MiR-942-3p as a Potential Prognostic Marker of Gastric Cancer Associated with AR and MAPK/ERK Signaling Pathway

Gastric cancer is a common tumor with high morbidity and mortality. MicroRNA (miRNA) can regulate gene expression at the translation level and various tumorigenesis processes, playing an important role in tumor occurrence and prognosis. This study aims to screen miRNA associated with gastric cancer prognosis as biomarkers and explore the regulatory genes and related signaling pathways. In this work, R language was used for the standardization and differential analysis of miRNA and mRNA expression profiles. Samples were randomly divided into a testing group and a training group. Subsequently, we built the five miRNAs (has-miR-9-3p, has-miR-135b-3p, has-miR-143-5p, has-miR-942-3p, has-miR-196-3p) prognostic modules, verified and evaluated their prediction ability by the Cox regression analysis. They can be used as an independent factor in the prognosis of gastric cancer. By predicting and analyzing potential biological functions of the miRNA target genes, this study found that the AR gene was not only a hub gene in the PPI network, but also associated with excessive survival of patients. In conclusion, this study demonstrated that hsa-miR-942-3p could be a potential prognostic marker of gastric cancer associated with the AR and MAPK/ERK signaling pathways. The results of this study provide insights into the occurrence and development of gastric cancer.


Introduction
Gastric cancer is one of the most common tumors and its overall survival rate is only about 10% [1]. Some treatments are developing rapidly, including surgery, radiotherapy, chemotherapy, and targeted therapy. However, the recurrence rate and poor prognosis remain a troubling issue. At present, some biomarkers related to the occurrence and prognosis of gastric cancer have been found [2] but their reliability has not been completely verified. Therefore, it is essential to screen new biomarkers or therapeutic targets for the prognosis of gastric cancer patients.
MicroRNA (miRNA) is a non-coding molecule, which can regulate gene expression at the translation level. Some studies have shown that miRNAs regulate various tumorigenesis processes (cell proliferation, cell differentiation, and cell apoptosis) by combining tumor suppressor genes or oncogenes. Yang L et al. found that miR-9-3p was a down-regulated gene of glioma cells. Its low expression resulted in increased levels of Herpud1 that could protect glioma cells from apoptosis [3]. Chen Z et al. showed that miR-143-5p could promote cadmium-induced apoptosis of LLC-PK1 cells by acting on the target gene AKT3 and inhibiting the Akt/Bad signaling pathway [4]. Ma R et al. verified that up-regulated miR-196b could induce a proliferative phenotype, leading to a poor prognosis in glioblastoma patients [5]. Chen M et al. showed that miR-135b could play the role of oncogenes by regulating the PI3K/Akt, HIF-1/FIH, Hippo, p53 signaling pathways, promote tumor

Detection of Differentially Expressed miRNAs and mRNA Combined with Clinical Information
Standardization and differential analysis of expression profiles were performed using R language (p < 0.05 and |logFC| > 1.0) [11]. Thereafter, clinical information on patients was combined with the disposed of miRNAs and mRNAs.

Construction of Sample Grouping and Prognostic Module
Samples were divided into training group and testing group randomly by R language package. Univariate Cox regression analysis was used to detect the miRNAs with p < 0.05 in the training group. Multivariate Cox regression was used to build the miRNA module prognostic biomarkers with different overall survival [12]. Then, we established the risk score of a prognostic miRNA signature and detected the Proportional Hazards Assumption of the Cox module. The module was used to assess the survival prognosis of patients in three groups by the Kaplan-Meier curve. Log-rank tests were classified into a high-risk and low-risk group according to the risk score of the median value grouping. R language ("survivalROC" package) was used to evaluate miRNA predictive power by receiver operating characteristic (ROC) curve [13].

Independent Prognostic Ability of miRNA
The univariate Cox regression was analyzed to test the relationship between the prognostic miRNA and the overall survival of patients in the training group. Clinical factors were also analyzed by multivariate Cox regression to serve as independent prognostic elements.

miRNA Target Genes Prediction and Functions Analysis
The miRNA information was downloaded from three prediction databases (targetScan, miRTarBase, and miRDB). The target genes of miRNA were obtained and crosschecked in at least two databases. Using the Cytoscape and Venn software to draw the relation between miRNAs and the target genes. Differentially expressed genes (DEGs) and target genes were taken at the intersection to test whether these target genes were involved in the progression of gastric cancer. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment pathway and Gene Ontology (GO) analysis displayed the potential function of all the intersection genes through R language ("org.Hs.eg.db" package and "clusterProfiler" package) [14].

Screening Core Target Genes and Survival Analysis
The protein-protein interaction (PPI) network between the target genes was obtained from STRING websites [15] while the medium confidence is 0.400. Then, the top ten hub genes were detected through Cytoscape plug-in CytoHubba. In addition, Kaplan-Meier curves were used to detect whether the intersection genes showed a relationship with overall survival.

Screening Core Target Genes and Survival Analysis
The protein-protein interaction (PPI) network between the target genes was obtained from STRING websites [15] while the medium confidence is 0.400. Then, the top ten hub genes were detected through Cytoscape plug-in CytoHubba. In addition, Kaplan-Meier curves were used to detect whether the intersection genes showed a relationship with overall survival.

Screening Core Target Genes and Survival Analysis
The protein-protein interaction (PPI) network between the target genes was obtained from STRING websites [15] while the medium confidence is 0.400. Then, the top ten hub genes were detected through Cytoscape plug-in CytoHubba. In addition, Kaplan-Meier curves were used to detect whether the intersection genes showed a relationship with overall survival.

Prediction and Assessment of Five miRNAs for Overall Survival in Three Groups
According to the median value grouping of risk score, the Kaplan--Meier curve displayed that the high-risk group had worse survival than the low-risk group in the training group (p = 1.417 × 10 −4 ), the testing group (p = 2.131 × 10 −2 ), and the whole group (p = 1.436 × 10 −5 ; Figure 4a-c). The area under curve (AUC) of ROC for the five miRNAs severally Figure 3. Five miRNAs associated with overall survival. (a) has-miR-135b-3p; (b) has (c) has-miR-9-3p; (d) has-miR-143-5p; and (e) has-miR-196b-3p.

Prediction and Assessment of Five miRNAs for Overall Survival in Three Groups
According to the median value grouping of risk score, the Kaplan-Meier curve displayed that the high-risk group had worse survival than the low-risk group in the training group (p = 1.417 × 10 −4 ), the testing group (p = 2.131 × 10 −2 ), and the whole group (p = 1.436 × 10 −5 ; Figure 4a-c). The area under curve (AUC) of ROC for the five miRNAs severally attained 0.719, 0.660, and 0.689 in the training group, the testing group, and the whole group (Figure 4d-f), which indicated that the five miRNAs perform well in predicting the overall survival of gastric cancer patients. Furthermore, patients with high-risk scores had a higher death rate than those with low-risk scores in the three groups (Figure 4g-i).

Independence of the Five miRNAs
Based on the univariate and multivariate Cox regression analysis, the five miRNAs were related to the overall survival of patients (HR = 1.726, 95% CI = 1.396−2.136, p < 0.001). They were also independent in overall survival considering other clinical elements (HR = 1.971, 95% CI = 1.557−2.494, p < 0.001). Other clinical features include age, gender, stage, T stage, metastasis, and lymph node stage (Table 4).

Independence of the Five miRNAs
Based on the univariate and multivariate Cox regression analysis, the five miRNAs were related to the overall survival of patients (HR = 1.726, 95% CI = 1.396−2.136, p < 0.001). They were also independent in overall survival considering other clinical elements (HR = 1.971, 95% CI = 1.557−2.494, p < 0.001). Other clinical features include age, gender, stage, T stage, metastasis, and lymph node stage (Table 4).

Independence of the Five miRNAs
Based on the univariate and multivariate Cox regression analysis, the five miR-NAs were related to the overall survival of patients (HR = 1.726, 95% CI = 1.396-2.136, p < 0.001). They were also independent in overall survival considering other clinical elements (HR = 1.971, 95% CI = 1.557-2.494, p < 0.001). Other clinical features include age, gender, stage, T stage, metastasis, and lymph node stage (Table 4).

Target Genes Functional Enrichment Analysis
The GO results in the top fifteen terms, including biological process (BP), cellular component (CC), and molecular function (MF) were displayed in dot plot (Figure 7a-c). BP mainly contained cell cycle G1/S phase transition, urogenital and renal system development; CC mainly contained transmembrane transporter complex, transporter complex, and apical part of cell; MF mainly contained ion channel and substrate-specific channel activity. KEGG pathways analysis results were mainly enriched in the neuroactive ligandreceptor interaction, cAMP signaling pathway, and the MAPK signaling pathway ( Figure  7d and Table 5).

Target Genes Functional Enrichment Analysis
The GO results in the top fifteen terms, including biological process (BP), cellular component (CC), and molecular function (MF) were displayed in dot plot (Figure 7a-c). BP mainly contained cell cycle G1/S phase transition, urogenital and renal system development; CC mainly contained transmembrane transporter complex, transporter complex, and apical part of cell; MF mainly contained ion channel and substrate-specific channel activity. KEGG pathways analysis results were mainly enriched in the neuroactive ligand-receptor interaction, cAMP signaling pathway, and the MAPK signaling pathway (Figure 7d and Table 5).

The Working Mechanism of AR and its Potential Relationship with the MAPK/ERK Signaling Pathway
From the above, we can conclude that the Androgen Receptor (AR) was not only a hub gene in the PPI network but also associated with excessive survival of patients. AR can regulate the transcription of genes and express new proteins, ultimately changing the function of cells. Figure 10 shows a typical AR working mechanism. AR usually forms a complex with heat shock proteins (HSPs) in the cytoplasm. The binding of AR to androgen (such as 5α-dihydrotestosterone, DHT) alters its conformation, and HSPs are subsequently released. Under the action of coactivators, androgen-AR complexes are

The Working Mechanism of AR and Its Potential Relationship with the MAPK/ERK Signaling Pathway
From the above, we can conclude that the Androgen Receptor (AR) was not only a hub gene in the PPI network but also associated with excessive survival of patients. AR can regulate the transcription of genes and express new proteins, ultimately changing the function of cells. Figure 10 shows a typical AR working mechanism. AR usually forms a complex with heat shock proteins (HSPs) in the cytoplasm. The binding of AR to androgen (such as 5α-dihydrotestosterone, DHT) alters its conformation, and HSPs are subsequently released. Under the action of coactivators, androgen-AR complexes are transferred to the nucleus and recognize androgen response elements in the form of homodimer to regulate downstream target gene expression.

The Working Mechanism of AR and Its Potential Relationship with the MAPK/ERK Signaling Pathway
From the above, we can conclude that the Androgen Receptor (AR) was not on hub gene in the PPI network but also associated with excessive survival of patients. can regulate the transcription of genes and express new proteins, ultimately changing function of cells. Figure 10 shows a typical AR working mechanism. AR usually form complex with heat shock proteins (HSPs) in the cytoplasm. The binding of AR to andro (such as 5α-dihydrotestosterone, DHT) alters its conformation, and HSPs are sub quently released. Under the action of coactivators, androgen-AR complexes are tra ferred to the nucleus and recognize androgen response elements in the form of homo mer to regulate downstream target gene expression. In the absence of androgen, AR may depend on the MAPK/ERK signaling pathw to play its role. Figure 11 shows the potential relationship between the AR a MAPK/ERK signaling pathways. In the cytoplasm, AR can interact with several signal molecules, including phosphoinositide 3-kinase (PI3K), Src family kinase (Src), GTPase (Ras), and protein kinase C (PKC), which in turn converge on the MAPK/E pathway. Then, the MAPK/ERK enters the nucleus, where it translocates and intera with transcription factors that regulate the expression of genes associated with cell pro eration. In the absence of androgen, AR may depend on the MAPK/ERK signaling pathway to play its role. Figure 11 shows the potential relationship between the AR and MAPK/ERK signaling pathways. In the cytoplasm, AR can interact with several signaling molecules, including phosphoinositide 3-kinase (PI3K), Src family kinase (Src), Ras GTPase (Ras), and protein kinase C (PKC), which in turn converge on the MAPK/ERK pathway. Then, the MAPK/ERK enters the nucleus, where it translocates and interacts with transcription factors that regulate the expression of genes associated with cell proliferation.

Discussion
Gastric cancer is one of the most common tumors with high morbidity and mortality. Therefore, the detection of sensitive specific biomarkers for gastric cancer is urgent. Many studies indicated miRNAs could regulate expression in vivo, and it plays an essential role in the biological process of human malignancy [16]. Currently, some miRNAs have been used as potential prognostic indicators for tumors, such as miR-191 [17], miR-1908 [18], miR-217 [19], and miR-200c [20]. Previously, a variety of miRNAs were discovered in many prognostic markers for tumors [21,22], especially for gastric cancer [23].
In this study, we obtained 267 DEmiRNAs. All samples were divided into training group and testing group randomly. Then, the five miRNAs were constructed in the training group. At the same time, based on the median grouping of risk score, these five miR-NAs were proved in the testing group and the whole group, respectively. Kaplan-Meier curves showed that overall survival was significantly lower in the high-risk group than in the low-risk group among the three groups. By ROC curve, the overall survival of the five miRNAs among the three groups showed better predictive ability. Subsequently, the Cox regression analysis indicated that the five miRNAs were independent of overall survival.
The target genes of five miRNAs were predicted in order to in-depth understand the regulatory mechanisms of these five miRNAs. GO analysis showed that the target genes were correlated with cell cycle G1/S phase transition, urogenital and renal system development, transmembrane transporter complex, transporter complex and apical part of the cell, ion channel, substrate-specific channel, and channel activity. The signaling pathways were enriched in the cAMP and MAPK signaling pathways and the Neuroactive ligandreceptor interaction. Park et al. pointed out that the cAMP signaling pathway inhibited the degradation of the HDAC8 and the expression of TIPRL in lung cancer cells, and also increased cisplatin-induced apoptosis [24]. Jagriti Pal et al. showed that the neuroactive ligand-receptor interaction pathway had a poor prognosis in patients with glioma [25]. The MAPK/ERK signaling pathway was essential in regulating cellular processes, such as cell differentiation, division, proliferation, and apoptosis.
The top ten hub genes (CCNA2, GRIA2, FOS, AR, RACGAP1, RBFOX1, LIN28A, DSCC1, GRID2, OPRK1) of target genes were detected by Cytoscape. Moreover, the Kaplan-Meier curve showed that eight target genes (AKAP12, AR, DEIP1, PCDHA11, PCDHA12, P115, SH3BGRL, TMEM108) were related to survival prognosis. Unexpectedly, AR was a hub gene in the PPI network, and it had a relationship with the excessive survival of patients. AR is a nuclear transcription factor, it can recognize and combine specific DNA sequences on target factors, thereby regulating the transcription of the gene and expressing new proteins, which ultimately changes the function of cells and promotes cell differentiation and the development of tissues and organs [26][27][28]. Salma S et al.

Discussion
Gastric cancer is one of the most common tumors with high morbidity and mortality. Therefore, the detection of sensitive specific biomarkers for gastric cancer is urgent. Many studies indicated miRNAs could regulate expression in vivo, and it plays an essential role in the biological process of human malignancy [16]. Currently, some miRNAs have been used as potential prognostic indicators for tumors, such as miR-191 [17], miR-1908 [18], miR-217 [19], and miR-200c [20]. Previously, a variety of miRNAs were discovered in many prognostic markers for tumors [21,22], especially for gastric cancer [23].
In this study, we obtained 267 DEmiRNAs. All samples were divided into training group and testing group randomly. Then, the five miRNAs were constructed in the training group. At the same time, based on the median grouping of risk score, these five miRNAs were proved in the testing group and the whole group, respectively. Kaplan-Meier curves showed that overall survival was significantly lower in the high-risk group than in the low-risk group among the three groups. By ROC curve, the overall survival of the five miRNAs among the three groups showed better predictive ability. Subsequently, the Cox regression analysis indicated that the five miRNAs were independent of overall survival.
The target genes of five miRNAs were predicted in order to in-depth understand the regulatory mechanisms of these five miRNAs. GO analysis showed that the target genes were correlated with cell cycle G1/S phase transition, urogenital and renal system development, transmembrane transporter complex, transporter complex and apical part of the cell, ion channel, substrate-specific channel, and channel activity. The signaling pathways were enriched in the cAMP and MAPK signaling pathways and the Neuroactive ligand-receptor interaction. Park et al. pointed out that the cAMP signaling pathway inhibited the degradation of the HDAC8 and the expression of TIPRL in lung cancer cells, and also increased cisplatin-induced apoptosis [24]. Jagriti Pal et al. showed that the neuroactive ligand-receptor interaction pathway had a poor prognosis in patients with glioma [25]. The MAPK/ERK signaling pathway was essential in regulating cellular processes, such as cell differentiation, division, proliferation, and apoptosis.
The top ten hub genes (CCNA2, GRIA2, FOS, AR, RACGAP1, RBFOX1, LIN28A, DSCC1, GRID2, OPRK1) of target genes were detected by Cytoscape. Moreover, the Kaplan-Meier curve showed that eight target genes (AKAP12, AR, DEIP1, PCDHA11, PCDHA12, P115, SH3BGRL, TMEM108) were related to survival prognosis. Unexpectedly, AR was a hub gene in the PPI network, and it had a relationship with the excessive survival of patients. AR is a nuclear transcription factor, it can recognize and combine specific DNA sequences on target factors, thereby regulating the transcription of the gene and expressing new proteins, which ultimately changes the function of cells and promotes cell differentiation and the development of tissues and organs [26][27][28]. Salma S et al. showed that the p14ARF tumor suppressor could restrain AR activity and prevent apoptosis in prostate cancer cells [29]. Peng L et al. verified that AR could be directly combined with LAMA4, and it was related to enhanced cisplatin resistance in gastric cancer, providing a new mechanism for the treatment of drug-resistant gastric cancer [30]. In addition, AR may depend on the MAPK/ERK signaling pathway to function. Specifically, AR can interact with a variety of signaling molecules (PI3K, Src, Ras, and PKC) in the cytoplasm, which in turn converge on the MAPK/ERK pathway [31,32]. MAPK/ERK then enters the nucleus, where it translocates and interacts with transcription factors to regulate the expression of genes involved in cell proliferation [33].
In a word, this study found that the MAPK/ERK signaling pathway may help AR signal transduction and promote the interaction between AR and transcription factors, leading to cell proliferation. At the same time, AR is a target gene of the has-miR-942-3p, which well verifies the important role of the has-miR-942-3p in the occurrence and prognosis of gastric cancer.

Conclusions
This study built the five miRNAs (has-miR-9-3p, has-miR-135b-3p, has-miR-143-5p, has-miR-942-3p, has-miR-196-3p) prognostic modules, also verified and evaluated the prediction ability of the five miRNAs by grouping. They can be used as an independent factor in the prognosis of gastric cancer. By predicting the target genes to explore the potential biological functions, our results could provide a deeper understanding of the occurrence and development. This study identified the AR gene regulated by has-miR-942-3p which may depend on the MAPK/ERK signaling pathway to promote the proliferation of cancer cells. In future experiments, we will further explore the regulatory mechanisms of other miRNAs (has-miR-9-3p, has-miR-135b-3p, has-miR-143-5p, has-miR-196-3p) to provide effective prediction and treatment targets for gastric cancer patients.