Lactate metabolism-related genes to predict the clinical outcome and molecular characteristics of endometrial cancer

Fragments per kilobase transcript(FPKM) data for EC were downloaded from the TCGA data portal(https://​portal.​gdc.​cancer.​gov/​), which includes 23 normal and 552 tumor samples. The somatic mutation profile, and corresponding clinical information of EC were as well obtained from the TCGA data portal. Data processing was performed using Perl software to generate a matrix of all differential genes in EC. Data analysis was performed with R software (version 4.2.1) and R Bioconductor packages.

By searching with the keywords “lactate” the Molecular Signatures database v7.4 (MSigDB; https://​www.​gsea-msigdb.​org/​gsea/​msigdb) [19], and 5 lactate-related pathways (GOBP_LACTATE_METABOLIC_PROCESS、HP_INCREASED_SERUM_LACTATE、HP_LACTIC_ACIDOSIS、HP_LACTICACIDURIA、HP_SEVERE_LACTIC_ACIDOSIS) were obtained, and 289 lactate metabolism-related genes were integrated by eliminating the duplicated genes. The mRNA expression matrix and LMRGs-related genes were analyzed by the limma package in R, and the expression matrix of LMRGs-related genes was extracted. 123 genes with a fold change (FC) > 1.5 and false discovery rate (FDR) < 0.05 were defined as differentially expressed LMRGs between tumor samples and adjacent tissues.

We further filtered these 123 lactate phenotype-associated genes for prognostic markers. Using univariate Cox regression analysis, we identified 18 prognostic markers were used as lactate-related markers with p-values < 0.05. The coefficients of each marker were determined by Cox regression models. Tumor mutation burden analysis was conducted on the 18 genes screened using the “maftools” package in R. To avoid overfitting, we further performed LASSO Cox regression (1000 iterations) with the “glmnet” package and the “survival” [20‐22]. Following the screening of the LASSO regressions, the selected LMRGs were used to establish the LMRGS via multivariate Cox regression analysis.

The LMRGS score was calculated as the following formula: LMRGs score = expression level of gene1 × coefficient of gene1 + expression level of gene2 × coefficient of gene2 +. . + expression level of genen ×coefficient of gene_n [23]. The risk score of each patient was predicted by using the ‘predict’ function included in the survival R package. Two subgroups according to the median LMRGs score of EC patients were classified, including the LMRGs-high and the LMRGs-low groups. To evaluate the accuracy of grouping, we performed principal component analysis for all differential genes, LMRGs and riskGenes. To assess the prognostic value of LMRGs, we performed Kaplan-Meier (KM) survival analysis to compare overall survival (OS) and Progression free survival (PFS) between the two groups of LMRGs. To explore the impact of LMRGS on EC progression, we elucidated the relationship between LMRGS and clinicopathological factors, including age, stage, pathological grade and risk score. To estimate the accuracy of the constructed models, multivariate independent prognostic analyses were performed and ROC curves for the above clinical characteristics were plotted using the “timeROC” package, “survival” package and, “survminer” package.

To investigate the prognostic effect of lactate fraction, the clinical characteristics were integrated in multivariate Cox regression analysis. The prognostic nomogram to predict the 1-year, 3-year, and 5-year OS of EC patients in the TCGA dataset were constructed [24]. Calibration curves were employed to assess the consistency between predicted survival and actual survival. The ROC curves over time were employed to evaluate the specificity and sensitivity of the model.

Thorsson etc. [25] have developed a new global immune classification of solid tumours based on the transcriptomic profiles of over 10,000 patients from all 33 non-haematological The TCGA cancer types in 2019. Six distinct immune subtypes (ISs) were identified, including C1 (wound healing), C2 (IFN-gamma dominant), C3 (inflammatory), C4 (lymphocyte depleted), C5 (immunological quiet), and C6 (TGF-beta dominant). The relationship between risk score and immunophenotyping were calculated using the “ggpubr” package. To detect immune cell infiltration in EC patients, the R package CIBERSORT (CIBERSORT, script v1.03) were utilized to estimate the abundance of 22 immune cell types based on gene expression data. The immune-associated signatures from previous studies were referenced and calculated their scores using genomic variance analysis (GSVA).

To evaluate the clinical application of LMRGs in EC therapeutics, the IC50 of chemotherapeutic agents commonly used in the EC dataset TCGA project was calculated using the algorithm developed by Geeleher et al. and the corresponding R package ‘pRRophetic‘ [26, 27]. This algorithm permits the user to predict clinical chemotherapy response by using only baseline tumor gene expression data, which is achieved by taking cell lines from the Cancer Genome Project’s Gene expression and drug sensitivity data from cell lines in the Cancer Genome Project to build statistical models. Differences in IC50s for common antineoplastic drugs between the high- and low-risk groups of LMRGs were compared using the Wilcoxon signed rank test.

For LMRGs, GO and KEGG pathway enrichment analyses (showing the top 20) were performed with the R package clusterProfiler and were plotted as box plots and bubble plots, respectively. GO analyses consisted of biological processes (BP), cellular components (CC), and molecular functions (MF). The KEGG database has been developed as a computer model of biological information systems represented in terms of molecular interaction and reaction networks [28]. We have got permission to use the KEGG software from the Kanehisa laboratory. GO and KEGG pathways were identified as significantly enriched with a P value < 0.05. For high and low-risk groups, the activity of signature pathways was analyzed using GSVA.

Total RNA from cells was extracted by using TRizol (#9108, RNAiso Plus, Takara, Japan) in accordance with the instructions. The cDNA was obtained with PrimeScriptRT Reagent Kit (#RR037A, Takara, Japan). qRT-PCR was conducted by Biosystem StepOne Plus PCR System (ABI) and Real-Time PCR Kit (Takara, Japan). The expression levels of RNA were calculated by the 2-ΔΔCT method using the expression of GAPDH as an internal reference. qRT-PCR with sense and antisense primers were as follows: GAPDH 5′-GGCAGAGATGATGACCCTTTT-3′ and 5′-AGATCCCTCCAAAATCAAGTGG-3′; TIMM50: 5′- ACTGTGCACGAGGTTGGCGA-3′ and 5′-GTCCACCGGGTTGTTTCCAAAG-3′.

EC cell lines consisted of HEC-1B, HEC-1 A, Ishikawa, RL-952, and KLE, which were purchased from the American Type Culture Collection (ATCC; Manassas, VA). Ishikawa, HEC-1B, RL-952, and KLE cells were cultured with DMEM/F12 medium (#11,330,032, Gibco), HEC-1 A cells were cultured with McCoy’s 5 A medium (#16,600,082, Gibco), complemented with 10% fetal bovine serum (#10,099,141, Gibco) and 1% streptomycin and penicillin (#PYG0016, Bosterbio, USA) in a 37 °C, 5% CO2 incubator.

The shRNA of TIMM50 was purchased from GenChem(Shanghai, China). EC cells were seeded in six-well plates and transfected with TIMM50 shRNA and control shRNA when the fusion level reached 60%-70%. TIMM50-sh1: 5’-CCGGACAUACAAAUAUUUUCAA-3’; TIMM50-sh2: 5’-GACACCAUGUAAAGGAUAUUU-3’; TIMM50-sh3: 5’-CCUCAAGACCAUUGCACUGAA-3’; TIMM50-Negative control: 5’-UUCUCCGAACGUGUCACGU-3’.

Cells(1*10^4 cells / well)were plated in 96-well plates and incubated overnight for EdU (5-ethynyl-2′ deoxyuridine) assay. Edu assay was performed using Edu kit (#C0071S, Beyotime, shanghai), according to the instructions. The EdU solution was prepared as a 1:1000 concentration of medium and added to the pre-prepared 96-well plates with 100 ul per well, which were incubated at 37 °C for 2 h and fixed in 4% paraformaldehyde for 15 min. The fixed cells were incubated with 50ul of reaction solution for 30 min sheltered from light, and then the nuclei were stained for 10 min by using DAPI. Cell proliferation was imaged by utilization of a 20x fluorescence microscope.

Cell Counting Kit-8 Assay (#34,302, CCK8, Bimake, USA) was applied to detect cell proliferation. Cells (4 × 10^3/well) were cultured in 96-well plates and CCK8 solution and cell culture medium mixture(1:9) were added to 96-well plates, shielded from light, and incubated at 37℃ for 1 h. The absorbance was measured at 450 nm at four time points: 24 h, 48 h, 72 and 96 h as instructed. The absorbance was measured by using an automated microplate reader (BioTek, VT, USA).

The 8 μm pore size chambers were used for migration experiments. 200 µl of serum-free cell suspension containing Ishikawa cells and HEC-1B cells (1 × 10^4/well) was added above the chambers, and 600 µl of medium containing 20% fetal bovine serum was added below the chambers. After 24 h of incubation, the chambers were removed, fixed with 4% paraformaldehyde for 10 min, stained with crystal violet, and photographed with a 10x microscope.

Total cellular protein was extracted utilizing RIPA lysate, and protein quantification was performed by BCA protein assay. Electrophoresis was performed using a 10% PAGE gel, followed by transfer to a 0.2 μm PVDF membrane on a Trans-Blot Turbo transfer system. After 1.5 h of skim milk closure and overnight incubation with primary antibody at 4 °C, the membranes were washed with TBST. The membranes were processed with diluted secondary antibodies for 1.5 h at room temperature. Proteins were visualized with ECL detection reagents and ChemiDoc imaging system. Since the molecular weight of TIMM50 protein was 40 kDa, the membrane was cut at around 40 kDa and incubated with antibody.

Sections were dissociated with xylene, rehydrated with ethanol, and antigen retrieval was performed in a microwave oven (121 °C, 15 min). Endogenous peroxidase activity was obstructed with 5% H₂O₂. Sections were then incubated in 10% normal mouse serum to block nonspecific binding of the antibody and incubated overnight at 4 °C with primary antibody TIMM50 (1:200, ABclonal, Wuhan, China). Sections were subsequently incubated with secondary antibodies for 2 h at room temperature, immersed in 0.1% diaminobenzidine tetrahydrochloride solution for 5 min, and counterstained with hematoxylin.

Dewaxing, 10 min each for dewaxed xylene I and II, coverslips prepared in advance. Overlay water, 100% (I, II), 90%, 80%, 70% alcohol for 5 min each, tap water rinse for 5 min × 3. Hematoxylin staining for 5 min, rinsed with running water. 5% acetic acid differentiation for 1 min, rinse with running water, add acetic acid by pipette drops, cloth can be covered with the tissue on the slide, the color became lighter and blue after differentiation. Eosin staining for 1 min. Dehydration: 70%, 80%, 90%, 100% alcohol for 10 s each, xylene for 1 min, can be dried naturally in the fume hood and then sealed piece, about 5 min or so. Drop on neutral gum, seal the film.

For cells in wall culture, the cells were scraped off with a cell scraper and the culture medium was centrifuged at 1000 rpm for 10 min at room temperature, and the supernatant was discarded and the cell sediment was left. Add 0.5ml of PBS to the cell precipitate and mix well, suspend the cells in PBS, transfer the cell suspension to a glass homogenization tube (2ml glass homogenization tube) with a pipette, place the glass homogenization tube in ice water mixture and homogenize manually for 3 min, then take the broken cell suspension for determination of lactate and glucose content.

Female BALB/c-nu nude mice (4–5 weeks old) (Charles River, China) were housed in a standard pathogen-free environment laboratory. Mice were randomly divided into two groups, sh-NC and sh-TIMM50. Treated HEC-1B (1 × 10⁶) cells were suspended in 100 µl serum-free MEM medium and subsequently injected subcutaneously into the right shoulder nail of mice. The subcutaneous tumor size was measured weekly, and after 28 days, the mice were treated with cervical dislocation and the subcutaneous graft tumors were removed, and the weight, length and width of the tumors were measured. Tumor volume = (length × width²)/2. Tumor samples were partially embedded in paraffin for histopathological analysis.

Quantitative analysis was performed using appropriate statistical methods. Student’s t-test or Wilcoxon signrank test was applied for comparisons between the two groups and one-way ANOVA for comparisons between more than two groups, depending on whether the samples conformed to parametric tests [29, 30]. KaplanMeier survival rates were compared across subgroups using the log-rank test. The Spearman correlation method was employed to test the correlation between the two groups of numerical variables. Cox proportional hazards models were applied to analyze hazard ratios (HR) with 95% confidence intervals (95% CI). A value of P less than 0.05 was considered significant. The statistical analyses were performed in R or Graphpad.

RNA-sequencing and clinical data of 552 EC and 23 normal endometrium tissues were downloaded from the TCGA database. To understand the metabolic characteristics of lactate in EC, 289 lactate genes from 5 lactate metabolic pathways were integrated in MSigDB. The workflow of the lactate metabolism-related gene signature analysis is demonstrated in Fig. 1. Based on RNA-seq profiling of the TCGA EC cohort, 122 were identified differentially expressed between EC tissue and normal endometrial tissue. Heat map and volcano map demonstrated 19 genes were low expression, whereas 103 genes high expression (Fig. 2a-b). To further explore the function of differentially expressed lactate metabolism-related genes (LMRGs) in EC, GO functional enrichment analysis and KEGG signaling pathway analysis on 122 lactate genes were performed. As shown in Fig. 2c, the BP is involved in generation of precursor metabolites and energy, energy derivation by oxidation of organic compounds, ATP metabolic process, cellular respiration, aerobic respiration, electron transport chain. The CC were mainly participated in the mitochondrial matrix, mitochondrial inner membrane, mitochondrial protein-containing complex, respiratory chain complex, NADH dehydrogenase complex, etc. The MF were mainly focused on electron transfer activity, oxidoreduction-driven active transmembrane transporter, NADH dehydrogenase activity and transporter activity. KEGG enrichment analysis was focused on Oxidative phosphorylation, Diabetic cardiomyopathy, and Thermogenesis (Fig. 2d).

To confirm the association between LMRGs and prognosis, univariate COX regression analysis was performed and 18 genes were identified to be associated with EC prognosis, including TP53, MRPL3, NDUFA11, NAXE, CYP27A1, PDSS1, RNASEH1, ACAT2, MECP2, NDUFB9, PDHA1, CARS2, HPDL, ATPAF2, TIMM50, FBP1, ALDOB, NDUFA6(Fig. 3a). After LASSO regression analysis and multifactorial COX regression analysis(Table 1), only 7 genes were eventually acquired (Fig. 3b-c). Tumor mutation load of EC patients were downloaded from the TCGA database and the 18 genes were screened univariately. The results indicated that TP53 was the most mutated gene (Fig. 3d-e). To further validate the screened genes, the online database TIMER was used for prediction and CYP27A1, MRPL3, NDUFB9, TIMM50, and FBP1 were associate with EC patient prognosis. Of these, CYP27A1 and FBP1 were positively associated with prognosis of EC patients, while the opposite was observed for MRPL3, NDUFB9, and TIMM50 (Fig. 3f).

To construct a reliable prognostic signature for EC patients, EC patients were categorized into a high-risk group and a low-risk groups based on the median risk score. Heat map analysis (Fig. 4a) shown seven genes were differentially expressed in the high- and low-risk groups, which were ALDOB, NAXE, NDUFA6, ACAT2, CYP27A1, CARS2, and TIMM50. Patients were categorized into two groups based on the optimal cutoff value for the risk of LMRGs for diagnostic accuracy (Fig. 4b). More mortality events were observed in the high-risk group, suggesting that the increased risk of LMRGs reflected an unfavorable prognosis for EC patients (Fig. 4c). Principal component analysis (PCA) at different levels was employed to verify whether the risk states among different influencing factors were relatively independent. PCA demonstrated that samples with two risk scores were divided into two independent groups (Fig. 4d). Kaplan–Meier survival curves analyses showed that the high-risk subgroup had a shorter OS and PFS than the low-risk groups (Fig. 4e-f). To further investigate whether the prognostic signature could be an independent predictor of survival in EC, univariate and multivariate Cox regression analyses were performed. Univariate Cox regression analysis revealed that risk score was an independent predictor of poor OS in EC patients (HR = 4.778, 95% CI: 2.870–7.955). The results of multivariate COX regression analysis were the same (Fig. 4g-h). Remarkably, the predictive capability of these traditional clinical parameters (age, grade, and stage) was significantly lower than that of the LMGRs risk score. ROC curves of clinically relevant factors were conducted to assess the performance of the risk prediction model, which described excellent predictive capability. The AUCs for risk, age, grading and staging were 0.774, 0.613, 0.684 and 0.732, respectively (Fig. 4i). To prove the stability of the model, 503 EC patients (entire set) were randomly allocated to the train set (n = 352) and test set (n = 151) by seven-to-three ratio. The distribution of risk score (Supplement Fig. 1a), the survival status (Supplement Fig. 1b), the survival outcome(Supplement Fig. 1c-d), univariate and multivariate Cox regression analyses and the ROC curve (Supplement Fig. 1e-f) of EC patients between two groups of the train set and test set were constructed, respectively. All results in the two sets show no difference. The risk curves and scatter plots for the train and test sets implied mortality was positively related to the risk score in two sets. Kaplan–Meier survival analysis indicated that patients in the low-risk groups demonstrated better OS and PFS than patients in the high-risk groups in train and test sets. Univariate and multivariate Cox regression analyses and ROC curves were performed. Remarkably, the predictive capability of these traditional clinical parameters (age, grade, and stage) was significantly lower than that of the LMGRs risk score of train and test sets. ROC curves, assessing the accuracy of this risk model in two sets, indicated that the LMGRs model is reliable and precise. Risk score could be used as an effective prognostic marker in EC.

To accurately predict the probability of OS, a nomogram integrating the LMRGs risk score and other clinicopathological features, including age, stage, and grad was constructed to assess survival at 1, 3 and 5 years. The results suggested that the OS of EC patients decreased along with increasing time (Fig. 5a). Moreover, the calibration curve was designed to assess the predictive accuracy of 1-, 3-, and 5-year survival rates. Interestingly, the predicted risk was similar to the actual risk, indicating the reliability of our nomogram (Fig. 5b). To further determine whether nomogram could be used as an independent prognostic indicator for EC patients, univariate and multivariate Cox regression analyses were performed. The outcomes showed that in the univariate Cox analysis, age, stage, grade, and nomogram were all significantly related to the probability of OS. However, in multivariate COX regression analysis, grade was as well an independent risk factor for prognosis in EC patients (Fig. 5c-d). ROC curves were plotted for nomogram versus clinical characteristics, and the results demonstrated that nomogram (AUC = 0.833) exhibited more favorable clinical predictive accuracy than any single factor (Risk, AUC = 0.774; age, AUC = 0.613; grade, AUC = 0.684; stage, AUC = 0.732) (Fig. 5e). Eventually, KEGG signaling pathway analysis was performed for the high- and low-risk groups, which were mainly enriched in Basal Transcription factors, Cell Cycle, Glycosylphosphatidylinositol gpi anchor biosynthesis, alanine aspartate and glutamate metabolism, citrate cycle TCA cycle, pyruvate metabolism, cysteine and methionine metabolism, et al. (Fig. 5f).

The analysis of the correlation between immunotyping and risk scores revealed that C3 was the most infiltrated in the low-risk groups and significantly different from C1, C2 and C4(Fig. 6a). To thoroughly analyze the immune microenvironment, CIBERSORT was used to calculate the degree of infiltration of 22 immune cell types. As shown in Fig. 6b, infiltration of Plasma cells, T cells regulatory (Tregs), and Dendritic cells activated were significantly different in the high and low-risk groups of EC patients. To know more about the activation of immune-related functions in EC, the correlation analysis of immune function and risk scores was conducted. The results(Fig. 6c) showed that APC co-stimulation, CCR, Check-point, Cytolytic activity, Inflammation-promoting, MHC class I, T cell co-inhibition, T cell co-stimulation and Type II IFN Response were more active in the low-risk groups compared to the high-risk group, while Type I IFN Response showed the opposite trend. Sensitivity to common anticancer drugs was compared between the high- and low-risk groups to determine potential treatment modalities for EC. The results showed that IC50s for Scr protein kinase inhibitors were higher in low-risk patients, suggesting that low LMRGs risk was more resistant to Scr kinase inhibitors compared to high-risk (Fig. 6d). These drugs might be the future guidance for the treatment of patients at the high-risk score of EC.

To explore the exact roles of TIMM50, qRT-PCR and Western Blot were performed to detect the expression of TIMM50 in different cell lines of EC in terms of RNA and protein level, respectively (Fig. 7a-b). TIMM50-si was transfected into Ishikawa and HEC-1B cells to verify its knockdown efficiency (Fig. 7c-d). CCK-8 assay showed that TIMM50 suppression repressed the viability of Ishikawa and HEB-1 C cells (Fig. 7e). EdU assay suggested that the inhibition of TIMM50 enhanced the apoptosis of Ishikawa and HEC-1B cells (Fig. 7f). The colony formation assay indicated that TIMM50 interference significantly restrained the colony formation ability of Ishikawa and HEC-1B cells (Fig. 7g). Transwell assay revealed that TIMM50 silencing repressed the invasion ability of Ishikawa and HEC-1B cells (Fig. 7h). Lactate and glucose assay assays (Fig. 7i) suggested that lactate synthesis was reduced in Ishikawa and HEB-1 C cells by silencing TIMM50, while glucose content showed the opposite trend. Hence, the ability of proliferation, migration and lactate synthesis of EC cells could be promoted by TIMM50 in vitro.

To explore the effect of TIMM50 on the growth of EC cells in vivo, the animal experiments were performed. HEC-1B cells were stably transfected with TIMM50-sh3 and TIMM50-NC vectors were injected subcutaneously into female BALB/c nude mice, and tumor-related indices were measured once a week and removed after 4 weeks. The size and weight of tumors in the TIMM50-silenced vector group were significantly smaller than those in the NC group (Fig. 8a-c). HE staining (Fig. 8d) showed the cellular morphology of subcutaneously transplanted tumors. In addition, immunohistochemical staining of these subcutaneous tumors demonstrated that TIMM50 protein levels were significantly decreased in the TIMM50-sh3 group compared to the NC group (Fig. 8d). Overall, the results of these experiments revealed that TIMM50 could regulate the proliferation of EC cells in vivo.