Pepsinogen C expression–related lncRNA/circRNA/mRNA profile and its co-mediated ceRNA network in gastric cancer

The expression of pepsinogen C (PGC) is considered an ideal negative biomarker of gastric cancer, but its pathological mechanisms remain unclear. This study aims to analyze competing endogenous RNA (ceRNA) networks related to PGC expression at a post-transcriptional level and build an experimental basis for studying the role of PGC in the progression of gastric cancer. RNA sequencing technology was used to detect the differential expression (DE) profiles of PGC-related long non-coding (lnc)RNAs, circular (circ)RNAs, and mRNAs. Ggcorrplot R package and online database were used to construct DElncRNAs/DEcircRNAs co-mediated PGC expression–related ceRNA networks. In vivo and in vitro validations were performed using quantitative reverse transcription–PCR (qRT-PCR). RNA sequencing found 637 DEmRNAs, 698 DElncRNAs, and 38 DEcircRNAs. The PPI network of PGC expression–related mRNAs consisted of 503 nodes and 1179 edges. CFH, PPARG, and MUC6 directly interacted with PGC. Enrichment analysis suggested that DEmRNAs were mainly enriched in cancer-related pathways. Eleven DElncRNAs, 13 circRNAs, and 35 miRNA–mRNA pairs were used to construct ceRNA networks co-mediated by DElncRNAs and DEcircRNAs that were PGC expression–related. The network directly related to PGC was as follows: SNHG16/hsa_circ_0008197–hsa-mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p–PGC. qRT-PCR validation results showed that PGC, PPARG, SNHG16, and hsa_circ_0008197 were differentially expressed in gastric cancer cells and tissues: PGC positively correlated with PPARG (r = 0.276, P = 0.009), SNHG16 (r = 0.35, P = 0.002), and hsa_circ_0008197 (r = 0.346, P = 0.005). PGC-related DElncRNAs and DEcircRNAs co-mediated complicated ceRNA networks to regulate PGC expression, thus affecting the occurrence and development of gastric cancer at a post-transcriptional level. Of these, the network directly associated with PGC expression was a SNHG16/hsa_circ_0008197–mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p – PGC axis. This study may form a foundation for the subsequent exploration of the possible regulatory mechanisms of PGC in gastric cancer.


Introduction
The latest global statistics show that gastric cancer is the fifth most common malignancy and the third most fatal tumor (Smyth et al. 2020). One reason for the high fatality rate lies in its unclear pathogenesis, and the lack of early diagnostic markers and molecular therapeutic targets. Therefore, it is of great importance to explore the potential pathogenesis of gastric cancer to achieve an early diagnosis and to improve the prognosis of each patient.
Pepsinogen C (PGC), a precursor of pepsin C, belongs to the aspartic enzyme family, and is mainly secreted by gastric mucosal principal cells into the gastric cavity (Hassan et al. 2010). PGC expression was found to gradually decrease or even become deficient in the progression of superficial Li-rong Yan and Han-xi Ding contributed equally to this work gastritis-atrophic gastritis-gastric cancer, suggesting it might be an ideal negative biomarker of gastric cancer (Ning et al. 2005). Several scholars have described positive PGC expression also gradually decreased in well differentiated, medium differentiated, and poorly differentiated gastric cancer cells (Fernández et al. 2000). PGC expression was associated with a poor prognosis and might be an independent prognostic factor (Fernández et al. 2000). Functional studies have shown that PGC synthetic fragments in the mouse could stimulate the growth of normal cell lines. PGC expression was increased in acetic acid-induced gastric ulcers, indomethacin-induced gastric lesions, andHelicobacter pyloriinfection, showing that PGC played an important role in gastric mucosal healing (Kishi et al. 1997). In general, PGC might have an important role in multiple gastric diseases, including gastric cancer. However, its pathogenetic mechanism is rarely studied. Current research is limited to PGC as a negative biomarker for gastric cancer, which has a certain clinical application. However, studies on the regulation of PGC expression are lacking.
Competing endogenous RNA (ceRNA) is a class of transcript that might cause mutual regulation by competing for shared microRNA (miRNA) at a post-transcriptional level (Salmena et al. 2011). Multiple molecules function in a similar manner to ceRNAs, such as mRNAs, long non-coding (lnc)RNAs, circular (circ)RNAs, and pseudogenic RNAs (Qi et al. 2015). The mechanism of ceRNAs is analogous to sponge-absorbing matched miRNAs by miRNA response elements (MREs), thus eliminating the effect of decreased miRNAs on the transcription and translation of target genes . The role of the ceRNA network is extensive and likely includes regulating the progression of different tumors, including gastric cancer. This provides an important clue for exploring pathogenetic mechanisms and to construct a ceRNA network mediated via lncRNAs and circRNAs.
MicroRNA can combine with the 3' untranslated region (3'-UTR) of PGC to regulate its expression (Shen 2017). At present, studies on interactions between PGC and multiple non-coding RNAs have been rare, and its regulatory mechanism at the post-transcriptional level is still unclear in the progression of gastric disease. This study aims to analyze and validate the differential expression profiles of PGC expression-related lncRNAs, circRNAs, and mRNAs, as well as the ceRNA network mediated via lncRNAs and cir-cRNAs. This might be an important foundation for exploring PGC pathogenetic mechanisms in gastric cancer.

Establishment of PGC-overexpressing cell line
AGS gastric cancer cells (purchased from National Biomedical Laboratory Cell Resource Bank, Beijing, China) were cultured in RPMI Medium 1640 (Solarbio, Beijing, China) containing 10% embryonic bovine serum (Biological Industries, USA) in a 37 °C, 5% CO 2 incubator. Treatment and control groups of cells were used in the study, with three replicates in each group. AGS cells were transfected with PCG-overexpressing and negative control plasmids. According to the protocol of a jet-PRIME transfection kit (Polyplus Transfection, USA), 2 μg of plasmid was added to 200 μL transfection buffer and mixed well, vortexed for 10 s, and then briefly centrifuged. Then, 4 μL of transfection reagent was added, the sample vortexed for 1 s, and briefly centrifuged. Finally, the premix was incubated at room temperature for 10 min and added to cells in a well of a 6-well plate containing 2 mL of 10% serum/ RPMI Medium 1640. Cells were then incubated for 24 h and then collected, RNA extracted, and RNA sequencing technology (RNAseq) used to detect PGC expression to verify successful transfection of the PGC plasmid.

Library construction and high-throughput sequencing
Total RNA was extracted using Trizol reagent (Invitrogen, CA, USA) following the manufacturer's procedure. The total RNA quantity and purity were analyzed using Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA) with RIN number > 7.0. Ribosomal RNA (rRNA) was isolated from the two cell groups using an Epicentre Ribo-Zero kit (Illumina, San Diego, CA, USA). The remaining rRNA-depleted RNA (polyA + and polyA-) was purified and recovered. The purified and recovered rRNA-depleted RNA with linear RNA molecules was then digested by 5U RNase R. The purified and recovered cir-cRNA was randomly interrupted into short fragments by Fragmentation Buffer. The fragmented rRNA-depleted RNA was used as a template to synthesize a strand of cDNA with random hexamers. Buffer, deoxyribonucleotide triphosphates, RNaseH, and DNA polymerase I were added for two-stranded cDNA synthesis. AMPure XP beads were used to purify the double-stranded product. T4 DNA polymerase and Klenow DNA polymerase activity were used to repair the sticky end of the DNA to a blunt end. Base A was added to the 3' end as well as a linker, and AMPure XP beads used for fragment selection. A USER enzyme was then used to degrade the second strand of cDNA containing U. Finally, PCR amplification was performed to obtain a final sequencing library. After the library was qualified, it was sequenced using an Illumina Hiseq2000/2500, with a sequencing read length of 2 × 125 bp (PE125) used.

Differentially expressed analysis
Fragments per kilobase of exon model per million mapped reads was used to measure the abundance of gene expression. A DESeq R package was utilized for differential analyses with intergroups. |log2FC|≥ 1 and P < 0.05 were considered statistically significant. Heatmaps and ggplot2 R packages were used to visualize the differential expression (DE) profiles of mRNAs, lncR-NAs, and circRNAs.

Construction of protein-protein interaction network and enrichment analysis of DEmRNAs
A protein-protein interaction (PPI) network for PGC expression-related mRNAs was constructed by a STRING online database (https:// www. string-db. org/). An interaction score > 0.4 was considered as a cut-off value. DEmRNAs were interpreted using the Gene Ontology (GO) system of classification, including biological processes, cellular components, and molecular functions classifications, and a Kyoto Encyclopedia of Genes and Genomes database. P < 0.05 was considered statistically significant.

Cis regulatory network construction of DElncRNAs
Long non-coding RNAs can modulate adjacent mRNA expression to generate biological roles by a cis-regulatory network. Therefore, the potential target mRNAs of DElncR-NAs were predicted. Long non-coding RNA cis-regulated target genes were mainly predicted based on location distribution, with the lncRNA cis-regulated network defined as constructed by DElncRNAs and DEmRNAs within 10 kbp of the chromosome. Pearson correlation analysis was performed on cis lncRNA-mRNA pairs (1 ≥ r ≥ -1, P < 0.05). The cis lncRNA regulatory network was visualized by Cytoscape 3.7.2.

Construction of lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA networks
DElncRNA-DEmRNA and DEcircRNA-DEmRNA pairs were carried out using Pearson correlation analysis by a ggcorrplot R package, respectively. Positive DElncRNA -mRNA pairs and DEcircRNA-circRNA pairs (r ≥ 0.8, P < 0.05) were chosen to construct a ceRNA network mediated by DElncRNAs and DEcir-cRNAs, respectively. The target miRNAs of DElncR-NAs and DEcircRNAs were predicted by StarBase (http:// starb ase. sysu. edu. cn/ starb ase2/ index. php). The target miRNAs of mRNAs positive with DElncRNAs and DEcircRNAs were predicted by miRTarBase and TarBase of NetworkAnalyst (https:// www. netwo rkana lyst. ca/). Finally, lncRNA-miRNA-mRNA and cir-cRNA-miRNA-mRNA were visualized by Cytoscape 3.7.2. The cytoHubba plug-in was used to calculate networks and to choose the top three lncRNAs and circRNAs with a degree score to an in-depth analysis, respectively.

Construction of co-regulated network mediated by DElncRNAs and DEcircRNAs
The mRNA-miRNA pairs that were shared with lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA were selected to construct a co-medicated ceRNA network by DElncRNAs and DEcircRNAs. Cytoscape 3.7.2 was used to visualize these. The cytoHubba plug-in was also used to calculate the network and select the top three lncRNAs and circRNAs with a degree score to an in-depth analysis, respectively.

Validation in vivo and in vitro
In an in vivo validation, we employed 46 pairs of gastric cancer and adjacent normal tissues collected from the First Affiliated Hospital of China Medical University. Our research was implemented according to the Declaration of Helsinki and supported by the research ethics committee of the First Affiliated Hospital of China Medical University. Written informed consent was obtained from all patients before samples were collected. For in vitro validation, a lentivirus was used to construct PGC stably transfected gastric cancer AGS and HGC-27 cells (purchased from National Biomedical Laboratory Cell Resource Bank, Beijing, China). The required amount of virus was calculated according to multiplicity of infection values for AGS and HGC-27 cells as recommended in each manufacturer's instructions. Opti-MEM (500 µL) was used and polybrene was added to a final concentration of 5 µg/mL. After 72 h, 2 μg/mL of puromycin was added to screen for transfected cells. The infection efficiency was verified by quantitative reverse transcription (qRT)-PCR.
Total RNA was extracted from AGS and HGC-27 cells overexpressing PGC, as well as 46 pairs of gastric cancer and adjacent normal tissues, and the relative expression level of genes was detected by qRT-PCR. Quantitative reverse transcription-PCR experiments were performed using a real-time PCR 480 system and SYBR-green PCR master mix. All curves were a single peak. Measurements were normalized using β-actin. Primer sequences are listed in Supplementary Table 1. SPSSv24.0 (IBM, Chicago, IL, USA) and GraphPad Prism V8.0 (GraphPad software, San Diego, CA, USA) were utilized for data analysis, and 2 −ΔCt was used to calculate relative expression. Differential expression profiles were assessed by Student's t-test for normally distributed data while a rank sum test was used for skewed distribution data. Spearman correlation analysis was used to calculate the correlation between the expression of genes. A chi-square test was utilized to analyze the association of gene expression with clinicopathological parameters. P < 0.05 was considered statistically significant.
In order to further examine the co-expression and change trends of PGC and its target regulators in the processes of superficial gastritis → atrophic gastritis → gastric cancer, we investigated the expression profiles of PGC regulators in this disease chain using GSE164166 (3 cases per group), which is the only transcriptional sequencing dataset available in the Gene Expression Omnibus database (https:// www. ncbi. nlm. nih. gov/ geo/).

Differential expression profile of mRNAs, lncRNAs, and circRNAs based on PGC expression
RNA sequencing showed that compared with a negative control, PGC expression was upregulated by a 111.58-fold change in the overexpressing group, verifying that the PGC plasmid was successfully transfected (Table 1). Differential expression profiles for PGC expression-related mRNAs, lncRNAs, and circRNAs indicated that a total of 637 DEm-RNAs (377 upregulated and 260 downregulated), 698 DEl-ncRNAs (376 upregulated and 322 downregulated), and 38 DEcircRNAs (15 upregulated and 23 downregulated) could be identified. The heatmap and volcano map of DERNAs are shown in Fig. 1.

PPI network and enrichment function of DEmRNAs
The PPI network of DEmRNAs was constructed by STRING and an interaction score > 0.4 was set as a cut-off value. It was found that the PPI network of DEmRNAs consisted of 503 nodes and 1179 edges (Fig. 2). In addition, this study found that PGC could directly interact with complement factor H (CFH), peroxisome proliferator activated receptor gamma (PPARG), and mucin 6 (MUC6); Table 2, Fig. 2). Gene Ontology enrichment analysis indicated that DEm-RNA might be involved in various biological processes, such as signal transduction, the G protein-coupled receptor signaling pathway, regulation of transcription, DNA-templates, and differentiation. These DEmRNAs might take part in cellular components, such as the membrane, plasma membrane, and integral components of the membrane. In addition, the molecular function of DEmRNAs mainly included protein binding, metal ion binding, DNA binding, and hydrolase activity (Fig. 3). Pathway enrichment analysis suggested that PGC expression-related DEmRNAs were mainly involved with cell adhesion molecules, and cAMP, chemokine, and oxytocin signaling pathways (Fig. 4).

The cis-regulatory pathway mediated by DElncRNAs
Differential expression genes were selected within 10 kbp upstream and downstream of each DElncRNA host gene as possible cis-regulated genes in order to analyze the regulation of each lncRNA and its neighboring genes. A total of 105 differentially expressed cis-regulatory genes were found upstream and downstream within 10 kbp of 94 DElncRNAs, which constituted 181 cis lncRNA-gene pairs. The Pearson Fig. 2 The protein-protein interaction network of DEmRNAs associated with PGC expression. The one circled in red represented protein that probably interact with PGC. DE, differentially expressed correlation coefficient between each DElncRNA and adjacent genes was calculated. It was found that 8.3% of gene pairs showed a strong positive correlation (r ≥ 0.9), 25% of gene pairs showed a moderate degree of positive correlation (0.9 > r ≥ 0.5), 50% of gene pairs had a weaker correlation (0.5 > r > − 0.5), and 16.7% of gene pairs had a moderately negative correlation (− 0.5 ≥ r > − 0.9). No gene pair showed a strong negative correlation (Fig. 5, Supplementary Table 2).

The ceRNA network mediated by DElncRNAs
Pearson correlation analysis was performed on DElncRNAs and DEmRNAs to select DElncRNA -DEmRNA pairs (P < 0.05 and r ≥ 0.8) in order to construct a ceRNA network. Next, DElncRNA-miRNA and DEmRNA-miRNA pairs were predicted. In addition, PGC-interacting miRNAs verified by our team in previous studies were also included in a subsequent analysis, including hsa-mir-98-5p, hsa-let-7f-5p, hsa-let-7c-5p, hsa-mir-365a-3p, and hsa-520a-5p. It was found that the ceRNA network mediated by DElncR-NAs consisted of 19 DElncRNAs, 85 DEmRNAs, and 76 miRNAs (Fig. 6). Further, a cytoHubba plug-in was used to select the top three DElncRNAs with the highest degree scores, including HCG18, SNHG16, and SNHG1. After an in-depth understanding of these DElncRNAs, it was found that only an SNHG16-hsa-mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p-PGC regulatory network was directly related to PGC, and that the remaining molecules were indirectly associated (Fig. 6).

The ceRNA network mediated by DEcircRNAs
Pearson correlation analysis was performed on DEcircR-NAs and DEmRNAs to select DEcircRNA-DEmRNA pairs (P < 0.05 and r ≥ 0.8) in order to construct a ceRNA network. Next, DEcircRNA-miRNA pairs and DEmRNA-miRNA pairs were predicted. Twenty-one circRNAs, 32 mRNAs, and 43 miRNAs were used to build a ceRNA network mediated by a DEcircRNA network (Fig. 7). Further, a cyto-Hubba plug-in was used to select the top three DEcircRNAs with the highest degree scores, including hsa_circ_0031583, hsa_circ_0008197, and hsa_circ_0036627. After an in-depth understanding of these DEcircRNAs, it was found that only a hsa_circ_0008197-hsa-mir-98-5p/hsa-let-7f-5p/hsalet-7c-5p-PGC regulatory network was directly related to PGC, and the remaining molecules were indirectly associated (Fig. 7).

Co-regulated ceRNA network mediated by DElncRNAs and DEcircRNAs
Based on a ceRNA network mediated by DElncRNAs and DEcircRNAs, DEmRNA-miRNA pairs shared by these were screened. A total of 11 DElncRNAs, 13 DEcircR-NAs, and 35 miRNA-mRNA pairs were used to construct a ceRNA network co-mediated by DElncRNAs and DEcircRNAs (Fig. 8). A similar method was used to screen out the top three molecules with the highest degree scores in a ceRNA network co-mediated by DEl-ncRNAs (SNHG14, HCG18, and SNHG16) and DEcir-cRNAs (hsa_circ_0031583, hsa_circ_0008197, and hsa_ circ_0036627). We further analyzed a ceRNA network co-mediated by these DEcircRNA and DElncRNA molecules, and found that an SNHG16/hsa_circ_0008197-hsamir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p PGC network was directly involved in PGC (Fig. 9); other molecules had an indirect relationship.

Validation results in vitro and in vivo
Compared with a negative control group, PGC-overexpressing AGS and HGC-27 cells showed 1566-and 3177.3-fold changes in upregulation, respectively (Fig. 10A), verifying that PGC-overexpressing cells were successfully constructed. Quantitative reverse transcription-PCR was used to perform in vitro validation including PGC-interacting proteins (PPARG, CFH, and MUC6), SNHG16, and hsa_ circ_0008197 directly related to PGC. Validation results using AGS and HGC-27 cells showed that compared with the negative control, PPARG, SNHG16, and hsa_ circ_0008197 were upregulated in PGC-overexpressing cells (Fig. 10A). There was no statistically significant difference in CFH and MUC6 levels (Supplementary Figure S1). Statistically significant molecules at a cellular level were identified by in vivo validation within 46 pairs of gastric cancer and adjacent normal tissues. Compared with adjacent normal tissues, PGC, PPARG, SNHG16, and hsa_circ_0008197 were downregulated in gastric cancer (Fig. 10B). Spearman correlation analysis showed that PGC was positively associated with PPARG (r = 0.276, P = 0.009), SNHG16 (r = 0.35, P = 0.002), and hsa_circ_0008197 (r = 0.346, P = 0.005); these results were consistent with our above analysis. The association of PGC, SNHG16, hsa_circ_0008197, and PPARG expression with clinicopathological parameters is detailed in Supplementary Tables 3-6. The result by GSE164166 analysis indicated that PGC expression was gradually decreased in the processes of superficial gastritis → atrophic gastritis + intestinal metaplasia → gastric cancer, with a statistical difference ( Figure   Fig. 5 The cis-regulatory network of DElncRNAs associated with PGC expression. The rectangle represents DEmRNAs and the equilateral triangle represents DElncRNAs. Blue color represents a medium intensity positive correlation, green color represents a high intensity positive correlation, gray color represents a weak correlation, and orange color represents a medium intensity negative correlation. DE, differentially expressed S2A). The expression of PPARG showed a decreased trend, but it was no statistically significant in the processes of superficial gastritis → atrophic gastritis + intestinal metaplasia, while its expression was continuedly downregulated with a statistical difference in the processes of atrophic gastritis + intestinal metaplasia → gastric cancer. In general, PPARG was gradually declined in the disease chain as the disease progresses ( Figure S2B). SNHG16 was gradually increased in the disease chain, but none of them reached a statistical difference ( Figure S2C). GSE164166 could not include the expression profile of hsa_circ_0008197 in these disease processes.

Discussion
As a marker of organ differentiation and maturation (especially gastric differentiation), PGC has unique physiological and pathological characteristics (Elabiad and Zhang 2011), (Feng et al. 2008). At present, research on PGC in gastric Fig. 6 DElncRNAs mediated ceRNA network associated with PGC expression. Red color represents upregulated lncRNAs, purple color represents downregulated lncRNAs, blue color represents upregulated mRNAs, orange color represents downregulated mRNAs and green color represents miRNAs. ceRNA, competitive endogenous RNA, DE, differentially expressed cancer has mainly focused on its expression level, with its upstream regulatory mechanism yet to be explored.
A ceRNA network mediated by DElncRNA/DEcircRNA plays an important role in the formation and development of tumors and has been used to explore the pathology, diagnosis, and prognosis of different tumors, including gastric cancer (Cheng 2020). In this study, RNA-seq technology was used to detect mRNAs, lncRNAs, and circRNAs based on PGC expression; an enrichment analysis of DEmRNAs was performed and a PPI network was constructed to explore its biological role and potential interacting molecules. A cis lncRNA regulatory network was predicted to analyze its potential role in gastric cancer. The positive correlation between lncRNA-mRNA pairs and circRNA-mRNA pairs was screened and their common target miRNAs were predicted. A ceRNA network was systematically constructed Fig. 7 DEcircRNAs mediated ceRNA network associated with PGC expression. Red color represents upregulated circRNAs, purple color represents downregulated circRNAs, blue color represents downregu-lated mRNAs, orange color represents upregulated mRNAs, and pink color represents miRNAs. ceRNA, competitive endogenous RNA, DE, differentially expressed

Fig. 8
DEcircRNAs and DEl-ncRNAs mediated co-regulated network associated with PGC expression. Blue color represents downregulated circRNAs, purple color represents downregulated lncRNAs, yellow color represents upregulated lncRNAs, gray color represents downregulated mRNAs, green color represents upregulated mRNA, and orange color represents miRNAs. DE, differentially expressed that was co-mediated by PGC expression-related DElncR-NAs/DEcircRNAs. The sequencing results were further verified by qRT-PCR using gastric cancer cells and tissues, which formed the basis for further research on the post-transcriptional regulatory mechanisms of PGC in gastric cancer.
RNA sequencing indicated a total of 637 mRNAs, 698 lncRNAs, and 38 circRNAs were differentially expressed, which suggested the association of these RNAs with the expression and function of PGC. Enrichment results showed that DEmRNAs related to PGC expression mainly focused on cell adhesion molecules, and chemokine, oxytocin and cAMP signaling pathways. These pathways were shown to be closely related to the occurrence and development of tumors (Rodríguez et al. 2014), (Fawcett et al. 1992), (Ehling et al. 2016). Recently, researchers have claimed that the ceRNA network mediated by noncoding RNA regulated these pathways and thus had an influence on the development of different tumors (Xian SNHG16, and hsa_circ_0008197 detected by qRT-PCR in gastric cancer tissues. (Measurement data were expressed as mean ± SEM, *P < 0.05, **P < 0.01 ***P < 0.001). OE, overexpressed. Ctrl, control. qRT-PCR, quantitative real-time polymerase chain reaction et al. 2019), (Kan et al. 2015). In other words, PGC regulated and modified the function of these pathways through a ceRNA network mediated by PGC expression-related DElncRNAs and DEcircRNAs, thus leading to a regulatory role in gastric cancer cells at the post-transcriptional level. The PPI network of DEmRNAs was made up of 503 nodes and 1179 edges. A direct interaction relationship exists between PGC-CFH (interaction score = 0.458), PGC-PPARG (interaction score = 0.452), and PGC-MUC6 (interaction score = 0.429). The qRT-PCR validation in vivo and in vitro indicated that only the differential expression of PPARG was statistically significant, which is in the same direction as PGC. Spearman correlation analysis also indicated that PGC was positive with PPARG (r = 0.276,P= 0.009). PPARG is a type of nuclear hormone receptor that can regulate various cell functions, including lipogenesis, lipid biosynthesis, energy consumption and storage, and inflammation (Desvergne and Wahli 1999). The study found that PPARG also promoted epithelial cell differentiation and inhibited tumor cell proliferation (Xin 2019), , which is consistent with the conclusion that PGC is a marker of organ differential maturation. Such conclusions suggested that PGC may interact with PPARG to co-regulate the differential maturation of gastric cancer cells, thus affecting tumor progression. But, the specific mechanism still needs to be further explored.
To understand the potential roles of PGC expression-related lncRNAs, adjacent target genes were predicted to construct a cis-regulatory network. A total of 105 differentially expressed cis-regulatory genes were found within 10 kbp upstream and downstream of 94 DElncRNAs, which constituted 181 cis lncRNA-gene pairs. The cis-regulation function of non-coding RNA might participate in various biological processes (Guttman et al. 2009) while the cisregulation function of lncRNA might affect tumor progression (Elcheva 2020), ). Our analysis indicated that DElncRNAs were mainly positive with adjacent proteincoding genes, with a negative correlation being relatively few in number. Therefore, we speculated that PGC expression-related DElncRNAs mainly positively regulated the expression of adjacent protein-coding genes. In addition, a ceRNA network mediated by DElncRNAs was constructed by 19 DElncRNAs, 85 DEmRNAs, and 76 miRNAs, of which HCG18, SNHG16, and SNHG1 showed the highest degree scores. Most research indicated that ceRNA mediated by DElncRNA regulated the progression of gastric cancer, colorectal cancer, and osteosarcoma (Wang 2019), ) . HCG18 was shown to promote gastric cancer progression by upregulating DNAJB12 via miR-152-3p . SNHG1 facilitated the growth and migration of gastric cancer cells via the miR-140/ADAM10 axis . This study initially found that these lncRNAs were associated with PGC expression and its mediated ceRNA network might have affected gastric cancer progression.
A ceRNA network mediated via PGC expression-related DEcircRNAs was constituted to explore a potential role. It was found that 21 DEcircRNAs, 32 DEmRNAs, and 43 miRNAs constituted a ceRNA network mediated via DEcircRNAs. Of these, the top three with the highest degree score of DEcircRNAs included hsa_circ_0031583, hsa_circ_0008197, and hsa_circ_0036627. A ceRNA network mediated via DEcircRNAs modulated histological classifications and gastric cancer progression (Cheng et al. 2018) and the stemness properties of colorectal cancer stem cells (Rengganaten 2020). No report has shown that hsa_ circ_0031583, hsa_circ_0008197, and hsa_circ_0036627 affected tumor progression. Using bioinformatics, this study initially revealed that the ceRNA networks mediated by PGC expression-related hsa_circ_0031583, hsa_circ_0036627, and hsa_circ_0036627 may play a major role in the progression of gastric cancer. This provides important clues for subsequent research of mechanisms in gastric cancer.
Based on a ceRNA network mediated by DElncRNAs and DEcircRNAs, a total of 11 DElncRNAs, 13 circRNAs, and 35 miRNA-mRNA pairs were used to construct their co-mediated ceRNA network. As is well known, ceRNA networks mediated via non-coding RNAs play an important role in gastric cancer (Luo and Liang 2020), (YiRen et al. 2017), . In this study, the co-regulated network included the SNHG16/hsa_circ_0008197-hsamir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p-PGC axis. This was directly related to PGC and was of great significance in exploring its post-transcriptional regulatory mechanisms in gastric cancer. Therefore, qRT-PCR validation for SNHG16 and hsa_circ_0008197 was performed in vivo and in vitro. It was found that SNHG16 and hsa_ circ_0008197 were differentially expressed at the gastric cell and tissue levels, and were positive for PGC, respectively (SNHG16: r = 0.35,P = 0.002; hsa_circ_0008197: r = 0.346, P = 0.005). This suggested the presence of a ceRNA network directly related to PGC. Small nucleolar RNA host gene 16 (SNHG16) was encoded by a 7571-bp region on chromosome 17q25.1 and was considered as a cancer-related lncRNA (Yang et al. 2019). A previous study showed that SNHG16 sponged miRNA by binding to MREs to modulate the expression of tumor-related target genes, thus regulating the proliferation, apoptosis, migration, and invasion of tumor cells (Gong et al. 2020). A dual luciferase reporter experiment and RNA binding protein immunoprecipitation were used to show that SNHG16 sponged hsa-mir-98-5p to regulate the progression of osteosarcoma (Liao et al. 2019). SNHG16 promoted the development of bladder cancer via a miR-98/STAT3/Wnt/β-catenin pathway (Feng et al. 2018). No report exists of SNHG16 sponging hsalet-7c-5p/hsa-let-7f-5p to modulate tumor progression, but several investigations showed that hsa-let-7c-5p and hsa-let-7f-5p also regulated proliferation, migration, and invasion of tumor cells (Chen et al. 2020), (Yan et al. 2015) (Chen et al. 2019). This study initially found and validated the SNHG16-mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p-PGC axis as having an influence on gastric cancer progression. hsa_ circ_0008197 is located on chr1:51,032,749-51,061,888. Although research on the association of hsa_circ_0008197 with human diseases is lacking, our study initially found and verified that hsa_circ_0008197 and SNHG16 co-mediated a PGC-related ceRNA network to transform the SNHG16/hsa_ circ_0008197-mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p-PGC axis. We speculated that this network might cause a change in PGC expression to generate a biological effect via a systemic post-transcriptional regulatory mechanism, thus influencing tumor progression. This yields a novel notion related to PGC pathogenetic mechanisms in gastric cancer.
In order to further examine the co-expression and change trends of PGC and its target regulators in the processes of superficial gastritis → atrophic gastritis → gastric cancer, we investigated the expression profiles of PGC regulators (PPARG, SNHG16 and hsa_circ_0008197) in this disease chain using GSE164166. The clues obtained so far have hinted us that these PGC regulators showed trends of gradual changes as the disease chain progresses. Because the sample numbers of per disease group were too small in the GSE164166 data set, we will further use in vivo experiments for in-depth exploration in the future research.
In conclusion, this study found that the expression of many types of mRNAs, lncRNAs, and circRNAs is related to PGC expression, which, in turn, is involved in cancer-related pathways. DElncRNA related to PGC expression might influence tumor progression by positively regulating the expression of adjacent protein-coding genes. DElncRNAs and DEcircRNAs related to PGC expression co-mediated a complicated ceRNA network to modulate its expression, of which the closest network is SNHG16/hsa_circ_0008197mir-98-5p/hsa-let-7f-5p/hsa-let-7c-5p-PGC. This study contributes to a foundation for the subsequent exploration of possible regulatory mechanisms related to PGC in gastric cancer. Author contribution Yuan Yuan and Qian Xu conceived and designed this study. Li-rong Yan, Han-xi Ding, Shi-xuan Shen, and Xiao-dong Lu were responsible for the data analysis and performed data interpretation. Li-rong Yan and Han-xi Ding wrote the paper. Qian Xu and Yuan Yuan revised the manuscript.

Funding
The authors received financial support from the National Natural Science Foundation of China (No.81772987).

Data availability
The data that support the results of this manuscript are available from the corresponding author upon reasonable request.
Code availability Not applicable.

Declarations
Ethics approval The study was supported by the research ethics committee of the First Affiliated Hospital of China Medical University ([2021]94).

Consent to participate
The written informed consents in this study were signed by all patients.
Consent for publication All co-authors have agreed to publish the paper.

Competing interests
The authors declare no competing interests.