Identification of DEGs in UCEC
Overall, we identified 6,268 DEGs, 410 candidate prognostic immune genes, and 100 differential TFs (Fig. 1 and 2). The differential expression of immune genes in all endometrial cancer samples is summarized in Supplementary File 1. Enrichment analysis of differentially expressed immunity genes showed that biological processes (BP), mainly chemotaxis migration of anti-inflammatory cells, including leukocyte and neutrophils, were primarily enriched (Fig. 3A). The enriched cellular components (CC) were mainly extracellular matrix whereas the main molecular function (MF) comprised of growth factor and cytokine activity. These findings implied that most differentially expressed immunity genes were associated with UCEC development, progression, and prognosis through immune cells. The enriched top 30 KEGG pathways are given in Fig. 3B. Notably, several signaling pathways involved in UCEC development, including PI3K-Akt, MAPK, Ras, and JAK-STAT, were identified.
Association between immunity genes and survival rates
Univariate Cox regression analysis of differentially expressed immunity genes revealed a significant correlation between survival rates and 21 of candidate prognostic immune genes (P < 0.01). In particular, the genes with a significant association were PDIA3, LTA, PSMC4, IL6, TNF, KCNH2, SYTL1, BACH2, PCSK1, BIRC5, SBDS, ANGPTL7, GPI, HDGF, ADCYAP1R1, HTR3E, NPR1, NR3C1, PGR, THRB, and CBLC. Among them, LTA, ADCYAP1R1, PGR, SYTL1, and PDIA3 were characterized as low-risk, while the remaining 16 were categorized as high-risk genes. Detailed information of all 21 genes is depicted in Fig. 4A.
Furthermore, to assess the relationship between the 21 prognosis-related immunity genes and TFs, a univariate Cox regression analysis was performed at | cor |>0.4 and P < 0.001, where a TF regulatory network was constructed (Fig. 4B). Notably, the regulatory network diagram that comprised of low- (PGR, SYTL1, and LTA), and high-risk (BIRC5, HDGF, HTR3E, THRB, NR3C1, BACH2) genes, as well as TFs (AR, BATF, CBX2, CENPA, E2F1, E2F3, ETS1, EZH2, FOXK1, FOXP3, GREB1, H2AFX, LMNB1, LYL1, NCAPG, NR3C1, RFX2, SNAI2, SOX17, SPDEF, SPIB, STAT5A, and WWTR1), elucidated a positive relationship between immunity genes and TFs. Lastly, BIRC5 was associated with several transcription factors namely CBX2, CENPA, E2F1, EZH2, FOXM1, H2AFX, LMNB1, and NCAPG.
The prognostic prediction signature
To establish a signature for predicting the prognosis of UCEC patients, we employed a Cox regression analysis and identified a ten-gene prognostic signature based on a training set. The genes in the signature included PDIA3, LTA, PSMC4, TNF, SBDS, HDGF, HTR3E, NR3C1, PGR, and CBLC (Table 1). We used the prognostic signature to calculate a risk score for each patient, while the median value was used to divide the patients into a high-risk (n=270) and low-risk groups (n=271) (Supplementary File 2 showed the risk score and immune gene expression per patient of the signature in the training set). The prediction power of the ten-gene prognostic signature for patients in training sets is outlined in Fig. 5, while the distribution of risk scores, gene expression levels, and patient survival status are displayed in Fig. 5A. Remarkably, AUC for the training set was 0.756, indicating good accuracy of the prognostic prediction-values across the ten-gene prognostic signature. From the Kaplan‐Meier curve, lower overall survival rates were recorded for patients in the high-risk compared to those in the low-risk group for the training set (P < 0.0001) (Fig. 5C). Besides, 5-year OS rates of 63.1 and 89.9%, were recorded for patients in the high- and low-risk groups, respectively, whereas 9‐year OS rates were 34.6 and 78.7% for patients in the high‐ and low‐risk groups, respectively.
Validation of the ten-gene prognostic signature in UCEC
To determine the feasibility and reliability of the ten-gene prognostic signature, we validated it using testing set A (n=270) and testing set B (n=271). In the testing sets A and B, a shorter overall survival rate was noted for patients in the high risk compared to those in the low-risk groups (P< 0.0001) (Fig6E/F). The AUC for the testing set A and B were 0.706 and 0.885 (Fig.6C/D), respectively, suggesting that the signature strongly predicts overall survival in UCEC patients. (Supplementary File 3 and File 4 showed the risk score and immune gene expression per patient of the signature in the testing set A and the testing set B, respectively.)
The ten-gene prognostic signature is an independent prognostic factor
To determine whether the signature risk score was an independent prognostic factor for patient survival, we employed univariate and multivariate Cox regression analyses. Results demonstrated P < 0.05, across both analyses, indicating that the risk score derived from the signature can be independent of other clinical traits, and thus an independent prognostic factor. In addition, univariate Cox regression analysis showed that age (P = 0.002, hazard ratio=1.035) and grade (P <0.001, hazard ratio=2.595) were significantly associated with prognosis. Of note, the prognosis of patients was worse with an increase in age and grade (Fig. 7).
Clinical parameters, immunohistochemical examination
The correlation between immune genes involved in the signature and clinical traits was assessed using Univariate Cox regression analysis. Here, patients were divided into two groups, based on clinical traits: Group 1 (comprised of patients aged <55 and ≥ 55) and Group 2 (G1 & G2 and G3). Results revealed a significant correlation between HDGF (P < 0.001), PGR (P = 0.04), PSMC4 (P < 0.001), TNF (P < 0.001), NR3C1 (P = 0.015), HTR3E (P = 0.033) and CBLC (P = 0.003) with age, whereas expression of HDGF, PSMC4, TNF, NR3C1, HTR3E, and CBLC increased with age. On the other hand, HDGF (P < 0.001), PGR (P < 0.001), PSMC4 (P < 0.001), TNF (P < 0.001), NR3C1 (P < 0.001), PDIA3 (P < 0.001), and SBDS (P < 0.001) were significantly associated with grade. Moreover, an increase in grade resulted in the upregulation of HDGF, PSMC4, TNF, NR3C1, and SBDS (Fig. 8). Immunohistochemical analysis based on The Human Protein Atlas database enumerated a significant upregulation of PSMC4, NR3C1, SBDS, and CBLC in endometrial cancer tissues, relative to normal tissues. On the other hand, immunohistochemical analysis of PGR and PDIA3 expression showed significant downregulation of these factors in endometrial cancer compared to normal tissues (Fig. 9).
Immune infiltrates analysis of the signature genes in patients with UCEC
Herein, a correlation analysis between risk scores in UCEC patients with abundance of six immune infiltrations indicated a significant positive association between B cells (P = 3.408e−10, cor=0.265) and neutrophils (P = 0.011, Cor = 0.109) with the patient's risk score (Fig. 10). To explain this relationship, we analyzed infiltration abundance, and noted a positive relationship between B cells and expression of LTA (Cor=0.594, P = 5.55e-29) (Fig. 11B), TNF (Cor=0.117, P = 4.60e-02) (Fig. 11D), and NR3C1 (Cor=0.301, P=1.85e-07) (Fig. 11H). Moreover, the infiltration abundance of neutrophils was positively correlated with expression of LTA (Cor=0.339, P = 2.65e-09) (Fig. 11B), PSMC4 (Cor=0.209, P = 3.23e-04) (Fig. 11C), TNF (Cor=0.408, P = 3.56e-13) (Fig. 11D), SBDS (Cor=0.418, P = 7.89e-14) (Fig. 11E), HDGF (Cor=0.309, P = 6.50e-08) (Fig. 11F), and NR3C1 (Cor=0.48 P = 2.70e-18) (Fig. 11H).