Participants in this study were either patients or healthy people who came for physical examination in the Affiliated Hospital of Nantong University (Nantong, China). Twenty pairs of GC and corresponding adjacent tissues were collected. In addition, peripheral blood serum samples of the following different populations were collected: newly diagnosed GC patients (100), superficial gastritis patients (30), atrophic gastritis patients (32), intestinal metaplasia patients (31), colorectal cancer patients (30), thyroid cancer patients (30), breast cancer patients (30) and age-matched healthy controls (80). Besides the above GC patients (100) who were pathologically or clinically diagnosed and had not been subjected neither to surgery related to the present study, nor chemotherapy or radiotherapy were collected, as well as peripheral blood serum samples of post-operative recurrence patients (65) and corresponding post-operative GC patients (16). All samples were stored at -80°C in a refrigerator until subsequent processing. The study was conducted under the approval of the Ethics Committee of Affiliated Hospital of Nantong University (ethical review report number: 2018-L055).

Human GC cell lines (MKN-45, SGC-7901, HGC-27, BGC-823, MKN-1, and AGS) and human gastric epithelial cells (GES-1) were purchased from the Stem Cell Bank of the Chinese Academy of Sciences (Shanghai, China), among which GES-1 was taken as the normal control. The cells were cultured in RPMI-1640 medium (Corning, Manassas, VA), which was subsequently supplemented with 10% Fetal bovine serum (FBS, Gibco, Grand Island, NY) and 1% penicillin and streptomycin, in a humidified incubator at 37°C and with 5% CO2.

The total RNA in cells and serum was extracted by TRIzol reagent (Invitrogen, Karlsruhe, Germany) and TRIzol LS reagent (Invitrogen, Canada), respectively. Then, their concentration and purity were detected by NanoDropTM One (Thermo Fisher Scientific, USA). Besides, genomic DNA (gDNA) was extracted from cells using the Revert Aid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, MA, USA).

The total RNA was reverse-transcribed into cDNA by a reverse transcription kit (Thermo Fisher Scientific, MA, USA) after extraction, under the following the protocols: 25°C for 5 minutes, 42°C for 60 minutes, and 70°C for 5 minutes. RT-qPCR was conducted using LightCycler ^® 480 (Roche, Switzerland) with the 2^−ΔΔCT relative quantitative calculation method. The housekeeping gene GAPDH and 18S, whose primers were synthesized by Sangon Biotech Corporation (Shanghai, China), were taken as the internal control for cells and serum samples, respectively. The specific divergent primers used in this study were synthesized by RiboBio Corporation (Suzhou, China). The sequences of the involved gene included: hsa_circ_0007507: AGTAGAAGAGGTGGTGGTATCA (forward) and GGCCTCCGATCACTCTCTCT (reverse); RASA1: CAGGGAAGTCTGGCAGTTATCT (forward) and TCTCCACACATAGCAATAATCCT (reverse); GAPDH: AGAAGGCTGGGGCTCATTTG (forward) and GCAGGAGGCATTGCTGATGAT (reverse); 18S: CGGCTACCACATCCAAGGAA (forward) and GCTGGAATTACCGCGGCT (reverse). Besides, PCR products were stored at -20°C in a refrigerator until subsequent Sanger sequencing and AGE.

Total RNA (10 μg) extracted from BGC-823 and MKN-1 cells was treated with Ribonuclease R (RNase R, 3-4U/μg) from Geneseed Biotech Co., Ltd (Guangzhou, China). The reaction system was then incubated at 37°C for 1 minute, followed by incubation at 70°C for 10 minutes, to inactivate the enzyme and terminate the reaction for subsequent reverse transcription. Next, 1 mg/ml actinomycin D was diluted into 2.5 μg/ml actinomycin D with a complete medium, and the ordinary medium in the six-well plate was replaced with the treated complete medium. Cellular RNA was extracted at 0h, 2h, 4h, 8h, and 12h, respectively.

The nuclear and cytoplasmic RNA of SGC-7901and BGC-823 cells were separated and isolated using a PARISTM Kit (Thermo Fisher Scientific) according to procedures in the manual. After trypsinization, the cells were counted to obtain a cell suspension of up to 10⁷, whose culture medium was discarded after centrifugation at low speed. The cell pellet was washed once with PBS, and subsequent operations on ice were performed. Besides, the cells with 100 to 500μl pre-chilled cell fractionation buffer were resuspended, followed by incubation on ice for 5-10 minutes. Then, they were centrifuged at 500g for 1-5min at 4°C, after which the cytoplasm was in the supernatant while the nucleus was in the pellet. The supernatant containing cytoplasm was transferred to a new tube for RNA extraction. The pellet containing nucleus was subsequently mixed with Cell Disruption Buffer, using the same volume as that of Cell Fractionation Buffer. The resulting mixture was vortexed until the lysate was uniform, thus obtaining the nucleus. The nuclear and cytoplasmic RNA were extracted according to the following instructions. An equal amount of 2× Lysis/Binding Solution was added and mixed at room temperature. Immediately gently pipet 3-4 times or invert the tube to mix well. Then, an equal amount of Ethanol absolute (ACS Grade) was added and fully mixed. The 700µl mixture was transferred into the adsorption column at a time and centrifuged at 12000 rpm for 1min. It was washed once with 700µl Wash Solution 1 and twice with 500µl Wash Solution 2/3, before getting centrifuged for 30s to remove any residue. The filter element was placed in the attached clean collection tube, while the RNA was eluted twice (40μl and 20μl respectively) with the Elution Solution preheated at 95-100°C and then centrifuged at 12000 rpm for 30s. Finally, the separated nuclear and cytoplasmic RNA solution were obtained, for storage at -80°C.

Selected for Luciferase reporter assay, MKN-1 and BGC-823 cells were inoculated in 24-well plates when grown into the third generation. Cell transfection was conducted when the fusion reached the range of 30%~40%. The hsa_circ_0007507 WT plasmid and several mimics of microRNA, with the predicted binding relationship with hsa_circ_0007507, were designed by Ribo. After 48 hours, Luciferase reporter assay was performed according to the protocols of the kit as follows. Firstly, 100μl of LAR II was added to the 96-well white plate, and then 20μl of the PLB lysed cell lysate was added to measure firefly fluorescence. Finally, 100μl Stop & Glo Reagent was added to detect Renilla fluorescence. Through a comparison of the relative fluorescence of the firefly to the Renilla fluorescence with the control well, the binding relationship between the molecules was verified.

As for the analysis of sequencing results, the reads were mapped to the latest UCSC transcript set by Bowtie 2 version 2.1.0, and the gene level of expression was estimated utilising RSEM v1.2.15, which was then normalized by the trimmed mean M-value (TMM). Genes which were differentially expressed were then identified using the edgeR program. Those showing altered expression with more than two-fold changes were considered to be differentially expressed. SPSS 26.0 (IBM SPSS Statistics, Chicago, USA) and GraphPad Prism 9.0 (GraphPad Software, La Jolla, CA) were used for statistical analysis. The analysis of relative hsa_circ_0007507 expression was conducted using GraphPad Prism 9.0 using Student’s t-test. The unpaired t-test was used to analyze the difference of expression between patients with different diseases and the healthy controls while the paired t-test was used to analyze the difference of expression among the same patients before and after surgery. Moreover, one-way ANOVA was used when more than two groups of data were being compared. The ROC curve was plotted using SPSS, also playing a role in conducting chi-squared testing and Spearman correlation testing for analyzing the correlation with clinical features. A P-value of less than 0.05 is considered statistically significant. OS rate was evaluated by employing the Kaplan‐Meier method. P value < 0.05 was considered to be statistically significant.

High-throughput sequencing of three GC tissues and their matched noncancerous tissues was carried out to identify circRNAs aberrantly expressed in GC tissues (Supplementary Files S1, S2), from which hsa_circ_0007507 was selected. According to the Human circRNA Database (circbank, http://www.circbank.cn/index.html), hsa_circ_0007507 was located on chr5_86564070_86687743, and the length of its mature transcript was 478 bp. As analyzed by circPrimer software, hsa_circ_0007507 was composed of exons 2, 3, 4, and 5 (Figure 1A). Considering that circRNAs usually form a closed loop through reverse splicing, a divergent primer that can reverse-amplify the circular molecule was designed based on the hsa_circ_0007507 circularization site. The level of expression of hsa_circ_0007507 was then detected by RT-qPCR, the product of which was detected by 2% AGE. The electrophoresis band was 112 bp, which was consistent with the size of the primer amplified product (Figure 1B). The reverse splicing site of hsa_circ_0007507 was further determined using Sanger sequencing (Figure 1C). Then, RT-qPCR was performed with genomic DNA (gDNA) and cDNA as templates, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as a negative control. Agarose electrophoresis revealed that hsa_circ_0007507 can be amplified from PCR products with cDNA as a template while negative results were observed in the control group with gDNA as a template (Figure 1D).

First, the expression of hsa_circ_0007507 was detected in 20 pairs of GC tissue and corresponding adjacent tissue samples by real-time quantitative PCR (RT-qPCR), and it was revealed to be significantly up-regulated in GC tissues (Supplementary Files S3). In order to explore its ability as a serum biomarker, the serum samples of 20 GC patients and 20 healthy donors were compared (Figure 2A). Many nucleases in cells can easily degrade RNA molecules while circRNAs have better stability and are not easily degraded. Therefore, BGC-823 and MKN-1 cells were treated with RNase R to verify the stability of hsa_circ_0007507. It was demonstrated that hsa_circ_0007507 was not easily degraded while its transcription gene RASA1 was severely degraded (Figure 2B). Besides, BGC-823 cells were cultured in a medium containing actinomycin D for 12 h, which can inhibit RNA production by inhibiting RNA polymerase. RT-qPCR showed that the half-life of hsa_circ_0007507 was significantly longer than the half-life of its transcription gene RASA1, further verifying the stability of hsa_circ_0007507 (Figure 2C). Furthermore, a large sample verification was conducted to verify the organ specificity of hsa_circ_0007507. Specifically, 30 serum specimens from patients with colorectal cancer were collected, and 22 normal human serum specimens were used as controls. The results of RT-qPCR indicated that the expression of hsa_circ_0007507 in colorectal cancer serum was not significantly different from that in normal human serum (p = 0.5319) (Figure 2D). Moreover, 30 serum samples from patients with thyroid cancer were collected, and 30 normal human serum samples were used as controls. The results of RT-qPCR proved that the expression of hsa_circ_0007507 in thyroid cancer serum was not significantly different from that in normal human serum (p = 0.5422) (Figure 2E). Moreover, 30 breast cancer serum samples were collected, and 27 normal human serum samples were used as controls. The results of RT-qPCR confirmed that the expression of hsa_circ_0007507 in breast cancer serum was not significantly different from that in normal human serum (p = 0.5178) (Figure 2F).

Considering the clinicopathological parameters of these patients, it can be concluded that the expression of hsa_circ_0007507 shows specificity in GC while the expression of colorectal cancer, thyroid cancer, and breast cancer is not statistically significant. Therefore, hsa_circ_0007507 has organ specificity in GC.

A large sample verification was conducted to assess the diagnostic significance of hsa_circ_0007507 in gastritis, intestinal metaplasia, and GC serum. Serum samples from 100 GC patients and 80 healthy donors were collected for detecting the difference in their level of expressions. Since the early diagnosis of many patients with GC is difficult, and a considerable number of cases were misdiagnosed as from gastritis and intestinal metaplasia, it is particularly important for us to distinguish GC from gastritis. Then we also collected the serum samples from 30 cases of superficial gastritis, 32 cases of atrophic gastritis and 31 cases of patients with early gastric cancer lesions (intestinal metaplasia). We performed statistical analysis on the expression of hsa_circ_0007507 in the serum samples of normal people, gastritis, gastric cancer and intestinal metaplasia. Statistically, the expression of normal subjects was significantly different from that of the other three groups (gastritis, intestinal metaplasia and gastric cancer) (p <0.0001), and there were statistically significant differences between gastritis and intestinal metaplasia (p =0.0158) and gastric cancer (p <0.0001), as well as between intestinal metaplasia and gastric cancer (p =0.0358). The results showed that, as gastritis developed towards gastric cancer, the expression of hsa_circ_0007507 showed an increasing trend (Figure 3A).

To understand whether hsa_circ_0007507 can be used as a potential GC diagnostic marker, ROC curves of hsa_circ_0007507, CEA, and CA199 were analyzed, and the area under the curve (AUC) was calculated based on the data obtained. The ROC curve demonstrated that the AUC value of hsa_circ_0007507 was 0.832 (95% CI: 0.771-0.892, p < 0.001) (Figure 3B), higher than CEA (0.765, 95% CI: 0.697-0.833, p < 0.001) (Figure 3C), and CA199 (0.587, 95% CI: 0.504-0.67, p = 0.044) (Figure 3D). Particularly, the combined use of hsa_circ_0007507 and CEA produced an AUC of 0.839; the combined use of hsa_circ_0007507 and CA199 resulted in an AUC of 0.835; the combined use of the three produced the largest AUC of 0.849 (Figure 3E). Besides, hsa_circ_0007507 was higher than CEA and CA199 in terms of sensitivity (73%), specificity (85%), overall accuracy (78%), positive predictive value (86%), and negative predictive value (72%) (Table 1). This analysis suggested that the combined use can make up for the limitations of a single marker, and hsa_circ_0007507 as a biomarker has potential in the diagnosis of GC. Meanwhile, 30 cases of superficial gastritis and 32 cases of atrophic gastritis were statistically analyzed by t-test, and the results showed that there was no significant difference between the two groups(p =0.3047). However, the expression of hsa_circ_0007507 in atrophic gastritis samples was higher than that of superficial gastritis (Figure 3F).

Then the 100 GC patients were divided into two groups [hsa_circ_0007507(high) and hsa_circ_0007507(low)] after the Youden index had been determined, which was calculated using SPSS software by importing the relative level of expressions of serum hsa_circ_0007507 in 100 GC patients and 80 healthy donors to evaluate the correlation between hsa_circ_0007507 expression and clinicopathological features in GC patients. As shown in Table 2, the level of expression of hsa_circ_0007507 was significantly related to tumor depth (p = 0.032*), lymph node metastasis (p = 0.028*), and TNM stage (p = 0.003**) using Pearson’s χ² test, while there was no evidence that hsa_circ_0007507 expression was correlated with gender (p = 0.828), age (p = 0.517), tumor size (p = 0.1), differentiation grade (p = 0.464), or nerve/vascular invasion (p = 0.122).

Post-operative serum samples of 16 patients were collected one week after surgery to evaluate hsa_circ_0007507 expression. The level of expression of hsa_circ_0007507 decreased significantly after surgery (Figure 4A). The results verified that hsa_circ_0007507 can be used for cancer surveillance. Besides, 65 relapsed patients and 36 non-relapsed patients serum samples were collected. The results showed that the serum hsa_circ_0007507 expression in relapsed patients was higher than that in non-relapsed patients, with statistical significance (p =0.0139) (Figure 4B). The survival rate of the hsa_circ_0007507 (low) group was higher than that of the hsa_circ_0007507 (high) group (Figure 4C).

To investigate the functional mechanism of hsa_circ_0007507 in GC cell lines, normal gastric mucosal epithelial GES-1 cells were used as controls to detect the level of expression of hsa_circ_0007507 in six GC cell lines (HGC-27, MKN-45, SGC-7901, MKN-1, AGS, and BGC-823). Besides, the expression of hsa_circ_0007507 was significantly up-regulated in BGC-823, and MKN-1, and SGC-7901, and down-regulated in HGC-27, MKN-45, and AGS (Figure 5A). Furthermore, RNA was extracted from BGC-823 and SGC-7901 through nuclear and cytoplasmic separation. It was revealed that hsa_circ_0007507 accounted for a higher proportion of the cytoplasm. The FISH assay also showed that hsa_circ_0007507 was mainly distributed in the cytoplasm of BGC-823 cell line (Supplementary Files S4). indicating that it may participate in the process of GC mainly through post-transcriptional regulation (Figure 5B). Next, the potential circRNA-miRNA-mRNA regulatory axis in GC was predicted using bioinformatics database analysis (miRanda, PITA, TargetScan). As illustrated in Figure 5C, 11 miRNAs (hsa-miR-676-5p, hsa-miR-544b, hsa-miR-7852-3p, hsa-miR-6878-3p, hsa-miR-4789-3p, hsa-miR-6747-3p, hsa-miR-4725-5p, hsa-miR-4743-3p, hsa-miR-4649-3p, hsa-miR-4312, and hsa-miR-6733-3p) and their corresponding target mRNAs were depicted. The interaction of hsa_circ_0007507 and these miRNAs was preliminarily verified by dual luciferase assay (Supplementary Files S5), indicating that hsa_circ_0007507, which was highly expressed in gastric cancer patients’ tissues, serum and gastric cancer cell lines, might be able to sponge functional mi-RNA and play a role in the occurrence and development of gastric cancer. This provides a new direction for future exploration of the regulatory network involved in hsa_circ_0007507 in GC.