2.1 Patients and endometrium samples

We selected a total of 20 EC patients from the First Hospital of Lanzhou University involved 20 cancer tissues (CT, n = 20). Patients were classified as low-grade (grade 1–2, n = 10) or high-grade (grade 3, n = 10) based on the degree of tumor differentiation. The participants satisfied the following inclusion criteria: women ≥18 years with primary endometrial cancer. Excluded participants: (1) stage III–IV and non-endometrioid histological type; (2) absence of detailed clinical characteristics; (3) previous neoadjuvant chemotherapy, preoperative radiation or endocrine treatment; (4) history of malignancies or chronic diseases.

2.2 Sample preparation

We collected tissue within 30 min of removing the uterine appendages, immediately frozen it and stored it at −80 °C. Each sample of 20 μg tissues was directly dissolved in sodium dodecylsulfate (SDS) with dithiothreitol (DTT) (SDT buffer). Following centrifugation at 14,000g × 40 min, the protein concentrations were determined using BCA protein assay kit (PC0020, Solarbo, Beijing, China), then extracted and quantified peptides via filter-aided sample preparation (FASP).

2.3 Spectral library generation

A Q-Exactive orbitrap mass spectrometer equipped with an Easy nLC 1200 chromatography system (Q Exactive-HFX, Thermo Scientific, Waltham, MA, USA) was used to perform data-dependent acquisition (DDA). The peptide was first loaded onto a 75 μm I.D. ×2 cm trap column (C18, 3 μm, Thermo Scientific) with 95% of buffer A (0.1% formic acid in water), then separated on a 75 μm I.D. ×25 cm analytical column (C18, 2 μm, Thermo Scientific, Waltham, MA, USA) with a linear gradient of buffer B (84% acetonitrile and 0.1% formic acid). The peptide separation was carried out in 65 minutes at a flow rate of 300 nL/min using an LC procedure that started at 8% buffer B and progressed linearly up to 100% buffer B. Mass spectrometry (MS) spectra was collected in DDA mode using Q-Exactive HFX, which had the following parameters: a 300–1800 m/z scan range, a resolution of 60,000 at 200 m/z for mass spectrometry 1 (MS1), an automatic gain control (AGC) target of 3 × 10⁶, a maximum injection time of 25 ms, and a dynamic exclusion of 30 s. After each MS1 scan, the top 20 precursor ions were selected for MS2. The maximum injection time was set to 25 ms, the AGC target value for the MS2 scan was 5 × 10⁴, and the normalized collision energy was 30 eV.

2.4 DIA-MS/MS analysis

The DIA-MS/MS analysis was carried out in the same way as the DDA-MS/MS. The DIA was carried out using 30 isolation windows, covering a mass range of 350–1800 m/z at a resolution of 60,000, with an AGC target of 3 × 10⁶ and a maximum injection duration of 50 ms. Spectronaut (SpectronautTM 14.4.200727, Biognosys, Switzerland) was used with default parameters to accomplish protein identification and quantification. All results were filtered based on a false discovery rate (FDR) of 1%.

2.5 DEGs identification and functional analysis

The criteria for identifying DEGs between high-grade cancer tissue (HG-CT) and low-grade cancer tissue (LG-CT) groups were “p < 0.05 and fold change (FC) >1.5 or <0.67”. The volcano plot, principal component analysis (PCA), and heatmap with hierarchical clustering analysis were conducted using R package (R Version 3.4, R Foundation for Statistical Computing, Vienna, W, Austria) to get an overview of the DEGs characteristics. We conducted a gene enrichment analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG), gene ontology (GO), and p < 0.05 was used as the limit for significantly enriched functional GO terms or KEGG pathways.

2.6 Screening of IRMGs

A total of 1793 genes associated to immunity and 1804 genes related to metabolism were obtained from the ImmPort database (https://www.immport.org/home) and previous report [13], respectively. Then, these genes were intersected with 285 differentially expressed genes (DEGs) which identified based on 20 fresh frozen cancer tissues to screen out 49 metabolic-DEGs and 20 immune-DEGs. To screen IRMGs, the Spearman correlation was utilized (p < 0.05 and |R| > 0.2).

2.7 Development of immune-related metabolic prognosis signature (IRMPS)

Based on TCGA-UCEC dataset, we utilized the Lasso-cox regression analysis to establish a prognostic model, 7 IRMGs and their regression coefficients were obtained. The samples were split into high-risk and low-risk group based on the risk score median value. Subsequently, Kaplan-Meier (KM) survival curves and time-dependent receiver operational feature (ROC) curve analysis were performed to evaluate the predictive power of the signature. In order to identify prognosis-related genes, we subsequently employ univariate cox regression analysis. Multivariate cox regression analysis is then utilized to confirm the independence of the IRMPS model. The predictive value of the IRMPS and other clinical variables was also evaluated using the decision curve analysis (DCA) curve.

2.8 Construct of an IRMPS-based nomogram

According to the risk score and other clinical traits, such as age, stage, race and tumor grade, an IRMPS-based nomogram was created. Calibration plot was utilized to confirm the effectiveness of the nomogram.

2.9 Gene set enrichment analysis (GSEA)

We performed GSEA to investigate potential biological pathway and immunological activity related IRMPS. Normalized “p < 0.05” as meaningful enriched terms.

2.10 Immune-related analysis

We used the Tumor Immune Estimation Resource 2.0 (TIMER 2.0) (http://timer.cistrome.org/) to investigate the associations among IRMPS and six immune checkpoints in TCGA-UCEC. Immune Cell Abundance Identifier (ImmuCellAI) online tool (https://guolab.wchscu.cn/ImmuCellAI//#!/) was used to estimate the abundance of 24 immune cells in UCEC. The correlations between IRMPS model and 24 immune cells were evaluated by Spearman. Via ESTIMATE (https://bioinformatics.mdanderson.org/estimate/), we downloaded the ImmuneScore, StromalScore and ESTIMATEScore of TCGA-UCEC and calculated using “estimate” package.

2.11 Treatment response prediction

Using the tumor immune dysfunction and exclusion (TIDE) website (http://tide.dfci.harvard.edu/), we calculated TIDE scores, a novel approach for assessing immune checkpoint inhibitor (ICI) efficacy. Additionally, we investigated the relationship between immunophenoscore (IPS) and two risk groups in order to predict the value of this score in predicting the effectiveness of immunotherapy (program death protein 1 (PD-1)/program death ligand 1 (PD-L1) blockade and/or cytotoxic T-lymphocyte-associated antigen-4 (CTLA-4) blockade) on TCGA-UCEC patients by the Cancer Immunome Atlas (TCIA) (https://tcia.at/home).

To assess the sensibility of chemotherapy drugs, we utilized an open accessible database named Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/), and estimate the half-maximal inhibitory concentration (IC50) for displaying the response to drugs.

2.12 Statistical analysis

Our data were tested for normality by using GraphPad Prism 9.0.0 software (GraphPad Software, San Diego, CA, USA). Comparisons between the two risk groups were assessed using Student’s t-tests (passed normality test) or Wilcoxon rank sum tests (do not pass normality test). Statistically significant differences were defined as *p < 0.05, **p < 0.01 and ***p < 0.001.

3.1 Identification of DEGs in EC and biological function

We used DIA quantitative proteomics to analyze fresh frozen cancer tissues of 20 EC patients, comprised 10 high-grade and 10 low-grade. A total of 285 DEGs were collected (p < 0.05; fold change >1.5 or <0.67), including 148 up-regulated and 137 down-regulated (Fig. 1A and Supplementary Table 1). Fig. 1B showed an expression heatmap for 285 DEGs. The principal component analysis (PCA) was performed to compress the protein abundances in different samples into 20 two-dimensional data points, which were then displayed using a scatter plot (Fig. 1B). The figure revealed differences of protein abundance patterns between several sample groups, the PCA-derived distances can distinguish different tumor grades of EC (Fig. 1C). Further analyses of KEGG and GO enrichment demonstrated that the DEGs mainly involved in metabolic pathways, especially oxidative phosphorylation and formed electron transport chain (Fig. 2A–D).

Fig. 1.Proteomics analysis of differentially expressed genes (DEGs) in HG-CT groups compared to LG-CT groups of 20 EC patients. (A) Volcano plot of 285 DEGs, as follows: proteins presenting a fold change of more than 1.5 or less than 0.67 and p < 0.05. (B) Hierarchical clustering of 285 dysregulated proteins in tumors of different grades, and labeled the Top 10 up/down-regulated protein names. (C) PCA illustrating moderate clustering of tumor samples within each group. HG-CT: high-grade cancer tissue; LG-CT: low-grade cancer tissue; DPY30: Dpy-30 homolog; TRIM24: Tripartite motif containing 24; NCF2: Neutrophil cytosolic factor 2; MRPL21: Mitochondrial ribosomal protein L21; GOLGA5: Golgin A5; DHRS7B: Dehydrogenase/reductase (SDR family) member 7B; SPART (SPG20): Spastic paraplegia 20; MRPL41: Mitochondrial ribosomal protein L41; EIF2B5: Eukaryotic translation initiation factor 2B, subunit 5; SERPINH: Serpin peptidase inhibitor, clade H (heat shock protein 47), member 1; CMC2: C-x(9)-C motif containing 2; EPX: Eosinophil peroxidase; IGKV3D-11: Immunoglobulin kappa variable 3D-11; C9: Complement component 9; GSTM5: Glutathione S-transferase mu 5; FGF2: Fibroblast growth factor 2; NDUFB8: NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 8; HMGN5: High mobility group nucleosome binding domain 5; ATP5F1D: ATP synthase, H+ transporting, mitochondrial F1 complex; MCFD2: Multiple coagulation factor deficiency 2.

Fig. 2.Function analysis of 285 DEGs. (A–C) Gene ontology (GO) functional annotation of 285 DEGs. (D) KEGG enrichment analysis of 285 DEGs. BP: biological process; CC: cellular component; MF: molecular function; ITGA4: Integrin alpha-4; PECAM1: Platelet endothelial cell adhesion molecule; GNAI3: Guanine nucleotide-binding protein G (I) subunit alpha-3; SIPA1: Signal-induced proliferation-associated protein 1; F11R: Junctional adhesion molecule A; GNG2: Guanine nucleotide-binding protein G (I); PRKAA2: 5′-AMP-activated protein kinase catalytic subunit alpha-2; GNG10: Guanine nucleotide-binding protein subunit gamma; MEF2D: MADS-box domain-containing protein; GNG12: Guanine nucleotide-binding protein subunit gamma; SMAD4: Mothers against decapentaplegic homolog 4; RAF1: RAF proto-oncogene serine; ACTG1: Actin, cytoplasmic 2; IGF2: Insulin-like growth factor II; SHC1: SHC-transforming protein 1; GSTM3: Glutathione S-transferase Mu 3; RB1: Retinoblastoma-associated protein; SMARCA4: SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4; NDUFA2: NADH dehydrogenase (ubiquinone) 1 alpha subcomplex subunit 2; MT-CYB: Mitochondrially encoded cytochrome b; MT-ND3: NADH-ubiquinone oxidoreductase chain 3; NDUFB9: NADH dehydrogenase (ubiquinone) 1 beta subcomplex subunit 9; NDUFB7: NADH dehydrogenase (ubiquinone) 1 beta subcomplex subunit 7; NDUFA4: NADH dehydrogenase (ubiquinone) 1 alpha subcomplex subunit 4; ALDOC: Fructose-bisphosphate aldolase C; ENO3: Enolase 3 (beta, muscle); ENO2: Enolase 2 (gamma, neuronal); IDH2: Isocitrate dehydrogenase 2 (NADP+), mitochondrial; ALDH18A1: Delta-1-pyrroline-5-carboxylate synthase; PYCR1: Pyrroline-5-carboxylate reductase; ARRB1: Arrestin, beta 1; COL4A1: Collagen alpha-1 (IV) chain; NNMT: Nicotinamide N-methyltransferase; CDS2: CDP-diacylglycerol synthase (phosphatidate cytidylyltransferase) 2; AKR1B1: Aldo-keto reductase family 1 member B1; HAAO: 3-hydroxyanthranilate 3,4-dioxygenase; ACAA2: Acetyl-CoA acyltransferase 2; INPP5A: Inositol-polyphosphate 5-phosphatase; CBR3: Carbonyl reductase (NADPH); PIP4K2B: Phosphatidylinositol 5-phosphate 4-kinase type-2 beta; RENBP: GlcNAc 2-epimerase; DPYD: Dihydropyrimidine dehydrogenase; CDIPT: CDP-diacylglycerol-inositol 3-phosphatidyltransferase; SRM: Spermidine synthase; LPCAT2: Lysophosphatidylcholine acyltransferase 2; MTMR1: Myotubularin-related protein 1; ALG6: Dolichyl pyrophosphate Man9GlcNAc2 alpha-1,3-glucosyltransferase; GLB1: Nitrogen regulatory protein P-II homolog; GPAA1: Glycosylphosphatidylinositol anchor attachment 1 protein; NUDT2: Nudix (nucleoside diphosphate linked moiety X)-type motif 2; IDUA: Alpha-L-iduronidase; CANT1: Calcium activated nucleotidase 1; SORD: Sorbitol dehydrogenase; FHIT: Fragile histidine triad; HADH: Hydroxyacyl-CoA dehydrogenase; UMPS: Uridine monophosphate synthetase; CERS2: Ceramide synthase 2; P4HA2: Procollagen-proline 4-dioxygenase; AACS: Acetoacetyl-CoA synthetase; GALNT12: Polypeptide N-acetylgalactosaminyltransferase 12; CA8: Carbonic anhydrase VIII; ACP2: Acid phosphatase 2, lysosomal; RDH10: Retinol dehydrogenase 10; ELOVL5: Elongation of very long chain fatty acids protein 5.

3.2 Identification of IRMGs in EC

We collected 1804 metabolic genes and 1793 immune genes in order to identify the immune-related metabolic genes. Venn diagrams were performed to find 49 metabolic-DEGs and 20 immune-DEGs (Fig. 3A,B). Then, 41 IRMGs were screen by Spearman correlation analysis (p < 0.05, |R| > 0.2; Fig. 3C).

Fig. 3.Screening immune-related metabolic genes (IRMGs). (A) Venn diagram of 49 metabolic-DEGs obtained based on the intersection between metabolic genes and DEGs. (B) Venn diagram of 20 immune-DEGs obtained based on the intersection between immune genes and DEGs. (C) The correlation between 49 metabolic-DEGs and 20 immune-DEGs in EC patients. *p < 0.05, **p < 0.01, ***p < 0.001. Cor: correlation; ACOT11: Acyl-coenzyme A thioesterase 11; NPC1: NPC intracellular cholesterol transporter 1; STXBP1: Syntaxin-binding protein 1; SERINC1: Serine incorporator 1; CYB5A: Cytochrome b5 heme-binding domain-containing protein; RPS27L: Ribosomal protein S27-like; VAMP2: Synaptobrevin-2; PLEKHA1: Pleckstrin homology domain-containing family A member 1; MED20: Mediator of RNA polymerase II transcription subunit 20; LSR: Lipolysis-stimulated lipoprotein receptor; ACP6: Lysophosphatidic acid phosphatase type 6; DDX31: DEAD (Asp-Glu-Ala-Asp) box polypeptide 31; GPD2: Glycerol-3-phosphate dehydrogenase, mitochondrial; RPS26: Ribosomal protein S26-like; SLC26A2: Sulfate transporter; PDK1: Protein-serine; IGKV3-15: Immunoglobulin kappa variable 3-15; HDGFL3: PWWP domain-containing protein; CD81: Tetraspanin; ANGPTL2: Angiopoietin-related protein 2; UBXN1: UBX domain-containing protein 1; CTF1: Cardiotrophin-1; GBP2: Guanylate-binding protein 2; TRIM5: Tripartite motif-containing protein 5; TMSB10: Thymosin beta-10; IRF9: Interferon regulatory factor 9; PGR: Progesterone receptor; MDK: Midkine.

3.3 Development of an IRMPS with good performance in the TCGA dataset

Afterwards, we investigated whether the IRMGs could predict the survival of EC patients. A total of 41 IRMGs were selected to fit a LASSO regression model based on TCGA-UCEC (Fig. 4A). Then, we found the best lambda values (λ = 0.0228) via 10-fold cross-validation (Fig. 4B). Finally, 7 IRMGs were constructed be a prognosis signature and the risk scores were calculated as the following formula: risk score = (−0.0264) × NDUFA2 + (0.2147) × PRKAA2 + (0.0146) × STXBP1 + (0.0075) × NDUFB9 + (−0.0401) × RPS27L + (−0.2055) × LPCAT2 + (0.049) × UMPS. In the TCGA-UCEC cohort, patients were divided into high-risk and low-risk categories. Fig. 4C showed the IRMPS can discriminate risk scores, survival status and expression of 7 hub genes. According to the results of the Kaplan-Meier survival analysis, the risk-high group had a significantly lower overall survival (OS) rate versus the risk-low group (p < 0.0001; Fig. 4D), demonstrating the significance of the IRMPS in predicting the prognosis of EC patients. For one, three and five years of survival, the corresponding areas under the ROC curves (AUC) were 0.692, 0.717 and 0.731 (Fig. 4E), indicating satisfactory prediction model accuracy. Furthermore, a univariate cox regression model of the expression levels of the 7 prognostic genes (Fig. 5A) demonstrated that high expression of PRKAA2 and UMPS, along with low expression of NDUFA2, RPS27L and LPCAT2 were associated with poorer prognosis (p < 0.05; Fig. 5B). Collectively, these findings implied that the risk-scoring model may be used to predict the survival of EC patients.

Fig. 4.Construction of the immune-related metabolic gene prognostic model in TCGA-UCEC datasets. (A) prognostic IRMGs were evaluated using LASSO analysis and 10-fold cross validation. (B) Coefficient profile plots of 7 prognostic IRMGs. Vertical dashed lines are plotted at the best lambda. (C) Distribution of risk score, survival status and 7 hub gene expression. (D) Kaplan-Meier curves of overall survival between two risk groups in EC patients. (E) Time-dependent receiver operational feature (ROC) curves analysis of the prediction model. HR: Hazard Rate; AUC: Area Under the Curve; CI: Confidence Interval.

Fig. 5.The hub genes of prognostic model. (A) The 7 hub genes screened out from LASSO in the TCGA-UCEC cohort were used to generate a forest plot of a univariate cox regression model. (B) Kaplan-Meier curves of NDUFA2, PRKAA2, RPS27L, LPCAT2 and UMPS. High exp: High expression; Low exp: Low expression; NDUFA2: NADH dehydrogenase (ubiquinone) 1 alpha subcomplex subunit 2; PRKAA2: 5′-AMP-activated protein kinase catalytic subunit alpha-2; STXBP1: Syntaxin-binding protein 1; NDUFB9: NADH dehydrogenase (ubiquinone) 1 beta subcomplex subunit 9; RPS27L: Ribosomal protein S27-like; LPCAT2: Lysophosphatidylcholine acyltransferase 2; UMPS: Uridine monophosphate synthetase; OS: Overall Survival; CI: Confidence Interval; HR: Harzad Rate.

3.4 The signature of seven immune-related metabolic genes might be an independent risk factor for EC

We used univariate and multivariate cox regression analysis to assess the independence of IRMPS for EC patients. The univariate cox analysis revealed that age, clinical stage, race, grade and risk score were important prognostic factors (p < 0.001; Fig. 6A). In addition, the risk score remained significant for prognosis in the multivariate cox analysis (p = 0.001; Fig. 6B), suggesting the IRMPS can be independent of other clinical features and should be incorporated into the establishment of prognostic models. A nomogram was created by merging other clinical parameters with the IRMPS model in order to further investigate the predictive capability of IRMPS (Fig. 6C). Afterward, the calibration curves showed powerful accuracy of the nomogram (Fig. 6D). DCA curve displayed that the prognostic value of IRMPS was superior than other clinical factors (Fig. 6E).

Fig. 6.Construction and validation of a nomogram. Univariate (A) and multivariate cox regression (B) to assess the independence of the immune-related metabolic gene prognostic signature (IRMPS). (C) Nomogram for predicting the likelihood of 1-, 3- and 5-year OS of UCEC patients. (D) Nomogram calibration curves to forecast the likelihood of OS at 1-, 3- and 5-year. (E) The decision curve analysis (DCA) curves of IRMPS and other clinical characteristics for 5-year OS in UCEC patients. CI: Confidence Interval; HR: Harzad Rate.

3.5 GSEA analysis and immune activity with different risk score

According to the GSEA analysis of IRMPS, several hallmarks were enriched, including kirsten rat sarcoma viral oncogene homolog (KRAS) signaling, xenobiotic metabolism, mitotic spindle and cell cycle checkpoints (G2/M) (Fig. 7A). Fig. 7B showed various of immunological pathways were related to the risk score, with substantial enrichment in the high-risk group.

Fig. 7.Gene set enrichment analysis. (A) Gene sets of hallmarks. (B) Gene sets of immune activity. L: low-risk group; H: high-risk group; ES: enrichment score; NP: nominal p-value.

Immune infiltration is a crucial factor related to tumor progression, and T cells are essential for the development, spread and treatment (especially immunotherapy) of cancer [14]. Through further comparison of the relationship between two risk groups and 24 tumor-infiltrating lymphocytes (TILs) abundances in EC via ImmuCellAI online tool, we observed that neutrophils, and Th2 cells were meaningfully positively associated with IRMPS (p < 0.001), while cytotoxic T cells (Tc), cluster of differentiation 8 (CD8)+ T cells, exhausted T cells (Tex), Th1 cells, T cell follicular helper (Tfh), natural killer (NK) cells and cluster of differentiation 4 (CD4)+ T cells were significantly negatively linked to IRMPS (p < 0.001; Fig. 8A). These findings proved that IRMPS is crucial for the immune infiltration of EC. Additionally, by comparing six immune checkpoint markers (PD1, CTLA4, PD-L1, PD-L2, lymphocyte activation gene-3 (LAG3) and hepatitis A virus-cellular receptor 2 (HAVCR2)) between the two risk groups, we discovered that IRMPS showed a remarkable negative correlation with CTLA4, PD-L1 and HAVCR2 (p < 0.001; Fig. 8B), indicating that an immune-suppressive microenvironment is more probable to form in low-risk groups.

Fig. 8.Immune activity analysis of the IRMPS model. (A) Correlation among the relative abundances of 24 tumor-infiltrating lymphocytes (TILs) and IRMPS model based on ImmuCellAI online tool. (B) Correlation between immune checkpoint and IRMPS model, including PD1, CTLA4, PD-L1, PD-L2, LAG3 and HAVCR2. *p < 0.05, **p < 0.01, ***p < 0.001. Th1: type I helper T cell; Th2: type II helper T cell; MAIT: mucosal-associated invariant T cell; NK: natural killer cell; NKT: natural killer T cell; DC: dendritic cells; Tgd: gamma delta T cell; Tcm: central memory T cell; Tem: effector memory T cell; Treg: regulatory T cell; Tr1: tregulatory1; Tc: cytotoxic T cell; CD8: cluster of differentiation 8; CD4: cluster of differentiation 4; Tex: exhausted T cell; Tfh: T cell follicular helper; PD-1: program death protein 1; PD-L1: program death ligand 1; PD-L2: program death ligand 2; CTLA-4: cytotoxic T-lymphocyte-associated antigen-4; LAG3: lymphocyte activation gene-3; HAVCR2: hepatitis A virus-cellular receptor 2; ^ns: non-significant.

3.6 Analysis of immunotherapy and chemotherapy response

Applying ESTIMATE methodology, we discovered that the IRMPS-high group had lower immune, stromal and ESTIMATE scores (Fig. 9A) while having higher tumor purity scores (Fig. 9B). The results implied that EC patients with high-risk IRMPS are associated with an environment that is more favorable for tumor growth. Through TCIA database, we found that the IRMPS-low group was significantly associated with higher IPS (p < 0.05) and IPS-CTLA4 (p < 0.05) scores (Fig. 9C). In the IRMPS model, we further generated the TIDE score to determine the effectiveness of immunotherapy. Lower TIDE scores suggest a lower likelihood of immune evasion, suggesting that patients may benefit from ICI therapy. In our analysis, low TIDE prediction scores were seen in the IRMPS-low group (p < 0.01; Fig. 9D), suggesting a better response to ICI therapy. However, the IRMPS-high group had higher exclusion (p < 0.001), myeloid-derived suppressor cell (MDSC) (p < 0.001), and tumor-associated macrophage (TAM) M2 (p < 0.05) scores, as well as a lower dysfunction score.

Fig. 9.Analysis of immunotherapy response of EC patients. (A) The correlation between IRMPS model and Stromal Score, Immune Score and Estimate Score in TCGA-UCEC. (B) The correlation between IRMPS model and tumor purity. (C) The correlation between IRMPS model and IPS, IPS of PD-1/PD-L1 blocker, IPS of CTLA4 blocker, as well as IPS of CTLA4 and PD-1/PD-L1 blocker via TCIA website. (D) TIDE signatures predict checkpoint inhibitors (ICIs) immunotherapy response, along with dysfunction score, exclusion score, MDSC score, CAF score, TAM M2 score in two risk groups. *p < 0.05, **p < 0.01, ***p < 0.001. TIDE: tumor immune dysfunction and exclusion; MDSC: myeloid-derived suppressor cell; TAM MS: tumor-associated macrophage M2; CAF: cancer-associated fibroblasts; pos: positive; neg: negative.

Chemotherapy is the most prevalent kind of treatment for EC patients. In our study, we forecasted the possibility that various chemotherapy medications would be effective via the GDSC database, then found that substantial variation in the IC50 between two risk groups among five chemotherapy medications. The IRMPS-low group were more sensitive to cisplatin, paclitaxel, vorinostat and vinorelbine, while the IRMPS-high group were more sensitive to bleomycin (Fig. 10).

Fig. 10.Analysis of chemotherapy response of EC patients.*p < 0.05, **p < 0.01, ***p < 0.001. IC50: half-maximal inhibitory concentration.

3.7 Correlation analysis of four molecular subtypes and IRMPS model in EC

via TCGA-UCEC dataset, 545 EC patients were divided into four molecular subtypes, viz. polymerase-epsilon (POLE), microsatellite instability (MSI), copy number low and copy number high. After analyzing the correlation between four molecular subtypes and two risk groups, the prognosis signature was constituted of 7 immune-related metabolic genes, we found that the low-risk group had more MSI cases than IRMPS-high group (Fig. 11A). Fig. 11B displayed that EC patients with copy number high mainly manifested as high-grade histological type, and were distributed in IRMPS-high group more than IRMPS-low group. Then, we used TIMER database to investigate the infiltration level of 6 TILs in four molecular subtypes. The infiltration level of CD8+ T cells and dendritic cells were significantly highest in MSI, compared to other three molecular subtypes (Fig. 12A), indicating that MSI type endometrial cancer had more TILs abundance. We further explored the relationship among four molecular subtypes with six immune checkpoint markers (PD-1, PD-L1, PD-L2, CTLA4, LAG3 and HAVCR2). The results demonstrated that the expression levels of PD-1, PD-L1, PD-L2, CTLA4 and LAG3 were meaningfully higher in MSI than other types (Fig. 12B), suggesting that immune checkpoint inhibitors may be effective against MSI type endometrial cancer. The four molecular subtypes did not significantly differ in their stromal, immune and ESTIMATE scores (Fig. 12C).

Fig. 11.The correlation of four molecular subtypes, tumor grade and two risk groups in EC patients. (A) Distribution of four molecular subtypes in high and low risk groups. (B) The Sankey diagram of four molecular subtypes, tumor grade and two risk groups. POLE: polymerase-epsilon; MSI: microsatellite instability.

Fig. 12.Immune activity analysis of the four molecular subtypes in EC. (A) The infiltration level of 6 TILs in four molecular subtypes based on TIMER database. (B) Correlation between immune checkpoint and four molecular subtypes, including PD-1, CTLA4, PD-L1, PD-L2, LAG3 and HAVCR2. (C) The correlation between four molecular subtypes and Stromal Score, Immune Score and Estimate Score in TCGA-UCEC. POLE: polymerase-epsilon; MSI: microsatellite instability; neg: negtive; pos: positive.