CircRNA_0000392 Promotes Colorectal Cancer Progression by Sponging miR-193a-5p

Background: Circular RNAs (circRNAs), an important member of the non-coding RNA family, have been revealed the role in the pathogenic progression of diseases in recent years, particularly in the malignant progression of cancer. With the application of high-throughput sequencing technology, a large number of circRNAs have been found in tumor tissues, and some circRNAs have demonstrated the role as oncogenic genes. In this study, we analyzed the circRNA expression prole in colorectal cancer (CRC) tissues and normal adjacent tissues by high-throughput sequencing, focusing on the circRNA_0000392, a circRNA with signicantly increased expression in colorectal cancer tissues, and further investigating its function in the progression of colorectal cancer. Methods: The expression prole of circRNAs in 6 pairs of CRC tissues and normal adjacent tissues was analyzed by RNA-sequencing. We veried the differential circRNAs with expanded samples by qRT-PCR, focused on circRNA_0000392, and evaluated its associations with clinicopathological features. Then we knocked down circRNA_0000392 in CRC cells and evaluated the effect in vitro and in vivo by functional experiments. The dual luciferase assay and RNA pull-down were performed to further explore the downstream potential molecular mechanisms. Results: CircRNA_0000392 was signicantly up-regulated in CRC compared with normal adjacent tissues and cell line. The expression level of circRNA_0000392 was positively correlated with the malignant progression of CRC. Functional studies revealed that reducing the expression of circRNA_0000392 could inhibit the proliferation and invasion of CRC both in vitro and in vivo. Mechanistically, circRNA _0000392 could act as a sponge of miR-193a-5p and regulate the expression of PIK3R3, then affect the activation of the AKT-mTOR pathway in CRC cells. Conclusions: The circRNA_0000392 has the function as an oncogene through miR-193a-5p/PIK3R3-Akt axis in CRC cells, implying that circRNA_0000392 is a total and by using on a denaturing agarose gel. After conrming the RNA is intact and pure, we using the Ribo-Zero rRNA Removal Kit (Illumina (cid:0) San USA) and the CircRNA Enrichment Kit (Cloud-seq (cid:0) USA) to remove the rRNA and enrich the circRNAs. The RNA-seq libraries were constructed by using pretreated RNAs with Stranded Total RNA Library Prep Kit (Illumina, San Diego, according to the manufacturer’s instructions. The libraries were denatured as single-stranded DNA molecules, captured on Illumina ow cells, amplied in situ as clusters and nally sequenced for 150 cycles on Illumina 4000 Sequencer (Illumina, according to the manufacturer’s instructions. Paired-end reads were harvested from Illumina HiSeq™ 4000 sequencer, and were quality controlled by Q30. The reads were aligned to the reference genome/transcriptome with STAR software and circRNAs were detected and annotated with DCC software. CircBase database and circ2Trait disease database were used to annotated the identied circRNAs. The differentially circRNAs between the two groups were identied T test statistical methods.

axis in CRC cells, implying that circRNA_0000392 is a potential therapeutic target for the treatment of colorectal cancer and a predictive marker for CRC patients.

Background
Colorectal cancer (CRC) threatens human health worldwide as the third most common cancers [1]. In recent years, the incidence of colorectal cancer has increased year by year [2]. With the continuous improvement of diagnosis and treatment, the ve-year survival rate of colorectal cancer has increased, but the ve-year survival prognosis is highly correlated with the stage of the disease. Patients with advanced colorectal cancer were usually accompanied by tumor metastasis, which ve-year survival rate was very low [3]. Therefore, it is urgent to further study the pathogenesis of colorectal cancer and the unknown molecular mechanism involved in tumor metastasis.
There is a large amount of non-coding RNA in the human genome, and the relationship between the existence of non-coding RNA and human diseases has always been a research hotspot, especially in malignant tumors. Circular RNAs (circRNAs) is an important member of the non-coding RNA family following microRNAs and lncRNAs. circRNAs are characterized by the absence of covalently closed loop structures at the 3' and 5' ends. Base on this closed structure, circRNAs are highly stable and not easily degraded [4]. The researchers have discovered the presence of circRNA in multiple organisms such as yeast, mitochondria and eukaryotes and detected more than 20,000 circRNAs in eukaryotes [5,6]. One study reported that exon rearranged circulating transcripts were rst discovered in leukemia cells, Hela cell lines and normal human primary blood cells, and approximately 80 circRNAs were identi ed [7]. Since then, more and more circRNAs have been identi ed in different tissues using high-throughput sequencing technology. The role of circRNA in the development of diseases such as encephalopathy and tumors were also gradually revealed [8,9]. The mechanism of circRNA as a competitive endogenous RNA has become the focus of research on its role in cancers. CircRNA adsorbs miRNA through the sponge action of miRNA to regulate the expression of its target genes [10,11]. With the continuous expansion of study on circRNAs, circRNAs have been shown to be involved in the development of almost all types of cancers [12][13][14][15][16][17]. All the studies on the relationship between circRNAs and cancer suggestted that circRNAs may be a novel potential biomarker and therapeutic target. However, since most circRNAs were still exist without exploration and the roles of circRNAs in the CRC progression are still largely unknown, further research is needed to nd the circRNAs associated with CRC tumorigenesis and to elucidate their functions.
In this study, we rst explored the expression pro les of circRNAs in the 6 paired CRC tissue and adjacent normal tissues by using high-throughput RNA sequencing. A total of 66855 circRNAs were detected, among them, 1687 circRNAs with signi cant differential expression were identi ed after CRC tissues was compared with adjacent normal tissues. After verifying some candidate circRNAs by qRT-PCR, we found that circRNA_0000392 was signi cantly up-regulated in the CRC tissues. The high expression of circRNA_0000392 was associated with the pathological stage and metastasis in CRC. We then focused on circRNA_0000392 and demonstrated that inhibition of its expression could signi cantly attenuated the proliferation and invasion in CRC cells. More importantly, we explored the mechanism of circRNA_0000392 in the progression of colorectal cancer and found that it could act as a sponge of miR-193a-5p, thereby releasing the inhibition of PIK3R3 by miR-193a-5p and promoting the phosphorylation of AKT / mTOR signaling pathway. Our ndings illustrated a new mechanism of CRC progression and provide new insights for its treatment and diagnosis.

Patient population and clinical data
The 40 pairs of CRC tissues and adjacent normal tissues were collected from patients who were diagnosed with CRC at the Longhua Hospital a liated to Shanghai University of Traditional Chinese Medicine (Shanghai, China). Tumor and normal adjacent tissue samples were obtained from the surgical treatment at the Department of General Surgery. The samples were snap frozen in liquid nitrogen after separated from the human body immediately and stored at − 80 °C before using. All the patients have signed informed consent prior to surgery and did not receive preoperative chemotherapy or radiotherapy. This study was approved by the Ethics Committee of Longhua Hospital.

RNA sequencing, identi cation and quanti cation of human circRNAs
The total RNA was isolated from the tissue samples using TRIzol reagent (Life Technologies, Carlsbad, CA) according to the manufacturer's instructions. Then we assessed the RNA integrity and DNA contamination by using electrophoresis on a denaturing agarose gel. After con rming the RNA is intact and pure, we using the Ribo-Zero rRNA Removal Kit (Illumina San Diego, CA, USA) and the CircRNA Enrichment Kit (Cloud-seq USA) to remove the rRNA and enrich the circRNAs. The RNA-seq libraries were constructed by using pretreated RNAs with TruSeq Stranded Total RNA Library Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer's instructions. The libraries were denatured as singlestranded DNA molecules, captured on Illumina ow cells, ampli ed in situ as clusters and nally sequenced for 150 cycles on Illumina HiSeq™ 4000 Sequencer (Illumina, San Diego, CA, USA according to the manufacturer's instructions. Paired-end reads were harvested from Illumina HiSeq™ 4000 sequencer, and were quality controlled by Q30. The reads were aligned to the reference genome/transcriptome with STAR software and circRNAs were detected and annotated with DCC software. CircBase database and circ2Trait disease database were used to annotated the identi ed circRNAs. The differentially expressed circRNAs between the two groups were identi ed T test statistical methods.
Analyses of circRNA-miRNA-mRNA interaction in CRC CircRNA-miRNA interaction were predicted by popular target prediction softwares including Circular RNA Interactome and RegRNA. Speci c predictions for the target gene of miRNA were based on miRanda, miRDB, miRWalk, RNA22 and Targetscan databases. All the circRNA-miRNA-mRNA networks were constructed by Cytoscape software.

RNA extraction and qRT-PCR
The total RNA was extracted by using TRIzol reagent (Life Technologies, Carlsbad, CA) and then reversetranscribed into cDNA using the SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA). The cDNA was used for qPCR using SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) with gene-speci c primers and the results were normalized with β-actin or U6 as a control. PCR primers are listed in Supplementary Table S1. circRNA RNase R resistance analysis and actinomycin D assay The SW620 and RKO cells were treated with 3 U/mg of RNase R (Epicenter, WI, USA) or 2 mg/L actinomycin D (Sigma, USA), then cultured at 37 °C. The cells were harvested at the indicated time points and the stability of circRNA_0000392 and YAF2 mRNA was detected by quantitative real-time PCR (qRT-PCR) assay.
Fluorescence in situ hybridization (FISH) SW620 and RKO cells were seeded in dishes and fused to 70-80%, then the cells were xed at room temperature with 4% paraformaldehyde, then treated with protease K. Then overlaid with FITC-labeled circRNA_0000392 probe (Gefanbio, China) at 65 °C for 48h. The signals of the probe were detected by a Fluorescent In Situ Hybridization Kit (Gefanbio, China) according to the manufacturer's protocol. Nuclei were counterstained with DAPI.

Luciferase reporter assay
The sequences of circRNA_0000392 and PIK3R3-3' UTR and their corresponding mutant versions without miR-193a-5p binding sites were synthesized and subcloned into luciferase reporter vector pmirGLO (Promega, Madison, WI, USA), named as circRNA_0000392 -WT, circRNA_0000392-Mut, PIK3R3 3' UTR-WT and PIK3R3 3'UTR-Mut, respectively. The plasmids were validated by sequencing and then cotransfected with the miRNA mimics or inhibitor with the negative controls, respectively. The relative luciferase activity was measured with a Dual Luciferase Assay Kit (Promega, Madison, WI, USA).

Transwell migration and matrigel invasion assays
The transwell chamber (Corning, Kennebunk, ME, USA) was using for the migration assays and the transwell chamber pre-coated with matrigel was using for the invasion assays. According to the protocol, the single cell suspensions were added to the upper chambers and incubated for 24 h. Then wash, x, and stain the cells with crystal violet. Base on the crystal violet staining, we calculate the migration and invasion rates through counting cells at least ve random elds.

RNA immunoprecipitation (RIP)
RIP assay was performed in SW620 and RKO cells. 1 x 10 7 cells were lysed by RNA lysis buffer completely, then incubated with RIP immunoprecipitation buffer containing magnetic beads conjugated with human anti-Argonaute2 (AGO2) antibody (Millipore, USA) or negative control mouse IgG (Millipore, USA). Add Proteinase K to the sample obtained from RIP and incubate at 55 °C for 30 mins. Then immunoprecipitated RNA was isolated and analyzed by qRT-PCR to quantify the enrichment of circRNA_0000392.

RNA pull-down
Biotin-labeled circRNA_0000392 probe or oligo probe (Gene-Pharma, China) were synthesized. SW620 and RKO cells were lysed with lysis buffer and incubated with speci c circRNA_0000392 probes. Then, SW620 and RKO cells were lysed with lysis buffer were lysed in lysis buffer and incubated with probecoated beads at 4 °C overnight. The beads were washed and the RNA complexes were extracted with TRIzol ( Life Technologies, Carlsbad, CA), then detected by qRT-PCR.

Immunohistochemistry
Detection of expression level of PIK3R3 by immunohistochemistry was performed on 5-μm thick para n sections based on patient tissue samples. Brie y, the sections were depara nized and rehydrated followed by antigen retrieving used 0.01M sodium-citrate buffer (pH 6.0) at a boiling temperature for 10 min. Then let the sections incubated with 3% hydrogen peroxide for 10 min, 5% bovine serum albumin for 1h and primary antibodies at 4 °C overnight. The sections were incubated with secondary antibodies after washing three times with PBS. Finally, the DAB system was used to display colors and the hematoxylin was to stain the nucleus. The immunostaining images were captured using Olympus FSX100 microscope (Olympus, Japan).
Xenograft tumor model BALB/c nude mice (male 3-to 4-week-old) were injected subcutaneously with 5 × 10 6 SW620 cells. Tumor volumes were measured with a caliper every 3 days, calculated from the length (a) and the width (b) by using the following formula: volume (mm3) = ab 2 /2. 30 days after injection, the animals were sacri ced, and the excised tumor tissues were removed to further assess tumor weight and pathological staining.

Statistical analysis
Statistical analyses were performed using GraphPad Prism 7.0 (GraphPad Software Inc., CA, USA), The Student's t-test, one-way ANOVA were used to compare differences between groups, as appropriate. The correlation between groups was analyzed by Pearson correlation. ROC curve analysis was performed to evaluate the diagnostic value. Data were presented as mean ± standard deviation (SD), p < 0.05 was considered statistically signi cant.

Additional methods
Cell transfection, Western blot, Cell proliferation, apoptosis assays are described in Supplementary Methods.

Results
Identi cation of circular RNAs by RNA-seq analyses in human CRC.
In order to obtain the expression pro les of circRNA and identify differentially expressed circRNAs in CRC patients, the secondary sequencing was used to pro le circRNAs expression in paired CRC tissues and adjacent normal tissues (ANT) from 6 patients with CRC. First, the scatter plot was showed the variation of circRNAs expression between the tumor tissues and adjacent normal tissues (Fig. 1A). Then under the cut-off of fold change > 2.0 and P < 0.05, the signi cant differentially expressed circRNAs between the two groups were shown in volcano plot ( Fig. 1B) and hierarchical cluster (Fig. 1C). In total, 66855 circRNAs were detected in the tissue samples by the sequencing, among which 19 circRNAs were signi cant up-regulated while 1668 were signi cant down-regulated ( Fig. 1D and Supplementary Table S2 and S3). GO and KEGG analyses of the host genes of differentially expressed circRNAs were shown in Supplementary Fig. S1 and S2. And base on the source of the circRNA formation, the pie chart showed the percentage of signi cant differentially expressed circRNAs from intronic, exonic, intergenic, antisense and sense overlapping (Fig. 1E). We then screened out the circRNAs that can be worked as miRNA sponges from the signi cant differentially expressed circRNAs and constructed a network map using cytoscape software ( Supplementary Fig. S3).
We overlapped the top 10 circRNAs that are signi cantly up-regulated or down-regulated (Supplementary Table S4) with circRNAs that function as miRNA sponges. From the overlap, we rst selected 4 circRNAs and veri ed the circRNA-seq results in 40 paired CRC tissues and adjacent normal tissues (ANT) by qRT-PCR. As shown in Fig. 1F, the circRNA_0000392 expression was signi cantly up-regulated in CRC tissues, which were consistent with the sequencing data. Of the two, circRNA_0000392 has approximately a 25fold change in sequencing and nearly a 13-fold change in 16 patients, meaning a more signi cant difference. Therefore, in this study, we will focus on the role of circRNA_0000392 in the tumorigenesis and progression of colorectal cancer.
circRNA _0000392 is up-regulated in CRC and associated with the progression.
To further con rm our results, we increased the number of tissue samples to 40 pairs, and detected the expression level of circRNA_0000392 by qRT-PCR. The results showed that in 29 of the 40 pairs of CRC tissues and adjacent normal tissues, the expression level of circRNA_0000392 was higher in tumor tissues than that of the adjacent normal tissues, and was signi cantly increased in the overall statistics of 40 pairs ( Fig. 2A and B). Then the ROC curve analysis was performed to assess the diagnostic value of circRNA_0000392 in CRC. It showed that circRNA_0000392 could discriminate CRC from adjacent normal tissues with AUC of 0.713 (95% CI: 0.598 -0.827, P < 0.01) (Fig. 2C). The associations between circRNA_0000392 expression level and clinical parameters of the CRC patients were listed in Table 1. There was no signi cant difference in the expression level of circRNA_0000392 between the two groups according to age, gender, tumor size and location. But the expression of circRNA_0000392 was signi cantly correlated with pathological stage (P = 0.0123), lymph node metastasis (P = 0.0055) and distant metastasis (P = 0.0084) (Fig. 2D -G, and Supplementary Fig. S4). The circRNA_0000392 expression level was also measured in a normal human colon mucosal epithelial cell line NCM460 and 7 CRC cell lines including HT29, HCT116, SW480, SW837, SW48, SW620, RKO. It showed that the circRNA_0000392 expression was markedly upregulated in human CRC cell lines compared with normal colon mucosal epithelial cell line (Fig. 2H). These results indicate that the circRNA_0000392 expression was elevated and may be involved in the tumorigenesis and development of CRC.
The characteristics of the circRNA _0000392 Before we delve into its speci c role in CRC, we have initially described the characteristics of circRNA_0000392. The genomic loci of circRNA_0000392 was shown in Fig.3A, and the spliced mature sequence length of circRNA _0000392 is 326 bp. Further the SW620 and RKO cells were treated with RNase R exonuclease and actinomycin D to verify the authenticity of circRNA_0000392. The circRNA_0000392 was resistant to RNase R (Fig. 3B) and actinomycin D (Fig. 3C and S4C), while YAF2 mRNA was signi cantly reduced after RNase R or actinomycin D treatment. RNA uorescence in situ hybridization (FISH) assay demonstrated that circRNA_0000392 was mainly localized in the cytoplasm (Fig. 3D). These data showed the circRNA_0000392 species was indeed circular.

Knockdown of circRNA _0000392 inhibits CRC cell proliferation and invasion
To explore the function of circRNA_0000392 in CRC cells, we rst designed two siRNAs targeting the back-splice region. Then the loss-of-function assays were performed in SW620 and RKO cells with relatively high expression of circRNA_0000392. After transfection the two siRNAs respectively, the expression of circRNA_0000392 was signi cantly reduced by siRNA #1 in both of the two cell lines (Fig.  4A). WST-1 assay demonstrated that down-regulation of circRNA_0000392 signi cantly inhibited the proliferation viability of SW620 and RKO cells ( Fig. 4B and C). We further analyzed whether circRNA_0000392 has an effect on the cell apoptosis of CRC cells by ow cytometry. By double staining Annexin V and PI, it showed that circRNA_0000392 knockdown signi cantly enhanced the cell apoptosis at 48 h post-transfection of the siRNA #1 or si-NC (Fig. 4D). Next, the abilities of cell migration and invasion after the siRNA transfection were performed using transwell assays with or without matrigel. As a result, the cell migration and invasion abilities of SW620 and RKO cells were signi cantly inhibited after knocking down the expression of circRNA_0000392 ( Figure. 4E and F). These results indicated that circRNA_0000392 was contributed to CRC cell proliferation and motility in vitro.
It is well known that function as miRNA sponge was one of the important mechanisms by which circRNA exerts its biological functions [10]. Since we previously predicted by prediction software that circRNA_0000392 was one of the circRNAs that could function as miRNA sponges, we mainly focus on the "miRNA sponges" to further explore its underlying mechanism in CRC cell proliferation. First, the RIP assay was performed in SW620 and RKO cells. The results showed that circRNA_0000392 was enriched in AGO2 immunoprecipitates, con rming that the circRNA_0000392 has the miRNA adsorption function (Fig. 5A). Then the circRNA-miRNA-mRNA interaction based on circRNA_0000392 was demonstrated by prediction and bioinformatics analysis using cytoscape software (Fig. 5B). From the prediction, we selected the top 5 candidate miRNAs to validated the speci c interaction by RNA pull down assay. The result showed that the miR-193a-5p has a dramatic difference of the pulled down level by circRNA_0000392 probe compared with oligo probe in both SW620 and RKO cells ( Figure. 5C and Supplementary Fig. S5A). In order to further con rm the interactions between circRNA_0000392 and miR-193a-5p, the dual-luciferase reporter assay was performed in 293T cells. The circRNA_0000392-wt or circRNA_0000392-mut plasmid was constructed with a luciferase reporter vector (Fig. 5D), and then cotransfected with miR-193a-5p mimic or NC in 293T cells, respectively. The dual-luciferase reporter assay results showed that miR-193a-5p mimics could signi cantly reduce the luciferase activity of circRNA_0000392-WT group, whereas had no effect on the mutant one (Fig. 5E). The expression levels of miR-193a-5p in 40 pairs of CRC tumor tissues and adjacent normal tissues were measured by qRT-PCR. The results showed that miR-193a-5p expression level in CRC tumor tissues was signi cantly lower than that in adjacent normal tissues ( Fig. 5F and Supplementary Fig. S6). And by Spearman correlation coe cient analysis, there was a negative correlation between miR-193a-5p and circRNA_0000392 expression in CRC tumor tissues (r = -0.365, P =0.021) (Fig. 5G). Overall, these results demonstrated that circRNA_0000392 acts as a sponge for miR-195-5p in the CRC. PIK3R3 is directly targeted by miR-193a-5p and indirectly regulated by circRNA _0000392.
According to our previous predictions, the EPHA2, PIK3R3, EGFR, USP22 and DDX58 are the most potential target genes for miR-193a-5p. Then we detected the mRNA expression level of these genes after transfected with the miR-193a-5p mimics or inhibitor in SW620 cells. The results revealed that the EPHA2 and PIK3R3 expression level were signi cantly downregulated by miR-193a-5p mimic and PIK3R3 expression level was upregulated after transfected with the miR-193a-5p inhibitor ( Fig. 6A and Supplementary Fig. S7A). The dual-luciferase reporter assay was following performed to con rm the binding relationship between PIK3R3 and miR-193a-5p (Fig. 6B). The PIK3R3 3'UTR WT and mutant plasmid was co-transfected with miR-193a-5p mimic in 293T cells. The results showed that cotransfection of PIK3R3 3'UTR WT plasmid and miR-193a-5p mimic signi cantly reduced the relative luciferase activity (Fig. 6C). Subsequently, we also tested whether miR-193a-5p affects the expression of PIK3R3. The qRT-PCR results revealed that miR-193a-5p mimic could markedly reduce the expression level of PIK3R3, while the expression level of PIK3R3 could signi cantly upregulated by miR-193a-5p inhibitor in both SW620 and RKO cell line (Fig. 6D). The protein level of PIK3R3 was signi cantly downregulated by the intervention of miR-193a-5p mimic (Fig. 6E). Then we detected the expression levels of PIK3R3 in 40 pairs of CRC tumor tissues and adjacent normal tissues. It showed that the PIK3R3 expression level in CRC tumor tissues was signi cantly higher than that in adjacent normal tissues and it negative correlated with miR-193a-5p expression in CRC tissues (r = -0.34, P =0.032) (Fig. 6F and Supplementary Fig. S7B -D). In order to explore whether the expression level of PIK3R3 could also regulated by circRNA_0000392, we detected the PIK3R3 expression after transfected with circRNA_0000392 siRNA. We found that knockdown of circRNA_0000392 signi cantly decreased the expressions of PIK3R3 (Fig. 6G). After the IHC staining of PIK3R3 in 40 CRC tissues, we found that PIK3R3 was positively correlated with circRNA_0000392 expression in CRC tissues (r = 0.385, P =0.014) ( Fig.6H and I). Collectively, these results demonstrated that the PIK3R3 was the target gene of miR-193a-5p and could regulated by circRNA_0000392.
To investigate whether circRNA_0000392 plays its role in promoting tumor progression through the /miR-193a-5p / PIK3R3 axis, rescue experiments were performed by transcription of miR-193a-5p inhibitors in circRNA_0000392 knockdown cells. The results of WST-1 and transwell assays indicated that the inhibition of proliferation and invasion by circRNA_0000392 knockdown could be rescued by inhibitor of miR-193a-5p in SW620 and RKO cells (Fig.7A -D). And the miR-193a-5p inhibitor could also rescue the effect of circRNA_0000392 knockdown on apoptosis (Supplementary Fig. S8). Then we detected the PIK3R3 mRNA expression by qRT-PCR and found that reduced PIK3R3 expression due to circRNA_0000392 siRNA could be alleviated by miR-193a-5p inhibitor (Fig.7E). In addition, the western blot assay revealed that knockdown of circRNA_0000002 decreased the protein levels of PIK3R3 and the phosphorylation level of AKT and mTOR, then the effects could be reversed by miR-193a-5p inhibitor (Fig.7F -I). Collectively, these results demonstrated that circRNA_0000392 might act as a regulator of miR-193a-5p, which further affects the expression of PIK3R3 to play a regulatory role in CRC.

Downregulation of circRNA _0000392 suppresses the growth of CRC cells in vivo
To explore the effects of circRNA _0000392 in vivo, the circRNA _0000392 knocking down SW620 cells and negative control cells were subcutaneously injected into the back of BALB/c nude mice. After 30 days of observation, the results showed that circRNA _0000392 knockdown reduced the volume and weight of SW620-derived tumors in vivo (Fig.8A -C). Then, the expression levels of Ki-67 and PIK3R3 in the two groups of tumor tissues were evaluated by immunohistochemical staining. As shown in the results, the expression levels of Ki-67 and PIK3R3 were both decreased in the tumor tissues which the expression of circRNA _0000392 was knocked down (Fig.8D -E). Therefore, inhibiting the expression of circRNA _0000392 could signi cantly inhibit the growth of CRC in vivo (Fig.8F).

Discussion
Colorectal cancer is one of the most common malignant tumors, and its incidence has increased year by year. In terms of treatment, although early CRC can be treated by endoscopic minimally invasive surgery and surgical eradication, since most CRC have no obvious clinical symptoms in the early stage, about 60% of CRC patients develop into the middle and late stages at the time of diagnosis, with lymph node and distant metastases [3]. For colorectal cancer, early detection and treatment can achieve a better prognosis, so it is important to screen for effective new biomarkers and to explore CRC pathogenesisrelated signaling pathways.
As a member of non-coding RNA, circRNAs form a closed continuous loop by covalent attachment of 3' RNA and 5' RNA [4,18]. As early as the 1970s, Sanger et.al [19] discovered the presence of singlestranded circular RNA in plant viruses. However, due to the limitations of detection, circRNA was considered to be a phenomenon of incorrect splicing during exon transcription [20], and thus its existence did not recieved su cient attention at that period. In recent years, with the development of highthroughput sequencing technology and bioanalysis, circRNA has become a research hotspot in the eld of biomedicine [21,22]. CircRNA is widely expressed in human cells and is tissue-speci c, with varying levels of expression in different type of tissues [23]. Due to the characteristics of circRNA, it has become a promising diagnosis a marker or therapeutic target for cancer. To date, many studies have identi ed circRNAs as diagnostic and prognostic biomarkers in distinct human cancers [24][25][26], and also reported the role of circRNAs in the progression of cancers [27][28][29][30]. However, the expression pro le and In our study, we performed the high-throughput sequencing of circRNA in cancer tissues and adjacent normal tissues of 6 colorectal cancer patients, and obtained the expression pro le of circRNA of colorectal cancer containing 66,855 circRNAs. Then the circRNAs differentially expressed in colorectal cancer tissues were identi ed by bioinformatics analysis. These circRNAs may become potential biomarkers and therapeutic targets for the diagnosis of colorectal cancer. Based on the data we analyzed, we selected some circRNAs with signi cant differences in expression and validated in extended samples. We found that circRNA_0000392 was signi cantly upregulated in colorectal cancer tissues and cell lines. And the expression level of circRNA_0000392 in colorectal cancer was markedly associated with clinical stage and malignant progression. The ROC curve analysis showed the diagnostic value of circRNA_0000392 in CRC, revealing that it may be a promising prognostic biomarker. Next, a series of functional experiments demonstrated that knockdown of circRNA_0000392 signi cantly inhibited cell proliferation and invasion of CRC cells, revealing its function as an oncogene.
In the research on circRNAs to date, the mechanism that could be used as a miRNA sponge has been one of the foundations for exploring its biological functions. Since Hansen [10] discovered that circRNA could function as a miRNA sponge and identi ed the ciRS-7 with a sponge function of miRNAs, a large amount of circRNAs with miRNA sponge function were revealed in human cancers [15,[31][32]. The RIP assay was used to con rmed that the circRNA_0000392 has the miRNA adsorption function. And based on circRNA_0000392, we rst predict circRNA-miRNA-mRNA interactions through target prediction software and constructed the networks. From the predicted candidates, we further con rmed that circRNA_0000392 could directly interact with miR-193a-5p by using RNA pull-down and dual luciferase reporter assays. The results of rescue experiments showed that the effect of decreased CRC cell proliferation and invasion caused by circRNA_0000392 knockdown was offset by inhibition of miR-193a-5p. It has been reported that miR-193a-5p mainly contribute as a tumor suppressor in a variety of cancers [33][34][35], and the interaction of circRNA_0000392 with miR-193a-5p attenuated the tumor suppressor e ciency of miR-193a-5p. Our results demonstrated that circRNA_0000392 serves as an oncogene by sponging the miR-193a-5p in CRC.
As a member of non-coding RNA, miRNAs were well known to exert biological effects by modulating their target genes. Those circRNAs with miRNA sponge function could adsorb miRNAs and release the targeted genes of miRNAs indirectly. After we determined that the miR-193a-5p could be adsorbed by circRNA_0000392, our next focus was to search for its effector target genes. Similarly, from the prediction we selected several of the most potential target genes, and by identi cation, we nally determined the PIK3R3. Our results showed that miR-193a-5p could speci cally bind to the 3' UTR region of PIK3R3, and regulated its expression level.
PI3K signaling is widely activated in human cancers and its role in tumor development and metastasis has been well-investigated. PIK3R3 was one of the mammalian gene from Class IA PI3Ks, encode p85a, p85b and p55g regulatory subunits [36]. It has showed that the PIK3R3 regulatory subunit was important for cell proliferation and tumorigenesis [37]. Also, PIK3R3 was overexpressed in some cancers and has been reported as an oncogene. Our data showed that the mRNA and protein expression levels of PIK3R3 in CRC tissues were elevated compared with those in adjacent normal tissues and had a signi cant positive correlation with the expression level of circRNA_0000392. Studies has showed that the expression level of PIK3R3 in CRC and ovarian cancer tissues has the same trend, which consisted with our results [38,39]. Our in vitro results showed that while knocking down the expression of circRNA_0000392, it could inhibit the CRC cell proliferation and invasion, while the expression level of PIK3R3 and phosphorylation levels of AKT1 and mTOR were also inhibited. Subsequently, rescue experiments also showed that the miR-193a-5p inhibitor restored the inhibition of cell proliferation by knocking down circRNA_0000392, and also restored the inhibition of this pathway by down-regulation of circRNA_0000392.

Conclusion
In conclusion, we revealed that circRNA_0000392 was upregulated in human CRC tissues by highthroughput sequencing, and it is a promising biomarker for CRC. Furthermore, we rst demonstrated that circRNA_0000392 acts as a sponge of miR-193a-5p to regulate the expression of PIK3R3, revealing the effect of the circRNA_0000392/ miR-193a-5p /PIK3R3 axis on the activation of the AKT-mTOR pathway, which is a novel mechanism for CRC progression. This study implied that circRNA_0000392 is a potential therapeutic target for the treatment of colorectal cancer and a predictive marker for CRC patients.

Consent for publication
The consent forms were signed by every participant, and will be provided upon request.

Availability of data and materials
The datasets used during the current study were available from the corresponding authors on reasonable request.

Competing interests
The authors declare that there was no con ict of interest. Authors' contributions HCX, YXX, and GJ discussed and designed this study; HCX, YJL, PQC and CYW performed the experiments; YL and YXX was responsible for collecting tissue specimen; HCX and WJZ conducted the data analyses; HCX drafted the manuscript. There are two corresponding authors in this manuscript. YXX has contributed to leading the surgery and collecting tissue specimen. GJ has contributed to study design, data interpretation, and editing the manuscript. All authors read and approved the nal manuscript.    Data represent mean ± SD. * P < 0.05, ** P < 0.01. assay with or without matrigel after circRNA¬_0000392 knocking down in SW620 and RKO cells. Data represent mean ± SD. * P < 0.05, ** P < 0.01. The correlation between circRNA_0000392 and miR-193a-5p in CRC tissues was analyzed by Spearman correlation coe cients. Data represent mean ± SD. * P < 0.05, ** P < 0.01, *** P < 0.001.