DOI: https://doi.org/10.21203/rs.3.rs-1057544/v1
Background: Immune infiltrated genes (IIGs) have been identified to associated with the prognosis of various cancers, but their expression and prognostic significance remain largely unclear in stomach adenocarcinoma (STAD).
Methods: Gene expression profiles and clinical data of STAD patients were downloaded from The Cancer Genome Atlas (TCGA) as a training dataset (n = 375) and Gene Expression Omnibus (GEO) databases as a validation dataset (n = 300). Construction of high and low immune cell infiltration groups was performed by single sample gene set enrichment analysis (ssGSEA) and evaluated by ESTIMATE algorithm-derived immune scores. The overlapping differentially expressed genes (DEGs) in tumor vs. normal and Immunity-H vs. Immunity-L were selected as differentially expressed immune infiltrated genes (DEIIGs), which were used to construct DEIIG prognostic signature and its performance was validated using validation dataset. Moreover, the association between clinical data and immune features were explored. Furthermore, ADH4 and ANGPT2 were selected for analyzing their expression and prognostic values in STAD patients.
Results: A total of 191 overlapping DEGs, including 6 lnRNAs and 185 mRNA were identified. Consecutively, 9 DEIIG prognostic signature (LINC00843, ADH4, ANGPT2, APOA1, ASLC2, GFRA1, KIAA1549L, MTTP and PROC) were identified as risk signature and Kaplan-Meier curve and ROC curve verified its performance in TCGA and GEO datasets. Total five clinical outcomes (age, pathologic T, radiotherapy, tumor recurrence and prognostic score model status) were identified to be associated with the survival prognosis of STAD patients. The TIMER algorithm revealed that B cell, T cell CD4+, neutrophil, macrophage and myeloid dendritic cell were positively correlated with STAD prognosis, while CD8+ was negatively correlated with STAD prognosis. Additionally, we validated that higher ADH4 and lower ANGPT2 predicted better survival prognosis in STAD patients.
Conclusion: We constructed and verified a robust signature of nine DEIIG prognostic signature for the prediction of STAD patient survival.
Stomach adenocarcinoma (STAD), as the most common pathology type of gastric cancer, is the fifth prevalent malignancies and leading causes among all malignancies, with estimated more than 100 million new cases and nearly 80 million deaths in 2018 [1, 2]. Despite the clinical outcomes have been improved under surgical therapies, chemotherapy and systemic treatments for STAD patients at early stage, most of 50% patients are identified as advanced stage, thereby causing less than 30% five-year survival rate [3–5]. Thus, it is great of importance to identify novel prognostic biomarkers and therapeutic targets for STAT patients.
As a well-recognized heterogeneous cancer, STAD is not only composed of cancer cells, but also by non-cancer cells, including endothelial cells, macrophages, stromal and immune cells [6]. Among these non-cancer cells, the tumor infiltrating immune cells (TIICs) have been reported to be closely associated with the clinical outcomes and response to immunotherapy for their crucial roles in pro- and anti-tumorigenic processes [7]. Increasing evidence indicates that immune cell infiltration plays a vital role in the prognosis of cancer, including breast cancer [8], colon cancer [9] and bladder cancer [10]. For instance, tumor-associated lymphocytes (TALs), primarily T cells, can regulate the proliferation and migration of cancer cells by releasing soluble cytokines, as well as participate in activating angiogenesis and host defense mechanism [11]. The infiltration of tumor-associated macrophages (TAMs) into tumor tissue has been reported to be significantly correlated with tumor vascularity, the depth of tumor invasion, lymph node status and clinical stages [12, 13]. As one of the primary effector cells of anticancer immunity, CD8+ T cells is identified as a potential prognostic indicator of gastric cancer [14]. Recent studies indicated that the alter gene expression levels exerts anti-tumor effects through regulating an immune suppression mechanism in TIICs and is correlated with favorable prognosis as follows: Chaudhary et al. [15] reported for the first time that neuropilin 1 (NRP1) is upregulated on tumor-infiltrating lymphocytes (TILs) and can be induced on peripheral blood mononuclear cells (PBMCs) from colorectal cancer liver metastases. Wang et al. [16] previously identified that SUPV3L1 and SLC22A17 as hub genes affect immune cell infiltration, result in the different prognosis of gastric cancer. In addition, immune infiltration revealed a significant correlation between JAK3/TYK2 expression and the abundance of immune cells as well as immune biomarker expression in STAD [17]. Nevertheless, the association among gene expression levels, tumor infiltrating immune and survival prognosis remains largely unclear.
In the present study, we evaluated the immune cell infiltration in STAD tumor samples obtained from TCGA database based on single sample gene set enrichment analysis (ssGSEA) algorithm and distinguished the high immune infiltration group from the low infiltration group. On the basis of immune grouping, Cox regression analysis and LASSO algorithm were combined to screen the prognostic marker RNAs factors of STAD, and the survival prediction model was constructed and verified based on the prognostic marker RNAs.
Gene expression profiles (lncRNAs and mRNAs) and corresponding clinical information from primary STAD tumors, uploaded up to the 20th October 2020, were obtained from The Cancer Genome Atlas (TCGA: https://cancergenome.nih.gov/) and Gene Expression Omnibus (GEO: https://www.ncbi.nlm.nih.gov/geo/). For TCGA datasets, total 407 samples containing 375 STAD samples and 32 normal samples were selected as training group. With the sample screening criteria (clinical follow-up information was retained and included samples at least 200), the gene expression assay GSE62254 (GP570, Affymetrix Human Genome U133 Plus 2.0 Array) as external validation dataset, including 300 STAD samples and their corresponding clinical information was retrieved from GEO database. Overlapping lncRNAs and mRNAs from these two datasets were selected for further analysis. The overall study design and the different samples that were included at every stage of the analysis were illustrated as a flowchart in Figure 1.
The immune cell infiltration levels of STAD tumor samples were quantified by single sample gene set enrichment analysis (ssGSEA) in R3.6.1 package Gene Set Variation Analysis (GSVA) Version 1.36.3 (http://www.bioconductor.org/packages/release/bioc/html/GSVA.html) [18]. The ccGSEA employed gene signatures expressed by immune cell populations to individual tumor samples. Subsequently, 375 STAD samples were divided into high immunity infiltration (Immunity-H) and low immunity infiltration (Immunity-L) groups according to the results from ccGSEA data. Moreover, the reasonableness of immune infiltration grouping was validated using ESTIMATE method [19] and CIBERSORT algorithm [20].
The samples of TCGA dataset were divided into two groups according to sample source (tumor vs. normal) and obtained immunity group (Immunity-H vs. Immunity-L). Differentially expressed genes (DEGs) between tumor and normal or between Immunity-H and Immunity-L groups were identified using limma package of R3.6.1 Version 3.34.7 [21] with the cut‐off value of FDR (false discovery rate) < 0.05 and log2 |fold change (FC)| > 1. These DEGs were visualized in a volcano plot in R. The overlapping DEGs in tumor vs. normal and Immunity-H vs. Immunity-L were selected as differentially expressed immune infiltrated genes (DEIIGs), which were visualized using Venn diagram.
Then, DEIIGs were analyzed by the Database for Annotation, Visualization and Integrated Discovery (DAVID) Version 6.8 bioinformatics tool (https://david.ncifcrf.gov/) [22, 23]. Gene Ontology (GO) biology process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were then performed to annotate the potential functions for DEIIGs with the cut-off value of FDR < 0.05.
Univariate Cox analysis in R3.6.1 survival package [24] was used to determine the association between the expression level of DEIIGs and patient's overall survival (OS) with the threshold of log-rank p value < 0.05. After filtration of prognostic DEIIGs, independent prognosis related DEIIGs were further screened via a multivariate Cox regression model with p-value < 0.05 as the cut-off criterion. Lasso-penalized Cox regression analysis [25] was performed to further reduce the number of independent prognosis related DEIIGs with the optimal lambda using 1000-time cross-validation likelihood based on penalized package Version 0.9.50 [26]. According to the best lambda value, a prognostic gene signature of STAD patients was constructed with the following formula: Prognostic score (PS) = ∑βDEIIGs × Exp DEIIGs. Here, βDEIIGs represent the regression coefficients (β) derived from the Lasso Cox regression model and Exp DEIIGs represent the expression levels of signature DEIIGs in training dataset.
Taking the median PS as the cutoff point, we divided the samples in training dataset into high-risk group (PS > median value) and low-risk group (PS < median value). Kaplan–Meier (KM) survival curves analysis was used to analyze the OS between the high-risk and low-risk groups. The accuracy and sensitivity of survival prediction based on the PS were verified by receiver operating characteristic (ROC) curve analysis and determined by the value of area under the curve (AUC). Meanwhile, the expression levels of signature DEIIGs in validation dataset GSE62254 were extracted and the PS was calculated according to the formula as above. Similarly, Kaplan-Meier curve and ROC curve analysis were performed to evaluate the predictive ability of the signature.
Next, univariate and multivariate Cox regression analyses Survival package (Version2.41–1, http://bioconductor.org/packages/survivalr/) [24] were performed in the TCGA dataset on the DEIIG prognostic signature and clinicopathological parameters including age, gender, pathologic-M, pathologic-N, pathologic-T, pathologic-stage, neoplasm histologic grade, radiation therapy and PS model status. Log-rank p value < 0.05 was considered statistically significant. Parameters with log-rank p value < 0.05 based on the univariate analysis were further included in the multivariate Cox regression analysis to obtain independent prognostic parameters. Visual presentation of independent prognostic parameters was performed with forestplot Version 1.10 in R3.6.1 language [27]. Subsequently, we constructed the clinical prognosis models based on these independent prognostic parameters alone, which were compared with PS prognostic model by drawing ROC curves with the quantitative indicator AUROC (0.5-1) [28].
Total six types of TICs, including B cell, T cell CD4+, T cell CD8+, neutrophil, macrophage and myeloid dendritic cell were retrieved from Tumor Immune Estimation Resource (TIMER: https://cistrome.shinyapps.io/timer/) [29] as a web server for comprehensive analysis of TICs. These six kinds of TICs abundance distribution from 351 tumor samples in the training cohort was estimated by CIBERSORT calculation method. The correlation between the TICs’ proportion and PS was calculated using Spearman coefficient test.
Total 59 paired tumor tissues and matched adjacent tissues were collected from STAD patients from July 2017 to October 2020. All participating patients gave their written informed consent and did not receive adjuvant chemotherapy or radiotherapy prior to surgery. Basic clinical information, including sex, age, tumor size and lymph node metastasis. Follow-up information for all participants was obtained every three months by telephone or via a postal questionnaire. During the follow-up period, overall survival was measured from diagnosis to death or to the last follow-up (at five years). This study was approved by the Ethical Committee of The First Affiliated Hospital of Zhengzhou University (Henan, China).
Total RNA sample was extracted from tissue specimens using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) and reverse transcription of mRNA was performed with PrimerScript RT reagent Kit (Takara Biotechnology Co. Ltd., Dalian, China) according to the manufactures’ instructions. SYBR Green quantitative PCR reaction was carried out in triplicate in a 20 μL reaction volume containing 2× PCR Master Mix (Applied Biosystems) with the cycling conditions as follows: 95 minutes or five minutes, followed by 40 cycles of 95 °C for 20 s and 60 °C for 30 s. Relative expression levels of ADH4 and ANGPT2 mRNA were calculated by the 2−ΔΔCq method [30].
Statistical analysis was performed using SPSS 22.0 (SPSS Inc.; Chicago, IL, USA). Continuous variables were analyzed using Student's t-test for paired samples. The association between gene expression levels and categorical variables were analyzed by the chi-square test. The relationship between gene expression levels and overall survival was analyzed through the Kaplan–Meier method, which was evaluated by the log-rank test. The univariate regression model was used to analyze the effects of individual variables on survival, and the multivariate Cox regression model was used to confirm the independent impact factors associated with survival. A p value < 0.05 was accepted as statistically significant.
The immune cell infiltration status was assessed by applying the ssGSEA approach to each tumor sample of TCGA STAD cohort. As shown in Figure 2, 375 tumor samples were distinctly divided into two clusters, including Immunity-H (n = 192) and Immunity-L (n = 183) groups based on the landscape of 28 immune cell subpopulations infiltrations in STAD. Detailed results from ssGSEA were presented in Table S1. Next, we calculated the stromal score, immune score and estimate score using ESTIMATE method (Table S2). As depicted in Figure 3A, there were significant differences in stromal score, immune score and estimate score between Immunity-H and Immunity-L groups, of which corresponding scores in Immunity-H group were notably higher than those in Immunity-L group. Moreover, the results from CIBERSORT algorithm on immune cell type showed that the fraction of some important immune cell subtypes varied distinctly between Immunity-H and Immunity-L groups (Figure 3B), which were summarized in Table S3. Collectively, the Immunity-H and L groupings obtained based on ssGSEA evaluation can be used for subsequent analysis.
We screened the DEGs between tumor and normal samples, and Immunity-H and Immunity-L samples in TCGA dataset. The volcano plots were drawn to visualize DEGs between tumor and normal group, and Immunity-H and Immunity-L group (Figure 4A). A total of 894 DEGs between tumor and normal group, and 592 DEGs between Immunity-H and Immunity-L group were screened, which were listed in Table S4. Moreover, the number of overlapping DEIIGs was 191, including 6 lncRNAs and 185 mRNAs (Figure 4B, Table S5).
GO function and KEGG pathway enrichment analyses were performed for 185 DEIIGs. These DEIIGs were significantly enriched in 11 biological processes, such as retinoid metabolic process, cell-cell signaling, collagen catabolic process, potassium ion transport and potassium ion transmembrane transport (Table 1, Figure 5). The KEGG pathway analyses showed that the DEIIGs were mainly concentrated in transcriptional misregulation in cancer, vitamin digestion and absorption, fat digestion and absorption, protein digestion and absorption, and gastric acid secretion (Table 1, Figure 5).
Univariate Cox regression analysis was performed for the 191 DEIIGs, of which 32 DEIIGs showed significant prognostic potential (log-rank p value < 0.05). Next, total 13 independent prognostic DEIIGs were further screened via a multivariate Cox regression model. After that, we performed the LASSO Cox regression analysis to reduce the number of independent prognosis related DEIIGs with the optimal lambda and finally obtained nine prognostic DEIIG signatures with corresponding coefficients for further study (Table 2).
According to the risk coefficient of each gene and the gene expression level, the PS of each patient in training dataset was calculated, which is a linear combination of the expression level of each gene weighted by its multivariate LASSO regression coefficient. The samples in training dataset were assigned into high-risk and low risk groups with the median PS as the cutoff value. The survival analysis indicated that the survival rate was remarkably lower in the high-risk group as opposed to low-risk group (p-value < 0.001, HR = 2.230, 95% CI = 1.583-3.142); whereas, the ROC curve analysis showed acceptable discrimination with AUC of 0.824, and high sensitivity and specificity in training dataset (Figure 6A). In addition, the external dataset GSE62254 was used to validate the prediction performance of the nine prognostic DEIIG signature. With the aforementioned formula, we calculated individual PS and classified the patients in validation dataset into high-risk and low-risk groups. Consistently, a significant separation was shown in the KM survival curve in validation dataset (p-value < 0.05, HR = 1.405, 95% CI = 1.020-1.935) and ROC curve analysis demonstrated accepted discrimination with an AUC of 0.766 (Figure 6B). In general, the nine prognostic DEIIG signature performed well at predicting OS of STAD.
Total 351 patients from the TCGA STAD dataset for which complete clinical information was provided, including age, gender, pathologic-M, pathologic-N, pathologic-T, pathologic-stage, neoplasm histologic grade and radiation therapy were included in the univariate and multivariate Cox regression analyses (Table S6). As shown in Table 3, univariate analysis revealed that age (p = 6.91E-03), pathologic M (p = 1.13E-02), pathologic N (p = 1.71E-03), pathologic T (p = 8.09E-03), pathologic stage (p = 1.68E-05), radiation therapy (p = 2.29E-04), recurrence (p = 3.44E-06) and PS model status (p = 2.56E-06) were significantly correlated with overall survival of STAD patients. Multivariate analysis further screened that age (p = 4.17E-03), pathologic T (p = 3.19E-02), radiation therapy (p = 1.05E-02), recurrence (p = 6.33E-04) and PS model status (p = 2.72E-03) were independent risk factors of overall survival. The results from forest map clearly described that age, pathologic T, tumor recurrence and PS model status were tumor risk factors, while radiotherapy was tumor protective factor (Figure 7). Subsequently, we performed ROC analyses to assess how these independent risk factors could behave in predicting prognosis. As shown in Figure 8, the AUC of PS model status performed on overall survival in the training cohort was 0.824, which was superior to those of age (0.545), pathologic T (0.537), radiotherapy (0.544) and recurrence (0.640), which may be the best performance in predicting overall survival.
Based on the expression levels of TCGA STAD samples, we used TIMER to analyze the proportion of six kinds of TICs (Table S7). Combining the results of correlation analysis (Figure 9), B cell, T cell CD4+, neutrophil, macrophage and myeloid dendritic cell were positively correlated with PS, whereas T cell CD8+ was negatively correlated with PS. Thus, the significant infiltration with these TICs may potential act as one of the critical factors that the nine DEIIG signature holds to influence the outcome of STAD pronounced.
As described above, we have identified nine-DEIIG prognostic signature baed on the TCGA database. To further verify our findings, 59 cases of STAD specimens were collected and performed with quantitative real time PCR. As expected, ADH4 was downregulated in tumor tissues compared with adjacent tissues (Figure 10A). Clinical analysis further demonstrated that decreased ADH4 was associated with TNM stage, lymph node metastasis (Table 4), and represented an independent risk factor for overall survival (Table 5, Figure 10B). Conversely, ANGPT2 was upregulated in tumor tissues compared with adjacent tissues (Figure 10C), which was correlated with TNM stage (Table 6) and worse prognosis (Table 7, Figure 10D).
As our best knowledge, gene expression and immune cell infiltration play a key role in the prognosis of tumors [31, 32]. Nevertheless, the association among gene expression levels, tumor infiltrating immune and survival prognosis remains largely unclear. Here, we used integrative bioinformatics to screen immune cell infiltration related genes based on the landscape of 28 immune cell subpopulations infiltrations in STAD derived from TCGA database. A total of 191 DEIIGs, including 6 lncRNAs and 185 mRNAs were obtained and used to apply for construction of the DEIIG prognostic signature. Total nine prognostic DEIIG signature (LINC00843, ADH4, ANGPT2, APOA1, ASLC2, GFRA1, KIAA1549L, MTTP and PROC) was identified to be associated with tumor cell immune infiltration. Alcohol dehydrogenases (ADHs), including class I (ADH1A, ADH1B, and ADH1C), class II (ADH4), class III (ADH5), class IV (ADH6), and class V (ADH7) [33], are huge family of dehydrogenase enzymes and associated with the prognosis of various cancers [34, 35]. A recent study by Wang et al. [36] identified that ADH4 was one of downregulated innate immunity genes in oral immune homeostasis. The presence of an ANGPT2-rich environment was associated with impairment of preexisting T-cell responses against tumor-associated antigens (TAA) and poor prognosis in patients with NSCLC [37]. In addition, APOA1 [38], GFRA1 [39], KIAA1549L [40] and MTTP [41] were all reported to be directly or indirectly associated with immune cell infiltration in cancer prognosis. There is little information concerning LINC00843, ASLC2 and PROC in immune infiltration related tumor prognosis, which need to be further explored.
Next, we evaluated the prognostic performance of DEIIG signature in TCGA and GEO datasets. Survival curves and time-dependent ROC and AUC analyses indicated that the nine prognostic DEIIG signatures have powerful predictive capacity for STAD. Moreover, the AUC of PS model status performed on overall survival in the training cohort was 0.824, which was superior to those of age (0.545), pathologic T (0.537), radiotherapy (0.544) and recurrence (0.640), which may be the best performance in predicting overall survival. Consistent with the analysis of multivariate prognostic modules, the hazard ratio (HR) value of risk score based on the nine DEIIGs was higher among the factors in the forest map. These outcomes further confirmed that the nine-DEIIG signature was the most effective signature for prognostic assessment of STAD patients when compared with other clinical features. Similar to our study, Wang et al. [42] collected clinical data of STAD patients from TCGA database and established a stromal-immune-score-based gene signature and risk stratification. Yang et al. [43] collected RNA-seq data of immune infiltrated-related genes (IRGs) of 372 STAD patients from TCGA database and established a 10 prognostic gene prognostic model. Wu et al. [44] integrated clinical data to identify seven hub IRGs and establish the IRG prognostic model associated with STAD. Compared with previous studies, our study used updated data from TCGA and included 375 STAD patients. We used different validation dataset from GEO database and identified different nine prognostic DEIIG signature. Moreover, these DEIIGs were significantly correlated with the clinical outcomes (age, pathologic T, radiation therapy and recurrence) of STAD patients.
Our study also clarified the correlation between the three useful prognostic indicators and six types of tumor-infiltrating immune cells using TIMER. The results showed that nine prognostic DEIIG signature was positively correlated with B cell, T cell CD4+, neutrophil, macrophage and myeloid dendritic cell, but was negatively correlated with T cell CD8+. In fact, CD8+ T cells is one of the primary effector cells of anticancer immunity, which has been identified as a potential prognostic indicator of gastric cancer [14]. Consistently, previous study suggested that CD8 T cells with APOA1 as an alternative cellular vaccine for highly-active antiretroviral therapy [45]. ANGPT2 is a well-studied potential prognostic marker in B cell related chronic lymphocytic leukemia [46]. Furthermore, our validation experiments further demonstrated that both ADH4 and ANGPT2 were aberrantly expressed in STAD tissues and correlated with poor prognosis in STAD patients. Therefore, the identified nine prognostic DEIIG signature may also exert a vital function in immunotherapy of STAD. In addition, there were some limitations in our study as follows: We performed analysis at mRNA and non-coding level but not protein level. Furthermore, lacking of in vitro and in vivo experiments used for validating our results.
In summary, we screened nine DEIIGs (LINC00843, ADH4, ANGPT2, APOA1, ASLC2, GFRA1, KIAA1549L, MTTP and PROC) with marked prognostic capability for STAD. These DEIIGs were further confirmed as independent prognostic factors associated with OS of STAD patients. The findings might provide a new perspective that will help to find potential novel targets for STAD immunotherapy.
STAD, stomach adenocarcinoma; ssGSEA, single sample gene set enrichment analysis; IIRGs, immune infiltrated-related genes; TCGA, The Cancer Genome Atlas; GEO, Gene Expression Omnibus; GSVA, Gene Set Variation Analysis; DEIIGs, differentially expressed immune infiltrated genes; DAVID, Database for Annotation, Visualization and Integrated Discovery; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; PS, prognostic score; TIMER, Tumor Immune Estimation Resource
This study was approved by the Committee on the Ethics of Ethical Committee of The First Affiliated Hospital of Zhengzhou University (Henan, China). All of the experiments were performed in accordance with the Declaration of Helsinki. All volunteers who donated tissues have provided their written informed consent.
Ya Yang and Xintan Zhang wrote the main text of the article and designed the experiments. Ya Yang and Tingxuan Li were responsible for data analysis work. Wang Yue Zhang prepared Figures and revised the figure style. Xiaoxiao Zuo revised the manuscript. All the authors approved the manuscript. All authors contributed to the article and approved the submitted version.
The analyzed data sets generated during the study are available from the corresponding author on reasonable request.
Not applicable.
Not applicable
This work was supported by the Medical Science and Technology Co-construction Project of Henan Province.
The authors declare that they have no conflict of interest.
Table 1
GO function and KEGG pathway analysis of DEIIGs
|
Category |
Term |
Count |
P-Value |
FDR |
Genes |
||||
|
Biology Process |
GO:0001523~retinoid metabolic process |
8 |
2.65E-06 |
1.47E-03 |
ADH4, RDH12, RBP2, APOC3, APOA1, APOA4, LRAT, APOB |
||||
|
GO:0030574~collagen catabolic process |
7 |
4.55E-05 |
1.26E-02 |
MMP12, MMP7, COL11A1, MMP3, COL4A6, COL10A1, MMP10 |
|||||
|
GO:0042158~lipoprotein biosynthetic process |
4 |
8.19E-05 |
1.50E-02 |
MTTP, APOA1, APOA4, APOB |
|||||
|
GO:0033344~cholesterol efflux |
5 |
1.09E-04 |
1.50E-02 |
ABCG8, APOC3, APOA1, APOA4, APOB |
|||||
|
GO:0006813~potassium ion transport |
7 |
1.83E-04 |
2.02E-02 |
ATP4B, ABCC8, KCNJ13, KCNMB2, KCNMB3, KCNA5, ATP1A2 |
|||||
|
GO:0071805~potassium ion transmembrane transport |
8 |
2.31E-04 |
2.12E-02 |
KCNE2, KCNB1, ABCC8, KCNH8, KCNMB2, KCNMB3, KCNA5, TRPM5 |
|||||
|
GO:0007267~cell-cell signaling |
11 |
2.68E-04 |
2.12E-02 |
IL11, BMP3, VIPR2, SST, CD80, ASIP, CCL3, TNFSF9, ADRB2, INHA, MLN |
|||||
|
GO:0007586~digestion |
6 |
4.39E-04 |
2.90E-02 |
CCKAR, CCKBR, SST, GKN1, MLNR, PGA3 |
|||||
|
GO:0042632~cholesterol homeostasis |
6 |
4.72E-04 |
2.90E-02 |
ABCG8, MTTP, APOC3, APOA1, APOA4, APOB |
|||||
|
GO:0042157~lipoprotein metabolic process |
5 |
5.72E-04 |
3.14E-02 |
ALB, APOC3, APOA1, APOA4, APOB |
|||||
|
GO:0001508~action potential |
4 |
6.24E-04 |
3.14E-02 |
KCNB1, KCNMB2, KCNMB3, AKAP6 |
|||||
|
KEGG Pathway |
hsa04977: Vitamin digestion and absorption |
5 |
1.34E-04 |
1.03E-02 |
RBP2, APOA1, APOA4, LRAT, APOB |
||||
|
hsa04974: Protein digestion and absorption |
7 |
7.19E-04 |
1.85E-02 |
COL11A1, KCNJ13, COL4A6, COL10A1, ATP1A2, PGA3, SLC8A2 |
|||||
|
hsa04975: Fat digestion and absorption |
5 |
7.93E-06 |
6.42E-04 |
ABCG8, MTTP, APOA1, APOA4, APOB |
|||||
|
hsa04060: Cytokine-cytokine receptor interaction |
9 |
6.15E-05 |
4.98E-03 |
IL11, CXCL8, CSF2, TNFRSF13B, OSM, LIF, CCL3, TNFRSF17, TNFSF9 |
|||||
|
hsa04971: Gastric acid secretion |
5 |
7.64E-05 |
6.19E-03 |
ATP4B, KCNE2, CCKBR, SST, ATP1A2 |
|||||
hsa00830: Retinol metabolism |
4 |
2.73E-04 |
2.21E-02 |
ADH4, CYP2B6, RDH12, LRAT |
||||||
|
hsa04022: cGMP-PKG signaling pathway |
6 |
2.80E-04 |
2.27E-02 |
KCNMB2, KCNMB3, CACNA1D, ATP1A2, ADRB2, SLC8A2 |
|||||
|
hsa05202: Transcriptional misregulation in cancer |
6 |
3.41E-04 |
2.76E-02 |
CXCL8, CSF2, ZBTB16, MMP3, FCGR1A, RXRG |
|||||
|
hsa00500: Starch and sucrose metabolism |
3 |
3.84E-04 |
3.11E-02 |
TREH, PYGM, GBA3 |
|||||
|
hsa04020: Calcium signaling pathway |
6 |
4.32E-04 |
3.50E-02 |
CCKAR, CCKBR, CACNA1D, ADRB2, RYR3, SLC8A2 |
|||||
|
hsa04512: ECM-receptor interaction |
4 |
5.69E-04 |
4.61E-02 |
COL11A1, CHAD, COL4A6, SPP1 |
|||||
Table 2
Information about the nine-DEIIR signature
Symbol |
Type |
Multi-variate Cox regression analysis |
LASSO coef |
||
HR |
95%CI |
P value |
|||
LINC00843 |
lncRNA |
0.618 |
0.369-0.933 |
6.59E-02 |
-0.3194 |
ADH4 |
mRNA |
1.166 |
1.034-1.314 |
1.22E-02 |
0.1550 |
ANGPT2 |
1.386 |
1.103-1.742 |
5.05E-03 |
0.2739 |
|
APOA1 |
0.799 |
0.889-0.979 |
4.67E-02 |
-0.0269 |
|
ASCL2 |
0.894 |
0.814-0.981 |
1.85E-02 |
-0.1004 |
|
GFRA1 |
1.217 |
1.034-1.435 |
1.86E-02 |
0.1002 |
|
KIAA1549L |
1.433 |
1.093-1.878 |
9.21E-03 |
0.3762 |
|
MTTP |
1.119 |
1.005-1.323 |
4.19E-02 |
0.1256 |
|
PROC |
1.265 |
1.016-1.574 |
3.57E-02 |
0.1655 |
Table 3
The independent prognostic clinical factors according to univariate and multivariate Cox regression analyses
Clinical characteristics |
TCGA (N=351) |
Uni-variables cox |
Multi-variables cox |
||||
HR |
95%CI |
P |
HR |
95%CI |
P |
||
Age (years, mean ± SD) |
65.49 ± 10.58 |
1.022 |
1.006-1.039 |
6.91E-03 |
1.033 |
1.010-1.056 |
4.165E-03 |
Gender (Male/Female) |
226/125 |
1.271 |
0.891-1.813 |
1.84E-01 |
- |
- |
- |
Pathologic_M (M0/M1/-) |
313/23/15 |
2.063 |
1.164-3.659 |
1.13E-02 |
1.138 |
0.423-3.062 |
7.986E-01 |
Pathologic_N (N0/N1/N2/N3/-) |
104/94/73/70/10 |
1.329 |
1.144-1.544 |
1.71E-04 |
1.157 |
0.875-1.530 |
3.061E-01 |
Pathologic_T (T1/T2/T3/T4/-) |
16/75/163/93/4 |
1.318 |
1.070-1.624 |
8.09E-03 |
1.472 |
1.034-2.095 |
3.187E-02 |
Pathologic_stage (I/II/III/IV/-) |
47/109/147/35/13 |
1.551 |
1.264-1.903 |
1.68E-05 |
1.144 |
0.699-1.872 |
5.902E-01 |
Neoplasm histologic grade (G1/G2/G3/-) |
9/127/206/9 |
1.366 |
0.991-1.883 |
5.63E-02 |
- |
- |
- |
Radiation therapy (Yes/No/-) |
62/265/24 |
0.429 |
0.262-0.703 |
2.29E-04 |
0.422 |
0.218-0.817 |
1.052E-02 |
H.pylori infection(Yes/No/-) |
18/142/191 |
0.642 |
0.276-1.494 |
3.00E-01 |
- |
- |
- |
Barretts esophagus (Yes/No/-) |
14/191/146 |
0.941 |
0.344-2.573 |
9.05E-01 |
- |
- |
- |
Recurrence (Yes/No/-) |
59/229/63 |
2.517 |
1.681-3.767 |
3.44E-06 |
2.348 |
1.439-3.832 |
6.330E-04 |
PS model status (High/Low) |
175/176 |
2.23 |
1.583-3.142 |
2.56E-06 |
2.059 |
1.284-3.301 |
2.722E-03 |
Death (Yes/No) |
144/207 |
- |
- |
- |
- |
- |
- |
Overall survival time (months, mean ± SD) |
20.37 ± 18.34 |
- |
- |
- |
- |
- |
- |
HR hazard ratio, CI confidence interval, SD standard deviation |
Table 4
Association of ADH4 expression with clinicopathological features of GC patients
Variables |
Cases (n = 59) |
ADH4 expression |
P value |
|
Low (n = 30) |
High (n = 29) |
(chi-square test) |
||
Sex |
|
|
|
0.506 |
Male |
32 |
15 |
17 |
|
Female |
27 |
15 |
12 |
|
Age |
|
|
|
0.083 |
< 60 |
25 |
16 |
9 |
|
≥ 60 |
34 |
14 |
20 |
|
Tumor size (cm) |
|
|
|
0.145 |
< 5 |
38 |
22 |
16 |
|
≥ 5 |
21 |
8 |
13 |
|
TNM stage |
|
|
|
0.041* |
I + II |
40 |
24 |
16 |
|
III + IV |
19 |
6 |
13 |
|
Lymph node metastasis |
|
|
|
0.005* |
Yes |
23 |
17 |
6 |
|
No |
36 |
13 |
23 |
|
Differentiation |
|
|
|
0.195 |
Well/moderate |
45 |
25 |
20 |
|
Poor |
14 |
5 |
9 |
|
Note: *Statistically significant; Abbreviations: GC, gastric cancer; TNM, tumor‑node‑metastasis |
Table 5
Univariate and multivariate analysis for overall survival in GC patients
|
Univariate analysis |
Multivariate analysis |
||
Characteristics |
HR (95% CI) |
P value |
HR (95% CI) |
P value |
Sex |
0.782 (0.489-1.252) |
0.305 |
NA |
NA |
Age (years) |
0.864 (0.588-1.269) |
0.455 |
NA |
NA |
Tumor size (cm) |
0.749 (0.508-1.105) |
0.006 |
NA |
NA |
TNM stage |
1.994 (1.398-2.846) |
1.12E-04 |
1.904 (1.062-3.412) |
4.58E-02 |
Lymph node metastasis |
1.438 (1.181-1.751) |
2.51E-04 |
1.025 (0.751-1.400) |
8.75E-01 |
Differentiation |
1.425 (0.719-2.823) |
3.07E-01 |
NA |
NA |
ADH4 expression |
2.456 (1.651-3.654) |
4.71E-03 |
2.205 (1.415-3.435) |
4.73E-02 |
Abbreviations: GC, gastric cancer; HR: hazard ratio; CI: confidence interval; NA, not analyzed |
Table 6
Association of ANGPT2 expression with clinicopathological features of GC patients
Variables |
Cases (n = 59) |
ANGPT2 expression |
P value |
|
High (n = 30) |
Low (n = 29) |
(chi-square test) |
||
Sex |
|
|
|
0.235 |
Male |
32 |
14 |
18 |
|
Female |
27 |
16 |
11 |
|
Age |
|
|
|
0.367 |
< 60 |
25 |
11 |
14 |
|
≥ 60 |
34 |
19 |
15 |
|
Tumor size (cm) |
|
|
|
0.207 |
< 5 |
38 |
17 |
21 |
|
≥ 5 |
21 |
13 |
8 |
|
TNM stage |
|
|
|
0.003* |
I + II |
40 |
15 |
25 |
|
III + IV |
19 |
15 |
4 |
|
Lymph node metastasis |
|
|
|
0.486 |
Yes |
23 |
13 |
10 |
|
No |
36 |
17 |
19 |
|
Differentiation |
|
|
|
0.942 |
Well/moderate |
45 |
23 |
22 |
|
Poor |
14 |
7 |
7 |
|
Note: *Statistically significant; Abbreviations: GC, gastric cancer; TNM, tumor‑node‑metastasis |
Table 7
Univariate and multivariate analysis for overall survival in GC patients
|
Univariate analysis |
Multivariate analysis |
||
Characteristics |
HR (95% CI) |
P value |
HR (95% CI) |
P value |
Sex |
0.989 (0.971-1.006) |
1.937E-01 |
NA |
NA |
Age (years) |
0.909 (0.537-1.540) |
7.226E-01 |
NA |
NA |
Tumor size (cm) |
1.594 (1.007-2.523) |
4.370E-02 |
1.225 (0.744-2.019) |
4.25E-01 |
TNM stage |
2.227 (1.431-3.467) |
2.620E-03 |
1.730 (1.524-3.141) |
7.19E-03 |
Lymph node metastasis |
1.811 (1.394-2.353) |
6.230E-03 |
1.052 (0.309-1.371) |
2.59E-01 |
Differentiation |
0.885 (0.503-1.559) |
6.710E-01 |
NA |
NA |
ANGPT2 expression |
0.830 (1.746-4.589) |
9.86E-03 |
2.508 (1.472-4.276) |
7.27E-02 |
Abbreviations: GC, gastric cancer; HR: hazard ratio; CI: confidence interval; NA, not analyzed |