EPO expression and its diagnostic value in HCC
We first compared EPO expression in TCGA datasets, and then correlated data in the GTEx dataset were added to supplement the healthy population for analysis to reduce bias. We found that EPO expression was overall dysregulated in human cancer (Fig. 1a, b). Notably, both results showed that EPO expression was downregulated in HCC (Fig. 1a, b, p < 0.001).
To further validate the results, we compared EPO expression in HCC tissues with that in both non-paired and paired normal tissues. As shown in Fig. 1c, d, both results validated the previous results, showing that EPO expression was downregulated in HCC. The diagnostic performance of EPO was examined by performing a ROC curve analysis. We also found that EPO was effective for diagnosing HCC as indicated by the AUC, which was 0.830 (p < 0.001) (Fig. 1e). With a cut-off value of 0.646, the sensitivity and specificity for diagnosing HCC was 66.6% and 91.2%, respectively.
Correlation of EPO with HCC patient survival
According to the clinical pathological stage, HCC can be divided into four groups. The clinical features of HCC patients are shown in table 1. Compared with normal tissue, EPO was downregulated in stage I (p < 0.001), stage II (p < 0.001), and stage III (p < 0.01), separately, but not in stage IV (Fig. 2a). As there were only five samples for stage IV, we reanalyzed after combining stages I and II as the early-stage and stages III and IV as the advanced stage. The results showed that EPO was downregulated in both early- and advance-stage HCC, while EPO expression in advanced-stage HCC (stages III and IV) was higher than that in early-stage HCC (stages I and II) (Fig. 2b).
EPO expression was significantly correlated with OS in patients with HCC. Kaplan–Meier survival analysis indicated that HCC patients with higher EPO expression had significantly reduced OS (hazard ratio [HR] = 1.65, confidence interval [CI] = 1.16–2.34, p = 0.005) (Fig. 2c). Univariate analysis using Cox regression revealed that some factors, including pathological stage (HR =2.090, p < 0.001), tumor status (HR = 2.317, p < 0.001), T stage (HR = 2.126, p < 0.001), and EPO (HR = 1.647, p = 0.005) were significantly associated with OS (Table 2). In multivariate analysis, high EPO expression (HR = 1.493, p = 0.43) and tumor status (HR = 1.844, p = 0.003) were independent prognostic factors of favorable prognosis (Table 2).
Based on the results, we concluded that EPO expression was correlated with HCC pathological stage and negatively correlated with HCC prognosis. Advanced-stage HCC patients with high EPO expression are more likely to have poor prognosis than those at the early stage.
Correlation and functional analyses of EPO in HCC
The volcano plot showed 731 DEGs, of which 201 were downregulated and 530 were upregulated (Fig. 3a). GO enrichment analysis of the DEGs predicted their functional roles based on biological processes, cellular components, and molecular functions. We found that GO:0006959 (humoral immune response), GO:0019724 (B cell-mediated immunity), GO:0016064 (immunoglobulin mediated immune response), GO:0002455 (humoral immune response mediated by circulating immunoglobulin), GO:0006958 (complement activation, classical pathway), GO:0006910 (phagocytosis, recognition), GO:0007218 (neuropeptide signaling pathway), GO:0006956 (complement activation), GO:0019814 (immunoglobulin complex). GO:0042571 (immunoglobulin complex, circulating), GO:0005179 (hormone activity), GO:0034987 (immunoglobulin receptor binding), has04974 (Protein digestion and absorption), and hsa04080 (Neuroactive ligand-receptor interaction) were significantly correlated with altered EPO expression (Fig. 3b). This indicated that EPO is closely related with the immune microenvironment in the development of diseases. Furthermore, correlation analysis revealed that EPO and EPOR expression in HCC were significantly positively correlated (r = 0.25, p < 0.001) (Fig. 3c).
Correlation of EPO expression with immune characteristics in HCC
Tumor-infiltrating lymphocytes are an independent predictor of cancer, and immune checkpoint blockade is a novel approach for cancer therapy [20], which has been shown to gradually improve outcomes for different cancer types [21,22]. Functional analysis revealed a potential correlation between EPO expression and immune responses. We compared the level of 24 lymphocyte types in HCC (Fig. 4a) and examined the relationship between EPO expression and enrichment of these tumor-infiltrating lymphocyte types in HCC (Fig. 4b). The results overlapped 11 types of cells. Among them, the enrichment scores of eight immune cells (activated DCs (aDC), B cells, immature DCs (iDC), macrophages, NK CD56bright cells, T cells, T follicular helper (Tfh), and Th2) were increased with higher EPO level in HCC, and the infiltration of all of them was positively correlated with EPO expression (Fig. 4c–j). The enrichment scores of the other three immune cells (central memory CD8+ T (T cm), T helper 17 (Th17), and T gamma delta (Tgd) cells) were decreased with higher EPO level in HCC, the infiltration of which was negatively correlated with EPO expression (Fig. 4k–m).
Next, we analyzed the correlation between EPO and 18 common immune-related genes in HCC (Fig. 5). EPO expression was associated with 11 of them: D80, CD86, VTCN1, HHLA2, TNFRSF14, NECTIN2, LGALS9, TNFSF9, TNFSF4, CD70 (all p < 0.001), and CD48 (p < 0.01). These results strongly indicated that EPO plays a crucial role in tumor immunity.