% and 82%, respectively [3]. However, the ve-year survival rate decreased to 12% among patients suffering distant metastasis [3]. Therefore, further clinical research is needed to improve the ve-year survival rate of CRC patients with distant metastasis as a means of enhancing treatment e ciency, and the molecular mechanisms of CRC tumorigenesis and its novel targets should be understood and identi ed, respectively.
The circular RNA (circRNA) is a non-coding RNA characterized by its stable circularized shape, and it is pervasively expressed in various cell types [4,5]. These circRNAs are resistant to exonuclease-mediated degradation and usually serve as competing endogenous RNAs (ceRNAs) to reduce downstream microRNA levels and function [6,7]. Recently, growing evidence has shown that circRNAs are involved in the regulation of tumorigenesis. Su et al. found that hsa_circ_0070269 suppresses hepatocellular carcinoma (HCC) cell proliferation and invasion and inhibits HCC tumor growth in vivo by sponging miR-182 to promote NPTX1 translation in HCC cells [8]. Li et al. showed that circ-ITCH, when decreased in esophageal squamous cell carcinoma, plays an antitumor role and acts as a ceRNA by sponging miR-17, miR-214, and miR-7 to increase ITCH mRNA [9]. Kong et al. revealed the hsa_circ_0085131/miR-654-5p/ATG7 axis as a potential therapeutic option among non-small cell lung cancer (NSCLC) patients who are resistant to cisplatin [10]. Moreover, emerging studies have shown that circRNAs are involved in the tumorigenesis and metastasis of CRC and identi ed as latent therapeutic targets for CRC treatment [11]. Chen et al. reported that the circ-001971/miR-29c-3p axis modulates CRC cell proliferation, invasion, and angiogenesis by targeting VEGFA [12]. Ma et al. showed that circ5615 exerts an oncogenic function as the ceRNA of miR-149-5p to release TNKS and activated Wnt/β-catenin pathways [13]. However, the detailed roles of circRNAs in CRC tumorigenesis and metastasis are only beginning to be revealed.
We identi ed a novel circRNA (also known as hsa_circ_0060967 but renamed as circRAE1 in this study) that was remarkably increased in CRC tissues. A high circRAE1 level was positively associated with advanced tumor stage, tumor size, and lymph node metastasis. Additionally, we con rmed that circRAE1 played an oncogenic role in the progression of CRC by serving as a ceRNA for miR-338-3p to increase TYRO3 expression. The ndings suggest that circRAE1 is a novel potential therapeutic target for CRC treatment intervene.

CRC patients and clinical tumor tissues
Eighty pairs of CRC tissues and the adjacent non-tumor tissues were collected from the Second A liated Hospital of Fujian Medical University in January-December 2019. The collection was approved by the Ethical Committee of the Second A liated Hospital of Fujian Medical University, and all patients were informed of the research details before admission. No additional treatment was given to the patients before surgery. The tissues were snap-frozen in liquid nitrogen and stored at −80 °C.

Cell culture and transfection
The human CRC cells HCT116, SW620, HT29, and SW480 and the normal colonic epithelial cell lines NM460 and HEK293T were obtained from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). All cells were cultured in DMEM-high glucose (Gibco) containing 10% fetal bovine serum (PAN Biotech) and 100 u/mL penicillin/streptomycin at 37 °C under 5% CO 2 in a humidi ed chamber.
Quantitative RT-PCR (qRT-PCR) assays Total RNA was isolated with the TRIzol reagent according to product speci cation. The cDNA was synthesized using a reverse transcription kit (Sangon, China) for circRNA and mRNA analysis. The RiboBio microRNA reverse transcription kit (Guangzhou, China) was used for the miRNA. Quantitative PCR was conducted using GoTaq ® qPCR Master Mix (Promega, USA). The relative expression levels of circRNAs, mRNAs, and miRNAs were calculated as the method of 2 −ΔΔCt and normalized to GAPDH for circRNAs and mRNAs or U6 for miRNAs. Each experiment was carried out in triplicate. The primer sequences are shown in Table 1.

Stability analysis of circRAE1
The SW620 and HT29 were treated by actinomycin D (2 μg/mL), then total RNA was extracted after 0, 4, 8, 12, and 24 h. The levels of circRAE1 and its linear subtype were detected by qRT-PCR. As for the RNase digestion assay, its total RNA was co-incubated with three units of RNase R for every 1 μg of RNA for 30 min at 37 °C. Then, the repuri ed RNA was subjected to qRT-PCR to determine the circRAE1 and its linear subtype.

Fluorescence in situ hybridization
In situ hybridization was employed to explore the intracellular location of circRAE1 and miR-338-3p in HT29 and SW620. The cells were grown on cover slips and xed with 4% paraformaldehyde (PFA). The procedure was carried out as previously reported [14]. Brie y, the slides were treated with CSK buffer (0.5% TritonX-100, 10 mM VRC) for 10-12 min, then treated with 70% alcohol for 10 min at 4 °C, followed by incubation with series alcohol for dehydration. After air drying, the slides were prehybridized for 1 h at 55 °C in a prehybridization solution. The Cy3-circRAE1 and/or digoxin-labeled miR-338-3p (DIG-miR338-3p) probe in the hybridization buffer were denatured at 76°C for 10 min, then added to each slide, and hybridized overnight at 37 °C in a dark humidi ed chamber. After washing, the anti-DIG [21H8]-FITC (Abcam, USA) was added to the slides, and the slides were incubated at 37 °C for 1 h in the humidi ed chamber. Finally, the slides were counterstained with 4,6-diamidino-2-phenylindole (DAPI) after washing.

Luciferase reporter assay
The PCR method was used to ampli cate the wild-type (WT) and mutant (MUT) 3'-UTR of the human TYRO3. Then, the WT and MUT of circRAE1 were cloned into the psiCHECKTM-2-luciferase reporter plasmid. The HEK293T cells were co-transfected with WT or MUT psiCHECKTM-2-circRAE1 plasmids or psiCHECKTM-2-TYRO3 (3'-UTR) and miR-338-3p mimics or miR-NC by using the Lipofectamine 2000 reagent. After 48 h, the cells were collected and subjected to the commercial Dual-Luciferase reporter assay system (Promega) following the manufacturer's instructions for measuring re y and Renilla luciferase activities.
Colony formation assay SW620 and HT29 cells (800 for each well) were seeded in six-well plates for two weeks. Subsequently, the colonies were stained with 1% crystal violet for 10 min after xation with 4% paraformaldehyde for 5 min. The colonies were microscopically examined and counted. The assays were repeated three times.
Wound-healing assay A wound migration model for the in vitro assay was used as previously described [15]. Culture inserts (Ibidi) were used in the wound-healing assay. Brie y, HT29 and SW620 cells were seeded to each well of the culture inserts. The cells were incubated at 37 °C for 24 h, and a cell-free gap of 500 μm was created after the removal of the culture insert. An inverted phase-contrast microscope was used to capture the images at 0, 24, and 48 h. Five randomly chosen elds were used to calculate the percentage of wound closure using the ImageJ software.

Transwell migration assay
Transwell chambers (Corning, USA) were used in the Transwell migration assay as previously described [9]. The SW620 and HT29 cells were seeded into the upper chambers with serum-free DMEM, while the lower wells were lled with DMEM containing 20% FBS. After 24 h, the cells in the top chamber were removed with cotton swabs, while those cells in the lower surface were xed using 4% PFA and stained with 0.1% crystal violet for 15 min. After washing twice with PBS, the cell numbers in ve randomly chosen elds were calculated under a microscope (Olympus, Japan).

Transwell invasion assay
The Transwell invasion assay was detected as previously described [16]. Brie y, the upper chambers (Corning, USA) were coated with diluted Martrigel (BD, USA). Afterwards, the SW620 and HT29 cells were seeded into the upper chambers with serum-free DMEM, then the Transwell chambers were incubated in wells lled with DMEM containing 10% FBS. After 24 h of incubation at 37 °C, the cells in the inner chambers and the remaining Matrigel were removed with cotton swabs, while the cells on the outside surface of the lower chambers were xed using 4% PFA and stained with 0.1% crystal violet for 15 min. After washing twice with PBS, the cell numbers in the ve randomly chosen elds were calculated under a microscope.

Statistical analysis
The SPSS 22.0 statistical software was used for the statistical analysis. The data were calculated as mean ± the standard deviation (SD). The signi cant differences between groups were estimated using one-way ANOVA. P < 0.05 was considered statistically signi cant.

Results circRAE1 was upregulated in CRC tissues and CRC cell lines
A previous study based on CapitalBio microarray data showed that numerous circRNAs were differentially expressed in CRC tissues unlike their normal adjacent tissues [17]; among them, the hsa_circ_0060967 (the corresponding gene symbol is RAE1 and thus renamed as circRAE1 in this study) was signi cantly upregulated. As the expression properties of circRAE1 in the CRC required validation, the circRAE1 expression in the 80 paired CRC tissues were detected using qRT-PCR. The circRAE1 expression was upregulated in the CRC tissues (Fig. 1A). On the basis of the analyzed clinical data and the expression levels of circRAE1, we can conclude that high circRAE1 expressions are positively related to advanced TNM stage, large tumor size, and lymph node metastasis ( Table 2).
We then explored the circRAE1 expression in CRC cells and found that the circRAE1 expression levels were signi cantly higher in CRC cell lines (SW620, HT29, and SW480) compared with the normal colonic epithelial cell line NM460 (Fig. 1B), especially for SW620 and HT29. Subsequently, we explored the stability of circRAE1 by adding actinomycin D, which can block new transcription, and detected the levels of circRAE1 and its linear control by qRT-PCR. The analyzed data showed that circRAE1 was more stable than the linear mRNA (Fig. 1C). Moreover, we observed that circRAE1 was resistant to RNase R, whereas its linear subtype was not resistant (Fig. 1D). In addition, the uorescence in the in situ hybridization results showed that circRAE1 was dominantly localized in the cytoplasm (Fig. 1E).

circRAE1 increased CRC cell migration and invasion ability
Here, we intended to select SW620 and HT29 for certain functional assays in consideration of the high-expression levels of circRAE1 in the two cell lines (Fig. 1B). We transfected the cells with various siRNAs by targeting circRAE1 and found that all the three siRNAs could effectively knock down circRAE1 expression, in which si-circRAE1#1 attained the best interference e ciency ( Fig. 2A). Thus, we selected si-circRAE1#1, which we named si-CircRAE1, for the succeeding circRAE1 knockdown assays. The CCK8 and colony assays presented a silencing of circRAE1 that could markedly inhibit proliferation in the two cell lines (Figs. 2B and 2C). The scratch tests showed a downregulation of circRAE1, which signi cantly reduced the wound healing e ciencies of the SW620 and HT29 cells (Fig. 2D). Subsequently, the Transwell assays showed that the downregulation of circRAE1 signi cantly inhibited SW620 and HT29 cell migration and invasion activities in vitro (Fig. 2E). Furthermore, we explored whether circRAE1 would be involved in the epithelial-mesenchymal transition (EMT) process. Thus, the expression of the EMT marker genes (E-cadherin and vimentin) were detected, and the circRAE1 expression was silenced (Fig. 2F).
Both the mRNA and protein levels of E-cadherin was remarkably upregulated by the knocking down of circRAE1, whereas vimentin presented the reverse (Fig. 2G). These results indicate that circRAE1 may serve as an oncogene in CRC progression by regulating the expression of EMT markers. circRAE1 targeted miR-338-3p The ceRNA, as a newly proposed mechanism, entails the crosstalk among lncRNAs, including circRNA, mRNAs, and their shared miRNAs. By using an online tool (CircInteractome), we speculate that the potential miRNAs may be bound by CircRAE1. Among them was miR-338-3p, which was dramatically reduced in CRC tissues (Fig. 3A), and it showed a negative relation with CircRAE1 in the expression (Fig. 3B). In addition, by using si-CircRAE1, we determined that downregulation of CircRAE1 could notably increase the miR-338-3p expression in SW620 and HT29 cells (Fig. 3C). Furthermore, the uorescence in situ hybridization results indicate that CircRAE1 and miR-338-3p were co-localized in the cytoplasm (Fig. 3D). The direct binding of miR-338-3p to circRAE1 was further con rmed. We constructed circRAE1 WT and circRAE1 MUT luciferase reporter plasmids. The luciferase reporter assay showed that the co-transfection with miR-338-3p mimics repressed the luciferase activity of circRAE1 WT, whereas the circRAE1 MUT was not affected by the co-transfection (Fig. 3E). This nding indicates that circRAE1 can directly sponge miR-338-3p.

circRAE1 functions by targeting miR-338-3p
First, in exploring the function of circRAE1 bound by miR-338-3p, we determined whether the circRAE1 overexpression could alter the miR-338-3p expression or whether the miR-338-3p mimic could affect the circRAE1 expression. The results suggest that the miR-338-3p mimic had no effect on the circRAE1 expression, while the circRAE1 overexpression remarkably reduced miR-338-3p (Fig. 4A), further indicating that circRAE1 could control miR-338-3p expression. Additionally, we employed wound CCK8, colony assays, healing assay, and Transwell migration and invasion experiments to determine if miR-338-3p in uenced CRC cell migration and invasion through the miR-338-3p. As shown in Figs. 4B-4E, the miR-338-3p mimic suppresses and inhibits the proliferation, and the wound closure in the SW620 and HT29 cells and the circRAE1 overexpression have partially reversed the effects caused by the miR-338-3p mimics on both cells. The migration and invasion capacity of the SW620 and HT29 cells, as determined by the Transwell assays, was also inhibited by the miR-338-3p mimic and rescued by the circRAE1 overexpression (Fig. 4E).We found that the TYRO3 protein level was upregulated by LV-CircRAE1 (Fig. 4F).

TYRO3 was directly targeted by miR-338-3p
The bioinformatics analysis, which was carried out using Targetscans (http://www.targetscan.org/), showed that TYRO3 was a putative target of miR-338-3p. The binding sites in between is shown in Fig. 5A. Then, the luciferase reporter plasmids of TYRO3-WT and TYRO3-MUT were constructed in our study (Fig. 5A). The luciferase reporter assay showed that adding miR-338-3p mimic could dramatically inhibit the reporter activity of TYRO3-WT rather than TYRO3-MUT (Fig. 5A), con rming that miR-338-3p could directly bind to the 3'-UTR of TYRO3. Moreover, the expression level of TYRO3 was also notably upregulated in CRC tissues (Fig. 5B) and showed a negative relation with CircRAE1 in the expression (Fig. 5C).

circRAE1 functions by regulating TYRO3 expression
We found that the expression of TYRO3 was upregulated by LV-CircRAE1, and it was signi cantly reduced by si-TYRO3 (Fig. 6A).Functionally, si-TYRO3 suppressed the proliferation, and the wound closure in the SW620 and HT29 cells and the circRAE1 overexpression partially blocked the effects of si-TYRO3 on both cells (Figs. 6B-6D). We also found from the Transwell migration and invasion experiments that si-TYRO3 remarkably reduced the migration and invasion capacity of the SW620 and HT29 cells, and the circRAE1 overexpression could rescue this effect (Fig. 6E).
The results suggest that circRAE1 can alter TYRO3 expression to regulate CRC cell migration and invasion.

Discussion
We validated in this study that circRAE1 was upregulated in CRC tissues, and it was positively associated with the much higher TNM stage, larger tumor size, and more lymph node metastasis in CRC patients. The loss-of-function assays showed that the silencing of circRAE1 could notably reduce the proliferation ability of CRC cells and cell wound closure e ciency, migration and invasion in vitro. The gain-of-function assays showed that the circRAE1 overexpression promoted CRC cell migration and invasion by acting as a ceRNA for miR-388-3p to regulate the TYRO3 levels. These data indicate that circRAE1 can serve as an oncogene in CRC tumorigenesis.
In accordance with remarkable advances in microarray and RNA sequencing technology, the growing numbers of circRNAs have been revealed to play vital roles in numerous disease processes, especially in cancers, including CRC [18]. For instance, circ_0026344 could suppress CRC cell growth and invasion in vitro and reduce CRC growth in vivo by sponging miR-21 and miR-31 [19]. Hsa_circ_0000069, when dramatically upregulated in CRC tissues, could promote proliferation, migration, and invasion of CRC cells [20]. The circRNA hsa_circ_000984 was also found to promote colon cancer growth and metastasis by binding miR-106b to increase CDK6 expression [21]. In this context, by analyzing the circRNA chips of four paired CRC tissues and their adjacent non-tumor control, as previously reported by Chen et al. [17], we further discovered a higher expression of circRAE1 in CRC tissue samples. The qRT-PCR analysis con rmed a similar pattern in CRC cell lines. A high circRAE1 expression indicates bad prognosis of patients with CRC. On the basis of the wound healing tests and Transwell assays, we can conclude that circRAE1 can function as an oncogene to promote CRC cell migration and invasion.
Recently, further research has revealed that circRNAs preferred to function as ceRNAs to sponge various miRNAs, leading to the increase in miRNA target expressions [22]. Hence, we conducted bioinformatics analyses via online (CircInteractome) and uncovered that miR-338-3p could be bound by CircRAE1. Previous studies have shown that miR-338 could suppress tumor progression in various human cancers. Zhang et al. found that miR-338 inhibited bladder cancer cell proliferation and invasion by reducing ETS1 [23]. He et al. con rmed that miR-338 inhibited cell proliferation and the expression of EMT markers of NSCLC cells by directly downregulating the NFATc1 expression [24]. By targeting MACC1, miR-338-3p could inhibit sw480 proliferation and migration and induce apoptosis [25].
Sun et al. demonstrated that miR-338-3p was remarkably downregulated in CRC compared with the adjacent nontumor tissues [26], a nding that accords with our current results. In the present study, our data indicate that miR-338-3p has an inverse impact on wound healing e ciency and migration and invasion in contrast to circRAE1 in CRC cells. The overexpression of circRAE1 could reverse the effect on CRC cells induced by miR-338-3p mimics and vice versa. We used TYRO3 as a direct target of miR-338-3pin CRC cells to further determine the mechanism of the circRNA-miRNA-mRNA network. TYRO3 is a protein tyrosine kinase. Previous literature has shown that TYRO3 is signi cantly increased in various cancers, and it promotes cancer cell proliferation and metastasis and enhances drug resistance, and thus is a potential therapeutic target [27,28]. In addition, TYRO3 was also con rmed to play vital roles in regulating the expression of EMT markers, chemical resistance, liver metastasis, cell proliferation, and apoptosis in CRC as an oncogene [29]. Here, we determined that knocking down the TYRO3 expression could rescue the increased cell wound healing e ciency and the migration and invasion induced by the overexpression of circRAE1 and vice versa; these expressions were also mediated by miR-338 mimics. Nonetheless, further in vivo assays are needed to further con rm the effectiveness in the clinical setting.

Conclusion
We identi ed the network consisting of circRAE1, miR-338-3p, and NPTX1 in CRC cells and determined that the increased circRAE1 can serve as an oncogene in CRC cells by functioning as a ceRNA to sponge miR-338-3p, resulting in an upregulated TYRO3 expression. These results can provide a novel potential therapeutic strategy to target the circRAE1/miR-388-3p/TYRO3 axis.       (E) Reversed the effects by Si-TYRO3 of CircRAE1 upregulation on the migration and invasion activities of SW620 and HT29 cells as determined by Transwell assays. *P < 0.05, **P < 0.01, ***P < 0.001.