Four samples from CRC patients, aged 50 to 62 years, were collected from Tianjin Union Medical Center, Nankai University Affiliated Hospital in 2016. The patients had not been treated with radiotherapy or chemotherapy before the surgery. The CRC tissue and their adjacent tissue samples were obtained after CRC surgeries, and recorded as treatment and control groups, respectively. The experimental study was approved by the hospital ethics committee, and the informed consent forms were obtained when the patients were accepted for the study by the hospital. All procedures performed in the studies involving human participants were in accordance with the ethical standards of Committee of the Tianjin Union Medical Center (Tianjin, China) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

The samples from the two groups were immediately washed with 0.9% NaCl (RNase-free) and quickly dipped in RNase inhibitor (Epicentre: Illumina, Inc., San Diego, CA, USA) according to the manufacturer's instructions. Then, total RNA was extracted from the CRC tissues using TRIzol (Invitrogen: Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to the manufacturer's instructions. Total RNA was digested with RNase R (Epicentre: Illumina, Inc.) to remove linear RNAs, and was then reversed transcribed as cDNA. The cDNA was purified with NucleoSpin^® Extract II kits (Macherey-nagel, Düren, Germany), and was labeled with a fluorescent probe (Arraystar Super RNA Labeling kit; Arraystar, Inc., Rockville, MD, USA). The fluorescent cRNA was purified with the the NucleoSpin^® Extract II kits. The concentration and specific activity of the labeled cRNAs (pmol Cy3/µg cRNA) were assessed using NanoDrop ND-1000 (Thermo Fisher Scientific, Inc.). A total of 1 µg of each labeled cRNA was fragmented by adding 5 µl of 10X blocking agent and 1 µl of 25X fragmentation buffer. Finally, 25 µl of 2X hybridization buffer was added to dilute the labeled cRNA. Then, a total of 50 µl mixtures were heated at 95°C for 3 min, and cooled 2 min on ice. The hybridization solution was dispensed into the gasket slide and assembled to construct the circRNA expression microarray slide. The slides were centrifuged for 16 h at 40 × g at 65°C in an Agilent Hybridization Oven (Agilent Technologies, Inc., Santa Clara, CA, USA). Finally, the hybridized arrays were washed, fixed and scanned using the Agilent scanner G2565CA (Agilent Technologies, Inc.).

The Agilent Feature Extraction 10.7 software (Agilent Technologies, Inc.) graph chip was used to read the values and to obtain the original data. The GeneSpring GX 12.5 software (Agilent Technologies, Inc.) was used for quintile normalization and subsequent data processing of the original data. After standardization, the cluster analyses of the eight samples were performed with hierarchical and average linkage algorithms of R package.

After single array and cluster analysis, the differentially expressed circRNAs were identified in CRC tissue samples compared with the adjacent tissue samples using P<0.05 and |log2(fold-change)|>1. Moreover, the volcano plot was plotted according to the P-value and log2(fold change). In addition, the Circos plot also was drawn.

KEGG orthology based annotation system (KOBAS) is a web server for gene/protein functional annotation (annotate module) and functional gene set enrichment (enrichment module) (1). In the present study, the related mRNAs of the differentially expressed circRNAs were annotated with the molecule annotation system (MAS) 3.0, and the diseases, pathways and functional enrichment analysis, in which they were enriched, were performed using KOBAS 3.0. The terms of enriched diseases included KEGG disease, FunDO, NHGRI GWAS catalog and OMIM; the enriched pathways contained Reactome, KEGG pathways, PANTHER and BioCyc; and BPs, molecular function (MF) and cellular component (CC) were involved in the Gene Ontology (GO) terms.

A circRNA can bind to a miRNA and indirectly regulate the translation of an mRNA, and miRanda is an algorithm used to find genomic targets for miRNAs. The related target miRNAs for the differentially expressed circRNAs were predicted with the miRanda algorithm. Subsequently, the key circRNAs with high |log2(fold-change)|, which were involved in the most circRNA/miRNA pairs were chosen. Ultimately, the circRNA/miRNA network for the key circRNAs was constructed and visualized using Cytoscape (7) software (Cytoscape 3.4.0, http://www.cytoscape.org/).

The tumor tissues and paracancerous specimens from four CRC patients were collected, after the patients had signed informed consent. The target gene of hsa_circ_0126897_CBC1 was found to be SLC4A4 from circBase (http://circrna.org/cgi-bin/singlerecord.cgi?id=hsa_circ_0126897). The SYBR^® Premix Ex Taq™ kit (Takara Biotechnology Co., Ltd., Shiga, Japan) and the Applied Biosystems™ QuantStudio™ 5 Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Foster City, CA, USA) were used to detect the mRNA levels of hsa_circ_0126897_CBC1 and SLC4A4 according to the manufacturer's instructions. All the primers were designed and synthesized by Takara Biotechnology Co., Ltd. (Beijing, China). The primers of hsa_circ_0126897_CBC1: 3′-CTCATGGTGATGCTGAACCTTCTTA-5′ and 3′-CTCATGGTGATGCTGAACCTTCTTA-5′ (139 bp); and SLC4A4: 3′-AGCACTCTATACGCCCCAAG-5′ and 3′-TTCCTTTTCTCTCACGCCCT-5′ (540 bp). In addition, β-actin was used as a reference, and the primer sequences were 5′-CTACAATGAGCTGCGTGTGG-3′ and 5′-AGGCATACAGGGACAACACA-3′ (308 bp).

SPSS v17.0 (SPSS, Inc., Chicago, IL, USA) was used for all statistical analyses, and data were expressed as the mean ± SD. To compare both two groups t-test was used, and P<0.05 was considered statistically significant.

After single array analysis, the cluster analysis between the two groups was performed. The cluster graph is displayed in Fig. 1A, which revealed that 4 CRC tissue samples (T1-T4) were clustered into one group, and the 4 adjacent tissue samples (C1-C4) were clustered into one class. Differentially expressed circRNAs were identified according to fluorescence signal intensity (Fig. 1B), and the volcano plot and the Circos plot are presented in Fig. 1C and D, respectively. Fig. 1 revealed that the differentially expressed circRNAs could separate tumor samples from paracancerous tissues, commendably. A total of 10,245 (6,264 up- and 3,981 downregulated) differentially expressed circRNAs were identified, and the top 30 most significant ones are presented in Table I. It was evident from Table I that hsa_circ_0027997_CBC1, hsa_circ_0060967_CBC1 and hsa_circ_0055546_CBC1 were the top 3 differentially expressed circRNAs.

The related mRNAs of the differentially expressed circRNAs were annotated by KOBAS 3.0, and they were enriched in some disease terms, GO terms and pathways. Furthermore, the enriched diseases included 462 KEGG diseases, 411 FunDO, 669 NHGRI GWAS catalog, and 845 OMIM, and the top 30 disease terms are shown in Fig. 2A-D. The KEGG diseases contained breast and ovarian cancer as well as CRC. The FunDO contained skin and stomach cancer as well as adenocarcinoma The NHGRI GWAS catalog contained lung and bladder cancer, and OMIM contained glioblastoma, gastric cancer and CRC. The enriched GO terms contained 8,485 BP; 2,026 MF; and 1,095 CC, and their hierarchy and top 30 GO terms are shown in Fig. 3A-D. The cell cycle process, the mitotic cell cycle and the cell cycle were the top 3 most significant BPs (Fig. 3D). A total of 1,334 Reactomes, 281 KEGG pathways, 117 PANTHER and 193 BioCyc were involved in the enriched pathways, and the top 30 terms of each are shown in Fig. 4A-D. Cell cycle, and ‘cell cycle, mitotic’ were the top 2 most significant Reactome pathways, cell cycle was the most significant KEGG pathway, and DNA replication was the most significant PANTHER pathway.

The related target miRNAs for the differentially expressed circRNAs were predicted. Then, the eight key differentially expressed circRNAs, which were associated with the most target miRNAs were chosen. They were hsa_circ_0126897_CBC1, hsa_circ_0052040_CBC1, hsa−circRNA6015−4_CBC1, hsa_circ_0126905_CBC1, hsa−circRNA15787_CBC1, hsa_circ_0138971_CBC1, hsa_circ_0126905_CBC1 and hsa−circRNA9432_CBC1, and were involved in 133 circRNA/miRNA pairs. Finally, the circRNA/miRNA network was established, and is presented in Fig. 5. In the network, hsa_circ_0126897_CBC1 was involved in the most pairs, which was significantly downregulated in the CRC samples compared with their adjacent tissue samples. In addition, the scatter plot figure of hsa_circ_0126897_CBC1 and its target gene (SLC4A4) is presented in Fig. 6. The mRNA level of hsa_circ_0126897_CBC1 was significantly decreased in the CRC samples compared to their adjacent tissue samples (P<0.0001, t=24.70). Moreover, the expression of target gene (SLC4A4) was also decreased in the CRC samples (P<0.0001, t=13.72).