Neurotransmitter release cycle-related genes predict the prognosis of lung adenocarcinoma

Abstract

Because of the limitations of therapeutic approaches, patients suffering from lung adenocarcinoma (LUAD) have unsatisfactory prognoses. Studies have shown that neurotransmitters participated in tumorigenesis and development. In LUAD, the expression of neurotransmitter release cycle-related genes (NRCRGs) has been reported to be disordered. This study aimed to study the correlation between NRCRGs and LUAD. In this study, based on the Cancer Genome Atlas cohort, consensus clustering analyses were performed on ten neurotransmitter release cycle-related (NRCR) differentially expressed genes. Neurotransmitter release cycle (NRC) scores were derived by the Least Absolute Shrinkage and Selection Operator-Cox regression model constituted by 3 NRCRGs. Univariate and multivariate Cox regression analyses were performed to evaluate the prognosis value of the NRC score. In addition, single-Sample Gene Set Enrichment Analysis and CIBERSORT were conducted in the Cancer Genome Atlas cohort. Finally, gene ontology and Kyoto Encyclopedia of Genes and Genomes analyses were also performed. As a result, the NRC-low group showed a good prognosis instead of the NRC-high group. NRC score was identified to be an independent prognosis factor for LUAD. In general, the NRC score based on the prognostic model was found to be closely correlated with immunotherapy-related anti-cancer immunity and inflamed tumor microenvironment. Functional enrichment results demonstrated that differentially expressed genes between 2 NRC groups were closely correlated with DNA replication, cell-substrate adhesion, Golgi vesicle transport, MAPK signal pathway, and many others. Novel biomarkers were offered for predicting the prognoses of LUAD patients. The NRC score might contribute to guiding LUAD patients with immunotherapy selection.

1. Introduction

According to a global report in 2020, new cases of lung cancer accounted for 11.4% of all cancer new cases. Lung cancer was the main cause of cancer deaths, accounting for 18% of all cancer deaths.^[1] Lung adenocarcinoma (LUAD) is the most popular type of non-small cell lung cancer (NSCLC) which accounts for 85% of lung cancer.^[2] Although a variety of therapies have been used, patients suffering from LUAD still had unsatisfactory prognoses. In consequence, it is pressingly needed to find valid biomarkers as therapeutic targets to improve the prognoses of patients with LUAD.

Neurotransmitters include amino acids (acetylcholine, glutamate, glycine, and γ-aminobutyric acid), biogenic amines (dopamine, norepinephrine, epinephrine, and serotonin), and neuropeptide (substance P, neuropeptide Y, vasoactive, and many others).^[3] Previous research has suggested that neurotransmitters in the tumor microenvironment (TME) can initiate signaling pathways by binding to the corresponding neurotransmitter receptors (NTRs) and afterward stimulate the growth and spread of cancer cells.^[4] Furthermore, neurotransmitters also can directly irritate the endothelial cells, immune cells, and stromal cells which exist in TME and then accelerate tumor progression.^[5] Researches have revealed that the malignant growth of ovarian cancer,^[6] pancreatic cancer,^[7] and lung cancer^[8] may ascribe to the activation of β-adrenergic receptors. Magnon et al^[9] demonstrated that the nervous system is correlated with cancer progression for the first time. Their study suggested that catecholamines and acetylcholine secretion of sympathetic and parasympathetic have effects on the growth and metastasis of prostate tumors, respectively. Additionally, Akt and MAPK signal pathways can be activated by muscarinic receptors (M3R) and promote cell proliferation in NSCLC.^[10]

Growing evidence demonstrated that neurotransmitter release cycle-related genes (NRCRGs) play a vital role in the occurrence and development of many cancers. SLC6A1 encodes gamma amino-butyric acid carrier (GAT-1) and the overexpression of SLC6A1 is correlated with a poor prognosis of prostate cancer.^[11] Moreover, STX1A acts as a neuronal soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) mediating synaptic vesicular fusion.^[12] STX1A is overexpressed not only in primary brain tumors, but also in colon and rectum, breast, and lung cancers.^[13–16] It also has been proved that the inhibition of ChAT which is responsible for synthesizing Ach can suppress the growth of cancer cells of LUAD.^[17] Besides, STXBP1 (Munc18-1) takes part in neurotransmitter release by regulating syntaxin and it has been determined as a biomarker for LUAD prognosis with overexpression in LUAD cases.^[18] Additionally, MAOA which catalyzes the oxidative deamination of dopamine, serotonin, and the other amines is overexpressed in NSCLC,^[19] while another study suggested that MAOA is downregulated in LUAD.^[20] Thus, we speculate that the NRCRGs model might exist prognostic value for patients with LUAD. In this study, we integrated the genomic information of LUAD patients from multiple datasets to develop and validate a 3 NRCRGs prognostic model named “Neurotransmitter release cycle (NRC) score.” Furthermore, the NRC score constructed based on the prognostic model was associated with immunotherapy-related anti-cancer immunity and inflamed TME characteristics.

2. Materials and Methods

2.1. Source and processing of raw data

RNA-seq data of 535 LUAD patients and 59 normal patients were obtained from the Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). The corresponding clinical data were acquired from UCSC Xena (http://xena.ucsc.edu/). Patients with incomplete clinical information were excluded from further analyses. Additionally, we downloaded the gene expression data of 226 LUAD patients as the validation dataset from the Gene Expression Omnibus (GEO) repository (GSE31210) (https://www.ncbi.nlm.nih.gov/geo/). Our study is based on public databases, so ethical approval and patient consent were not necessary.

2.2. Gene expression analysis

Fifty-one NRCRGs (Table S1, Supplemental Digital Content 1, http://links.lww.com/MD/H226) were obtained from the PathCards module in GeneCards (https://www.genecards.org/), a comprehensive resource for gene-related information.^[21] The Reactome website (https://reactome.org/PathwayBrowser/) serves as a free database of human biological pathways. According to the Reactome Knowledgebase, the 51 NRCRGs are involved in 6 neurotransmitter release cycle-related pathways, including gamma amino-butyric acid synthesis, release, reuptake, and degradation, and acetylcholine, norepinephrine, serotonin, dopamine, glutamate neurotransmitter release cycle. The protein–protein interaction network was conducted in STRING (https://www.string-db.org/) with the highest confidence at 0.9. After that, hub genes were filtrated by Cytoscape_v3.8.2. Differentially expressed genes (DEGs) between tumor samples and normal samples were identified by “limma” packages with the absolute values of |logFC| ≥ 1 and false discovery rate (FDR) < 0.01.

2.3. Consensus clustering analysis

For the sake of determining whether the expression levels of neurotransmitter release cycle-related (NRCR) DEGs were correlated with the subtypes of LUAD, the patients from the TCGA cohort with complete survival information were divided into different groups in the light of the expression value of these genes by “ConsensusClusterPlus” R package. The overall survivals (OSs) of different clusters were compared by Kaplan–Meier method with the log-rank test.

2.4. Building and validation of the prognostic model

To evaluate the value of the NRCR DEGs in predicting the survival outcomes, univariate Cox regression analyses were applied to preliminarily screen prognosis-related genes by setting a cut-off P value = 0.2 to prevent omissions. The Least Absolute Shrinkage and Selection Operator (LASSO) model serves as a shrinkage method that can effectively select genes from a large and potentially multicollinear set of genes in the regression, bringing about more relevant and interpretable predictive genes.^[22] Ultimately, by performing LASSO-Cox regression analysis, 3 genes and their coefficients were reserved, based on the minimum value of λ through the R package “glmnet”. The formula with coefficients and gene expressions as main elements was generated by R (NRC Score = ${\displaystyle \sum \text{Xi×Yi}}$, X: coefficients, Y: values of gene expression). Then, LUAD patients from TCGA were divided into NRC-high and NRC-low groups based on the median NRC score. The “survivalROC” R package was applied to execute the receiver operating characteristic curves of 1 to 3 years. According to the formula derived from the TCGA cohort, NRC scores of patients in the GSE31210 cohort were calculated to further validate the results from the TCGA cohort. To validate the independent prognostic value of the NRC score, univariate and multivariate Cox regression analyses were performed in TCGA and GEO cohorts.

2.5. Functional enrichment analysis

The “limma” R package was applied for identifying DEGs in high and low NRC-score groups (|log FC|≥ 1 and FDR < 0.05). Subsequently, gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were carried out to excavate the function of these DEGs by applying “clusterProfiler” R package.

2.6. Association of NRC score with immunotherapy-related anti-cancer immunity and TME

We resorted to several indicators to characterize the immunotherapy-related anti-cancer immunity, including 6 immune checkpoint genes (ICPGs) (CD274, HAVCR2, IDO1, LAG3, PDCD1, PDCD1LG2), IFN-γ signaling score, immunologic constant of rejection (ICR) score, inflammation signature score, antigen processing machinery (APM) score and CD8 T effector score. As displayed in a recent study, TME was characterized by several TME signatures containing Pan_F_TBRs score quantizing TGF-β activity, epithelial to mesenchymal transition (EMT1-3) scores, and stromal scores. Moreover, DNA damage and repair related-signature scores including base excision repair score, DNA damage response score, DNA replication score, mismatch repair score, and nucleotide excision repair score were compared between NRC-low and NRC-high groups. CIBERSORT was employed to quantify 22 tumor-infiltrating immune cells between NRC-low and NRC-high groups. The related signatures score was calculated by the method of single-Sample Gene Set Enrichment Analysis. All signatures used in our study were listed in Table S2, Supplemental Digital Content 2, http://links.lww.com/MD/H227.

2.7. Statistical analysis

Pearson chi-square test was applied for comparing clinical information between different groups. Kaplan–Meier method with the log-rank test was employed for comparing the OSs in diverse groups. Independent influencing factors of prognosis were determined by univariate and multivariable Cox regression analyses. The differences in anti-cancer immunity score and TME features between subgroups were examined by the Mann–Whitney U test. All statistical analyses were carried out with R software (version 3.6.2). The workflow chart is presented in Figure 1.

Figure 1.

Detailed research design ideas. DEGs = differentially expressed genes, FDR = false discovery rate, GEO = Gene Expression Omnibus, GO = gene ontology, KEGG = Kyoto Encyclopedia of Genes and Genomes, LASSO = Least Absolute Shrinkage and Selection Operator, LUAD = lung adenocarcinoma, NRC = neurotransmitter release cycle, PPI = protein-protein interaction, TCGA = The Cancer Genome Atlas.

3. Results

3.1. Differential expression analysis of NRCRGs

The protein–protein interaction network of 51 NRCRGs which were derived from STRING was presented in Figure 2A. Then, the first 20 genes were determined to be hub genes in line with the degree value by Cytoscape_3.8.2 among the 51 NRCRGs (Fig. 2B). The expression level differences of 51 NRCRGs were tested between 535 tumor samples and 59 normal samples from the TCGA cohort. At the cutoff (|logFC| ≥ 1 and FDR < 0.01), ten genes were identified as DEGs including downregulated 7 genes (LIN7A, SLC1A1, MAOA, VAMP2, SLC18A2, UNC13B, SLC6A12) and 3 upregulated genes (STX1A, GAD1, PPFIA4). Figure 2C demonstrates the expression levels of these genes between tumor and normal samples (blue: low; yellow: high). And LIN7A, VAMP2, SLC18A2, UNC13B, and STX1A were also hub genes.

Figure 2.

PPI network of 51 NRCRGs and the expression levels of NRCR DEGs. (A) PPI network demonstrating the interactions of the NRCRGs (interaction score = 0.9). (B) Hub genes filtrated by Cytoscape, the darker the color, the higher the connection degree. (C) Heatmap (presenting the expression levels of NRCR DEGs between tumor samples (Tumor, blue) and normal samples (Normal, red). All P < .01. DEGs = differentially expressed genes, NRCR = neurotransmitter release cycle-related, NRCRGs = neurotransmitter release cycle-related genes, PPI = protein–protein interaction.

3.2. Consensus clustering analysis for LUAD patients based on NRCRGs

The k = 2 was determined as optimal clustering stability based on the similarity shown by the expression levels of NRCR DEGs and on the scale of the fuzzy clustering measure. All patients with LUAD from the TCGA cohort were divided into 2 subtypes (Fig. 3A), in other words, cluster A (n = 272) and cluster B (n = 250). The expression levels of NRCR DEGs between 2 clusters are displayed in Figure 3B. The heatmap shows that the expression levels of GAD1, PPFIA4, and STX1A were relatively lower in cluster B, while the expression levels of MAOA, UNC13B, VAMP2, and SLC18A2 were relatively higher in cluster B. Patients in cluster B appeared greater prognoses than cluster A (log-rank P < .001), the OS curves are shown in Figure 3C. Additionally, the clinical features including status, age, gender, stage, and tumor node metastasis stage were compared between cluster A and cluster B. As shown in the pie chart (Fig. 3D), proportions of age (P = .017), N stage (P = .008), status (P = .0075) showed obvious statistical differences between cluster A and cluster B.

Figure 3.

Tumor patients clustering based on the expression of NRCR DEGs. (A) 522 LUAD patients were divided into 2 subtypes based on the consensus clustering analysis (k = 2). (B) A heatmap showing the expression levels of NRCR DEGs between 2 clusters. (C) Kaplan–Meier curves of OSs in 2 clusters. (D) A pie chart respectively presenting the distribution of clinical features in 2 clusters, including status (Alive or Dead), age (>60 or ≤60), gender (female or male), stage (I–IV), and TNM stage. DEGs = differentially expressed genes, LUAD = lung adenocarcinoma, NRCR = neurotransmitter release cycle-related, NRCRGs = neurotransmitter release cycle-related genes, OSs = overall survivals, PPI = protein–protein interaction, TNM = tumor node metastasis.

3.3. Building and evaluation of prognostic model

Firstly, 513 patients with LUAD were included to execute a univariate cox regression analysis to screen genes referring to survival. A total of 6 genes (LIN7A, MAOA, SLC18A2, SLC1A1, STX1A, UNC13B) that meet the threshold of P < .2 were maintained for further analysis, with 5 genes (SLC18A2, UNC13B, MAOA, SLC1A1, LIN7A) determined to be protective genes (HRs < 1) and STX1A identified to be a risky gene (HRs > 1). The forest map is presented in Figure 4A. Secondly, the LASSO and multivariate Cox regressions were performed to construct a 3-gene signature model to predict survival outcomes of LUAD patients (Fig. 4B and C). The NRC scores were computed via formula: NRC score = (−0.149 * SLC18A2 exp.) + (0.040 * STX1A exp.) + (−0.107 * UNC13B exp.). In the light of the median NRC score, individuals suffering LUAD in the TCGA cohort were grouped into the NRC-high group (N = 256) and NRC-low group (N = 257). Figure 4D presents the distribution of NRC scores, OSs, and gene expression levels of 3 genes in 2 groups. Markedly, in the NRC-high group, SLC18A2 and UNC13B are downregulated, while STX1A is upregulated. Patients in the NRC-high group have a poorer prognosis instead of the NRC-low group (log-rank P < .001) (Fig. 4E). To appraise the sensitivity and specificity of predictions of the identified NRC signature, 1 to 3-year time-dependent receiver operating characteristic curves were plotted and area under curve (AUC) values were calculated. The 1 to 3-year survival AUCs were 0.636, 0.627, and 0.633 respectively (Fig. 4F).

Figure 4.

Building of NRC signature in TCGA cohort. (A) The forest map showing univariate Cox regression analysis results for NRCR DEGs. (B) LASSO regression of 6 neurotransmitter release cycle survival-related genes. (C) The selection of the parameter in the LASSO regression. (D) The distribution of NRC scores, OSs, and the gene expression levels of 3 genes in 2 NRC groups (NRC-high group: red, NRC-low group: green). (E) The Kaplan–Meier curves of OSs in 2 NRC groups. (F) ROC curves assess the prognostic sensitivity and specificity of the NRC score. DEGs = differentially expressed genes, LASSO = the Least Absolute Shrinkage and Selection Operator, NRC = neurotransmitter release cycle, NRCR = neurotransmitter release cycle-related, OSs = overall survivals, ROC = receiver operating characteristic, TCGA = The Cancer Genome Atlas.

3.4. Validation of prognostic value for the 3-gene signature

226 LUAD patients in the GEO database (GSE31210) were employed as a validation set. Under the formula from the TCGA training set, NRC scores of samples from the validation set were calculated. And similarly, 226 LUAD patients were grouped into 2 NRC groups (each, N = 113). NRC scores, OSs, and the gene expression levels of 3 genes in 2 NRC groups are demonstrated in Figure 5A. Patients of the NRC-high group appeared shorter survival time and lower expression levels of UNC13B and SLC18A2, and higher expression levels of STX1A. Patients in the NRC-high group had poorer prognoses than those in the NRC-low group, consistent with the result of the training set (log-rank P < .0001) (Fig. 5B). The values of AUC were respectively 0.658, 0.704, and 0.702 for 1-3-year survival and it indicated that this model had great predictive efficacy (Fig. 5C).

Figure 5.

Validation of prognostic gene model in GEO cohort. (A) The distribution of NRC scores, survival status, and the expression levels of 3 genes in 2 NRC groups (NRC-high group: red, NRC-low group: green). (B) The Kaplan–Meier curves of OSs in 2 NRC groups. (C) 1- to 3-year ROC curves assessing the prognostic sensitivity and specificity of NRC score. DEGs = differentially expressed genes, GEO = Gene Expression Omnibus, NRC = neurotransmitter release cycle, OSs = overall survivals, ROC = receiver operating characteristic.

3.5. NRC score was determined as an independent influencing factor for the prognosis of LUAD patients

Independent influencing factors for the prognoses of LUAD patients were determined via Cox regression analyses. According to univariate Cox regression analysis (Fig. 6A and C), NRC score and stage were significantly associated with the prognoses of LUAD patients. After excluding other confounding factors in multivariable Cox regression, as shown in Fig. 6B and D), NRC score and stage were still determined to be independent influencing factors for prognoses of individuals suffering from LUAD. Furthermore, in the TCGA cohort, as Figure 6E shows, survival status (P < .001), stage T (P < .01), stage, and stage N, gender (P < .05) of patients were diversely distributed between 2 NRC groups. Meanwhile, for the GEO cohort, there were statistical differences in gender (P = .0033), status, and stage (P < .001) of patients between the 2 NRC groups (Fig. 6F).

Figure 6.

Univariate and multivariate Cox regression analysis. (A) Univariate Cox regression analysis in TCGA cohort (stage: I–IV; smoking: 1: never smoke, 2: ever smoke, 3: smoke now; NRC: high or low; gender: female and male; age: ≥60 or <60). (B) Multivariate Cox regression analysis in TCGA cohort (smoking: 1: never-smoker, 2: ever-smoker, 3: smoker now). (C) Univariate Cox regression analysis in the GEO cohort (stage: I or II; smoking: 0: never smoke, 1: ever smoke; NRC: high or low; gender: female and male; age: >60 or ≤60). (D) Multivariate Cox regression analysis in GEO cohort (smoking: 0: never-smoker, 1: ever-smoker). (E) Comparison of clinical features between 2 NRC groups for TCGA cohort. (F) Comparison of clinical features between 2 NRC groups for the GEO cohort. GEO = Gene Expression Omnibus, NRC = neurotransmitter release cycle, TCGA = The Cancer Genome Atlas.

3.6. GO and KEGG enrichment analyses

8893 genes were identified to be DEGs between 2 NRC groups using “limma” package with criteria |logFC| ≥ 1 and FDR < 0.05. Among them, 2865 genes were upregulated while the other 6028 genes were downregulated. GO analysis result (Fig. 7A)) manifested that, at biological process level, these genes were significantly enriched in small GTPase, Wnt and Ras protein signal transduction, cell cycle, mitosis, DNA replication, cell-substrate adhesion, cell migration, and Golgi vesicle transport. And at the molecular function level, genes mainly took part in GTPase activator activity and some kinase regulator activity. Furthermore, the cellular component revealed that these DEGs were related to the cell-substrate junction, focal adhesion, early endosome, and endocytic vesicle. KEGG analysis (Fig. 7B) shows that these genes are mostly enriched in DNA replication, cell cycle, endocytosis, focal adhesion, NSCLC, p53, sphingolipid, and MAPK.

Figure 7.

Enrichment analysis for the DEGs between the 2 NRC groups in the training data. (A) Bar chart showing GO analysis result (the higher the column, the higher the reliability of this pathway ID). (B) Bar chart presenting KEGG analysis result (the higher the column, the higher the reliability of this pathway ID). DEGs = differentially expressed genes, GO = Gene Ontology, KEGG = Kyoto Encyclopedia of Genes and Genomes, NRC = neurotransmitter release cycle.

3.7. Association of NRC score with immunotherapy-related anti-cancer immunity and TME

In terms of immunotherapy-related anti-cancer immunity, 6 ICPGs, namely CD274, HAVCR2, IDO1, LAG3, PDCD1, and PDCD1LG2, overexpressed in the NRC-high group rather than the NRC-low group (all P < .001). And scores of IFN-γ, ICR, inflammation, APM, and CD8 T effector signatures were all higher in the NRC-high group than in the NRC-low group (all P < .001). Moreover, TME signatures were also analyzed in 2 NRC groups, and the heatmap suggests that the NRC-high group had higher TGF-β activity which was quantified by Pan_F_TBRs, and higher EMT1-3, stromal scores (all P < .01) (Fig. 8A). Concerning DNA damage and repair signatures, the results visualized in Figure 8B show that base excision repair, DNA damage response, DNA replication, and nucleotide excision repair were higher in the NRC-high group (all P < .001). Last but not least, 22 types of tumor-infiltrating immune cells enrichment scores in 2 NRC groups were calculated by CIBERSORT. Box plots presented in Figure 8C show that B cells memory (P < .05), macrophages M0 (P < .001), M1 (P < .01), T cell CD4 memory activated (P < .001) and T cells gamma delta (P < .001) were more abundant in the NRC-high group, while it had lower B cells naive (P < .01), mast cells resting (P < .01), monocytes (P < .05), plasma cells (P < .05) and T cells CD4 memory resting (P < .001).

Figure 8.

Comparison of the enrichment scores of immune-related signatures. (A) Heatmaps presenting ICPGs expression, anti-tumor immunity, and TME signature levels in 2 NRC groups. (B) Box plots showing differences in DNA damage and repair related-signatures levels in 2 NRC groups. (C) Box plots showing differences of 22 types of TIICs enrichment scores in 2 NRC groups. ICPGs = immune checkpoint genes, NRC = neurotransmitter release cycle, TIICs = tumor-infiltrating immune cells, TME = tumor microenvironment.

4. Discussion

In the past time, a vast number of researchers have clarified the regulatory role of neurotransmitters for functions concerning physiology and pathology in various tissues and organs.^[3,23,24] There are new studies indicating that neurotransmitter-initiated signaling pathways can activate the uncontrolled proliferation and spread of cancer cells. On the one hand, neurotransmitters can bind to NTRs and thereby initiate signaling pathways that regulate cell growth and invasion. EGFR, Wnt, and YAP signal pathways can be stimulated by M3R in gastric cancer.^[25,26] On the other hand, the released neurotransmitters and growth factors can promote tumor inflammation and new blood vessel formation by acting on endothelial cells and immune cells.^[5] mAChRs and nAChRs which can be activated through either autocrine or paracrine mechanisms exist in many immune cells such as macrophages, dendritic cells, neutrophils, and monocytes.^[27] Our study was the first to construct NRCR model to predict OSs of LUAD patients. In this study, a signature featuring 3 NRCRGs (SLC18A2, STX1A, UNC13B) that can predict survival outcomes of LUAD patients was generated.

Vesicular monoamine transporters are preconditions for the transportation of biogenic amines to organelles.^[28] SLC18A2 (VMAT2), which serves as an integral membrane protein for secreting vesicles, encodes vesicular monoamine transporter 2. It can transport monoamines in the cytosol, including dopamine, serotonin, epinephrine, and other monoamines into vesicles either stored in vesicles or exocytotic release.^[29] Moreover, SNARE proteins take part in transporting neurotransmitters, growth factors, and recycling receptors, and also play a role in secreting matrix proteases. Therefore, it endows cells with the ability to invade and migrate.^[30] A previous study has found that the abnormal expression of SNARE protein serves as a key part in the development of cancer by taking part in vesicle fusion as a core signaling protein and has become an effective target for cancer treatment.^[31] STX1A acts as neuronal SNAREs mediating synaptic vesicular fusion.^[12] Studies have shown that STX1A is overexpressed not only in primary brain tumors, but also in colon and rectum, breast, and lung cancers.^[13–16] Similarly, in this study, there were higher expression levels of STX1A in LUAD patients. Conversely, tumor growth of glioblastoma can be inhibited by blocking STX1A expression.^[32] UNC13B, also named MUNC13 and it participates in regulating exocytosis of vesicles.^[33] SNARE-mediated membrane fusion depends on regulation of accessory proteins and then regulates neurotransmitter release exquisitely. MUNC13, which serves as one of the accessory proteins, adjusts the formation of SNARE complex.^[34] It could be seen that 3 NRCRGs (SLC18A2, STX1A, UNC13B) in a signature are all related to vesicle transport. It can be speculated that the disorder of vesicle transport might affect the release of neurotransmitters and thus have an impact on the progression of tumors.

In the light of the results of GO and KEGG analyses, DEGs were correlated with mitosis, DNA replication, Golgi vesicle transport, cell-substrate adhesion, focal adhesion, cell migration, and small GTPase, Wnt, Ras, p53, sphingolipid, MAPK signaling pathway, and the other signaling pathways. It has also been proved that interfering with mitosis for cancer treatment in the clinics is successful.^[35] Additionally, cell-matrix adhesion and cell migration play vital roles in controlling the invasion and metastasis of cancer cells.^[36] Research on microglia has shown that dopamine might affect cell adhesion.^[37] Abnormal vesicle transport might promote the formation of cancer cells, alteration of cell adhesion and extracellular matrix, and thereby facilitate tumor growth.^[38] Last but not least, it has been already discussed that neurotransmitters can bind to NTRs and activate the corresponding signal pathways. For example, Akt and MAPK signal pathways can be activated by M3R and promote cell proliferation in NSCLC.^[10] In most cancers, the imbalance of signal transduction was proved to be crucial in triggering cell survival under malignant conditions.^[39] Therefore, according to GO and KEGG analyses in this study, the hypothesis that these DEGs might initiate signal pathways by regulating vesicle transport and thereby influence the development of LUAD is reasonable.

ICPs are mainly responsible for preventing excessive inflammatory response of T cells which can launch an attack on normal cells. However, if ICPs are overexpressed, the immune function of the human will be suppressed and then tumors are prone to occur and metastasis. Clinically, blocking antibodies of ICP are usually used for cancer treatment. In this study, 6 ICPGs were overexpressed in the NRC-high group. For TME signatures, EMT can improve the invasion and migration capabilities of cancer cells by inducing changes in TME, while TGF-β^[40] and inflammation^[41] serve as drive factors of EMT. Meanwhile, changes in tumor stroma which serve as a vital component of TME act as a stimulant or depressant in tumors.^[42] The activation of stroma in TME was regarded as immune-suppressive.^[43] Therefore, although immune cells including B cells memory, macrophages M0, M1, T cell CD4 memory T cells gamma delta were abundant in the NRC-high group, these cells might merely stay in the stroma rather than penetrate the tumor. By analysis from this study, Pan_F_TBRs, EMT1-3, and stromal-related signatures were higher in the NRC-high group than in the NRC-low group. Meanwhile, NRC-high patients had higher DNA damage and repair levels as well. The higher activation of base excision repair means the more effective the anti-tumor effect. In addition, higher IFN-γ, APM, inflammation, ICR, and CD8 T effector suggest immune activation. Inflammation can also lead to DNA damage and there exists a positive feedback loop between inflammation and DNA damage.^[44] The inflamed TME can lead to a great immunotherapeutic effect. These results proved that patients in the NRC-high group might have better immunotherapy effects than those in the NRC-low group.

In a word, after the above discussion, neurotransmitters are closely correlated with LUAD. And by building a LASSO-Cox model, the signature featuring 3 NRCRGs can predict prognosis effectively, and the NRC score has been proved to be an independent risk factor whether in the training data or the validation data. This is the first study on the prognostic value of NRCRGs in LUAD. There is significant clinical significance in our model with 3 genes. However, the results in this study still need follow-up experiments to support. The mechanism of the role of NRCRGs in cancer needs further research and verification.

5. Conclusions

The study offered reliable biomarkers for predicting the survival outcomes of patients suffering from LUAD and the NRC score might contribute to guiding LUAD patients with immunotherapy selection. Meanwhile, GO and KEGG analyses provide ideas for the mechanism of NRCRGs on LUAD.

Acknowledgments

We thank all the researchers involved in the consolidation and submission of the data from the TCGA database, which may provide convenience and the possibility of tumor studies in a large cohort. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. In addition, the authors have no conflicts of interest to disclose.

Author contributions

Conceptualization: You Ge, Han Li.

Data curation: Pingmin Wei.

Formal analysis: Han Li, Zemin Wang.

Methodology: Yangyang Liu.

Project administration: Pingmin Wei.

Writing – original draft: Han Li.

Writing – review & editing: You Ge.