Effect of the differential expression of EPC1 on the prognosis and clinical outcomes of patients with HNSCC
In the TISIDB platform, Spearman correlation analysis was performed to study the associations of EPC1 expression with HNSCC subtypes. EPC1 expression levels were not equal or completely equal between different subtypes (Fig. 1A). The Kaplan-Meier plotter platform was used to analyze survival and EPC1 expression (Fig. 1B). The median survival time for the low-EPC1 expression group was 33.10 months, and the median survival time for the high-EPC1 expression group was 61.27 months; the difference was statistically significant (HR < 1, P < 0.01), suggesting that EPC1 is a protective factor against HNSCC and that patients with high EPC1 expression have a better prognosis.
Based on TCGA samples and the UALCAN website, EPC1 expression in HPV-positive HNSCC tumors was not only significantly higher than that in paracancerous tissues (P < 0.01) but also significantly higher than that in HPV-negative HNSCC samples (P < 0.01) (Fig. 1C). Using HPV-positive HNSCC samples, we further explored the relationship between EPC1 expression and patient prognosis. The results showed that patient prognosis was significantly better with higher EPC1 expression (Fig. 1D) (P < 0.01). However, no significant effect of EPC1 expression on patient prognosis was found when analyzing HPV-negative HNSCC samples. In addition, compared with that in TP53-mutated HNSCC, EPC1 expression in wild-type TP53 HNSCC was significantly higher (P < 0.05) (Fig. 1E). In HPV-positive HNSCC samples, the expression of wild-type TP53 EPC1 was relatively higher. In HPV-negative HNSCC samples, no significant difference in EPC1 expression was identified between wild-type TP53 and TP53-mutated samples (Fig. 1F). Therefore, we hypothesize that EPC1 and TP53 may be correlated.
Screening and functional prediction of genes associated with the differential expression of EPC1 in HNSCC
LinkedOmics was used to screen genes that were significantly positively correlated with the EPC1 gene and genes that were significantly negatively correlated with the EPC1 gene. A total of 20,164 related genes were obtained, including 8208 genes with negative correlations and 11,956 genes with positive correlations, and volcano plots were drawn (Fig. 2A). The notable positively correlated genes included ZNF41, NR2C2, and CEP350. The notable negatively correlated genes included MRPL28, C14orf156 and TMEM280. After obtaining the gene dataset, we further performed GSEA. The rank criteria were a P-value < 0.05, an FDR ≤ 0.05, a minimum number of genes (size) = 5, and simulations = 500. KEGG pathway analysis was conducted. We selected “Redundancy reduction: Weighted set cover” and screened 5 positively correlated KEGG pathways (labeled blue, Fig. 2B): phosphatidylinositol signaling system, cell adhesion molecules, cGMP-PKG signaling pathway, Rap1 signaling pathway, and pathways in cancer. We also screened 5 negatively correlated KEGG pathways (labeled orange, Fig. 2B): purine metabolism, thermogenesis, spliceosome, proteasome, and ribosome. Using pathways in cancer as an example, a total of 190 genes were enriched (enrichment score = 0.58; normalized enrichment score = 1.56; P < 0.01), and the difference was statistically significant (Fig. 2C). The above steps were repeated for the GO analysis (biological process). Five positively correlated biological processes were screened (labeled blue, Fig. 2D): protein autophosphorylation, covalent chromatin modification, regulation of GTPase activity, positive regulation of cell motility, and Ras protein signal transduction. Additionally, 5 negatively correlated biological processes were screened (labeled orange, Fig. 2D): protein folding, nucleoside triphosphate metabolic process, protein targeting, ribonucleoprotein complex biogenesis, and mitochondrial gene expression. The above analyses showed that the differential expression of EPC1 at the gene level is related to cancer pathways. EPC1 may regulate the metastasis and spread of tumor cells by positively enhancing the function of cell adhesion molecules, regulating cell migration, and reducing the expression of mitochondrial genes, thereby improving patient prognosis.
Construction of a lncRNA-miRNA-mRNA network based on the differential expression of EPC1 in HNSCC samples
A total of 52 EPC1 gene-related miRNAs that were experimentally verified were identified using the TarBase V. 8 database, and 201 miRNAs differentially expressed in HNSCC were identified using the YM500v2 platform. The two datasets were intersected, resulting in 10 overlapping miRNAs (Fig. 3A). The LncBase v.2 database was used to predict the related lncRNAs for the 10 miRNAs, and the Lnc2Cancer database was used to obtain experimentally verified HNSCC-related lncRNAs; 27 lncRNAs differentially expressed between head and neck cancer and HNSCC were identified, and 6 lncRNAs overlapped between the 2 (Fig. 3B). From 6 lncRNAs and 10 miRNAs whose associations had been experimentally verified, 3 lncRNAs were screened (ENSG00000130600 (H19), ENSG00000234741 (GAS5), and ENSG00000205592 (MUC19)), and 7 miRNAs were screened (hsa-miR-26a-5p, hsa-miR-26b-5p, hsa-miR-454-3p, hsa-miR-130b-3p, hsa-miR-301a-3p, hsa-miR-182-5p, and hsa-miR-101-3p). Highcharts software was used to construct lncRNA-miRNA-mRNA network diagrams, with the thickness of the line widths indicating the degree of possible association between nodes (Fig. 3C).
Screening and pathway enrichment of proteins related to differential EPC1 expression in HNSCC
LinkedOmics was used to screen 13 genes that were positively related to EPC1 gene expression (Fig. 4A) and 7 proteins that were negatively related to EPC1 expression (Fig. 4B), all satisfying P < 0.05. Corresponding heat maps were drawn. Using the STRING database, an interaction network consisting of 21 proteins was constructed (Fig. 4C), and protein enrichment analysis was used to obtain the top 10 related pathways in terms of gene ratios (Fig. 4D). The PPI network suggested that proteins co-expressed with EPC1 may be involved in various cancer-related signaling pathways such as HPV infection, endocrine resistance, cell cycle disruption, plaque adhesion, breast cancer, gastric cancer, hepatocellular carcinoma, pancreatic cancer, and small cell lung cancer.
Differential expression of EPC1 among all HNSCC samples, HPV-positive HNSCC samples, and HPV-negative HNSCC samples and the association with immunity
Immune molecules associated with EPC1 expression were screened using the TISIDB platform. EPC1 expression in HNSCC samples was positively correlated with the immune enhancer TNFSF15 and two chemokine receptors CCR4 and CCR8, and high expression of these three proteins was predictive of the prognosis of patients with HPV-positive head and neck cancer; the difference was statistically significant (Fig. 5).
At the cellular level, the HNSCC samples, HPV-positive HNSCC samples, and HPV-negative HNSCC samples were subjected to immunological analysis using the Timer platform. Data from EPIC, CIBERSORT, XCELL, and other tool websites were integrated and used in the analysis. After adjusting for tumor purity, 12 EPC1-related immune cells were screened out from the HNSCC samples with the restriction condition that all three sample types reached |rho|>0.3 and P < 0.05 (Table 1). These results suggest that EPC1 may alter the tumor microenvironment of HNSCC by affecting the expression of immune-related molecules and immune proteins.
Table 1
Correlation between EPC1 expression and the level of immune infiltration
CELL TYPE | HNSCC (n = 522) | HNSCC-HPV− (n = 422) | HNSCC-HPV+ (n = 98) |
B cell | 0.460 | 0.440 | 0.442 |
Endothelial cell | 0.430 | 0.436 | 0.450 |
Mast cell | 0.441 | 0.324 | 0.345 |
Myeloid dendritic cell | 0.429 | 0.429 | 0.368 |
Neutrophil | 0.387 | 0.397 | 0.424 |
NK cell | 0.403 | 0.392 | 0.352 |
T cell CD4+ | 0.424 | 0.387 | 0.596 |
T cell CD4+ (nonregulatory) | 0.361 | 0.302 | 0.521 |
T cell CD4 + memory resting | 0.360 | 0.329 | 0.519 |
T cell CD4 + Th1 | -0.349 | -0.362 | -0.417 |
T cell CD8+ | -0.324 | -0.307 | -0.361 |
T cell regulatory (Tregs) | 0.547 | 0.495 | 0.682 |