The GEPIA web server is a vital and highly quoted resource for analyzing gene expression on the basis of tumor and normal samples in The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov) and Genotype-Tissue Expression (GTEx, https://www.gtexportal.org/home/datasets) databases (18). The GEPIA2 (http://gepia.cancer-pku.cn) was herein used to compare the differential expression levels of APO genes and the survival prediction of immune signatures in EC.

TNMplot (19) is an open-source online network tool for comparing gene expressions among normal tissues, tumor tissues, and metastatic tissues. Gene arrays from the Gene Expression Omnibus of the National Center for Biotechnology Information (NCBI-GEO) or RNA-seq data from TCGA, Therapeutically Applicable Research to Generate Effective Treatments (TARGET), and GTEx databases were analyzed. In the present research, TNMplot (https://tnmplot.com) was used to analyze differential gene expression in tumor tissue, normal tissue and metastatic tissue. P < 0.05 was considered statistically significant.

UALCAN (20, 21) is an integrated interactive network for analysis of OMICS data developed by The University of Alabama at Birmingham Cancer Institute in 2017 and updated in 2022. In UALCAN, researchers can access level 3 RNA-seq data from TCGA. Using about 20,500 protein-coding genes in 33 different tumor types in DRUGBANK, GTEx, and Open Targets, gene expression, gene correlation, survival analysis, promoter methylation and pan-cancer analysis can be performed. In the present study, the clinical data of EC patients, such as disease stage and p53 mutation, were collected from UALCAN (http://ualcan.path.uab.edu/index.html).

Kaplan Meier plotter (22) can assess the correlation between the expression and survival data of all genes (mRNAs, miRNAs, proteins) in 21 different tumor types of 30 K+ samples, including gastric cancer, ovarian cancer, lung cancer, and breast cancer. It includes GEO, TCGA, and European Genome-phenome Archive (EGA) databases. In the present study, we analyzed the overall survival/progression-free survival (OS/RFS) of EC patients based on differentially expressed APOs by Kaplan-Meier survival curve (http://kmplot.com/analysis).

TIMER 2.0 (23) is a web server, providing a more reliable immune infiltration level estimation using cancer genome map (TCGA) or six cutting-edge algorithms. TIMER 2.0 web server facilitates comprehensive analysis and visualization of tumor-infiltrating immune cells. we studied the relationship between tumor-infiltrating immune cells and APOs using TIMER 2.0 web server (http://timer.cistrome.org).

TISIDB (24) is a portal site that explores the interaction between the tumor and the immune system, and it is based on a variety of databases, such as TCGA, UniProt, GO, DrugBank, etc. TISIDB has completely integrated 4176 records in 2530 copies and reported 988 antitumor immunities. The relationships among the differentially expressed APOs genes, immune signatures (immune types), and clinical characteristics (grade and stage) were developed in TISIDB (http://cis.hku.hk/TISIDB).

The ESTIMATE algorithm was used to estimate the number of stromal cells and immune cells in malignant tumor tissues, and immune score, stromal score, and estimated score were used to reflect the abundance of immune cells, stromal cell infiltration level, and tumor purity, respectively (25). The infiltration level of 24 immune cells was indicated by enrichment scores based on corresponding markers. Single-sample gene set enrichment analysis (ssGSEA) implemented by “GSVA” R package was employed to calculate enrichment scores (26). The cancer immune cycle is reflection of the anticancer immune response and consists of seven steps. The activity of these steps, determining the fate of tumor cells, was evaluated using ssGSEA (27).

Fresh peripheral whole blood was collected from EC patients and promptly processed within 2 hours. Following centrifugation, the peripheral blood lymphocyte pellet was obtained and resuspended to a final cell concentration of 1×10⁶ cells per 100 μl. The antibodies (Supplementary Table S1) were used at a pre-optimized 1:100 (v/v) working concentration (1.0 µl per 100 µl blood sample), and antibody-specific immune cell subset markers are listed in Supplementary Table S2. The antibodies and blood sample were incubated in the dark for 15 min with gently stirred, and then, supplemented with 450 µL 1× hemolysins (M-30PCFL; Mindray Inc., New York, NY, USA) and vortexed gently for 5 min of room temperature incubation. Finally, these tubes were placed and tested on a BD FACSCantoTM Flow Cytometer (BD Biosciences, San Jose, CA, USA). The data were analyzed using the FlowJo 10.8.1 software (BD Biosciences).

Human endometrial Ishikawa cell line (Shanghai Bohu Biological Technology Co., Ltd., China) was cultured in a phenol red-free Dulbecco’s modified Eagle’s medium/F12 medium (Gibco, Waltham, MA, USA) supplemented with 10% carbon-stripped fetal bovine serum (S181-F500; Biowest, Paris, France), 100 µg/ml streptomycin, and 100 units/ml penicillin in the presence of 5% CO₂, 95% humidity at 37 °C. 17β-estradiol (E2; 10^-8 M; 50-28-2; Sigma Aldrich, St. Louis, MO, USA) and 0.1% dimethyl sulfoxide (DMSO) (control; Con) for treating Ishikawa cells for 48 h, and the cells of the two group were collected for western blotting.

Proteins were extracted from treated Ishikawa cells (E2 and Control), which were lysed for 30 min on ice, centrifugation was performed at 10,000 g for 30 min at 4 °C, and the supernatant was removed for analysis. The BCA protein assay kit (Cat. No. 23225; Pierce, Rockford, IL, USA) and the Bradford protein method were used to determine total protein content. Proteins (15 µg) were loaded onto 4% stacked, 10% TrIS-glycine prefab gels, separated by SDS-PAGE, and were then transferred to polyvinylidene difluoride membranes (Bio-Rad Laboratories, Hercules, CA, USA; 100v, 1 h). After washing with TBS, the membranes were sealed and incubated overnight at 4 °C with rabbit anti-human APOE (1:2000, ab271944; Abcam, Cambridge, UK) and rabbit anti-human APOD (1:1000, ab256496; Abcam) antibodies. The membranes were then washed and incubated with peroxidase-labeled goat anti-Rabbit IgG (H+L) antibody (1:25, 074-1506, KPL Co., Ltd., Paris, France). The expression level of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), as a loading control protein, was assessed using horseradish peroxidase (HRP)-labeled murine GAPDH monoclonal antibody (Cat. No. KC-5G4; Aksomics Co., Ltd., Beijing, China). Immunoblots were observed using the enhanced chemiluminescence kit (Cat. No. 34095; Pierce) according to the manufacturer’s instructions. We performed densitometric analysis using ImageJ software (1.53t, NIH, USA).

Ishikawa cells (5×10⁵ cells per well) were inoculated into 6-well plates for 12 h. The cells were then incubated overnight. The lentivirus of siRNA targeting APOD (si-APOD; TL314722; OriGene, Rockville, MD, USA) and APOE (si-APOE; TR314721; OriGene), and the corresponding control (negative control; TR30023; OriGene) were transfected with reagents (TF81001; OriGene) for 12 h. The culture was then continued with a fresh medium. After transfection of Ishikawa cells, a stable cell line was established, and quantitative reverse transcription polymerase chain reaction (RT-qPCR) was used to verify its effectiveness. The cell monolayer was scratched using a sterile pipette tip, the floating cells were rinsed off with phosphate-buffered saline (PBS), and the medium was replaced with a fresh medium. Cells were then incubated for 48 h and photographed for the second time. ImagePro PLUS software (Media Cybernetics, New York, NY, USA) was used to analyze and calculate the cell migration distance.

The results of western blotting and flow cytometry were analyzed by SPSS 25.0 software (IBM, Armonk, NY, USA). P < 0.05 was considered statistically significant.

First, we analyzed the differential expressed APO genes (APOA-APOJ) in EC and control tissues (TCGA and GTEx databases) by the GEPIA2 (Figure 1A). |log2FC|>1 and adjusted P < 0.05 were considered as significant and were selected for subsequent analysis. Totally, 9 differentially expressed APO genes, including APOC1, APOC2, APOC4, APOD, APOE, APOL3, APOL4, APOLD1, and APOO were found. The violin plots were obtained from the TNMPlot database (Figure 1B). Then, we queried these differentially expressed APO genes in cBioPortal (https://www.cbioportal.org), and APOD, APOC1, and APOE genes were found in 10%, 5%, and 4% of patients, respectively. At the same time, estrogen receptor 1 (ESR1) gene was identified in only 7% of patients (Figure 1C). Although APOC4 had the differential expression between EC and control tissues (Figure 1B), and APOC4 was found in 4% of patients (Figure 1C), it still very lowly expressed in tissues (Figure 1D). Therefore, due to its low expression, APOC4 was not included in the analysis.

We analyzed the relationships between expression levels of APOs and clinicopathological characteristics (stage and tumor grade) of EC by UALCAN and TISIDB databases (Figure 2). APOE expression was significantly higher in tumors than controls across all four stages, with higher mRNA levels positively correlating with advanced stages. Although APOLD1 was lower in EC than controls, it similarly showed a positive correlation with tumor stage. APOD was downregulated in tumors versus controls across all stages but had no significant correlation with stage. Regarding EC grade, APOC1, APOE, and APOO expression positively correlated with grade, while APOC2, APOD, and APOL4 showed a negative correlation.

In 2013, integrated genomic characterization of EC was established for 4 subgroups: distinct subgroups, POLE (ultra-mutated), microsatellite instability (MSI; hypermutated), low copy number (endometrioid), and high copy number (serous-like) (28). As shown in Figure 3A, expression levels of all 8 APOs were significantly differed from different molecular subtypes (high copy number: 160; low copy number: 144; MSI: 124, POLE: 79).

TP53 and PTEN are among the most frequently mutated genes in EC (28, 29), while ESR1 and ESR2 are established EC pathogenic factors. We compared APOs expression between p53-mutant and p53-non-mutant, PTEN-mutant and PTEN-non-mutant, and low and high ESR1/ESR2 expression groups. Results showed APOC1, APOE, APOL3, and APOLD1 were significantly upregulated in p53-mutant EC, whereas APOO and APOD exhibited the opposite pattern (Figure 3B). Additionally, PTEN-mutant EC showed higher APOL4/APOO and lower APOE/APOLD1 expression compared with PTEN-non-mutant EC (Figure 3C).

Additionally, high expression levels of APOD, APOL4, and APOO were detected in high ESR1 group (Figure 3D), and ESR1 expression positively correlated with these three genes (Supplementary Figure S1A). In contrast, APOE expression was reduced in the high ESR1 group (Figure 3D) and negatively associated with ESR1 (Supplementary Figure S1A). Meanwhile, high expression levels of APOC1, APOD, APOE, APOL3, and APOL4 were detected in high ESR2 group (Figure 3E). GPR30 (encoded by GPER1) is an estrogen transmembrane receptor, and the estrogen/GPR30 signaling pathway in EC has been characterized in our prior work (30, 31). We further evaluated APO expression differences between high and low ESR1 groups, and analyzed their associations with ESR2 and GPER1 (Supplementary Figures S1–S3).

Therefore, APOs were significantly correlated with clinicopathological characteristics (stage, grade, molecular subgroups, p53/PTEN mutation status, and ESR1/ESR2 expression levels). Of the 8 APOs, only APOE mRNA was significantly associated with all these characteristics.

The prognostic value of differentially expressed APOs in EC patients was determined by Kaplan-Meier analysis of overall survival (OS) using the Kaplan-Meier plotter (Figure 4). According to the results, higher expression levels of APOD and APOL3 were significantly correlated with a longer OS. In contrast, higher expression levels of APOC1, APOE, and APOLD1 were associated with a poor OS. Therefore, APOs expression may considered as predictors for better or poorer OS, indicating that APOs may be promising prognostic biomarkers for EC patients.

APOE showed the strongest correlation with EC risk factors (tumor stage/grade, TP53/PTEN mutation, ESR1/ESR2 expression), followed by APOD (4 factors: tumor grade, TP53 mutation, ESR1/ESR2 expression), APOC1 (3 factors: tumor grade, p53 mutation, ESR1 expression), and APOLD1 (3 factors: tumor stage, p53/PTEN mutation). Thus, among the 8 APOs, APOE, APOD, APOC1, and APOLD1 may be the most clinically relevant in EC.

TME is closely associated with cancer prognosis and therapeutic effects. TILs were found as integral components of the TME and were correlated with prognosis and therapeutic response. Then, we analyzed the differences of TILs between EC and adjacent noncancerous tissues by ssGSEA. It was found that 21/24 (87.5%) of TILs in EC differed significantly from those in adjacent noncancerous tissues (Figure 5).

Next, we investigated the prognostic value of TILs. The Kaplan-Meier analysis revealed that EC patients with high CD4+ naïve T cells, CD8+ naïve T cells, CD8+T cells, central memory T cells, cytotoxic T cells (CD3⁺CD8⁺CD28⁺T cells, Tc), mucosal-associated invariant T (MAIT) cells, and T helper type 1 (Th1) cells were correlated with a longer OS (Figure 6).

Finally, we compared the differences of 14 TILs in a detection panel, including CD4+ naïve T cells, CD8+T cells, CD8+ naïve T cells, and Tc in EC patients (20 cases) to control uterine fibroids (controls; 20 cases) (Figures 7A–N). We found that CD4+ naïve T, CD8+ T cells, and Tc in EC patients were significantly lower than controls (Figures 7F, I, L), however, there was no significant difference in CD8+ naïve T cells between EC and control groups (Figure 7J).

We further explored the relationship between the expression levels of APOE or APOD and the level of immune cell infiltration, which would be correlated with prognosis, using Spearman correlation analysis. The results showed that the expression levels of both APOD and APOE were positively correlated with TILs of CD4+ naïve T cells (Figures 8A, E), CD8+T cells (Figures 8B, F), CD8+ naïve T cells (Figures 8C, G), and Tc (Figures 8D, H). Other TILs, including central memory T cells, MAIT cells, and Th1 cells, which were related to prognosis, were also positively correlated with the expression levels of APOD or APOE (Supplementary Figure S4).

To clarify the regulatory effects of E₂ on APOE and APOD expression, we first detected the protein levels of these two genes in Ishikawa cells treated with or without E₂. As shown in Figure 9A and Supplementary Figure S5, E₂ stimulation significantly upregulated the protein expression of APOE (0.072 ± 0.003 vs 0.056 ± 0.009, p=0.0298), whereas it exerted an inhibitory effect on APOD protein expression (0.027 ± 0.007 vs 0.061 ± 0.005, p=0.0003). Furthermore, in vitro wound healing assay confirmed that APOE promoted the migratory potential of EC cells (Figure 9B).