Targeting the CD47-HCK-LGALS9 axis disrupts proliferation-immunosuppression coupling in early-stage endometrial cancer

Abstract

Backgrounds

Significant heterogeneity was observed in the epithelial-derived endometrial cancer (EC), which contributes to different prognostic outcomes. While the mechanisms involved in this heterogeneity in the tumor microenvironment (TME) of early-stage EC remain poorly defined.

Methods

Single-cell RNA sequencing, spatial transcriptomics, and multiplex immunofluorescence (mIF) were applied to characterize cellular subpopulations in early-stage EC. EC cell proliferation was assessed by CCK-8 and flow cytometry (FC). Furthermore, the epithelial dynamics and macrophage interactions were also investigated in vitro by molecular docking and GST pull-down mass spectrometry. Phagocytic activity was also evaluated by FC and immunofluorescence. Moreover, the upstream transcription factors were identified using CUT&Tag. For in vivo study, an organoid–macrophage co-culture model was established to mimic the EC microenvironment, and altered organoid proliferation and macrophage phagocytosis were confirmed. CD47 was evaluated as a potential prognostic marker by proteomic analysis and immunohistochemistry.

Results

We identified a previously unrecognized epithelial subpopulation (CD47⁺CDK1⁺) that was strongly linked to cell-cycle signaling and malignant proliferation. This subpopulation interacts closely with macrophages through the CD47–Galectin 9 (LGALS9) receptor-ligand pair. Functionally, CD47 inhibition suppressed EC cell proliferation and induced apoptosis, while CD47 overexpression enhanced the proliferative capacity of EC patient-derived organoids (PDOs). EC cells and PDOs engaged macrophages through the CD47–HCK (Hemopoietic Cell Kinase) axis, driving the secretion of immunosuppressive molecules LGALS9, IL-10, and TGF-β1 and forming an immunosuppressive TME. In turn, macrophage-derived LGALS9 reinforced EC cell and PDO proliferation via CD47, establishing a positive feedback loop involving CD47–HCK–LGALS9. Furthermore, estrogen-related receptor gamma (ERRγ) was identified as an upstream transcriptional regulator of CD47, and its expression was suppressed by progesterone.

Conclusions

Regulated by ERRγ, an optimistic therapeutic target, the CD47-HCK-LGALS9 axis modulated cellular proliferation-immunosuppression coupling in early-stage EC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12943-025-02534-0.

Introduction

Endometrial cancer (EC) is one of the most common gynecological malignancies, with approximately 420,000 new cases and 97,704 deaths reported worldwide in 2022, corresponding to an age-standardized mortality rate of 0.9% [1]. Recent epidemiological data further indicate that EC remains a growing public health concern. For instance, cancer statistics in the United States for 2024 estimate that EC accounts for 6.98% of new cancer cases (67,880/972,060) and 4.6% of cancer-related deaths (13,250/288,920) among women, representing increases of 2.5% and 1.7% compared with 2022 [2, 3]. Similarly, in China, 84,520 new EC cases and 17,543 deaths were reported in 2022, reflecting increases of 3.1% and 5.6% since 2020 [4, 5]. Although the majority of patients are diagnosed at an early stage and generally have a favorable prognosis, both the incidence and mortality of EC continue to rise. Previous studies have suggested that obesity and prolonged estrogen exposure are among the most significant risk factors contributing to this upward trend [6].

Despite early-stage EC often being associated with good outcomes, recent evidence indicates that its incidence is shifting toward younger women, thereby posing substantial challenges for fertility preservation and psychological well-being during reproductive years. However, early diagnosis remains problematic, as reliable biomarkers are lacking and EC is frequently misdiagnosed as atypical endometrial hyperplasia (AEH). Fertility-preserving treatment with high-dose progesterone has been applied in young patients, but clinical outcomes remain suboptimal: therapeutic response rates range from 70.7% to 81.1%, while recurrence rates reach 21.0% to 42.4% [7]. Moreover, obesity has been associated with increased treatment failure and recurrence, further limiting the effectiveness of this therapeutic approach [8]. Thus, the limitations of current strategies underscore the urgent need for more accurate diagnostic tools and targeted therapies.

In parallel, the underlying pathophysiological mechanisms of early-stage EC remain insufficiently elucidated, particularly with respect to the tumor microenvironment (TME) and immune cell dynamics. Growing evidence suggests that the TME plays a pivotal role in tumor initiation, progression, and therapeutic response. For example, single-cell RNA sequencing (scRNA-seq) by Chen and colleagues in 2022 revealed previously unrecognized oncogenic subpopulations in early-stage EC, characterized by specific markers such as LCN2 and SAA1/2, and further suggested that EC originates from non-ciliated glandular epithelial cells. Their work also provided initial insights into the features of the tumor immune microenvironment (TIME) [9]. Nevertheless, the mechanisms governing the crosstalk between oncogenic subpopulations and immune components were not fully delineated. Building on this, our group recently identified a malignant epithelial subpopulation characterized by TRAP1^low/CAMSAP3^low using an integrated proteomic and scRNA-seq. Notably, our findings revealed close interactions between epithelial cells and myeloid-derived cells within the TIME, thereby offering a theoretical framework for further exploration of the role of the TIME in driving malignant proliferation in early-stage EC [10]. These observations laid the foundation for further investigation into how TIME regulates tumor growth in early-stage EC and highlighted potential therapeutic targets.

To gain deeper insight into molecular mechanisms underpinning epithelial-immune crosstalk in early-stage EC, we performed a comprehensive scRNA-seq analysis on early-stage EC specimens and identified a previously unreported epithelial subpopulation, termed CD47⁺CDK1⁺epi. Mechanistic studies demonstrated that this subpopulation sustains its proliferative capacity through receptor–ligand interactions with macrophages and establishes bidirectional communication that fosters an immunosuppressive TME. Although CD47-targeted therapies have shown promise, their clinical utility is limited by treatment-associated toxicities. To address this, we further explored upstream regulatory pathways of CD47 and found that estrogen-related receptor gamma (ERRγ/ESRRG) directly activates CD47 transcription, whereas progesterone effectively suppresses ERRγ activity. Together, these results offer a detailed mechanistic framework for understanding how specific epithelial subpopulations contribute to shaping the early-stage EC TIME and highlighting a therapeutic strategy that combines selective targeting of CD47⁺CDK1⁺epi cells with progesterone-mediated repression of ERRγ as a promising approach for early-stage EC.

Methods

Specimens and clinical data

Patients with International Federation of Gynecology and Obstetrics (FIGO) stage I–II EC were included in this study. All tissue samples were obtained from Fujian Maternity and Child Health Hospital, and ethical approval was granted by the institutional review board. Written informed consent and authorization forms were signed by all participants before sample collection. A total of 24 formalin-fixed paraffin-embedded (FFPE) samples were used for proteomic analysis, 8 fresh tissue samples were collected for scRNA-seq, and 1 FFPE sample was processed for spatial transcriptomics. For control tissues, normal endometrial samples paired with para-cancerous tissues from patients with EC were analyzed, and the absence of malignant lesions was confirmed by histopathological examination.

A proteomic study was performed on patients who underwent surgery for EC between May 2013 and May 2023. The inclusion criteria were as follows: (1) availability of complete clinical information, (2) preservation of well-maintained pathological wax blocks, (3) no history of hormone therapy within six months before surgery, and (4) no concurrent tumors or infectious diseases. After matching patients by follow-up duration, age, body mass index (BMI), and FIGO stage, 15 surviving patients and 9 deceased patients were selected. Clinical outcomes were classified as poor prognosis if death occurred within 10 years post-surgery, and favorable prognosis if survival exceeded 10 years.

Additionally, organoids were established from tissues collected from patients with early-stage EC at Fujian Maternity and Child Health Hospital between July 2025 and September 2025. Organoid derivation was performed under standardized protocols to ensure reproducibility.

Single-cell collection and library construction

Tumor specimens and matched paracancerous tissues were dissected into 1–3 mm³ fragments under sterile conditions. Single-cell suspensions were prepared, and cell viability was quantified using a Count Star analyzer, maintaining a final concentration of 300–600 viable cells/µL. Single-cell capture was performed on the Chromium Single-Cell Controller (10x Genomics) using the Single-Cell 3′ Library and Gel Bead Kit v3.1 (10x Genomics, 1000121) and the Chromium Single-Cell G Chip Kit (10x Genomics, 1000120), according to the manufacturer’s instructions. cDNA synthesis and library construction were carried out using the v3.1 kit. Sequencing was performed on an Illumina NovaSeq 6000 platform (CapitalBio Technology, Beijing) with a paired-end 150 bp (PE150) strategy, achieving a minimum depth of 100,000 reads per cell.

Quality control

Raw sequencing data were processed with Cell Ranger (10x Genomics) to generate a unique molecular identifier (UMI) count matrix. The resulting matrix was converted into a Seurat object for downstream analysis using the Seurat package in R. Cells were excluded if they expressed fewer than 200 genes, if their detected gene counts ranked within the top 1%, or if their mitochondrial gene content exceeded 25%.

Unsupervised clustering, data integration, and cell type annotation

Single-cell data were analyzed using the Seurat R package(v4.3). The NormalizeData function was applied with a scaling factor of 10,000, followed by log-transformation for data normalization and scaling. Highly variable genes were identified using the FindVariableFeatures function with the variance stabilizing transformation (VST) method, selecting the top 2,000 most variable genes. Expression levels of all genes were then Z-score normalized with the ScaleData function, ensuring a mean of 0 and a variance of 1 across cells. Principal component analysis (PCA) was performed, and the top 50 principal components (PCs) were retained.

To integrate single-cell data from eight samples and correct for technical variation arising from batch effects, we employed the Harmony algorithm which builds upon the PCA-derived embeddings. Harmony identifies shared biological features across batches by maximizing local cell similarity in low-dimensional space while simultaneously modeling the joint distribution of cells across batches within a Bayesian framework.

For two-dimensional visualization of cellular distributions, dimensionality reduction was performed using the RunUMAP functions, projecting cells into a low-dimensional embedding based on the top 20 PCs to preserve local neighborhood structure. A shared nearest neighbor (SNN) graph was constructed with the FindNeighbors function using the first 20 PCs as input. Clustering was then performed with the FindClusters function, which applies the Louvain community detection algorithm, with the resolution parameter set to 0.3 to adjust cluster granularity and define a discrete number of cell populations.

To identify cluster-specific marker genes, differential expression analysis was conducted using the FindMarkers and FindAllMarkers functions, with thresholds set at log2 fold change (avg_log2FC ≥ 0.25) and minimum expression fraction (min.pct ≥ 0.1). Final cell type annotations were determined based on the expression patterns of differentially expressed genes (DEGs) in combination with canonical markers curated from the literature, thereby assigning biological identities to each cluster.

Intercellular communication

Processed data, including the count matrix and cell type annotations, were formatted for CellPhoneDB(v4.0) analysis [11]. The program was executed with default parameters to statistically infer ligand–receptor interactions between predefined clusters. Interaction probabilities were estimated by permutation testing (n = 1,000), and interactions with p < 0.05 were considered statistically significant.

Copy number variation (CNV) Estimation

To distinguish malignant epithelial cells, CNV inference was performed using CopyKAT on rigorously quality-controlled, normalized, and integrated expression profiles from the eight samples. Core parameters were set following the developer’s recommendations and further optimized based on preliminary analyses. Specifically, a minimum of five genes per chromosome was required for CNV estimation (ngene.chr = 5). A Kolmogorov–Smirnov test threshold of 0.1 was applied to filter out cells with expression distributions significantly deviating from the reference. Euclidean distance was used to compute CNV similarity between cells.

Malignancy scoring of tumor cells

Construction of a reference normal expression profile

Following CopyKAT-based classification, the set of predicted normal cells was denoted as = {, , …,}, where 𝑁 represents the total number of normal cells. The gene expression vector of each cell was defined as , with 𝐺 denoting the total number of genes. The mean reference expression profile of normal cells was then computed as:

To enhance baseline stability, only genes expressed in at least 80% of normal cells were retained, thereby filtering out low-coverage genes.

Calculation of CNV scores for malignant cells

For each malignant tumor cell with expression vector , a CNV score was computed as the normalized Euclidean distance between and the reference normal profile :

Where denotes the expression value of gene 𝑘 in cell and _{k, normal} a is the mean expression of gene k across normal cells. A larger CNV score indicates greater deviation from the reference baseline, reflecting a higher degree of CNV-associated abnormality and malignancy.

Pseudotime trajectory analysis

Pseudotime trajectories were reconstructed using the Monocle2 algorithm. Processed expression data and cell cluster annotations were imported, and dimensionality reduction was performed with the DDRTree method. Cells were ordered along a trajectory represented by a minimum spanning tree, capturing the progression from initial to terminal states. Branching points were analyzed to identify critical transcriptional transitions.

Spatial transcriptomics

Spatial transcriptomics analysis was performed on an FFPE sample from one EC patient whose tissue was matched with scRNA-seq data. The Visium Spatial Gene Expression for FFPE workflow (10x Genomics, PN-1000336) was followed according to the manufacturer’s demonstrated protocol (CG000407). Libraries were sequenced on an Illumina NovaSeq 6000 (CapitalBio Technology, Beijing) using a paired-end 150 bp (PE150) strategy, with a sequencing depth of at least 50,000 reads per spot. Data processing, including alignment, filtering, barcode counting, and UMI counting, was performed using the Spaceranger (v1.1.0) count module to generate feature-barcode matrices and clusters. Normalization was performed with the SCTransform function in Seurat. Dimensionality reduction was performed using PCA. To infer the cellular composition of each spot, mapping was conducted by integrating scRNA-seq reference data using the FindTransferAnchors and TransferData functions. Each spot was annotated according to the cell type with the highest predicted score.

TCGA database analysis

All statistical analyses were performed using the R statistical environment. Transcript per million (TPM) expression data for uterine corpus endometrial carcinoma (TCGA-UCEC) were downloaded using the TCGAbiolinks R package. A total of 574 samples were obtained, comprising 539 EC samples and 35 normal endometrial samples. Corresponding clinical data were acquired from the UCSC database. After integrating gene expression with clinical information, 167 samples corresponding to FIGO stage IA were identified for further analysis. Differential expression analysis and enrichment analysis were subsequently performed to compare expression levels across groups.

Gene function enrichment analysis

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed to investigate the biological functions of differentially expressed genes (DEGs). The DAVID online platform (https://david.ncifcrf.gov/) was used to identify significantly enriched pathways (P < 0.05) [12, 13]. Additionally, KEGG enrichment analysis was carried out using the enrichKEGG function in the clusterProfiler R package.

Cell line construction and culture

HEC-1 A, KLE, and THP-1 cell lines (KeyGEN BioTECH, China) were used in this study. HEC-1 A cells were cultured in McCoy’s 5 A glucose-free medium supplemented with glucose (PM150718, Procell Life Science & Technology, China). KLE and THP-1 cells were maintained in RPMI-1640 medium (PM150122, Procell Life Science & Technology, China) containing glucose. All media were supplemented with 10% fetal bovine serum (FBS, #10091148, Gibco, USA) and 1% antibiotic-antimycotic solution (#B120901, Basal Media, China). Cells were incubated at 37 °C in a humidified atmosphere containing 5% CO₂. All cell lines were authenticated by short tandem repeat (STR) DNA profiling.

HEC-1 A and KLE cells were treated with the following agents: RRX-001 (#HY-16438, MCE, USA) at 5, 6.25, and 7.5 µM; recombinant LGALS9 (#11147-H01H, Sino Biological, China) at 50 ng/mL; GSK5182 (#SML3150, Germany) at 10, 20, and 30 µM; progesterone (#HY-N0437, MCE, USA); and estradiol (#HY-B0141, MCE, USA). All treatments were applied for 24 h.

Regulation of ERRγ expression was achieved through lentiviral transduction. The LV-si-ESRRG vector (designated ERRγ knockdown) was synthesized using the siRNA sequence 5′-AGAGGAGTATGTTCTACTAAA-3′. A lentiviral vector for CD47 overexpression (ovCD47) and ERRγ overexpression (ovERRγ) and a negative control vector were also obtained from GeneChem. Lentivirus-mediated siRNA transduction effectively reduced ERRγ expression levels.

CCK-8 assay

Adherent cells were digested, resuspended, and seeded into 96-well plates at 100 µL per well. Cells were treated with RRX-001, RRX-001 combined with recombinant LGALS9, recombinant LGALS9 alone, or GSK5182. Each condition was tested in triplicate, and blank and untreated control groups were included. After treatment, 10 µL of CCK-8 reagent (APExBIO, China) was added to each well. Absorbance was measured at 450 nm using a microplate reader after incubation.

Cell cycle analysis

Cells were harvested, fixed in ethanol, and stored overnight at − 20 °C. The fixed cells were washed, resuspended, and incubated with RNase A (100 µL) at 37 °C for 30 min. Propidium iodide (PI, 50 µg/mL; 400 µL) was added, and samples were analyzed immediately using flow cytometry with excitation at 488 nm to determine DNA content and cell cycle distribution.

Apoptosis assay

Apoptosis was evaluated by Annexin V/7-AAD staining. After 24 h of treatment, cells were harvested and resuspended in 1× binding buffer. A 100 µL aliquot of cell suspension was transferred to flow cytometry tubes and stained with Annexin V–APC and 7-AAD (5 µL each) in the dark at room temperature for 15 min. After staining, 400 µL of 1× binding buffer was added, and samples were analyzed within 1 h using a flow cytometer.

Western blot analysis

After the cells were treated, cells were lysed to extract total protein. Equal amounts of protein were subjected to SDS-PAGE and transferred to PVDF membranes. Membranes were blocked and subsequently incubated with primary antibodies against ERRγ (1:1000, R&D, USA), CD47 (1:1000, CST, USA), Cyclin D1 (1:1000, Thermo Fisher Scientific, USA), PCNA (1:1000, ABclonal, China), Bcl-2 (1:1000, CST, USA), Bax (1:1000, CST, USA), and GAPDH (1:3000, ImmunoWay, USA). After washing, membranes were incubated with goat anti-rabbit (1:10000, Proteintech, China) or goat anti-mouse (1:10000, Proteintech, China) secondary antibodies for 1 h at room temperature, followed by extensive washing for 30 min. Protein bands were visualized using an enhanced chemiluminescence reagent, and signals were detected by exposure to X-ray film in a darkroom.

RT‑qPCR

Total RNA was extracted by lysing cells directly in culture dishes, followed by chloroform–isopropanol extraction. Briefly, cell lysates were vortexed and centrifuged, and the aqueous phase was collected after chloroform addition. RNA was precipitated with isopropanol, washed with 75% ethanol, and dissolved in RNase-free water. First-strand cDNA was synthesized using the Reverse Transcription System (#A2791, Promega, USA), and qPCR amplification was performed with Eastep qPCR Premix (Promega, USA). GAPDH served as the internal reference, and relative gene expression was calculated using the 2^−ΔΔCt method.

ELISA

According to the manufacturer’s instructions, we used an ELISA kit (ABclonal, China; Huabio, China) to quantitatively determine the levels of LGALS9, IL-10, and TGF-β1 in cell culture supernatants.

Co-culture of EC cells and macrophages

THP-1 cells were seeded into the upper chambers of six-well Transwell inserts and differentiated into M0 macrophages by incubation with 100 ng/mL phorbol 12-myristate 13-acetate (PMA) for 24 h. Differentiated macrophages were then seeded into the lower chamber of six-well plates. Tumor cells were seeded into 0.4 μm pore-size Transwell inserts placed above macrophages to establish an indirect co-culture system. The system was maintained in medium supplemented with 10% FBS for 24–48 h to allow paracrine interaction without direct cell–cell contact.

In vitro phagocytosis assay

For flow cytometry–based phagocytosis assays, 5 × 10⁴ macrophages were seeded into each well of 24-well plates and incubated in serum-free medium for 2 h. Tumor cells (2 × 10⁵) were then added and co-cultured with macrophages at 37 °C for 2 h. Cells were harvested, centrifuged at 300 × g for 5–10 min at 4 °C, and stained with anti-CD45 antibody to identify macrophages. Phagocytosis was quantified as the percentage of CD45⁺GFP⁺ macrophages among total CD45⁺ cells by flow cytometry.

For immunofluorescence–based assays, tumor cells were labeled with Dil dye (Beyotime, China) at 37 °C for 15 min, washed twice with PBS, and co-incubated with macrophages for 6 h. After incubation, wells were washed twice with serum-free medium to remove non-phagocytosed tumor cells. Cells were then imaged with a fluorescence microscope to assess phagocytosis.

GST-pull down mass spectrometry (MS)

Proteins interacting with CD47 in macrophages were analyzed by GST pull-down followed by mass spectrometry. Macrophages were harvested, lysed in buffer for 30 min, and centrifuged at 12,000 × g for 30 min to obtain protein lysates. Lysates were incubated with purified GST-tagged CD47 at 4 °C overnight, and GST magnetic beads were used to capture interacting proteins. Bound proteins were eluted and subjected to 4D label-free MS analysis. For sample preparation, 4 µL of 0.05 M TCEP was added to lysates and incubated at 60 °C for 1 h for reduction, followed by alkylation with 55 mM MMTS at room temperature for 45 min in the dark. Proteins were then transferred to 10 kDa ultrafiltration tubes, washed by centrifugation with urea buffer (8 M, pH 8.5), and exchanged into 0.25 M TEAB solution. Proteins were digested overnight at 37 °C with trypsin at a 1:50 enzyme-to-protein ratio, followed by an additional digestion step (1:100 ratio) for 4 h. Peptides were collected, vacuum dried, resuspended in 0.1% formic acid/2% acetonitrile, centrifuged at 13,000 rpm for 20 min at 4 °C, and analyzed by mass spectrometry.

Cleavage under targets and tagmentation (CUT&Tag)

CUT&Tag assays were performed using an anti-ERRγ antibody conjugated to Tn5 transposase. The antibody–enzyme complex selectively bound to protein targets through affinity interactions. Upon binding, the transposase cleaved DNA near the target protein and simultaneously inserted DNA fragments containing sequencing adaptors at the cleavage sites. The resulting fragments were subjected to high-throughput sequencing to obtain genome-wide profiles of protein-bound DNA. Peak-associated genes were identified and annotated to genomic regions using ChIPseeker, which was also employed to determine the nearest genes to each peak.

Molecular docking

Molecular docking experiments were carried out using PyMOL and AutoDock. Amino acid sequences of CD47, HCK, and LGALS9 were prepared by adding hydrogen atoms, repairing structural defects, and converting files into docking-compatible formats. Active sites were identified in PyMOL, and the coordinates and size of docking boxes were defined. Docking parameters were set in AutoDock, and docking simulations were performed to predict potential binding conformations. The docking results were visualized and analyzed in PyMOL. Binding modes with the lowest free energy and structurally reasonable conformations were considered optimal.

Immunohistochemistry (IHC)

FFPE specimens were obtained from patients with AEH and EC at Fujian Provincial Maternity and Child Health Hospital between 2021 and 2023. Five AEH cases were selected to match the median age and BMI of the scRNA-seq cohort. IHC analysis was performed on 5 EC specimens, 3 adjacent non-tumorous tissues, and 5 AEH specimens. Tissue sections were deparaffinized, subjected to citrate-based antigen retrieval, and blocked with 5% goat serum. Sections were incubated overnight at 4 °C with anti-CD47 (1:400, CST, USA) and anti-LGALS9 (1:100, Proteintech, China). Detection was performed using DAB chromogen, followed by hematoxylin counterstaining and ethanol dehydration. Slides were evaluated independently by two pathologists in a double-blind manner. IHC scores were calculated by multiplying staining intensity (0–3) with the proportion of positive cells (1–4).

Multiplex Immunofluorescence (mIF)

mIF was performed on pathological sections derived from three EC specimens and matched adjacent non-tumorous tissues used in scRNA-seq analysis. Sections were deparaffinized, rehydrated, and subjected to antigen retrieval in citrate buffer (10 mM, pH 6.0). After blocking with 5% bovine serum albumin for 1 h at room temperature, sections were incubated overnight at 4 °C with primary antibodies, including anti-CD47 (1:100, Proteintech, China) and anti-CDK1 (1:100, Proteintech, China). Fluorescent secondary antibodies were applied for 1 h at room temperature. Stained sections were imaged using confocal microscopy.

Patient-derived organoids (PDOs) generation and cultivation

Tumor tissues were obtained from three patients with early-stage EC who were pathologically diagnosed at Fujian Maternal and Child Health Hospital within the past three months. Fresh tumor specimens (0.5 × 0.5 cm) were retrieved from preservation solution, and necrotic and non-tumorous areas were removed. Tissues were minced into 1–2 mm³ fragments, washed with Buffer P1, and digested at 37 °C on a shaker for 1–2 h until flocculation and visible cell clusters appeared. Digestion was terminated with basal medium, and single-cell suspensions were prepared by gentle pipetting, followed by filtration through a 100 μm cell strainer and centrifugation. Red blood cells were lysed, and the remaining cells were washed with pre-chilled basal medium. Cell pellets were resuspended in organoid culture medium, mixed with Matrigel at a 1:1.5 ratio, and seeded into pre-warmed 24-well plates (50 µL per well). After polymerization, cultures were overlaid with fresh organoid medium and maintained in a humidified incubator. Organoid growth was monitored daily, and medium was refreshed every other day.

Organoids virus transfection

For viral transduction, organoids were harvested by digesting Matrigel with 1 mL organoid passage digestion solution per well. Suspensions were transferred to 15-mL centrifuge tubes, incubated on ice with gentle shaking for 10 min, and diluted with four volumes of pre-chilled DMEM/F12. After centrifugation at 300 × g for 5 min at 4 °C, supernatants were discarded and pellets were resuspended in organoid culture medium. Cells were mixed with lentivirus at a 1:4 ratio and incubated at 37 °C for 1–2 h. Following centrifugation and supernatant removal, pellets were resuspended in organoid culture medium mixed with Matrigel (1:1) and seeded into pre-equilibrated 24-well plates (50 µL per well). After solidification at 37 °C for 30 min, 500 µL of culture medium was added. Organoid morphology and fluorescence were monitored every other day, and medium was refreshed every 2–3 days. On day 6 post-transfection, neomycin was added for selection, and surviving organoids were collected 48 h later for downstream experiments.

Organoid proliferation assay

Organoid proliferation was assessed using the Cell Counting-Lite 3D Luminescent Cell Viability Assay (CCL 3D) according to the manufacturer’s instructions. Organoids transduced with either ov-CD47 or con-CD47 lentiviruses were seeded into 96-well plates and cultured overnight. Organoid spheroid formation was first observed under an optical microscope. An equal volume of CCL 3D reagent was added to each well, followed by shaking for 5 min to ensure mixing. Plates were incubated at room temperature for 25 min to allow complete lysis. Organoid morphology was re-examined microscopically, and luminescence was measured to evaluate cell viability.

Phagocytosis in vivo

Peripheral blood (20 mL) was collected from a healthy male donor into anticoagulant tubes, and PBMCs were isolated using the SepMate™ system. Isolated PBMCs were seeded into six-well plates at ~ 4 × 10^6 cells per well. After overnight culture, non-adherent cells were removed, and adherent cells were maintained in RPMI-1640 medium supplemented with 10% FBS and 20 ng/mL macrophage colony-stimulating factor (M-CSF) to induce differentiation into macrophages. The medium was refreshed every three days, and adherent cells were identified as macrophages after six days.

Organoids labeled with CFSE (Beyotime, China) and transfected with ov-CD47 or control viruses were digested into single-cell suspensions. Macrophages were collected, counted, and mixed with organoid-derived cells at a 10:1 ratio. The mixture was resuspended in 100 µL co-culture medium containing 5% Matrigel and seeded into ultra-low attachment 96-well plates. After 72 h of co-culture at 37 °C, organoid spheres were harvested. Flow cytometry was used to assess phagocytosis, quantified as the percentage of CD14⁺CFSE⁺ cells among CD14⁺ macrophages. Additionally, the proportion of CD47⁺ cells within CFSE⁺ cells was measured to evaluate the efficiency of ov-CD47 transfection.

Statistical methods

The experiments were not randomized, and blinding was not applied during experimental procedures or outcome evaluation. Data collection and analysis were not constrained by experimental conditions. Statistical analyses were performed using GraphPad Prism 9.5.1 (GraphPad Software, USA). Statistical parameters, including the definition and exact value of n, the statistical tests used, and levels of significance, were reported in the corresponding figure legends. Data from at least three independent experiments were expressed as mean ± SD. Comparisons between two groups were performed using an unpaired, two-tailed Student’s t-test, whereas multiple-group comparisons were conducted using one-way ANOVA. The difference was considered statistically significant when *P < 0.05, **P < 0.01, ***P < 0.001; ns means no significant difference, that is, P > 0.05.

Results

Transcriptomic analysis identified proliferation-associated alterations in early-stage EC

Differential expression analysis was conducted using TCGA data, comparing FIGO IA stage EC samples (n = 167) with normal endometrial tissues (n = 35). A total of 8,985 genes were upregulated and 240 genes were downregulated (|log2 fold change| > 1, P < 0.05) (Fig. 1A). KEGG enrichment analysis showed that upregulated genes were significantly enriched in the cell cycle pathway (Fig. 1A). Proliferation-related markers, including CDK1, MKI67, TK1, CCNB1, MCM6, and PCNA, were markedly increased in tumor tissues compared with normal controls (Fig. 1B). These results indicated that dysregulated cell proliferation was a central molecular feature of early-stage EC.

Fig. 1.

ScRNA-seq analysis of early-stage EC. A The differentially expressed genes (DEGs) and their KEGG pathway. Left: Volcano plot showing DEGs between FIGO IA stage EC samples and normal samples from the TCGA database. Red dots represent upregulated genes, while blue dots represent downregulated genes. Right: KEGG pathway enrichment analysis of the upregulated DEGs. B Box plots illustrating the expression levels of proliferation markers in FIGO IA stage EC samples compared to normal samples. C UMAP plot displaying the clustering of 7 major cell types from 8 samples, including tumor tissues and adjacent normal tissues of FIGO IA stage EC. D Expression patterns of canonical markers and DEGs for each cell type. Each dot represents a gene, with color saturation indicating the average expression level and dot size representing the percentage of cells expressing the gene. E Percentage distribution of different cell types across samples and between tumor and normal tissues. F UMAP plot of epithelial cells in normal and tumor samples. G DEGs between epithelial subpopulations cluster_0, 3, and 9. Red and blue colors denote up-regulated and down-regulated genes, respectively. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

ScRNA-seq identified early-stage EC-specific epithelial subpopulation cluster _9

To investigate the dynamic changes in the cellular composition during early-stage EC, we performed scRNA-seq on samples from five FIGO IA stage EC patients and matched adjacent normal tissues from three of them. Based on canonical cell markers as defined in the literature and Cell Marker 2.0 [14] (epithelial cells: EPCAM, KRT8 [15, 16]; stromal fibroblasts: DCN [15]; endothelial cells: ADGRL4, PLVAP [15]; lymphocytes: CD3D, CD4, CD8B [17–19]; neutrophils: CD11b, CD16, CSF3R [20, 21]; myeloid-derived cells: CD68, CD86, CD14 [22, 23]; smooth muscle cells: RGS5, ACTA2 [15, 24]), seven major cell types were identified (Fig. 1C, D). Comparative analysis of tumor versus normal tissues revealed a marked increase in epithelial and immune cell populations, including myeloid-derived cells, lymphocytes, and neutrophils, accompanied by a significant reduction in endothelial cells and stromal fibroblasts (Fig. 1E). Each major cell type was further subdivided into distinct subpopulations according to their transcriptional signatures (Fig. S1A, B). A total of 34,377 epithelial cells were extracted and reclustered, leading to the identification of a tumor-specific epithelial subpopulation (cluster_9) (Fig. 1F, Fig. S1C). Since EC arises from non-ciliated epithelium [9], clusters_0, 3, and 9 (each > 500 cells) were retained for further analysis, while cluster_7 (ciliated epithelium) was excluded. Differential expression analysis showed that cluster_9 was enriched for proliferation-associated genes, including TOP2A and MKI67, whereas cluster_0 was characterized by upregulation of LCN2 and SAA1, previously implicated in EC-associated epithelial subsets (Fig. 1G). Together, these findings pinpoint cluster_9 as a distinct epithelial subpopulation specifically enriched in early-stage EC, which we prioritized for subsequent mechanistic investigations.

Molecular functional characteristics of EC-specific epithelial oncogenic subpopulation cluster_9

To further delineate the molecular and functional features of carcinogenic epithelial subpopulations, CNV analysis revealed that cluster_9 exhibited high copy number variations on chromosomes 8 and 12, partially consistent with TCGA dataset findings [25], indicating strong malignant potential (Fig. 2A). Tumor score analysis further confirmed that cluster_9 had the highest tumor score among all epithelial subpopulations (Fig. 2B). Tumor epithelial cells exhibited significant differentiation heterogeneity along the inferred trajectory, forming two distinct evolutionary branches. The trajectory originated from the subpopulation (cluster_3) exhibiting the lowest malignancy score. One branch progressively differentiated into a tumor-specific subpopulation (cluster_9) characterized by a high-malignancy phenotype, while the other branch evolved into cluster_0, which is densely populated in both tumor and normal cell (Fig. 2C–E, Fig. S1D). A heatmap of the top 50 temporally differentially expressed genes along the pseudotime axis highlighted transcriptional dynamics among clusters_3, 0, and 9 (Fig. 2F, G). Notably, proliferation-associated markers such as TK1, CDK1, and MKI67 were significantly upregulated in cluster_9. A scatter plot demonstrated a strong correlation between pseudotime progression and CNV burden, further supporting the high malignancy of cluster_9 (Fig. 2H). Moreover, upregulated DEGs in cluster_9 when comparing with cluster_0 showed significant enrichment in the cell cycle pathway (Fig. 2I, J), consistent with the transcriptomic data. Compared to the highly malignant cluster_9, cluster_0 predominantly exhibited differential expression of genes associated with cellular stress responses and redox regulation, with enriched pathways primarily linked to inflammatory processes (e.g., IL-17 signaling pathway and TNF signaling pathway) (Fig. 2I, J). These results collectively suggest that dysregulated cell cycle activity and uncontrolled proliferation underlie the development of early-stage EC, with cluster_9 representing a key oncogenic epithelial subpopulation driving this process.

Fig. 2.

Identification of malignancy and molecular functional characteristics of the oncogenic subpopulation cluster_9. A CopyKAT-based heatmap to map inferred CNV for the different samples and epithelial cell subpopulations. Rows represent genomic regions ordered by chromosomal position, and columns represent individual cells grouped by their respective clusters. B Violin plot of tumor scores across epithelial subpopulations. C Epithelial cell trajectory constructed using Monocle2, showing changes along the pseudotime axis. The gradient from light to dark indicates the direction and progression of pseudotime. D Annotation of the cell trajectory based on the initially defined epithelial subpopulations. E Faceted visualization of the cell trajectory for the three epithelial subpopulations between normal and tumor samples. F Heatmap of the top 50 pseudotime-related genes in tumor epithelial cells. G Heatmap of dynamic expression changes of key gene sets along pseudotime branches, generated using BEAM analysis: the blue module (cell fate 1) represents the trajectory from cluster_3 to cluster_9, while the red module (cell fate 2) represents the trajectory from cluster_3 to cluster_0. H Scatter plot showing the correlation between pseudotime progression and tumor malignancy scores in epithelial subpopulations. I Volcano plot of DEGs between cluster_9 and cluster_0. J Bar plot of KEGG enrichment pathway of DEGs and top marker genes. Red represents pathways significantly upregulated in cluster_9, while blue represents pathways significantly upregulated in cluster_0

Cell-cell communication analysis and spatial localization of the oncogenic subpopulation

Our previous multi-omics studies reported extensive interactions between epithelial and myeloid-derived cells in the TIME of early-stage EC [10]. Building on this foundation, we systematically examined receptor-ligand pairs between epithelial cells and myeloid-derived cells in the TME. Specifically, we analyzed interactions among tumor cells, normal cells (clusters_0, 3, and 9), macrophages, neutrophils, and dendritic cells (DCs). Chord diagrams illustrated the number of interacting cells, while heatmaps reflected interaction strengths. These analyses demonstrated that cluster_9 interacted most strongly with macrophages (Fig. 3A, B, Fig. S1E). Mapping of receptor–ligand pairs revealed that LGALS9-CD47 and SIRPA-CD47 interactions were particularly enriched between cluster_9 and macrophages (Fig. 3C). Further expression profiling showed that LGALS9 was markedly up-regulated in tumors (Fig. 3D), supporting the hypothesis that the LGALS9-CD47 axis is the dominant mediator of cluster_9-macrophage crosstalk. Pseudotime analysis further indicated that CD47 expression significantly increased during the transition from cluster_3 to cluster_9 (Fig. 3E) and was positively correlated with pseudotime progression (Fig. 3F). Consistent with the molecular characteristics of the carcinogenic subpopulation, the cell cycle marker CDK1 also showed a positive correlation with pseudotime progression (Fig. 3G). Spatial transcriptomics further confirmed the co-localization of CD47 and CDK1 (Fig. 3H). Based on these findings, we defined cluster_9 as the CD47⁺CDK1⁺ epi and mapped the spatial distribution of this subpopulation using spatial transcriptomics (Fig. 3I). Notably, the CD47⁺CDK1⁺ epi was surrounded by dense immune cell infiltration, underscoring its potential role in shaping the TME. These results collectively suggest that the LGALS9-CD47 axis plays a critical role in the interaction between the carcinogenic subpopulation and macrophages, driving malignant proliferation and promoting immunomodulatory remodeling of the TME.

Fig. 3.

Cell-cell communication analysis and spatial localization of the oncogenic subpopulation. A Chord diagram showing the number of interactions between epithelial cells and myeloid-derived cells, where line thickness indicates the interaction count. B Heatmap depicting the interaction strength between epithelial cells and myeloid-derived cells. C Dotplot of receptor-ligand pairs in intercellular communication between epithelial cells and myeloid-derived cells, with red circles representing P < 0.05. D The expression of LGALS9 and SIRPA in macrophages in tumor and normal samples. E Distribution of CD47 expression levels across all cells along the pseudotime axis, with the color gradient (blue to red) representing low to high expression. F, G Scatter plot showing the correlation between pseudotime progression and CD47 (F) and CDK1 (G) expression levels. H Spatial localization of CD47 and CDK1. I Spatial localization of cell subpopulations. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

In vitro experiments confirm that CD47 inhibition suppresses EC cells proliferation

To assess the role of CD47 in EC cells, we treated the cells with the CD47 inhibitor RRX-001 and compared cell proliferation between the treated and untreated groups. After 24 h of treatment, EC cells exposed to RRX-001 displayed a significant reduction in optical density (OD) values (P < 0.05) (Fig. 4A). Over a 5-day observation period, RRX-001 treatment exhibited a pronounced, time-dependent inhibitory effect on EC cell proliferation (P < 0.05)(Fig. 4B). Flow cytometry analysis revealed a significant increase in apoptosis in EC cells following CD47 inhibition (Fig. 4C, D). Additionally, in HEC-1 A cells, the proportion of cells in the G0/G1 phase increased, accompanied by a decrease in the S phase population, while no significant change was observed in the G2/M phase. In KLE cells, the proportions of cells in both the S and G2/M phases decreased as the G0/G1 phase population increased (Fig. 4E, F). The expression levels of proliferation markers Cyclin D1 and PCNA were significantly reduced (Fig. 4G, H). Furthermore, the protein levels of Bax were significantly increased, while Bcl-2 expression was significantly decreased after RRX-001 treatment (Fig. 4H, I). Taken together, these findings suggest that CD47 inhibition triggers apoptosis and induces cell cycle arrest at the G0/G1 phase. Overall, our results demonstrate that CD47 is a key driver of EC cell proliferation, and its inhibition represents a promising therapeutic strategy for EC.

Fig. 4.

Inhibition of CD47 suppresses EC cell proliferation. A Changes in cell viability after 24 h of treatment with RRX-001 detected by the CCK-8 assay. B Time-dependent changes in cell viability over 5 days of treatment with RRX-001 detected by the CCK-8 assay. C-F Effects of RRX-001 treatment for 24 h on cell cycle and apoptosis were analyzed by flow cytometry. G Expression levels of cyclin D1 and PCNA at the mRNA levels in EC cells treated with RRX-001 for 24 h. H, I After treating with RRX-001 for 24 h in EC cells, the expression of CD47, cyclin D1, PCNA, Bcl-2 and Bax was measured by western blotting (WB), with GAPDH used as a loading control. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

The interaction between CD47⁺CDK1⁺ epi and macrophages promotes an immunosuppressive TME

To investigate the LGALS9-CD47 interaction in the TME, we first performed molecular docking, which predicted potential binding between LGALS9 and CD47 (Fig. 5A). Using a transwell co-culture system (Fig. 5B), we established an indirect co-culture model of EC cells and macrophages. Interestingly, CD47 inhibition in EC cells for 24 h markedly reduced LGALS9 secretion in this co-culture system (Fig. 5C). Our studies revealed that co-culture of CD47⁺CDK1⁺ tumor cells with macrophages significantly enhanced macrophage secretion of immunosuppressive cytokines IL-10 and TGF-β1 (Fig. 5D, E), suggesting that this oncogenic subpopulation actively drives the formation of an immunosuppressive TME. Further functional assays demonstrated that macrophages exhibited significantly reduced phagocytic activity toward CD47⁺CDK1⁺ tumor cells compared to CD47⁻CDK1⁻ cells (Fig. 5F, G, Fig. S2A). To elucidate the mechanism underlying CD47-mediated LGALS9 secretion by TAMs, we employed GST-pull down MS and identified 283 macrophage proteins that potentially interact with CD47 (Fig. 5H, I). Correlation analysis with CD47/LGALS9 expression in TCGA data narrowed these candidates to six membrane proteins (ANXA6, AGTRAP, HCK, GNAI2, MSN, and SLC16A3) (Table S1, Fig. S2B). Among them, HCK showed the most pronounced expression changes in CD47-overexpressing EC cells co-cultured with macrophages (Fig. 5J), and molecular docking confirmed potential CD47-HCK binding (Fig. 5K). These results support a model in which CD47⁺CDK1⁺ tumor cells interact with HCK on TAMs, inducing LGALS9 secretion. Elevated LGALS9, in turn, suppresses macrophage phagocytosis while stimulating IL-10 and TGF-β1 production, thereby reinforcing an immunosuppressive TME and facilitating tumor immune evasion.

Fig. 5.

The interaction of the LGALS9-CD47 axis and the formation of an immunosuppressive microenvironment in the co-culture model of EC cells and macrophages. A Molecular docking predicting the interaction between LGALS9 and CD47. B The co-culture model of EC cells and macrophages. C ELISA analysis of LGALS9 secretion levels in the supernatant of macrophages under different conditions: EC cells, macrophages (M), EC cells + M, and EC cells + M + RRX-001. D, E ELISA analysis of IL-10 and TGF-β1 secretion levels in the supernatant of macrophages under different conditions: M, CD47⁺CDK1⁺ EC cells + M, and CD47⁻CDK1⁻ EC cells + M. F Represents the phagocytic images of macrophages incubated with GFP-labeled EC cells, with a scale bar of 50 μm; G Depicts the FACS-based phagocytosis diagram, where Q1 represents unphagocytosed GFP⁺ EC cells, Q2 represents phagocytosed GFP⁺CD45⁺ macrophages, and Q3 represents unphagocytosed CD45⁺ macrophages. H WB results of GST-pull down MS. Input represents the whole cell lysate before the experiment, used to confirm the expression of the bait protein GST-CD47 and target proteins. Lanes: 1, GST; 2, GST-CD47; 3, GST + macrophage sample; 4, GST-CD47 + macrophage sample. I Silver staining results of GST-pull down MS. Control1 represents GST + sample, Control2 represents GST-CD47 + lysate, and Sample represents GST-CD47 + macrophage sample. J Changes in membrane protein expression levels in the co-culture model. K Molecular docking predicting the interaction between HCK and CD47. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

The LGALS9-CD47 axis promotes EC cell proliferation

Our previous studies established that the CD47 ⁺ CDK1 ⁺ subpopulation interacts with macrophages to enhance LGALS9 secretion. To explore the functional consequences of this secreted LGALS9 on EC cells in the TIME, we conducted a series of assays using recombinant LGALS9 (rLGALS9). When EC cells were treated with rLGALS9 for 24 h and maintained for 5 days, we observed a significant increase in proliferation, which was effectively counteracted by the CD47 inhibitor RRX-001 (Fig. 6A, B). Apoptosis assays revealed that rLGALS9 treatment markedly reduced the apoptotic rate of EC cells, whereas CD47 inhibition not only reversed this effect but further enhanced apoptosis beyond baseline levels (Fig. 6C, D). Cell cycle analysis revealed that treatment with rLGALS9 led to an accumulation of cells in the S and G2/M phases, accompanied by a decrease in the G0/G1 population (Fig. 6E, F). Inhibition of CD47 partially or completely reversed this effect. Supporting these findings, Edu proliferation assays confirmed that rLGALS9 significantly enhanced the fluorescence rate of EC cells, indicating a direct pro-proliferative effect (Fig. S2C). Furthermore, rLGALS9 treatment upregulated Cyclin D1, PCNA, and Bcl-2, while CD47 inhibition partially reversed this effect. Conversely, the expression of pro-apoptotic Bax was significantly suppressed by rLGALS9 (Fig. 6G-I). Collectively, these findings indicate that LGALS9 enhances EC cell proliferation and survival through a CD47-dependent mechanism, thereby reinforcing the oncogenic role of the LGALS9-CD47 axis in the TIME.

Fig. 6.

The LGALS9-CD47 axis promotes EC cell proliferation. A, B. Changes in cell viability after 24 h and 5 days of treatment with rLGALS9 and RRX-001, detected by the CCK-8 assay. C-F. Effects of rLGALS9 and CD47 inhibitor treatment for 24 h on cell apoptosis and cell cycle, analyzed by flow cytometry. G Expression levels of CD47, cyclin D1 and PCNA at the mRNA levels in EC cells treated with rLGALS9 and RRX-001 for 24 h. H, I After treating with rLGALS9 and RRX-001 for 24 h in EC cells, the expression of CD47, cyclin D1, PCNA, Bcl-2 and Bax was measured by WB, with GAPDH used as a loading control. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

ERRγ is the transcription factor of the CD47⁺CDK1⁺ epi and positively regulates CD47 transcription

Given the substantial side effects associated with current CD47-targeting therapies, identifying upstream transcription factors of CD47 is essential for developing alternative strategies. By intersecting the upregulated DEGs between the CD47⁺CDK1⁺ subpopulation and cluster_3, with the upregulated DEGs in tumor versus normal samples across all epithelial cells and transcription factor–related genes [26], we identified eight transcription factors significantly upregulated: CENPX, NME2, SOX6, ERRγ, CHCHD3, CREB5, MEIS1, and DACH1 (Fig. 7A, B). We further examined the correlation between these transcription factors and macrophage infiltration in scRNA-seq data. Among the candidates, ERRγ showed the strongest correlation with M2 macrophages (cor = 0.237) (Fig. 7C, Fig. S3C). Consistently, transcript levels across EC cells supported that ERRγ may be the dominant regulator in the CD47⁺CDK1⁺ subpopulation (Fig. 7D). Functional validation further confirmed this regulatory role: CUT&Tag assays identified ERRγ binding within the CD47 promoter region, while dual-luciferase reporter assays demonstrated direct transcriptional activation of CD47 by ERRγ (Fig. 7E, F). Additionally, experimental modulation of ERRγ expression revealed parallel changes in CD47 levels (Fig. 7G-I), further supporting that ERRγ positively regulates CD47.

Fig. 7.

ERRγ is the major upstream transcription factor in CD47⁺CDK1⁺ epi. A Venn diagram showing the intersection of upregulated DEGs between the CD47⁺CDK1⁺ oncogenic subpopulation and the benign epithelial subpopulation cluster_3, upregulated DEGs in tumor versus normal samples of epithelial cells, and all transcription factors. B Bar plot illustrating the Log2 Fold change trends of 8 transcription factors in the CD47⁺CDK1⁺ oncogenic subpopulation. C Correlation analysis between the 8 transcription factors and macrophage infiltration (M1 + M2) using the TIMER database. D The mRNA expression levels of eight transcription factors. E, F Validation of the binding site between ERRγ and CD47 using CUT & Tag and dual-luciferase reporter assays. G-I Expression levels of CD47 at the mRNA and protein levels after modulating ERRγ, with GAPDH used as a loading control. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

Targeting the CD47⁺CDK1⁺ epi suppresses EC cell proliferation

To evaluate whether inhibiting ERRγ could suppress EC cell proliferation, we treated EC cells with increasing concentrations of the ERRγ inhibitor GSK5182 for 24 h and for 5 consecutive days. Both short- and long-term treatments led to a significant reduction in proliferation (Fig. 8A, B). Consistently, transfection with siERRγ virus led to a significant decrease in EC cell viability (Fig. 8C), accompanied by downregulation of the proliferation markers Cyclin D1 and PCNA at both mRNA and protein levels (Fig. 8D-F). These findings suggest that inhibiting ERRγ effectively suppresses EC cell proliferation. Given that estrogen stimulates ERRγ expression in EC cells [28], we treated EC cells with various concentrations of estrogen and progesterone for 24 h, selecting optimal concentrations for subsequent experiments (Fig. S4A, B). When EC cells were treated with the optimal hormone concentrations, we observed an increase in the expression levels of ERRγ, CD47, and proliferation markers, while progesterone exerted the opposite effect (Fig. 8G-J). These results suggest that progesterone inhibits the expression of ERRγ and CD47 and suppresses EC cell proliferation, supporting a potential therapeutic strategy combining ERRγ inhibition with progesterone supplementation in early-stage EC. Additionally, we evaluated the effects of estrogen and progesterone on LGALS9 secretion by macrophages. No significant changes in LGALS9 secretion were observed after treatment with the optimal hormone concentrations (Fig. S4C), suggesting that the regulatory effects of hormones are largely confined to EC cells rather than macrophages.

Fig. 8.

Targeting the CD47⁺ CDK1⁺ epi suppresses EC cell proliferation. A Changes in cell viability after 24 h of treatment with GSK5182, detected by the CCK-8 assay. B  Time-dependent changes in cell viability over 5 days of treatment with GSK5182, detected by the CCK-8 assay. C Changes in EC cell viability after transfection with si-ERRγ. D-F Expression levels of proliferation-related genes at the mRNA and protein levels in EC cells after overexpression (ov-ERRγ) or knockdown (si-ERRγ) of ERRγ, with GAPDH used as a loading control. G-J. After treating with estrogen and progesterone for 24 h in EC cells, the expression of CD47, ERRγ, cyclin D1 and PCNA was measured by WB, with GAPDH used as a loading control. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

Clinical data confirm the association of ERRγ/CD47/HCK/LGALS9 with EC development

We analyzed pathological sections from three patients with paired tumor and adjacent non-tumor tissues from scRNA-seq. We performed mIF and three regions of interest (ROIs) were selected from both the tumor epithelium and matched normal epithelium. Quantitative analysis revealed a significant increase in the proportion of CD47⁺CDK1⁺ cells in the tumor epithelium compared to the normal epithelium (Fig. 9A, B, Fig. S5A-D), consistent with the scRNA-seq findings and supporting their reproducibility. Additionally, we performed IHC on five EC samples, three adjacent endometrial tissues, and five AEH samples. Both CD47 and LGALS9 expression were upregulated as early as the AEH stage, indicating their involvement in the initiation of EC and potential utility as biomarkers for early diagnosis (Fig. 9C–E). Importantly, the expression level of CD47 progressively increased with the progression of endometrial lesions (Fig. 9D), indicating that CD47 is a key marker in the early carcinogenic progression of EC.

Fig. 9.

Correlation between CD47 and EC in clinical samples. A, B. Immunofluorescence was performed to identify CD47⁺CDK1⁺ positive cells in normal and EC epithelium (3 ROIs per group). Scale bar = 100 μm (A). Bar chart of the proportion of CD47⁺CDK1⁺ cells in tumor epithelium and adjacent normal epithelial ROI (B). C-E. Representative images of the IHC of CD47 and LGALS9 in cancer tissues and normal tissues and a statistical plot of CD47 and LGALS9 expression in different groups. F Brief flow chart of proteomics. G. Expression of CD47 in 24 tumor samples and adjacent normal endometrial tissues. H. The expression of CD47 in 15 cases of favorable prognosis and 9 cases of poor prognosis EC tissues. I. Prediction of survival changes in EC patients with different expression levels of CD47 using proteomics data. EC patients were stratified into low-risk and high-risk groups based on the median expression value. Survival curves were compared using the Log-rank test and visualized with the Kaplan-Meier method. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

To further validate the clinical relevance of the CD47-HCK-LGALS9 axis and ERRγ in the onset and progression of EC, we analyzed an independent external dataset (GSE17025) from the GEO database, comprising early-stage EC samples. Differential expression analysis between tumor and normal tissues revealed significant upregulation of ERRγ, CD47, HCK, and LGALS9 in early-stage EC (Fig. S6A). Receiver operating characteristic (ROC) analysis showed strong diagnostic performance for all four genes, with AUC values > 70% (CD47: AUC = 82.0%; LGALS9: AUC = 78.3%; HCK: AUC = 78.4%; ERRγ: AUC = 75%) (Fig. S6B, C). Stratifying tumors based on pathological grade, 91 tumors were divided into early grade (G1–G2) and advanced grade (G3). CD47 expression was significantly elevated in advanced-grade tumors, while no significant differences were observed for ERRγ, HCK, and LGALS9 (Fig. S6D). We further stratified the TCGA-UCEC cohort, classifying G3 or FIGO stage III/IV samples as advanced grade and advanced stage. CD47 expression was significantly higher in both advanced grade and advanced stage tumors (Fig. S6E, F). HCK and LGALS9 were also significantly upregulated in advanced-grade tumors, but no significant increases were observed in advanced-stage samples (Fig. S6E, F). Molecular subtyping plays a crucial role in guiding individualized therapy, predicting prognosis, and understanding tumor biology in EC. Traditionally, EC is categorized into four molecular subtypes: POLE ultramutated, microsatellite instability-high (MSI), copy number-low (CNL), and copy number-high (CNH). Notably, p53 mutation status is a critical determinant in molecular classification. Analysis of TCGA-UCEC samples with available molecular subtyping (n = 223) or TP53 mutation data (n = 519) revealed that CD47 expression was significantly elevated in the CNH and p53-mutant subtypes (Fig. S6G, H). Furthermore, high expression of CD47, HCK, and ERRγ was associated with poor prognosis (Fig. S2I).

We also conducted proteomic analysis on 24 early-stage EC tissues with varying prognoses, paired with adjacent normal endometrial tissues. This cohort included samples from 15 alive patients and 9 deceased patients [10] (Fig. 9F, Table S2). Our results showed that CD47 protein levels were significantly higher in tumor samples compared to normal endometrial tissues (Fig. 9G). In contrast, no significant differences were found in the expression of HCK and LGALS9 between tumor and normal samples (Fig. S7A). Furthermore, patients with poor prognosis exhibited stronger CD47 expression compared to those with favorable outcomes (Fig. 9H). No significant differences in HCK and LGALS9 expression were observed between patients with different prognoses (Fig. S7B). Elevated CD47 expression was significantly associated with poorer prognosis in early-stage EC patients (Fig. 9I), while HCK and LGALS9 showed no such association in proteomic samples (Fig. S7C). Overall, our findings highlight the critical role of CD47 in the development and prognosis of EC, suggesting its potential value for diagnosis and molecular subtyping.

Validation of the critical role of CD47 in EC using 3D organoid models in vivo

To further validate the functional role of CD47 in early-stage EC, we established PDOs using tumor samples from three early-stage EC patients, thereby recapitulating the in vivo microenvironment. Successful construction of EC PDOs was confirmed by optical microscopy and H&E staining (Fig. S8A). IHC staining for the tumor markers EPCAM, Ki67, and pan-CK further verified the formation of PDOs (Fig. S8B). After optimizing the viral transduction concentration (ov-CD47 and empty vector) (Fig. S8C, D), we transduced EC organoids with the lentivirus. Flow cytometry confirmed the overexpression efficiency of CD47 in ov-CD47-transduced PDOs (Fig. 10A, B). Following five days of culture post-transduction, we performed a CCL-3D proliferation assay, which revealed that the ov-CD47 group exhibited accelerated growth and higher viability compared to controls, indicating that CD47 promotes the proliferation of early-stage EC PDOs (Fig. 10C, D). Subsequently, PDOs with varying CD47 expression levels were co-cultured with macrophages derived from PBMCs. After 72 h, flow cytometric analysis showed that macrophage phagocytosis of ov-CD47 PDOs was reduced (Fig. 10E, F). Moreover, ELISA analysis of culture supernatants indicated significantly increased secretion of LGALS9, IL-10, and TGF-β1 in the ov-CD47 group (Fig. 10G-I), suggesting that CD47 contributes to the establishment of an immunosuppressive microenvironment in EC. In addition, treatment of PDOs with rLGALS9 for 72 h promoted their proliferation (Fig. 10J, K). Collectively, these results highlight the pivotal role of the LGALS9-CD47 axis in promoting early-stage EC proliferation and shaping an immunosuppressive TME.

Fig. 10.

The role of the LGALS9-CD47 axis in early-stage EC was confirmed in PDOs. A Flow cytometric analysis of CD47 expression in PDOs transfected with ov-CD47 virus (n = 3). B Bar graph showing the expression levels of CD47 in PDOs transfected with control and ov-CD47 virus. C The morphological changes in PDOs at 0 h, 72 h, and 120 h were evaluated using light microscopy after transfection with control and ov-CD47 virus (n = 3). D The cell viability treated with different viruses was evaluated by CCL 3D with three replications. E, F Phagocytosis of indicated PDOs was represented by the percentage of FITC⁺CD14⁺ cells in total CD14⁺ cells (n = 3). G-I. The levels of LGALS9, IL-10 and TGF-β1 secreted by PBMC-derived macrophages (M) were detected by ELISA. J The morphological changes in PDOs at 0 h, 24 h, and 72 h were evaluated using light microscopy with or without rLGALS9 treatment (n = 3). K. The cell viability treated with different treatments was evaluated by CCL 3D with three replications. Significance levels: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

Disscussion

The malignant proliferation of tumor cells is a hallmark feature of tumorigenesis, both biologically and clinically. It not only drives tumor initiation but also promotes progression, invasion, metastasis, and drug resistance. In this study, we analyzed the transcriptomic characteristics of early-stage EC and identified a strong association between early-stage EC and the dysregulation of the cell cycle. Dysregulation of the cell cycle is a well-established feature of uncontrolled tumor cell proliferation, contributing to the development of early-stage EC [27]. While early-stage EC patients are traditionally considered to have a favorable therapeutic response to surgical resection and long-term survival, a subset of patients with early-stage, low-grade disease experiences rapid progression. Therefore, it is crucial to gain a deeper understanding of the biological and molecular mechanisms underlying this disease. Such insights may contribute to the development of more effective, timely, and individualized treatment strategies for early-stage EC patients.

The advent of scRNA-seq, spatial transcriptomics, and mIF technology has enabled a deeper exploration of the heterogeneity of the TME and its role in tumor biology. Previous studies have demonstrated that epithelial subpopulations in EC play a significant role in its development and poor prognosis [9, 28], while interactions between immune cells and tumor cells within the TME are also critical in the initiation and progression of EC [29]. Jin et al. utilized a co-culture system incorporating EC cells and macrophages, including direct, indirect, and conditioned medium-based co-culture methods, to demonstrate that tetrabromobisphenol A stimulates TAMs toward M2 polarization, thereby promoting EC proliferation, migration, and invasion [30]. Additionally, the accumulation of TAMs in EC tissues significantly increases interleukin-8 (IL-8) secretion and downregulates estrogen receptor alpha expression, which ultimately contributes to poor prognosis in EC patients [31]. However, most current studies primarily focus on prognosis-related mechanisms in EC, with a notable gap in systematic investigations of the heterogeneous epithelial subpopulations in early-stage EC. In our study, we characterize the changes in the TME of early-stage EC and identify the CD47⁺CDK1⁺ epi as a critical driver of malignant proliferation through integrated molecular profiling. CDK1, a key regulator of the cell cycle, is well-known for its role in driving tumor cell cycle progression. It is closely linked to the initiation and progression of various cancers, where it accelerates proliferation and increases invasiveness [32].

CD47 (also termed IAP, MER6, or OA3) is an integrin-associated glycoprotein expressed on cell surfaces [33]. Its expression is significantly upregulated in multiple malignancies—including glioblastoma, colon cancer, and gastric cancer—where it facilitates tumor immune evasion [34–36]. Zhang et al. [37] showed that CD47 promotes craniopharyngioma cell proliferation, migration, and invasion by activating the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) signaling pathway. Beyond MAPK/ERK activation, CD47 receptor signaling stimulates the phosphatidylinositol 3-kinase/protein kinase B (PI3K/Akt) pathway through interactions with PLIC-1 and Gβγ dimers, driving proliferation in human astrocytoma cells [38]. Studies indicate that CD47 expression is substantially elevated in EC tissues and positively correlates with advanced pathological staging, myometrial invasion, and lymph node metastasis, positioning it as a potential diagnostic and prognostic EC marker [39]. In this study, in vitro functional experiments demonstrate that CD47 inhibition markedly reduces EC cell and EC PDOs proliferation and downregulates proliferation-related markers, confirming CD47’s pro-proliferative role in EC. Clinically, IHC revealed a progressive upregulation of CD47 across adjacent normal endometrium, AEH samples, and tumor tissues, correlating with disease progression. Moreover, proteomic analysis confirmed that CD47 serves as an unfavorable prognostic factor in early-stage EC.

Furthermore, this study identifies a CD47⁺CDK1⁺ epithelial subpopulation whose ligand-receptor crosstalk not only suppresses macrophage phagocytosis but also drives malignant proliferation of EC cells. We experimentally confirmed, for the first time, the interaction between LGALS9 and CD47. LGALS9 functions as an immunoregulatory inhibitor that promotes regulatory T cell (Treg) differentiation and reduces helper T cell subsets, including Th1 and Th17 cells, thereby restraining excessive immunity and inflammation [40, 41]. Studies have shown that in macrophages overexpressing LGALS9, the expression levels and activity of CD206, transforming growth factor-beta (TGF-β), interleukin-10 (IL-10), and signal transducer and activator of transcription 3 (STAT3) are significantly elevated, which is closely associated with M2-type macrophages, suggesting that LGALS9 overexpression can induce M2 macrophage polarization and inhibit M1 macrophage differentiation [42]. This further underscores the role of LGALS9 in tumor immune suppression. Our study observed that co-culturing macrophages with EC cells significantly upregulates LGALS9 secretion, an effect reversed by CD47 inhibitor treatment. Given that CD47 is a membrane-bound protein, we hypothesize that its engagement with macrophage surface proteins triggers increased secretion of LGALS9. Using GST pull-down MS analysis, we identified hematopoietic cell kinase (HCK)—a SRC family tyrosine kinase predominantly expressed in myeloid-derived cells [43]—as the macrophage membrane protein that binds to CD47. Prior studies reported that HCK is significantly up-regulated in a variety of malignant tumors, such as mantle cell lymphoma [44] and colorectal tumors [45], which promotes the occurrence and proliferation of tumors and is closely related to the poor prognosis of patients. Ashleigh et al. revealed that HCK-dependent signaling in macrophages exerts tumor-extrinsic effects in colorectal cancer, promoting polarization of macrophages toward an immunosuppressive alternatively activated phenotype and enhancing expression of STAT3-activating cytokines (e.g., IL-6 and IL-11) in both tumor cells and TAMs. HCK-dependent signaling aligns with elevated STAT3 activity in tumor epithelia and accelerated tumorigenesis [46]. Therapeutically targeting HCK activity may therefore stimulate immune cell activation while suppressing the immunosuppressive TME, potentially improving responses to immunotherapy. However, the role of HCK in the EC immune microenvironment remains unexplored. Our study predicts that CD47-HCK interaction promotes secretion of the immunosuppressive molecule LGALS9; the subsequent rise of LGALS9 in the TME facilitates binding to CD47 on tumor cells, thereby driving EC proliferation. These findings elucidate the mechanistic basis of protein interactions between epithelial subpopulations and macrophages that trigger tumor proliferation and position CD47 as a pivotal node in the malignant proliferation of early-stage EC cells, informing the design of more targeted early-intervention strategies.

CD47-targeting antibodies, as immune checkpoint agents, have shown promising therapeutic effects. However, because CD47 is widely expressed on non-malignant cells—including normal and senescent red blood cells and platelets—many CD47 antibodies cause severe adverse events such as hemolytic anemia [47]. To mitigate these on-target toxicities, we investigated upstream transcriptional regulation of CD47, aiming to indirectly suppress CD47 expression by targeting its regulator ERRγ and thereby develop safer, more precise therapies. ERRγ, an orphan nuclear receptor critical for mitochondrial biogenesis and cellular energy homeostasis, is markedly upregulated in multiple cancers, including estrogen receptor–positive breast cancer [48], and is increased in chemotherapy-resistant cancer cells [49]. Sun et al. reported that in EC cells, ERRγ activates the Akt and ERK1/2 pathways, mediating estrogen-driven proliferation [50]. Consistent with this, our in vitro data show that ERRγ regulates EC cell proliferation and its expression is modulated by estrogen and progesterone. Clinically, for early-stage, well-differentiated, localized EC—particularly progesterone receptor–positive cases—combining ERRγ-targeted therapy with high-dose progesterone may improve outcomes for patients seeking fertility preservation.

In this study, we firstly identified a specific oncogenic CD47⁺CDK1⁺ subgroup in early-stage EC through scRNA-seq and validated these findings with spatial transcriptomics. Molecular functional assays and TME analyses indicate that this subgroup closely interacts with macrophages via the CD47-HCK-LGALS9 axis, promoting an immunosuppressive microenvironment. Additionally, we explored upstream transcription factors of this oncogenic subgroup and found that progesterone suppresses ERRγ expression. These findings provide novel insights into the molecular mechanisms driving early-stage EC and suggest that targeting the CD47-HCK-LGALS9 axis may offer a promising therapeutic approach. Moreover, understanding the regulatory role of ERRγ could contribute to the development of more effective treatment strategies, particularly in combination with hormone therapy.

However, this study has limitations. First, the sample size of the scRNA-seq analysis was relatively small; future research could address this limitation by expanding cohorts and validating the CD47⁺CDK1⁺ subgroup in larger, more diverse patient populations, both in vivo and in clinical settings. Second, the spatial transcriptomics approach was limited to a single sample (n = 1), and the FFPE processing may have compromised cellular integrity. To overcome this limitation, future studies could include a broader set of samples and employ alternative tissue preservation techniques to enhance data reliability. Despite these limitations, our study lays the groundwork for further exploration into the molecular characteristics of early-stage EC and its potential clinical applications.

Conclusions

In summary, this study is the first to identify an oncogenic CD47⁺CDK1⁺ epithelial subpopulation in early-stage EC. This subset suppresses macrophage phagocytosis and, together with macrophage-derived LGALS9, drives malignant proliferation. In addition, this study further confirmed the ability of CD47 to promote malignant proliferation of EC and promote the formation of an immunosuppressive microenvironment through the EC organoids. It shows that early-stage EC patients can be treated by targeting CD47. We further identify ERRγ as an upstream regulator of CD47, offering an avenue for indirect targeting (Fig. S9). Based on these findings, we propose combining progesterone with ERRγ inhibitors as a potential therapeutic strategy for early-stage EC.

Authors’ contributions

Jingxuan Ye, Yitong Yan, and Xiaoqi Sun conducted the experiments, analyzed the data, and wrote the manuscript. Qibin Wu and Yashi Shi collected and analyzed the data and provided valuable discussions. Maotong Zhang, Xite Lin, and Yuan Ren performed the data management. Xiaodan Mao, Mingyan Lin, and Pengming Sun contributed to the critical revision of the article for important intellectual content, supervised the research and provided financial support. All authors gave final approval for the submitted version and agreed to take responsibility for all aspects of the work.

Funding

This research was funded by the National Key Research and Development Program of China (No.2022YFC2704303), the National Nature Science Foundation of China (No.82203739), the Fujian Province Central Government-Guided Local Science and Technology Development Project (No.2023L3019) and the Startup Fund for scientific research, Fujian Medical University(No.2023QH2045 and No.2024QH1211).

Data availability

The data presented in this study are available upon request from the corresponding author. The data are not publicly available due to privacy concerns.

Declarations

Ethics approval and consent to participate

All subjects gave written or in the form of recorded phone-informed consent to participate in the study. Informed consents were waived by the Ethics Committee as all clinical data were extracted from the medical system and personal information of any patient was masked. The study was conducted in accordance with the Declaration of Helsinki and approved by Fujian Maternity and Child Health Hospital Ethics Commission (Grant NO. 2021KRD001).

Consent for publication

All authors have approved the manuscript for publication.

Competing interests

The authors declare no competing interests.