The relative expression level of PHGDH in endometrial cancer
The flow chart of this study is shown in Figure1. First, we summarize the expression levels of PHGDH mRNA in pan-cancer (33 cancers) from the TCGA database (Figure 2). Among them, the expression level of PHGDH in tumor tissues was significantly higher than that in normal tissues, including endometrial cancer tissues (P<0.001).
Next, we explored the relationship between PHGDH expression and clinicopathological features of endometrial cancer. Next, to investigate the role of PHGDH in endometrial cancer progression, we explored the relationship between PHGDH expression and clinicopathological characteristics of endometrial cancer. The expression pattern of PHGDH gene in 552 cases of endometrial cancer tissues and 35 cases of normal endometrial tissues was predicted using the TCGA database (Figure 3A). The results showed that the expression level of PHGDH gene in primary endometrial cancer tissues was significantly higher than that in normal endometrial tissues (P<0.001). The expression of PHGDH in normal tissues adjacent to cancer and UCEC tissues were further compared. PHGDH was highly expressed in UCEC tissues compared with paraneoplastic tissues (P<0.001) (Figure 3B). In addition, PHGDH expression was significantly upregulated in 23 cases of endometrial cancer tissues compared with paired paraneoplastic tissues (P<0.01) (Figure 3C).
To evaluate the diagnostic value of PHGDH expression levels in normal GTEx/UCEC tissues and paraneoplastic/UCEC tissues, we also plotted the receiver operating characteristic (ROC) curves separately. The area under the curve (AUC value) of PHGDH expression levels in normal GTEx/UCEC tissues was 0.745 (CI=0.656-0.834), and the area under the curve (AUC) value for the expression level in paraneoplastic/UCEC tissues was 0.787 (CI=0.732-0.841), suggesting a high diagnostic potential (Figure 3D, E).
The protein expression of PHGDH was analyzed using the UALCAN and HPA databases. The protein expression level of PHGDH was also significantly upregulated in endometrial cancer tissues compared to normal endometrial tissues (Figure 3F, G), indicating that PHGDH protein and mRNA had similar expression profiles in different databases.
Correlation between PHGDH expression and clinical features
We then analyzed the relationship between PHGDH expression and clinical characteristics in 552 patients with endometrial cancer in the TCGA-UCEC dataset (Table 1). Patients were divided into PHGDH high and low expression groups according to the mean value of PHGDH expression. The relationship between PHGDH expression and clinical characteristics was assessed using Wilcoxon signed rank test and logistic regression analysis. The results showed statistically significant differences in PHGDH expression between stage I and stage II-IV tumors (P=0.014) (Figure 4A) and between histological grade G1/2 and G3 (P<0.001) (Figure 4C). PHGDH expression was also higher in serous endometrial carcinoma than in endometrioid carcinoma (P<0.001) (Figure 4B).
Univariate logistic regression analysis showed the correlation between PHGDH expression and clinicopathological features of endometrial cancer (Table 2). A comparison of baseline information between the high and low PHGDH expression groups showed that PHGDH expression was associated with the clinical stage (Stage I vs. Stage II-IV, OR=1.448, P=0.036), primary therapy outcome (CR vs. PD&SD&PR, OR=0.433, P=0.021), histological type (Serous vs. Endometrioid, OR=2.716, P<0.001), and histological grade (G1&G2 vs. G3, OR=2.546, P<0.001) were significantly associated.
The independent diagnostic value of PHGDH expression in endometrial carcinoma
Survival analysis demonstrated that high PHGDH expression was correlated with poor OS (P<0.001, HR=2.21) as well as poor DFS (P = 0.019, HR=2.20) (Figure 5A, B). Univariate Cox regression analysis showed that high PHGDH expression was significantly correlated with poor OS (HR=1.698, 95% CI=1.112-2.592) and DFI (HR=1.614, 95% CI=1.132-2.300). Moreover, multivariate regression analysis and visualized forest plots further confirmed that PHGDH expression was an independent prognostic factor for DFI in patients with endometrial carcinoma (HR=1.614, 95% CI=1.027-2.474, P = 0.038) (Table 3, Figure 6).
Subsequently, nomogram models predicting the survival of patients with endometrial cancer were constructed using age, clinical stage, histological grade, tumor invasion, histological type, and PHGDH levels (Figure 7A). The calibration curves provided ideal nomogram predictions for clinical outcomes at 1, 3, and 5 years (Figure 7B). The above results suggest that PHGDH could be a valuable biomarker for predicting survival in patients with endometrial cancer.
PHGDH expression-related signaling pathway based on GSEA and KEGG
GO, KEGG pathway analysis, and GSEA were used to identify possible cellular mechanisms for the role of PHGDH in endometrial cancer. As shown in Figure 8A, KEGG enrichment analysis showed that Neuroactive ligand-receptor interaction (hsa04080) was the most relevant pathway to the PHGDH high expression group. At the same time, Cytokine-cytokine receptor interaction (hsa04060) and Staphylococcus aureus infection (hsa05150) were also associated with the role of PHGDH in endometrial cancer. The GO pathways related to the position of PHGDH in endometrial cancer include receptor ligand activity (GO:0048018), external side of plasma membrane (GO:0009897), humoral immune response (GO:0006959), etc.
Meanwhile, the GSEA analysis of estrogen response, glycolysis, hypoxia, K-Ras signaling, epithelial mesenchymal transition, mTOR signaling, etc., were the most abundantly differential pathways in the PHGDH high expression phenotype (Figure 8B, C).
PHGDH Expression Is Associated With Immune Signatures in UCEC
Related studies have shown that tumor immune cell infiltration can be an independent predictor of tumor anterior lymph node status and prognosis[28]. Here, we used TIMER to analyze whether PHGDH expression was associated with the level of immune infiltration in UCEC. As shown in Figure 9A, PHGDH expression was negatively correlated with the levels of macrophage (p=0.004), dendritic cells (p<0.001), B cells (p=0.043), and CD8+ T cells (p<0.001). These results suggest a crucial role of PHGDH in the immune infiltration of UCEC. In addition, a significant correlation was found between PHGDH CNV and the level of infiltration of CD8+ T cells and neutrophils cells (Figure 9C). PHGDH expression was also significantly associated with the immune markers CTLA4 (p<0.001) and PDCD1 (PD-1, Programmed cell death protein-1) (p<0.001) (Figure 9B).
In addition, we sought to analyze the differential expression of 24 immune cells between different PHGDH expression groups by the CIBERSORT package to determine whether there are differences in the tumor immune microenvironment between high and low PHGDH expression levels in UCEC. The CIBERSORT results showed that the expression of T cells, CD8+ T cells, Neutrophils, Mastcells, Cytotoxic cells, Eosinophils, iDC, NK CD56 bright cells, NK CD 56 dim cells, pDC, Tcm, TFH, Th17 cells, and Th2 cells differed more between the high and low PHGDH expression groups. Among them, Th2 cells were increased in the PHGDH high expression group compared to the low expression group, while other cells were decreased in expression (Figure 9D). These results suggest that PHGDH expression in UCEC is associated with immune cell infiltration in different ways.
Drug Sensitivity Analysis of PHGDH
Due to the possible drug resistance role of PHGDH in tumors, we further investigated the analysis of potential correlations between drug sensitivity and PHGDH expression using the CellMinerTM database. Our results showed that PHGDH expression was negatively correlated with the sensitivity of Dasatinib (p=0.002), Pluripotin (p=0.004), BMS-690514 (p=0.004), 6-Thioguanine (p=0.005) and BMS-599626 (p=0.007)(Figure 10A). Furthermore, there was a significant difference in the expression of Dasatinib (p=0.007), 6-Thioguanine (p=0.007), and BMS-599626 (p=0.002) in the high versus low PHGDH expression groups (Figure 10B).