3.1 Expression profile of circRNAs in GC tissues and paracancer tissues
To study the expression profile of circRNAs in GC tissues, we conducted circRNA sequencing in three pairs of GC tissues and paracancerous normal tissues and identified 27,870 circRNAs (GSE131414) (Figure 1A). According to the principle of expression difference multiple >2.0 and P<0.05, 2007 circRNAs were further screened (Figure 1B). Excluding 936 undefined circRNAs, 663 circRNAs were up-regulated and 408 circRNAs were down-regulated in GC patients. In combination with CircRNA Db, CircBase and CircBank databases, we selected circ_0004771 with length of 203bp for subsequent research. First, we detected the expression level of circ_0004771 in 20 pairs of tissue samples, and the results showed that the expression level of circ_0004771 in GC tissues is higher than that of the adjacent normal tissues (Figure 1C), which was consistent with the circRNA sequencing results. Next, we detected the expression level of circ_0004771 in GC cell lines (HGC-27, MKN-45, BGC-823 and MKN-1) (Figure 1D). The results showed that the expression level of circ_0004771 in MKN-1 has the highest fold change. In recent years, researchers have found that more and more circRNAs can be detected in human body fluids. For example, plasma circARS can be used as a biomarker to monitor tumor development and progression in prostate cancer[23]. Serum circSETDB1 can be used as a non-invasive biomarker for detecting the tumor progression of high grade serous ovarian cancer (SOC) patients and predict the response and recurrence to chemotherapy[24]. Serum circFoxO3a can act as a potential noninvasive prognostic biomarker and therapeutic target[25]. Therefore, we want to explore the possibility of plasma circ_0004771 as a biomarker. First, we detected the expression level of circ_0004771 in the plasma of 20 GC patients, revealing that the relative expression of circ_0004771 in the plasma of GC patients is higher than that of healthy group (Figure 1E). Next, we observed a well correlation of the circ_0004771 expression between tissues and plasma of GC patients (P<0.01, R2=0.6706) (Figure 1F). In order to further explore that plasma circ_0004771 originate from GC cells, we found that with the longer culture time of GC cells in the culture flasks, the relative expression of circ_0004771 in the supernatant of SGC-7901 and MKN-1 increased gradually, especially in SGC-7901. While there was no significant change in the relative expression of circ_0004771 in the supernatant of GES-1. It suggested that circ_0004771 may be released into the blood by GC cells. (Figure 1G).
3.2 Advantages of plasma circ_0004771 as a diagnostic biomarker
Circ_0004771 is originated from the NRIP1 gene, which locates at chromosome21. It consists of the backsplicing of exon2 and exon3 (Figure 2A). Moreover, Sanger sequencing confirmed that the cyclization site of the amplification products is consistent with that provided in CircBase database (Figure 2B). The above experiments verifed that circ_0004771 has a cyclic structure. Due to its circular structure, circ_0004771 is not easily degraded by Rnase R, while the liner NRIP1 expression decreased. (Figure 2C). Meanwhile, the half-life of circ_0004771 after being treated with actinomycin D is longer than that of its parent gene NRIP1 (Figure 2D).
3.3 Plasma circ_0004771 as a biomarker for the detection of large sample size
Because the detection of circ_0004771 in plasma have no standard internal reference, we compared a series of the most commonly used internal reference such as 18S rRNA, GAPDH, β-actin, β2M, Tub and RPⅡ in 20 GC patients and 20 healthy examiners. Since 18S rRNA has preferable stability in both patients and healthy donors (Supplementary Table 1), we finally chose 18S rRNA as the endogenous reference for detecting the relative expression of circ_0004771. In order to verify that circ_0004771 is suitable for clinical laboratory analysis, we used qRT-PCR method to detect the relative expression of circ_0004771. First, we constructed a standard recombinant plasmid for circ_0004771. The regression equation of the standard curve of circ_0004771 is y=−3.4424x + 34.732, and R2 is 0.9983, indicating that qRT-PCR is an effective method to detect the plasma circ_0004771 at different concentrations (Figure 3A). Next, we selected mixed plasma for precision determination of plasma circ_0004771. The results showed that the coefficient of variation (CV) of circ_0004771 in inter-assay is 3.59% and CV of circ_0004771 in intra-assay is 2.16% (Table 1). We placed the mixed plasma samples at room temperature for 0, 6, 12, 18 and 24 hours and repeated freeze-thaw for 0,1,3,5 and 10 times, and then detected the relative expression level of circ_0004771. The results revealed that there is no significant difference in the expression level of circ_0004771 in plasma (P>0.05), indicating that the detection of circ_0004771 had good repeatability and stability (Figure 3B, 3C). Finally, we correctly designed the divergent primer for circ_0004771, and the length of the amplification product of qRT-PCR is expected to be 125bp. AGE (Figure 3D) and melting curves (Figure 3E) further verified the accuracy and specificity of the amplified products. All the above results demonstrated that the detection of plasma circ_0004771 by qRT-PCR is suitable for clinical laboratory analysis.
Under this test model, we collected plasma samples from 120 GC patients, 40 precancerous lesions, 40 gastritis patients and 120 healthy donors. We found that plasma circ_0004771 expression level in GC patients was significantly higher than that in healthy controls (P<0.0001) as well as gastritis patients (P<0.0001). But there is no significant difference in the expression level of circ_0004771 between gastritis patients and healthy donors (P=0.1639). In addition, the expression level of circ_0004771 in precancerous patients is higher than that in GC patients (P=0.0356), suggesting the possibility of drug intervention before GC (Figure 3F).
3.4 Diagnostic utility of plasma circ_0004771 in GC
ROC curves and the AUC of ROC curves were performed on data from the 120 GC patients and 120 healthy controls to investigate the ability of plasma circ_0004771 as a diagnostic biomarker for GC. The results showed that the plasma circ_0004771 effectively distinguished the primary GC patients from the healthy subjects, and the AUC is 0.831 (95%CI: 0.779-0.883, P<0.0001) (Figure 4A). In addition, the AUC of plasma circ_0004771 in differentiating primary GC from gastritis patients is 0.845 (95%CI: 0.772-0.917, P=0.0356) (Figure 4B). Meanwhile, plasma circ_0004771 could also effectively distinguish primary GC patients from precancerous patients, with an AUC of 0.623 (95%CI: 0.524-0.722, P<0.0001) (Figure 4C).
Both Figure 4D and Table 2 showed that the combined detection of plasma circ_0004771, CEA and CA199 in plasma is superior to any of the biomarkers detected separately in the diagnosis of GC patients and healthy subjects (AUC=0.864, sensitivity =0.79, specificity =0.78). Plasma circ_0004771 combined with CEA, a commonly used clinical diagnostic marker for GC, could improve diagnostic specificity in distinguishing GC patients from gastritis patients (AUC=0.846, sensitivity=0.66, specificity=0.85) (Figure 4E, Table 3).
Combined with the clinicopathological data of 120 patients with GC, we found that the high expression level of plasma circ_0004771 is related to the histological classification (P<0.0001), lymph node metastasis (P=0.009), T stage (P<0.0001) and TNM stage (P=0.016), and has little correlation with age, sex and tumor size (Table 4). All the above indicated that plasma circ_0004771 has high diagnostic efficiency.
3.5 Effect of plasma circ_0004771 on tumor dynamic monitoring in GC patients
To verify the dynamic relationship between plasma circ_0004771 expression and tumor progression, we compared the expression level of circ_0004771 in 20 pre-treatment GC patients with 20 post-treatment patients. As is shown in Figure 5A, plasma circ_0004771 in primary GC patients was abnormally expressed and dynamic decreased after surgical treatment. Meanwhile, the expression level of plasma circ_0004771 was still up-regulated in patients with postoperative recurrence. All the above suggested the potential of plasma circ_0004771 in dynamic monitoring of GC patients.
3.6 Forecast the downstream regulatory network of circ_0004771 in GC
CircRNAs in the nucleus mainly regulates the transcription of parent genes, while circRNAs in the cytoplasm mainly play the role as competitive endogenous RNA (ceRNA)[26]. Using nucleo-cytoplasmic separation assay, we found that circ_0004771 mainly located in the cytoplasm (Figure 5B), suggesting that circ_0004771 may regulate the progression of GC after transcription. Then, we predicted the potential circRNA-miRNA-mRNA axis through circRNA sequencing and bioinformatics analysis. As is shown in Figure 5C, ten miRNAs (miR-4298, miR-339-5p, miR-4687-3p, miR-548, miR-148p-5p, miR-3921, miR-7974, miR-1267, miR-1200, miR-4277) and their corresponding mRNAs were presented, which may provide a new direction of circ_0004771 in the regulation of GC progression. However, more experiments are needed to confirm the in-depth regulation mechanism of circ_0004771 in GC.