LGR5, a prognostic stem cell target, promotes endometrial cancer proliferation through autophagy activation

Highlights

• •A ten stemness-related gene prognostic model was constructed in TCGA cohort, which was validated in PKUPH cohort. The Kaplan-Meier plot and ROC curves indicated higher accuracy.
• •LGR5 was high expressed according to TCGA cohort and single-cell sequencing analysis, which also could significantly influence the prognosis of EC patients and involved in Wnt signaling pathway.
• •LGR5 expression was related to age, stage, histological type, molecular subtypes, and menopause status.
• •The GSEA analysis showed the autophagy signaling was activated in the LGR5 high-expression group. Immunofluorescence staining and western blot showed autophagy activation in overexpression LGR5 group.

Abstract

Endometrial cancer (EC) is a common malignant tumor in women worldwide. Although early EC has a good prognosis, advanced endometrial cancer is still associated with the risk of drug resistance and recurrence. Cancer stem cells (CSCs), a category closely related to drug resistance and recurrence, are rarely studied at present. Here, we constructed a risk model containing ten stemness-related prognostic genes. Compared with patients in the low-risk group, patients in the high-risk group had a shorter overall survival time. The accuracy of this model was verified by ROC in the TCGA (AUC = 0.779) and Peking University People's Hospital (PKUPH, AUC = 0.864) cohorts. The risk score and stage were independent risk factors in the multivariate regression analysis, which was subsequently used to construct the nomogram and verified in the TCGA cohort. LGR5 was significantly correlated with overall survival and involvement in the Wnt signaling pathway. In addition, LGR5 was highly expressed in EC tissues and was related to age, stage, histological type, and menopause status in the TCGA database. Overexpression of LGR5 accelerated the proliferation rate of EC cells, which may be related to autophagy activation. Taken together, our study established a prognostic model based on transcription sequencing data from the TCGA database and verified it in the PKUPH cohort, which has prospective clinical implications for the prognostic evaluation of EC. We systematically studied the code gene LGR5 in EC, which may help clinicians make personalized prognostic assessments and effective clinical decisions for EC.

Introduction

Endometrial cancer (EC) is the most common malignant tumor of the female reproductive system worldwide and the second most common in China after cervical cancer [1,2]. The incidence rate for EC has been stable in recent years, and there will be approximately 84,520 new cases and 17,543 deaths in the Chinese mainland in 2022, according to the same methodology as GLOBOCAN 2020 [3]. However, although 70–80 % of EC patients are diagnosed at an early stage with a 5-year survival rate of 81 % overall, patients with advanced and recurrent EC have a poor 5-year survival rate under 20 % [1]. Thus, a full understanding of EC and the development of a novel therapeutic target are needed.

Cancer stem cells (CSCs) are a small subpopulation of cells within tumors that can initiate tumors and exhibit self-renewal and differentiation properties. The first evidence of CSCs in EC was derived from Hubbard's study [4]. Subsequently, increasing evidence suggests that CSCs contribute to tumor recurrence and metastasis. Furthermore, CSCs are intrinsically chemoresistant, which results in a low response rate to chemotherapy [5]. CSCs are of enormous research value since they may be responsible for initiating tumor as well as tumor growth, metastasis, relapse, and treatment resistance. Until now, endometrial cancer stem cells (ECSCs) have been proposed to express a variety of markers, including CD44 [6], CD133 [7], CD55 [8], Musashi-1 [9], BMI1 [10], and SOX9 [11]. There are many pathways involved in the regulation of ECSCs, such as the Wnt/β-catenin, Hedgehog, NOTCH, and PI3K/AKT signaling pathways [12]. However, recent clinical studies have shown that the effects of drugs that target CSC biomarkers and pathways are limited. Therefore, validation of diagnostic ECSC biomarkers is essential for the development of successful, biology-oriented therapeutic strategies.

Accumulating evidence has proven the critical role of a prognostic model in identifying diagnostic biomarkers. Recently, most studies have used mRNAsi together with weighted gene coexpression network analysis (WGCNA) to identify prognosis-related stemness genes [13], [14], [15]. Liu et al. identified six genes related to stemness in endometrial carcinoma [16]. Nevertheless, the role of these prognostic biomarkers in endometrial cancer is not well documented. Due to the heterogeneity of tumor cells, single-cell sequencing technology provides a good tool for a comprehensive understanding of cancer cells [17]. In this study, we used the stemness-related gene list from the CancerSEA database to construct the prognostic model. This database explored distinct functional states (stemness, differentiation, cell cycle, etc.) of cancer cells at the single-cell level [18].

Leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5), also known as GPR49, is a member of the G protein-coupled receptor family, which is involved in the activation of the canonical Wnt/β-catenin signaling pathway [19,20]. Recently, more evidence has indicated that LGR5 is a cancer stem cell marker in various tumors that exhibit multiple biological functions. In general, LGR5 has been demonstrated to be upregulated in cancers, such as colorectal cancer [21], glioma [22], hepatocellular carcinomas [23,24], and ovarian cancer [25]. LGR5-positive stem cells in the intestinal tract were proven to be the original cells of intestinal cancer [26]. In terms of biological functions, LGR5 has been demonstrated to promote epithelial-mesenchymal transition (EMT) in glioma by activating the Wnt/β-catenin pathway [27]. However, LGR5 suppressed colon cancer metastasis by activating TGF-β signaling [28,29]. The above research indicated that LGR5 plays a paradoxical role in different cancers. The human endometrium is a highly regenerative tissue that is thought to have a stem cell basis. Many studies have highlighted increased LGR5 expression in endometrial epithelial cells, which indicates that LGR5 serves as an endometrial epithelial stem cell marker [30]. However, the role of LGR5 in EC remains uncertain. Thus, it is critical to determine the exact role of LGR5 in endometrial cancer.

In this study, we constructed a prognostic model based on transcription sequencing data from the TCGA database and verified it in the PKUPH cohort. Simultaneously, LGR5 was identified as a novel diagnostic biomarker of ECSCs that promoted the proliferation of EC cells through autophagy activation.

Material and methods

Data source

We downloaded the original transcription data (in the FPKM format) of EC patients from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov) for the training cohort, containing 552 EC samples and 35 normal samples. The related clinical data were obtained from the UCSC Xena website (https://xenabrowser.net). The stemness-related gene list was obtained from the Cancer Single-cell State Atlas website (CancerSEA, http://biocc.hrbmu.edu.cn/CancerSEA/goDownload), which contained 166 stemness-related genes, as shown in Table S1. A total of 24 specimens for the testing cohort were collected from patients who underwent surgery between January 2008 and December 2012 at the Peking University People's Hospital (PKUPH), and total RNA was extracted to construct the transcription sequencing, which had been mentioned before [31]. 10x single-cell data of five EEC samples and five samples from the website (https://www.ncbi.nlm.nih.gov/sra?linkname=bioproject_sra_all&from_uid=786266).

Identification of stemness-related DEGs

The “Edge R” package was used to identify differentially expressed genes (DEGs) between normal and tumor samples with criteria of |log2-fold change (FC)|>1.0 and adjusted to p < 0.05. We next used a Venn diagram tool (http://jvenn.toulouse.inra.fr/app/example.html) to identify the intersecting genes between DEGs and stemness-related genes for further analysis.

Establishment of a prognostic model in the training cohort

Univariate regression analysis, least absolute shrinkage and selection operator (LASSO) regression analysis, and multivariate regression analysis were applied to explore the correlation between the expression levels of stemness-related genes and patient overall survival (OS). The “survival” and “glmnet” R packages were employed to complete the analysis. The formula of the risk score was calculated as follows:

$$\text{Risk}\text{score}=\Sigma \text{coefficient}\left(\text{gene}\right)\times \text{expression}\left(\text{gene}\right).$$

Subsequently, the TCGA cohort patients were divided into high- and low-risk groups based on the median value of the risk score.

Internal and external validation of the prognostic model

The risk score and survival status distribution in the low- and high-risk groups were analyzed with the “pheatmap” package in the TCGA and PKUPH cohorts, respectively. The receiver operating characteristic (ROC) curve and Kaplan‒Meier survival curve were drawn to evaluate the accuracy of the predictive model.

Construction and assessment of a nomogram

To further evaluate whether the risk score was an independent risk factor for OS, we carried out univariate and multivariate Cox regression analyses with risk score, age, stage, grade, and histological type. We next chose the independent risk factor (p < 0.05) from multivariate Cox analysis to construct a nomogram using the R package “rms”. Three-year and 5-year calibration curves were drawn, and the concordance index (C-index) was calculated to assess the accuracy of the nomogram.

Functional enrichment analyses

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of 58 stemness-related DEGs were conducted using the “clusterProfilter” and “enrichplot” packages in R software. Only terms with both p values and adjusted p values < 0.05 were considered significantly enriched, and the top ten terms were drawn by the “ggplot2” package. GSEA (http://www.gsea-msigdb.org/gsea/index.jsp) was performed to study the functions associated with the prognostic model and the expression of LGR5. The enriched KEGG pathways were considered statistically significant with the criteria of |NES| ≥ 1.0, NOM p-val ≤ 0.05, and FDR q-val ≤ 0.25.

Single-cell clustering and analyses

We downloaded 10x single-cell data of five EEC samples and five Normal endometrial samples from the open-access database. Then, we followed the basic Seurat single-cell data analysis process to filter, standardize, mutate genes, reduce dimensionality, and annotate cell types. We used the t-distributed stochastic neighbor (t-SNE) algorithm to explore and visualize cluster classifications across cell samples, and marker genes (EPCAM, KRT8, KRT18) were used to define epithelial cells. The expression of stemness-related DEGs was compared between the EC (EEC sample) and NC (Normal endometrial sample) groups by using Student's t-test.

Immunohistochemistry analysis

LGR5 expression in clinical EC tissues was analyzed by immunohistochemistry (IHC) staining. The tissue microarray (TMA, HUteA060CS01) containing 34 EC samples and 26 adjacent normal tissues was purchased from Shanghai Outdo Biotech Co., Ltd. A total of 47 EC tissues with clinicopathological features were obtained from patients who underwent primary surgery at PKUPH between 2020 and 2021. The tissue samples were fixed with 4 % paraformaldehyde, embedded in paraffin, and then cut into 4 µm sections. After xylene dewaxing and a series of ethanol dehydrations, the paraffin sections were placed in sodium citrate solution and heated in a microwave for antigen retrieval. Other steps were performed according to the SP-9000 (ZSGB-BIO, Beijing, China) manufacturer's recommendations. Representative images of each tissue were captured by microscopy and scored according to the intensity of the dye color and the area of positive cells. The staining intensity was divided into 4 grades, including 0 (no color), 1 (light yellow), 2 (light brown), and 3 (brown). The area of positive cells was graded as 0 (<5 %), 1 (5–25 %), 2 (25–50 %), 3 (51–75 %) and 4 (>75 %). The two parts were added together to obtain the final score. Anti-LGR5 antibody (1:300, bs-20747R, Bioss, Beijing) was used.

Cell culture and transfection

The human EC cell lines Ishikawa, AN3CA and HEC-50B were obtained from the Department of Gynecology Laboratory of Peking University People's Hospital. Ishikawa cells were cultured in DMEM/F12 supplemented with 10 % fetal bovine serum, while AN3CA and HEC-50B cells were cultured in MEM. Cells were cultured at 37 °C in the presence of a humidified atmosphere with 5 % CO₂. All cell lines were free of mycoplasma and were verified by short tandem repeats (STRs).

The lentiviral vectors encoding LGR5 gene short-hairpin RNA (sh-LGR5), LGR5 overexpression (OE-LGR5), and its negative controls were purchased from Shanghai Genechem Co., Ltd.

Western blotting

Cells were lysed in RIPA lysis buffer with cocktail for 30 min on ice. The supernatant was obtained by centrifugation and quantified by the BCA method. A total of 30 µg proteins were separated on polyacrylamide gel (10 %, 12.5 %) and then transferred to a polyvinylidene fluoride (PVDF) membrane. After blocking with 5 % nonfat milk, the membranes were incubated with diluted primary antibodies overnight at 4 °C. After washing with TBST, the membranes were incubated with secondary antibodies for 1.5 h at room temperature. Bands were detected by a chemiluminescent imaging system (ChemiDoc MP, Bio-Rad, USA) and quantified by ImageJ software. Information on the antibodies used is listed in Table S2.

RNA extraction and quantitative real-time PCR

Total RNA was extracted by TRIzol reagent (Invitrogen, Thermo) and detected using a spectrophotometer (Nano500, ALLSHENG, China). A total of 1000 ng of mRNA was reverse-transcribed into cDNA and then used as a template for quantitative real-time PCR (qRT‒PCR). Primer sequences for the human genes are listed in Table S2.

Spheroid formation assay

A total of 1 × 10³ cells were plated into ultralow attachment six-well plates, and the cells were cultured at 37 °C for 10 days in DMEM supplemented with 20 ng/ml EGF, B27 (1:50) and 20 ng/ml basic FGF. The size of spheres (diameter > 70 µm) was analyzed under a microscope.

CCK8 assay

CCK8 assays were used to analyze cell proliferation; 0.4 × 10⁴ cells were seeded into 96-well plates and cultured overnight. Then, 10 µl of CCK8 solution was added to each well at different time points (0, 24, 48, 72 h) and incubated at 37 °C for 2 h. Absorbance was then detected at a wavelength of 450 nm by an infinite M200 multifunctional enzyme-labeling instrument.

Colony formation assay

A total of 1 × 10³ cells were seeded into each well of a six-well plate, cultured at 37 °C for 14 days, fixed with 4 % paraformaldehyde for 20 min, and stained with crystal violet for 30 min.

Immunofluorescence staining

A total of 7 × 10³ cells were seeded into 96-well plates and cultured overnight for adhesion. The cells were fixed with 4 % paraformaldehyde for 15 min and then permeabilized with 0.25 % Triton for 30 min at room temperature. After blocking with 1 % BSA for 30 min, the cells were incubated with primary antibody (ab192890) overnight at 4 °C. Goat anti-rat Alexa Fluor 633 (A-21,094, Invitrogen) was used as the secondary antibody to track the expression of LC3B. The nuclei of cells were stained with DAPI, and fluorescence images were continuously taken by the High Content Imaging System (Operetta CLS™).

Statistics

All experiments were performed under the same conditions in triplicate. Statistical analyses were performed with GraphPad Prism 8.0, and significant differences were evaluated by Student's t-test, chi-square test, and one-way ANOVA. p < 0.05 was considered statistically significant. *, p<0.05; **, p<0.01; ***, p<0.001. R 4.1.0 software was used for bioinformatics analysis.

Results

Construction of a stemness‑related gene prognostic model based on the TCGA cohort

The bioinformatics analysis design is depicted in Fig. S1. After “Edge R” package analysis, we obtained a total of 3826 DEGs, including 2411 upregulated genes and 1415 downregulated genes. The volcano map is shown in Fig. 1A. After the intersection with stemness-related genes, 58 genes remained (Fig. 1B), and detailed information on 58 genes is listed in Table S3. We next selected 14 stemness-related genes with prognostic value using univariate regression analysis (Fig. 1C). Ten genes were identified as independent prognostic factors after LASSO and multivariate regression analysis and were used to establish the prognostic model (Fig. 1D-F). Detailed information on multivariate Cox regression is presented in Table S4. The risk score was calculated as follows:

$$\begin{array}{c}\text{Risk}\text{score}=(\text{TCF}4*-0.255+\text{ENG}*-0.016+\text{QPCTL}*0.019+\text{ANPEP}*0.004+\text{CD}44*0.005+\hfill \\ \text{LY}6\mathrm{D}*0.016+\text{MESP}2*0.024+\text{LGR}5*0.008+\text{EPHB}2*0.093+\text{MALAT}1*0.003).\hfill \end{array}$$

Fig. 1.

Construction of the Stemness‑Related Gene Prognostic Model in the TCGA cohort. A, Volcano plot of DEGs between tumor and normal tissues of TCGA database. Red dots represent lower expressed genes in tumor tissues and the blue dots represent higher expressed genes. B, Venn plot showing intersecting genes shared by DEGs and stemness-related genes. C, Univariate Cox analysis of 14 intersecting genes in the training cohort. D, The LASSO regression model of the prognostic genes. E, Multivariate Cox analysis for the selecting stemness-related genes. F, The coefficient of selected genes for constructing a prognostic model (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Internal and external validation of the prognostic model

Patients were divided into high-risk and low-risk groups based on the median risk score. The risk score distribution and survival status of the TCGA and PKUPH cohorts are shown in Fig. 2A-B, which indicated that most patients who died were distributed in the high-risk group. The Kaplan‒Meier plot showed that patients with high-risk scores had shorter OS times (p < 0.001; Fig. 2C), and the same trend was detected in the PKUPH cohort (p < 0.001; Fig. 2D). ROC curves indicated that the prognostic model (AUC = 0.779) had higher accuracy than age, stage, grade, and histological type (AUC = 0.682, 0.694, 0.614, and 0.583, respectively), which is shown in Fig. 2E. Similarly, the accuracy of the model was demonstrated in the PKUPH cohort (AUC = 0.864; Fig. 2F). The expression levels of ten genes and their corresponding clinicopathological features in the high-risk and low-risk groups are presented in the heatmap (Fig. S2).

Fig. 2.

Internal and External Validation of the ten-gene prognostic model. A-B, Prognostic model gene expression, risk curve, and survival status were displayed in the TCGA and PKUPH cohorts, respectively. C, Kaplan-Meier plot of the high- and low-risk groups in TCGA cohort. D, Kaplan-Meier plot of the high- and low-risk groups in PKUPH cohort. E, The ROC curve analyses of the variable clinicopathological features and risk score in the TCGA cohort. F, The ROC curve analyses the risk score in the PKUPH cohort.

To reveal the prognostic model-related signaling pathways, the GSEA tool was applied to analyze the enriched KEGG pathways between the high- and low-risk groups. Compared with the low-risk group, many tumor-related and metabolism-related pathways, including the cell cycle, DNA replication, citrate TCA cycle, and glycolysis gluconeogenesis, were enriched in the high-risk group (Fig. S3).

Construction and assessment of nomogram

We next used univariate and multivariate Cox regression to verify whether the risk score was an independent risk factor for OS in the TCGA cohort. The results showed that the risk score served as an independent prognostic factor in both univariate (HR = 1.075, 95 % CI = 1.056–1.093; p < 0.001) and multivariate (HR = 1.054, 95 % CI = 1.034–1.074; p < 0.001) Cox regression (Fig. 3A and B). However, only stage (HR = 1.672, 95 % CI = 1.247–2.240; p < 0.001) and risk score were statistically significant in the multivariate analysis. Subsequently, we established a nomogram to predict the survival of EC patients using previously screened risk factors (stage and risk score), and the survival rate of each patient could be predicted by calculating the sum of all variable scores (Fig. 3C). In addition, the results of the calibration curve analysis showed good concordance between the actual and nomogram-predicted OS at 3 and 5 years (Fig. 3D).

Fig. 3.

Construction and Assessment of Nomogram. A, Univariate Cox regression analysis of the variable clinicopathological features and risk score in TCGA cohort. B, Multivariate Cox regression analysis of the variable clinicopathological features and risk score in TCGA cohort. C, Nomogram constructed by the prognostic features in TCGA cohort. D, Calibration curve of 3-year and 5-year survival in EC patients.

Screening of the hub genes

The expression of ten genes between normal and tumor samples in the TCGA group is shown in Fig. 4A, and nine genes had significantly differential expression after log conversion. TCF4, ENG, and CD44 were lower in tumor samples, while QPCTL, ANPEP, LY6D, MESP2, LGR5 and EPHB2 were higher in tumor samples. The Kaplan‒Meier plot of ten genes was analyzed, and only four genes had significant differences for OS (Fig. 4B), including LY6D, LGR5, EPHB2 and ENG.

Fig. 4.

Identification of hub gene in endometrial cancer. A, Expression of prognostic model genes in normal and tumor tissues. The red bars represent tumor tissues, and the blue bars represent normal tissues. student's t-test. B, Kaplan-Meier plot of prognostic genes, p<0.05. C, GO enrichment analysis (molecular function) of prognostic genes. D, KEGG Enrichment analysis of prognostic genes. E-G, Single-cell RNA sequencing analysis of normal and EEC samples. E, Distribution of epithelial cells, defined by EPCAM, KRT8, and KRT18 positive expression. F, Distributions of single cells from EEC and NC samples. G, Distributions of LGR5 expression in each cell are represented. t-SNE, t-distributed stochastic neighbor (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

To further explore the biological functions of stemness-related DEGs, we conducted GO and KEGG pathway annotation analyses. The top ten enriched GO terms are shown in Fig. 4C, which suggested that cell development was enriched. In addition, KEGG analysis revealed that classical stem-regulatory pathways, including the Wnt and Hedgehog signaling pathways, were enriched (Fig. 4D). Detailed information on KEGG enrichment is shown in Table S5, LGR5 is involved in the Wnt signaling pathway.

We next observed stemness-related DEG expression in the single-cell RNA sequencing data of endometrial cancer, which showed an almost completely different trend from the TCGA cohort. The results indicated that EPHB2, LY6D, MESP2, ANPEP, and QPCTL expression was extremely low, and the difference was not significant. The expression of CD44, LGR5, and MALAT1 in tumor tissues was higher than that in normal tissues, while the expression of ENG and TCF4 in normal tissues was higher than that in tumor tissues (Figs. 4E–G and S4). Only LGR5 showed a consistent trend in the TCGA cohort and single-cell sequencing data. Overall, LGR5 could significantly influence the prognosis of EC patients and be involved in the Wnt signaling pathway.

The relationship of LGR5 and clinicopathological factors in EC patients

To further clarify the role of LGR5 in the development of EC, we first confirmed the expression of LGR5 in GSE17025, and the results showed that LGR5 was more highly expressed in tumors than in normal endometrium (Fig. 5A). We next confirmed the relationship between the expression of LGR5 and various clinicopathological features in TCGA. The results showed that LGR5 expression was related to age, stage, histological type, menopause status and subtypes but not histological grade (Fig. 5B–G). Additionally, we analyzed the correlation between clinicopathological factors and the other three significant stemness-related genes, including LY6D, ENG and EPHB2. The results are shown in Fig. S5. Age over 55, early stage, endometrial serous carcinoma, post menopause period and copy-number high type had higher LGR5 expression, which was consistent with previous reports that progesterone reduces LGR5 expression. We next detected the protein expression of LGR5 in the TMA by immunohistochemical staining, and the results showed that the proportion of strongly positive staining for LGR5 was significantly higher in EC tissues than in adjacent normal tissues (44.1 % vs. 23.1 %), while the proportion of negative or poor positive staining was lower (17.6% vs. 30.8 %) (Fig. 5H). In summary, the expression of LGR5 seems to be tightly related to hormone levels and poor prognosis. To verify the function of progesterone on LGR5, we used different doses of medroxyprogesterone acetate (MPA) treated EC cells (Ishikawa and AN3CA), and the results showed that MPA could significantly decrease the expression of LGR5 in a dose-dependent manner (Fig. 5I–K).

Fig. 5.

Verification of the Relationship between LGR5 and Clinicopathological Factors in EC Patients. A, The LGR5 expression of GSE17025 dataset. B-G, Association between the risk score and clinicopathological features in TCGA, including age (B), stage (C), histological type (D), grade (E), molecular classification (F), and menopause period (G) by student's t-test. H, Representative IHC images of LGR5 expression in TMA, containing endometrial cancer tissues (n = 34) and adjacent normal tissues (n = 26). The statistical result of IHC was displayed in the right panel. I, LGR5 mRNA expression after different doses of MPA treatment (0, 30, 60 µM) by qRT-PCR. J, Immunoblot analysis of LGR5 protein expression after different doses of MPA treatment (0, 30, 60 µM). K, The relative abundance of the proteins was quantified against β-actin.

LGR5 induced stemness of EC cells

We first detected the mRNA and protein expression of LGR5 in three EC cell lines (Ishikawa, AN3CA and HEC-50B). The results showed that LGR5 was highly expressed in Ishikawa cells and expressed at low levels in AN3CA and HEC-50B cells (Fig. 6A-B). Thus, we established a stable LGR5 knockdown model in Ishikawa cells and a stable LGR5 overexpression model in AN3CA cells, and the western blot results confirmed successful construction (Fig. 6C-D). Next, a sphere formation assay was performed to assess the stemness traits of LGR5, and the results indicated that the sphere sizes were markedly increased after LGR5 overexpression and decreased after LGR5 knockdown (Fig. 6E-F). We next examined the protein expression of the stem cell markers NANOG and SOX2 by western blot, and the results showed that NANOG and SOX2 were increased after LGR5 overexpression and decreased after LGR5 knockdown (Fig. 6G-H).

Fig. 6.

LGR5 Induced Stemness of EC Cells A, LGR5 protein expression in EC cell lines (Ishikawa, AN3CA, and HEC-50B) which detected by western blot. B, LGR5 mRNA expression in EC cell lines by qRT-PCR. C-D, Knockdown LGR5 in Ishikawa cells and overexpression LGR5 in AN3CA cells, which was verified by western blot. E, Images of floating spheres of sh-LGR5 and its negative control in Ishikawa cells after 10 days. The statistical result of sphere size was displayed in the right panel. F, Images of floating spheres of OE-LGR5 and its vector in AN3CA cells after 10 days. The statistical result of sphere size was displayed in the right panel. G, Immunoblotting was carried out with antibodies against NANOG, SOX2, and β-actin. H, The relative abundance of the proteins was quantified using ImageJ software against β-actin, and data were presented as mean ± SD. student's t-test.

LGR5 promoted the proliferation of EC cells

Then, we verified the relationship between LGR5 and clinicopathological features in tissues from PKUPH, which indicated that LGR5 expression was positively correlated with age (Table 1, Fig. 7A). To explore the biological function of LGR5 in EC cells, we detected the proliferation of EC cells using the CCK8 assay and colony formation assay. As shown in Fig. 7B, the CCK8 assay indicated that knockdown of LGR5 could significantly inhibit the growth of Ishikawa cells, and LGR5 overexpression in AN3CA cells could obviously promote cell growth. The results were then further validated by the colony formation assay (Fig. 7C).

Table 1.

Clinicopathological analysis of LGR5 low and high expression.

	Total	LGR5 expression		p-value
		Low(n = 17)	High(n = 30)
Age(years)				0.029*
≤55	18	10(55.6 %)	8(44.4 %)
>55	29	7(24.1 %)	22(75.9 %)
FIGO stage				0.692
I+II	39	15(38.5 %)	24(61.5 %)
III+IV	8	2(25 %)	6(75 %)
Grade				0.509
G1	20	7(35 %)	13(65 %)
G2	18	8(44.4 %)	10(55.6 %)
G3	9	2(22.2 %)	7(77.8 %)
LVSI				0.925
Negative	30	11(36.7 %)	19(63.3 %)
Positive	17	6(35.3 %)	11(64.7 %)
Lymph node metastasis				1.000
Negative	41	15(36.6 %)	26(63.4 %)
Positive	6	2(33.3 %)	4(66.7 %)

FIGO, Federation International of Gynecology and Obstetrics; LVSI, Lymph-vascular space invasion.

Fig. 7.

LGR5 Promoted the Proliferation of EC Cells through Autophagy Activation. A, Representative images of LGR5 expression (high and low) in EC tissues from PKUPH (n = 47). Chi-squared test. *, p<0.05. B, CCK8 assay detected the proliferation rate in sh-NC or sh-LGR5 of Ishikawa cells and Vector or OE-LGR5 of AN3CA cells. C, Cloning formation assay for detecting the proliferation abilities. The statistical results were displayed in the right panel. D, GSEA analysis of LGR5 high- and low-expression. The interested differentially KEGG enriched pathways between LGR5 high- and low-expression were displayed. E, The cells were stained with anti-LC3B antibody (Alexa Fluor 633-labeled) and then visualized by the High Content Imaging System. DAPI was used to stain nuclei. Scale bar = 50 µm. F, Immunoblot analysis of LC3B-I/II, ATG5, and SQSTM1/p62 in Ishikawa sh-NC or sh-LGR5 cells and AN3CA Vector or OE-LGR5 cells. G, Immunoblots shown are representative of the results obtained from three independent assays. Values represent means ± SEM.

LGR5 induced autophagy activation

To reveal the LGR5-related signaling pathways, GSEA was applied to analyze the enriched KEGG pathways between the LGR5 high and low-expression groups. Compared with the low-expression group, many tumor-related pathways, including the MTOR signaling pathway and regulation of autophagy, were enriched in the high-expression group; oxidative phosphorylation was inactive (Fig. 7D, Fig. S6). Since autophagy signaling was activated in the LGR5 high-expression group, we believed that LGR5 might promote tumor progression through autophagy activation, and we next monitored autophagosome formation in EC cells using immunofluorescence staining and western blotting. The results showed that knockdown of LGR5 decreased autophagy, and overexpression of LGR5 increased autophagy (Fig. 7E). Immunoblot analysis showed that overexpression of LGR5 enhanced the conversion of LC3B-I to LC3B-II while increasing ATG5 and SQSTM1/p62 protein expression (Fig. 7F-G). Collectively, LGR5 induced autophagy, while autophagy flux was blocked or impaired at the late stages of the process.

Discussion

EC is a common malignant gynecological tumor with the characteristics of early detection and a good prognosis. However, the dilemma of EC treatment is also very prominent. The prognosis of advanced and recurrent EC is worse, and the optimal treatment for these patients remains controversial. CSCs were demonstrated to initiate tumors, which may lead to recurrence and drug resistance. In the present study, we developed a ten stemness-related gene signature to predict EC prognosis. The risk score was further revealed to be associated with poor prognosis in the TCGA and PKUPH cohorts. Additionally, its role as an independent prognosis predictor has been proven.

Gene Ontology analysis revealed that the differentially expressed stemness-related genes were enriched in cell development. Moreover, the KEGG analysis showed that Wnt/Hedgehog signaling pathways were enriched, which has been revealed as the canonical pathway in stem cell biology in previous studies [12]. In addition, further GSEA revealed enrichment of the cell cycle and metabolism engineering. Emerging evidence indicates that CSCs have a distinct metabolic phenotype that can be highly glycolytic or OXPHOS dependent [32,33], which is consistent with our findings.

We next selected four prognostic predictors by KM plot analysis for EC, including LY6D, LGR5, EPHB2, and ENG, and found that LGR5 was involved in the Wnt signaling pathway. Overall, only LGR5 was involved in the classical stem cell regulating pathway and correlated with poor overall survival. Previous studies have demonstrated that LGR5 is an endometrial stem cell marker that binds R-spondins to regulate the canonical Wnt signaling pathway in the endometrium [34], [35], [36]. More interestingly, LGR5 was downregulated by progesterone but independent of the menstrual cycle phase [36], [37], [38]. Thus, LGR5 may play a complicated role in endometrial cancer.

We further explored the relationship between the expression of LGR5 and clinicopathological features, and the results indicated that the increased expression of LGR5 was significantly associated with age (>55), stage II, endometrial serous carcinoma, postmenopausal period, and copy-number high subtype. Endometrial cancer is known to be associated with long-term estrogen stimulation without progesterone resistance. The patients over 55 years old were mostly in the postmenopausal period and had low-level progesterone, which was consistent with the results of LGR5 reduction by progesterone. Additionally, previous studies have noted that LGR5 is highly expressed in the initial stages of endometrial tumorigenesis and remarkably downregulated in developed tumors [39], which is consistent with our findings of the highest LGR5 expression in stage II EC. Overall, LGR5 may be related to the initiation of endometrial cancer.

To investigate the functions of LGR5 in EC, we detected the proliferation rate after overexpression or knockdown of LGR5 in EC cell lines. The results confirmed that LGR5 plays an oncogenic role in the progression of EC. GSEA between the high and low LGR5 expression groups was performed, and the results showed that regulation of autophagy and oxidative phosphorylation were enriched. Autophagy is a form of cell death that provides energy through lysosomal degradation and the release of cellular material, which maintains bioenergetic needs for cells [40]. Recent studies have demonstrated that many types of cancer cells rely on mitochondrial respiration and upregulate oxidative phosphorylation activity to fuel tumorigenesis [41]. In addition, studies have shown that mitochondria can regulate many cellular physiological processes, such as autophagy and DNA repair [42]. Oxidative phosphorylation downregulation in the high LGR5 expression group may indicate mitochondrial damage. Mitochondrial autophagy is a type of autophagy, which is a cellular metabolic pathway that mediates the selective elimination of dysfunctional mitochondria [43]. SQSTM1/p62, as an autophagy receptor involved in selective autophagy pathways, especially participates in organelle degradation [44]. Therefore, it is critical to understand the relationship between autophagy, mitochondria, and cell proliferation. We next confirmed that LGR5 induced autophagy activation, and upregulated SQSTM1/p62 may be related to mitochondrial autophagy, which suggested that LGR5 may play a role in promoting tumor progression through autophagy activation and mitochondrial dysfunction.

Inevitably, several limitations need to be addressed in the future. First, the prognostic model needs to be validated in a database with large samples. Second, other genes in this model need to be studied in the future. In addition, approximately 7 % of EC cases occur in women under 45 years old, and 70 % of them are nulliparous [45,46]. The current strategies for fertility-sparing therapy aim to reverse endometrial lesions with high doses of progesterone, but high-quality evidence on efficacy and safety is lacking [45,47]. Considering the relationship between progesterone and LGR5, LGR5 may play a pivotal role in progesterone resistance, which needs to be further investigated.

In conclusion, we constructed a ten stemness-related gene prognostic model in EC based on the TCGA database and verified it with data from the PKUPH cohort, which was an independent risk factor, and it can predict overall survival combined with clinical stage. We further explored the functions of LGR5 as an EC oncogenic gene, accelerating the proliferation rate of EC cells and activating autophagy, which may be a novel approach for cancer therapy.

Disclosure

Not applicable.

Funding information

This work was supported by the National Natural Science Foundation of China (Grant Nos. 82072861 and 82203568), Beijing Natural Science Foundation (Grant No. 7202213), Peking University People’s Hospital Research And Development Funds (Grant No.RDJP2022-09, RS2023-04), and Peking University Medicine Fund of Fostering Young Scholars’ Scientific & Technological Innovation (Grant No. BMU2022PYB028).

Ethics statement

• • -
    Approval of the research protocol by the ethics committee of PKUPH (Approval number: 2020PHB063-01)
  • -
    Informed Consent. N/A.
  • -
    Registry and the Registration No. of the study/trial. N/A.
  • -
    Animal Studies. N/A.

CRediT authorship contribution statement

Chengcheng Li: Methodology, Formal analysis, Writing – original draft. Xiao Yang: Software, Investigation, Data curation, Writing – review & editing, Funding acquisition. Yuan Cheng: Validation, Resources, Visualization. Jianliu Wang: Conceptualization, Resources, Supervision, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

• The data that support the findings of this study are available from the corresponding author upon reasonable request.