GC patient baseline clinical characteristics
The clinical characteristics, including gender, age, race, TNM stage, histologic grade, primary therapy outcome, pathologic stage , and H. pylori infection, were collected (Table 1). In this study, 134 female and 241 male patients were analyzed in total, including 18 H. pylori infection patients and 145 non-H. pylori infection patients. Concerning the GC stage, 53(15%), 111(31.5%), 150(42.7%), and 38(10.8%) patients were in stages I, II, III, and IV, respectively.
High INHBA expression in cancer patients
First, we assessed INHBA expression in pan-cancer data from The Cancer Genome Atlas(TCGA) and Genotype-Tissue Expression(GTEx). The analysis showed that INHBA expression was higher and lower in 17 and 7 tumors, respectively(Figure 1A). In order to clarify the INHBA expression differences between GC and healthy tissues, the INHBA expression level in 375 and 32 GC and adjacent GC tissues were examined, respectively, by scatter plot and a potentially high INHBA expression was verified in the GC tissues(P < 0.001, Figure 1B). In addition, the INHBA expression in 32 GC tissues and their matched adjacent tissues were also analyzed. The results indicated that GC tissues highly expressed INHBA(P < 0.001, Figure 1C). Moreover, the INHBA expression of healthy samples from GTEx combined with data from the TCGA database and GC samples from the TCGA database were compared, showing the same results as the aforementioned analysis, indicating that INHBA was significantly overexpressed in the GC samples (P < 0.001, Figure 1D).
Correlation between INHBA expression and clinical features
The correlation between INHBA expression and clinical features in patients with GC was shown in Table 2. The high INHBA expression in GC patients significantly associated with the T stage (P < 0.01), histological type (P < 0.01), histologic grade (P < 0.01), OS event (P < 0.01), pathologic stage (P < 0.01, Figure 2). The logistic regression analysis indicated that the expression of INHBA definitely as a dependent variable was associated with poor prognostic clinical characteristics. In GC patients, the high INHBA expression was significantly associated with the T stage(T1 vs. T2&T3&T4: P = 0.003), histological type(Diffuse Type &Mucinous Type & Signet Ring Type &Not Otherwise Specified vs. Tubular Type &Papillary Type: P < 0.001).
High INHBA expression with poor GC patient prognosis
In order to identify the prognostic value of INHBA in GC, we performed Kaplan–Meier curves analysis and showed that high expression of INHBA correlates with poor prognosis (P = 0.037), similar to pathologic stage Ⅱ (P = 0.014), PR&CR (P = 0.001), T4 (P = 0.03), N2 (P = 0.045),M0 (P = 0.019), Diffuse Type &Mucinous Type &Signet Ring Type &Not Otherwise Specified (P = 0.021), and R0 (P = 0.018) in the subgroup analysis(Figure 3). Univariate Cox analysis confirmed that high INHBA expression was significantly associated with poor OS (P = 0.037). Interestingly, multivariate Cox analysis indicated that the INHBA expression was an independent risk factor for OS in GC patients(P = 0.004, Table 3). Therefore, a nomogram was constructed based on the result of the Cox multivariate analysis to predict the different time periods survival probability of the GC patients by combining the INHBA expression levels with independent clinical variables(Figure 4).
INHBA as a potential new diagnostic biomarker in GC
We conducted a ROC curve analysis to evaluate the diagnostic value of INHBA in GC. The INHBA area under the curve(AUC) was 0.961 (Figure 5A), indicating a high INHBA diagnostic value. The subgroup analysis demonstrated the INHBA gene expression diagnostic value in different clinical features such as T1/T2, T3/T4, tubular type/mucinous type, Barretts esophagus and stage I/II (Figure 5B–5F).
Functional enrichment analyses of INHBA and associated genes
To predict the function of INHBA, including associated pathways, we performed a correlation analysis between INHBA and other GC-related genes using TCGA data and displayed the results as heatmaps (Figure 6). The top 300 genes associated with INHBA were derived and analyzed for enrichment analysis. The GO analysis revealed that INHBA was associated with extracellular matrix organization, response to growth factor, cell-substrate adhesion, and negative regulation of cell differentiation. In addition, the KEGG pathway analysis indicated that protein digestion and absorption, proteoglycans in cancer, Wnt, Hippo as well as p53 signaling pathways comprised the top 300 enriched genes and were involved in crosstalk (Table 4). Meanwhile, INHBA-associated Reactome pathways were screened by GSEA, revealing that IL4 and IL13 signaling, signaling by PDGF, collagen formation and glycosaminoglycan metabolism were significantly enriched (Figure 7). These results suggest that INHBA could be associated with multiple malignancy-related pathways in GC and might promote GC development by altering the cancer microenvironment.
Correlation between INHBA expression and immune cells infiltration in GC
Different immune infiltration levels in the tumor microenvironment were significantly associated with overall patients survival. The above-described findings suggested that INHBA was an independent risk factor and correlated with OS in GC. Therefore, investigating the relationship between INHBA expression and immune infiltration would be reasonable. We used the TIMER database to analyze the correlation between INHBA and the immune infiltration level. The results showed that INHBA was significantly associated with tumor purity, as well as B cells, macrophages, or neutrophil and dendritic cells (Figure 8A). Furthermore, we also performed Kaplan–Meier analysis to assess the association between INHBA expression and immune cell infiltration in GC. As figure shows, except for the INHBA expression, macrophage infiltration also correlated with GC prognosis (Figure 8B). This indicated that INHBA has a regulatory function on immune cell infiltration in GC, especially on macrophage infiltration.
Association between INHBA expression and immune markers
To further pursue the interplay between INHBA expression and immune cell infiltration in GC, we used the TIMER and GEPIA databases to explore the relationship between INHBA expression and several immune cell markers. Briefly, these included B cells, CD8+ T cells, macrophages, monocytes, dendritic cells, natural killer cells, tumor-associated macrophages, neutrophils, and T cell subsets, such as Th1, Th2, Th17, Treg, Th9, Th22, Tfh, and T cell exhaustion. In TIMER, 35 immune cell markers were significantly associated with INHBA expression before and after tumor purity correction. Most of these markers belong to subsets of T cells, such as Th1/Th2 or Th17/Treg, and macrophages, such as tumor-associated macrophages (TAMs) or M1/M2 macrophages, monocytes, neutrophils and DCs (Table 5 and Figure 9). This predicted that INHBA plays a key role in the tumor microenvironment to affect immune infiltration.
Since macrophages were the most significant immune cells infiltrating in GC (Figure 9), and the macrophage markers clearly correlated with INHBA expression (Table 5), we further pursued the correlation between INHBA expression and macrophage-associated markers in GEPIA(Table 6). As the results show in TIMER, the correlation could be observed between INHBA expression and the markers of M1/M2 macrophages, monocytes, and TAMs. All of this suggested that INHBA might drive tumor-associated macrophage polarization in GC but it needs further experiments to explore the underlying mechanism.