Gastric cancer (GC) is the third major cause of cancer-related death and the fourth most common cancer worldwide; thus, GC is a key public health concern (Bray et al., 2018). Although early diagnosis and several different treatment strategies have improved the survival rate of patients with GC, the prognosis of patients with advanced GC remains unsatisfactory. The 5-year survival rate of advanced GC is approximately 20%, which is largely attributable to uncertain histopathological behaviors and metastatic characteristics in the early stages (Coburn et al., 2018). During GC development, tumor-induced immune suppression leads to imbalanced immune activity (Wang et al., 2017; Lazãr et al., 2018). GC immunotherapy, including immune-checkpoint therapy and cancer vaccines (which depend on the pertinent immunological mechanisms that moderate cancer development and occurrence), has improved in recent years (Martin et al., 2020). Nevertheless, immune heterogeneity, i.e., different immune cell banks and immune profiles, is associated with different responses to immunotherapy (Li J. et al., 2018; Ho and Tan, 2019). Recent progress brought about by omics research has enhanced our perception of the fundamental molecular mechanisms of GC and revealed a wide range of molecular heterogeneity between individuals, which emphasizes the importance of applying molecular biomarkers for clinical diagnosis and prognosis.

Differences in the survival rates of patients with GC are often caused by different genetic or molecular backgrounds. Gene-expression regulators play important roles in regulating immune responses and can mediate posttranscriptional gene modifications (Bradner et al., 2017). Therefore, research on tumor immunogenomics can broaden our understanding of precision medicine and tumor immune mechanisms.

Recently, a study focused on a newly discovered type of noncoding RNAs (ncRNAs), termed microRNAs (miRNAs), whose transcripts are approximately 20 nucleotides in length, has gained significant attention (Bavelloni et al., 2017). During the posttranscriptional regulation of gene expression, miRNAs play crucial functional roles by binding to 3′-untranslated regions of target messenger RNAs (mRNAs), thereby affecting their stabilities (Gebert and Macrae, 2019). Bioinformatic analysis of the predicted binding spectra of miRNAs showed that they directly influence the expression of most protein-coding genes (PCGs) in mammals (O’brien et al., 2018). Interestingly, miRNA genes are often located near cancer-related chromosomal regions and fragile locations in the genome (Georgakilas et al., 2014). Dysregulated biogenesis and expression of miRNAs are involved in tumor progression and initiation, therapy resistance, and metastasis formation, suggesting the existence of both tumor-suppressive and oncogenic miRNAs (Rupaimoole and Slack, 2017). Previous data have demonstrated that miRNAs may also exert cancer-causing or tumor-suppressing functions by affecting tumor immunogenicity or antitumor immune responses (Lin and Gregory, 2015; Hata and Kashima, 2016; Cortez et al., 2019; Verduci et al., 2019). Immune-related miRNAs have been associated with the occurrence, prognosis, and development of GC (Wu et al., 2014; Zhang et al., 2019b). Recent results have shown that some miRNAs belonging to the miR-148 family (including miR-152, miR-148B, and miR-148A) can inhibit histocompatibility antigen, class I, G (HLA-G) expression, as well as the occurrence and development of GC (Zhai et al., 2014). Although the mechanisms of miRNAs in interferon-c-mediated HLA-G upregulation are unclear, an association has been documented between transforming growth factor-β-induced HLA-G expression and decreased miR-152 expression in GC (Guan et al., 2015). miRNAs such as miR-20a, miR-10b, miR-25, miR-93, miR-17-5p, miR-433, and miR-106b can promote tumor growth; regulate the expression of genes encoding the major histocompatibility complex class I chain-related proteins A and B; increase tumor angiogenesis, cell proliferation, and invasion; and negatively interfere with natural killer cell-mediated cytotoxicity (Jasinski-Bergner et al., 2014; Xie et al., 2014).

In this study, we analyzed the expression profiles of immune-related genes and miRNAs in 389 patients with GC using data from The Cancer Genome Atlas (TCGA) project to systematically identify immune-related miRNAs and investigate their prognostic value. Using machine-learning methods, we identified and validated an immune-related nine-miRNA signature (ImmiRSig) that can help determine the prognosis of patients with GC in different cohorts.

Although aggressive and multimodal treatments (e.g., surgery, oxaliplatin-based chemotherapy, and immunotherapy) have improved the survival of patients with GC, the outcomes of patients with advanced GC remain unacceptable. At an advanced stage, patients with similar clinical risk factors for treatment have significantly different tumor responses and a different prognosis. Heterogeneous diseases are characterized by various pathological forms and molecularly heterogeneous changes, which (in addition to existing clinical risk factors) lead to an urgent need for determining other molecular prognostic indicators. GC is also a heterogeneous disease. Several molecular markers, such as mRNA or long ncRNAs, have been proposed over the past decade to support the prognosis of patients with GC (Xia et al., 2014; Sun et al., 2016; Cheng et al., 2018; Wang et al., 2020).

Some GC-specific biomarkers, including blood carcinoembryonic antigen CA 72-4 and carbohydrate antigen 19-9, have been studied through noninvasive testing. Newly identified prognostic biomarkers have primarily been used for the molecular analysis of tumors, including human epidermal growth factor receptor 2 (HER2) expression; the statuses of microsatellite instability/deficient mismatch repair (MMR) expression, programmed death-ligand 1 (PD-L1) expression, and Epstein–Barr virus infection; and detection of the neurotrophic tropomyosin receptor kinase fusion gene (Gravalos and Jimeno, 2008; Duffy et al., 2014; Shinozaki-Ushiku et al., 2015; Zhang et al., 2016; Cocco et al., 2018). The National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology for GC have incorporated genetic biomarkers for the complex clinical management of patients with GC (NCCN, 2018), and the statuses of MMR protein expression, HER2 immunohistochemical staining, and PD-L1 expression have emerged as biomarkers. Based on the clinical-translation potential of miRNAs in controlling the expression of immune checkpoints, a previous study was conducted that revealed a correlation between the plasma levels of PD-1 and PD-L1, and the expression levels of immune-related miRNAs (Incorvaia et al., 2020). Some biomarkers are still being studied, including gene-expression profiles in tumorous tissues and circulating tumor cells in the peripheral blood, which have shown significant associations with clinical endpoints and were included in a randomized clinical study (Li T.t. et al., 2018).

Notably, validated biomarkers specific for GC are still limited and insufficient for supporting diagnostic and therapeutic requirements. Recent studies have shown that disturbances in miRNA-expression levels in various types of cancer have been widely observed, indicating the primary roles of miRNAs in cancer biology (Farazi et al., 2011; Shin et al., 2015).

In a previous study of a feature-selection procedure and classification model, five miRNAs (i.e., miR-106b, miR-135b, miR-141, miR145, and miR-20a) were selected as optimal biomarkers for diagnosing GC (Zhang et al., 2019a). Another study conducted using Cox regression and a robust likelihood-based survival model explored the prognostic signature of GC, and a three-miRNA signature (i.e., miR-145-3p, miR-125b-5p, and miR99a-5p) was established to help determine the prognosis of patients with GC (Zhang et al., 2018).

Our present findings highlighted immune-related miRNAs. Forward- and backward-variable selection and multivariate Cox regression analyses were performed to screen for survival-associated immune-related miRNAs. A survival-associated signature comprising the ImmiRSig, which included miR-145-3p, miR-133a-5p, miR-328-3p, miR-125b-5p, miR-942-3p, miR-1292-5p, miR-99a-3p, miR-675-3p, and miR-92b-5p, was constructed for patients with GC. A formula was established to obtain the risk score for evaluating the survival risks of patients.

Furthermore, among nine immune-related miRNAs, miR-328-3p, miR-125b-5p, miR-942-3p, miR-1292-5p, miR-675-3p, and miR-92b-5p were considered independent prognostic indicators of survival. These immune-related miRNAs may have distinct effects on the occurrence, progression, and metastasis of GC.

Among these six miRNAs, miR-942-3p showed the most connections with the mRNAs included in the regulatory network. We also explored the roles and functions of miR-942-3p, which was previously found to be upregulated in GC tissues (Chen et al., 2020), which is consistent with our current results. High levels of miR-942-3p expression increased the ImmiRSig score in patients with GC, suggesting that it may be an oncogenic factor in hepatocellular carcinoma (HCC) cells by targeting mannose-binding lectin 2 (Xu et al., 2019). MiR-942-3p promotes breast cancer progression by downregulating the expression of forkhead box protein A2 (Zhang J. et al., 2019). A new positive-feedback loop was identified between miR-942-3p and a transcriptional coactivator with a PDZ-binding motif, which regulated the biological function of bladder cancer cells and illustrated that these targets might have therapeutic potential as targets for prostate cancer treatment (Wang et al., 2021). The results of that study also suggest that miR-942-3p regulates the development of a variety of tumors. A few reports have discussed the association between miR-675-3p and cancer, and the literature shows that it is the target of several cancer-related mRNAs and interacts with multiple PCGs. These genes may participate in the competitive-endogenous RNA network in lung adenocarcinoma (Fan et al., 2020). Currently, the role of miR-92b-5p in GC remains undetermined; however, miR-92b-5p has been found to promote the proliferation and invasion of cervical cancer cells both in vitro and in vivo. MiR-92b-5p affects angiogenesis and the radiosensitivity of cervical cancer cells by targeting suppressor of cytokine signaling proteins (Wei et al., 2019). Platinum-based chemical resistance in metastatic cancer cells is associated with MiR-92b/glutathione S-transferase M3 (Ma et al., 2019). MiR-1292-5p showed a negative correlation with the risk-score formula. In a previous study, the miR-1292-5p–DEK axis was enhanced by the circular RNA circ0000039, thereby promoting GC progression (Gayosso-Gómez et al., 2014). These results differ from our current results and warrant further verification in a clinical trial. MiR-328-3p is also frequently described as a miRNA with a tumor-suppressor role in cancer. miR-328-3p can inhibit osteosarcoma cell proliferation and migration but accelerates programmed death by inhibiting matrix metalloproteinase-16 (Shi et al., 2019). MiR-328-3p acts as a tumor suppressor in HCC by inhibiting thioredoxin reductase 1 (TXNRD1), suggesting that it is a potential target for the clinical treatment of HCC (Li et al., 2019). In addition, miR-328-3p mimics are useful for treating colorectal cancer and can effectively predict tumor recurrence and treatment outcomes in patients with colorectal cancer (Wang et al., 2016; Ji et al., 2018). MiR-125b-5p is a commonly discussed immune-related miRNA that plays anticancer roles in several types of cancer. MiR-125b-5p was included in the three-miRNA-based risk score evaluation model for GC prognosis (Zhang et al., 2018). Li Y. et al. (2018) showed that by targeting KIAA1522, miR-125b-5p can inhibit tumor progression and regulate breast cancer progression. The miR-125b-5p–hexokinase 2 pathway has been reported to inhibit the progression of squamous cell carcinoma by regulating laryngeal squamous cell carcinoma glycolysis (Hui et al., 2018). Additionally, by inhibiting TXNRD1, miR-125b-5p was found to act as a tumor suppressor in HCC, making it a potential target in the clinical treatment of HCC (Hua et al., 2019).

The prognostic efficacy of the ImmiRSig was well validated in this study. Survival analysis was performed for both the training and test groups. Kaplan–Meier-survival curves indicated that the survival probability of patients in the high- and low-risk groups could be predicted with statistical significance.

The results of our GO and KEGG analyses indicated that cancer-related pathways, including the MAPK, Rap1, RAS, and other signaling pathways, were the most enriched pathways. The MAPK pathway leads to malignant phenotypes such as autonomic cell proliferation, which is often activated in human cancers (Sumimoto et al., 2006). Rap1 serves several roles in the invasion and metastasis of different cancers by interacting with other proteins (Zhang et al., 2017). Additionally, abnormal activation of the RAS protein is the main oncogenic factor that controls the function of the main signaling pathways involved in the occurrence and development of human malignant tumors (Zenonos and Kyprianou, 2013; Khan et al., 2019).

Intravenous chemotherapy is a very common method for treating patients with GC. Previous data showed that the correct administration time, dose intensity, and drug-effect ratio of paclitaxel combined with other chemotherapeutic drugs can safely be used as first-line treatment for patients with advanced GC, with high activity and a curative effect (Bruera et al., 2018). Our current findings confirmed that low-risk patients were more sensitive to paclitaxel and sorafenib than high-risk patients. Because of high immunosuppression and low immunoreactivity in the tumor microenvironment, these differences promote tumor growth, progression, invasion, and metastasis, which can contribute to a poor prognosis in high-risk patients. These differences suggest that high-risk patients may have lower benefits from immunotherapy and chemotherapy than low-risk patients.

The immune cell-infiltration levels were estimated using various software programs, and the results were compared across 389 patients. We identified significant differences in the infiltration of CD4⁺ T cells, macrophages, and myeloid dendritic cells in the high- and low-risk groups, suggesting a potential cellular mechanism. The limitation of this study is that, although the functional implication of the ImmiRSig was inferred by in silico functional analysis, careful functional characterization of the miRNAs should be performed in biological experiments, such as with RNA interference or in vitro cell-based assays. A well-designed and controlled clinical study should be conducted to verify the prognostic efficacy of the ImmiRSig.

In conclusion, our results established the ImmiRSig for predicting the OS of patients with GC. MiRNAs were identified sequentially using univariate Cox analysis, multivariate-regression analysis, and forward- and backward-variable selection. The ImmiRSig score can be calculated using the expression levels of immune-related miRNAs and their respective coefficients. Furthermore, the identified immune-related miRNAs may be promising candidates as therapeutic targets.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.

The studies involving human participants were reviewed and approved by The Second Affiliated Hospital, Wenzhou Medical University, Wenzhou, China. The patients/participants provided their written informed consent to participate in this study.

JX and JW contributed to the study concepts and study design. JX and SL contributed to data acquisition and reconstructions. JX, XS, and TY contributed to the data analyses and interpretation. YH contributed to the statistical analysis. CX and YZ contributed to the manuscript preparation and manuscript editing and reviewing. All authors read and approved the final manuscript, contributed to the article, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.