3.1 FSTL1 expression and potential prognosis in GC
TIMER and Oncomine databases were used to assess the differential expression of FSTL1 based on cancer type. The majority of data sets from the TIMER and Oncomine database showed that FSTL1 was highly expressed in most malignancies, especially in GC (Figure 1A, B). In representative Chen Gastric data sets from the Oncomine database, FSTL1 expression was significantly high expression in GC tissues compared with normal tissues, and was elevated in multiple pathological typing, including adenocarcinoma, intestinal carcinoma, diffuse carcinoma and mixed carcinoma (Figure 1C, D). Same result was confirmed in GEPIA database, and the expression of FSTL1 increases with the stage of GC (Figure 1E, F). In addition, we further confirmed that FSTL1 was highly expressed in GC through RT-qPCR and Western blot (Figure 2A, F). Moreover, the mRNA expression of FSTL1 in the Nx group (P=0.037) and stage III group (P=0.018) was significantly higher than N0 group and the I/II group in GC patients, respectively (Figure 2C, D). But there was no significant change in FSTL1 expression in stage T grouping and differentiation grouping (Figure 2B, E).
Meanwhile, KM-plotter and GEPIA database survival analysis indicated that highly FSTL1 expression indicated poorly prognosis in GC patients (Supplementary Figure 1A, B). To confirm the predicted prognostic value of FSTL1 in GC, FSTL1 were analyzed in random selected 240 primary GC tissues and paired adjacent non-tumor tissues by immunohistochemistry. The staining of FSTL1 protein ranged from weak to strong (Figure 2H), and the result showed that FSTL1 positive staining was increased in GC tissues than adjacent non-tumor tissues (Figure 2I). The positive expression rate of FSTL1 (43.3%, 104/240) in GC samples was significantly higher than that in adjacent non-tumor tissues (15%, 36/240) (P<0.001) (Figure 2G).
Kaplan-Meier analysis and log-rank test were performed to analyzed the prognostic value of FSTL1 expression in GC patients. The result showed that patients with positive FSTL1 expression (41.1±3.8 month) had worse overall survival than negative FSTL1 expression (70.9±3.1 month) (P<0.001) (Figure 2J).
3.2 Correlation of FSTL1 expression and clinical prognosis in GC with different clinicopathological features
To further understand the effect of FSTL1 in the prognosis of GC, we studied the relationship between FSTL1 expression and the clinicopathological features of these cancers using the Kaplan-Meier plotter database and immunohistochemistry assay.
In Kaplan-Meier plotter database, the expression of FSTL1 has a significant correlation with the prognosis of GC patients with various clinicopathological characteristics. It is worth noting that increased FSTL1 expression was associated with worse OS and PFS in stage III-IV and stage N1-3, but not in stage I/II and stage N0 GC patients (Table 1). The result means that FSTL1 expression level can impact the prognosis in the advanced and lymph node metastasis GC patients, which consistent with the increased expression of FSTL1 in III group and Nx group compared with I/II group and N0 group in Figure 2C, D.
Next, correlation between FSTL1 expression and clinicopathological was been investigated (Table 2). The result showed that expression level of FSTL1 was correlated with tumor size (P<0.001), lymph node metastasis (P<0.001), and tumor-node-metastasis (TNM) stage (P=0.001). No significant correlations between FSTL1 expression and ages (P=0.48), gender (P=0.519), depth of invasion (P=0.147), differentiation (P=0.151), nerve invasion (P=0.809) and vascular invasion (P=0.317) were detected. The results are consistent with the results in the KM-Plotter database.
Univariate and multivariate analyses were conducted to investigate the independent prognostic factors of GC patients (Table 3). Univariate analysis indicated that tumor size (P<0.001), depth of invasion (P<0.001), lymph node metastasis (P<0.001), TNM stage (P<0.001), vascular invasion (P=0.003) and FSTL1 expression (P<0.001) was significantly correlated with OS of GC patients. These factors were subjected to multivariate analysis, which indicated that tumor size (P=0.006), depth of invasion (P=0.007), lymph node metastasis (P<0.001), TNM stage (P<0.001) and FSTL1 expression (P=0.002) were independent prognostic factors in GC patients.
3.3 Protein-protein interaction of FSTL1 network analysis
Genes co-expressed with FSTL1 were analyzed by LinkedOmics database, which analyzed mRNA sequencing data of 415 GC patients from TCGA. All the genes associated with FSTL1 GC was showed in Figure 3A. And Figure 3B revealed the most 50 significant genes positively correlated with FSTL1 in GC. The 50 genes differentially expressed associated with highly FSTL1 expression were primarily enriched in extracellular region/exosome/space/matrix/matrix organization, cell adhesion (Figure 3C). KEGG pathway analysis found enrichment in Pathway in cancer, Focal adhesion, PI3K-Akt signaling pathway, ECM-receptor interaction, which indicated that elevation of FSTL1 was significantly correlated with tumor development and immune response (Figure 3D).
In addition, we applied GSVA and clustering on a set of immune-specific signatures to further validate the enrichment of FSTL1 expression in the immune-related pathway. As FSTL1 expression increased, most immune-related pathways were activated in GC, CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION, INNATE_IMMUNE_SYSTEM, NATURAL_KILLER_CELL_MEDIATED_CYTOTOXICITY, and others in STAD (Figure 4). Hence, FSTL1 plays a crucial role in immune cell infiltration and tumor-immune system interactions in GC.
3.4 Correlation between FSTL1 and tumor environment in GC
TILs, which are essential components of the TME, have been shown to reflect the host antitumor immune response and are prognostic factors for cancer [26, 27]. As mentioned, GO analysis revealed that FSTL1 has functions associated with extracellular region/exosome/space/matrix/matrix organization, cell adhesion, while KEGG analysis indicated that FSTL1 involved in the ECM-receptor interaction pathway, PI3K-Akt signaling pathway. It is well-known that remodeling the extracellular matrix (ECM), disordered PI3K-Akt signaling pathway can exert a great influence on the TME[28, 29]. Thus, we analyzed the correlation between FSTL1 expression and TILs in GC using the TIMER database. The result showed that FSTL1 expression was found to be correlated with high immune infiltration of CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells in GC (Figure 5).
Immune cells are frequently identified based on the expression of cell surface receptors and intracellular markers. To further assess the relationship between FSTL1 and each tumor-infiltrating immune cell, we analyzed the correlation between FSTL1 expression and immune infiltrating cells, including CD8+ T cells, T cells (general), B cells, neutrophils, NK cells and dendritic cells (DCs), monocytes, TAMs, and M1/M2 macrophages, in GC using the TIMER and GEPIA databases. In addition, the functional T cells were also been investigated, including T helper type 1 (Th1), Th2, Th17, follicular helper T cell (Tfh), and regulatory T (Treg) cells, as well as exhausted T cells. Even if the correlation was adjusted based on tumor purity, the majority of tumor-infiltrating immune cell markers were positively correlated with FSTL1 expression in GC (Table 4).
3.5 Correlation between FSTL1 expression and macrophage polarization in GC
Studies have shown that polarization from the antitumor M1 (classically activated) macrophage to protumor M2 (alternatively activated) macrophage phenotype is correlated with tumor development [30, 31]. Interesting, we found that the markers for monocytes, TAMs, and M2 macrophages showed moderate-to-strong correlations with FSTL1 expression, while the M1 macrophage markers were weakly correlated with FSTL1 expression (Figure 6A-D, Table 4). To confirm the results with the TIMER database, the correlation between FSTL1 expression and the cell markers for monocytes, TAMs, M1 macrophages, and M2 macrophages in GC was assessed using GEPIA database. Findings from the GEPIA database were consistent with the TIMER database (Supplementary Table 1). Therefore, we using CIBERSORT method to further explored whether FSTL1 expression was correlated with the polarization of macrophages in GC. The results showed that high FSTL1 expression had a higher ratio of M2 macrophage (Figure 6E-G).