A three-step metastatic ecological model of endometrial cancer lymphovascular space invasion and its construction method

Abstract

This invention discloses a three-step metastatic ecological model of endometrial cancer lymphovascular space invasion and its construction method, belonging to the field of biomedical technology. This invention constructs single-cell atlases of endometrial cancer under different lymphovascular space invasion states through single-cell RNA sequencing, performs cell subset analysis, functional scoring, and cell communication network inference, and combines multiplex immunofluorescence spatial validation to construct a three-step metastatic ecosystem model consisting of three consecutive stages: tumor epithelial invasion initiation, microenvironmental synergistic remodeling, and vascular invasion. This invention expands lymphovascular space invasion from a static pathological description to a dynamic multicellular ecological evolution process, clarifying key cell subsets, differential ligand-receptor interaction axes, and spatial validation biomarker combinations, providing a technical foundation for early risk assessment, molecular biomarker development, and therapeutic target screening of lymphovascular space invasion.

Description

Technical Field

[0001] This invention belongs to the fields of biomedicine, tumor single-cell omics and tumor microenvironment analysis technology, and specifically relates to a method for constructing a three-step metastatic ecosystem model for endometrial cancer lymphovascular space invasion based on single-cell transcriptome sequencing, cell communication inference and multiplex immunofluorescence spatial verification. Background Technology

[0002] Lymphovascular space invasion (LVSI) is an important pathological feature in the staging of endometrial cancer, referring to the invasion of tumor cells into the lymphatic or vascular-like spaces lined by endothelial cells, forming tumor cell clusters or tumor emboli. This pathological phenomenon is closely related to the local invasion, lymph node metastasis, distant dissemination, recurrence risk, and patient prognosis of endometrial cancer, and is one of the important indicators in clinicopathological assessment and risk stratification. In traditional clinical practice, LVSI is mainly interpreted as a postoperative pathological marker, and its diagnosis relies on the observation of tumor cell clusters located within lymphatic or vascular-like structures in the surgically resected specimen. Currently, the existing diagnostic model is based on the histological results after LVSI has formed, while the specific biological changes in its early occurrence and evolution are still unknown.

[0003] Tumor cells typically undergo multiple sequential steps to enter the vascular system, including enhanced malignant epithelial cell invasion phenotype, basement membrane degradation, local migration, extracellular matrix remodeling, hypoxic adaptation, immune escape, endothelial cell activation, and altered vascular barrier permeability. During this process, changes in the transcriptional state of the tumor cells themselves, along with the synergistic effects of surrounding immune cells, stromal cells, and endothelial cells, intertwine to propel tumor cells from local infiltration into the vascular lumen. Therefore, lymphovascular infiltration is not a single, static pathological outcome, but rather a multicellular process with temporal sequence and spatial tissue characteristics. This process may involve a dynamic interplay of tumor cell acquisition of invasive capabilities, tumor microenvironment remodeling, alterations to the vascular or lymphatic barrier, and local immunosuppression.

[0004] Currently, some studies have attempted to predict LVSI status preoperatively using radiomics methods. For example, fusion models are constructed based on multi-parameter MRI radiomics features combined with clinicopathological factors, or preoperative models are established based on clinical biomarkers to predict lymph node and intervascular space involvement. However, these predictive models are mostly based on macroscopic imaging or clinical statistical parameters, and are essentially still indirect determinations of the presence or absence of LVSI, failing to reveal the mechanisms of LVSI at the cellular and molecular levels.

[0005] At the level of LVSI mechanism research, existing technologies mostly analyze the relevant mechanisms of LVSI from the perspectives of clinicopathological factors, single cell types, single molecular pathways, or bulk transcriptome. Bulk transcriptome analysis reflects the average expression signal of the entire tissue and cannot distinguish the specific cellular origin of molecular changes, thus making it difficult to construct evolutionary models with cell type resolution. Although single-cell RNA sequencing technology has been used to map the cellular atlas of endometrial cancer development, and some studies have integrated multi-omics data to analyze the immunotherapy response of endometrial cancer, none of these studies have focused on the multi-cell co-evolutionary process of LVSI as a specific pathological event. In particular, they lack a systematic research method that combines single-cell transcriptome cell subset analysis, ligand-receptor cell communication inference, and spatial validation techniques.

[0006] Therefore, the existing technology has at least the following shortcomings: Current techniques mostly treat lymphovascular space infiltration as a static interpretation of postoperative pathological morphology, lacking dynamic, staged analysis of its occurrence process, making it difficult to clarify the continuous links from local invasion to entry into the blood vessel lumen of tumor cells.

[0007] Existing technologies make it difficult to determine which specific malignant epithelial cell states initiate lymphovascular space infiltration, and what invasive molecular characteristics these malignant epithelial cells possess.

[0008] Existing technologies struggle to explain how the microenvironmental components, such as immune cells, macrophages, fibroblasts, and endothelial cells, undergo synergistic remodeling during lymphovascular space infiltration.

[0009] Current technologies lack a modeling method for lymphovascular space infiltration mechanisms that combines single-cell transcriptome cell subset analysis, ligand-receptor cell communication inference, and multiplex immunofluorescence spatial verification.

[0010] Current technologies have not yet verified the immunosuppressive and metastatic ecological characteristics of the invasion front and the interior of vascular tumor emboli at the spatial tissue structure level.

[0011] In recent years, advancements in technologies such as single-cell RNA sequencing, cell communication inference, and multiplex immunofluorescence have provided significant technical support for resolving cellular composition at single-cell resolution, inferring intercellular signaling interactions, and validating molecular spatial distribution in situ within tissues. Therefore, this invention combines single-cell RNA sequencing, cell communication inference, and multiplex immunofluorescence to systematically elucidate the mechanisms of lymphovascular invasion from three levels: cellular composition, signaling interactions, and spatial localization. The ultimate goal is to construct a multicellular evolutionary model of lymphovascular invasion in endometrial cancer, providing a technical foundation for screening molecular markers related to lymphovascular invasion, exploring auxiliary risk assessment methods, and developing targeted therapy strategies. Summary of the Invention

[0012] To address the problems mentioned in the background art, this invention provides a three-step metastasis-promoting ecological model of endometrial cancer with lymphovascular space invasion and its construction method. The invention constructs an atlas of endometrial cancer cells in different LVSI states through single-cell RNA sequencing, analyzes the changes in malignant epithelial cells, immune cells, macrophages, fibroblasts and endothelial cell subsets, and further combines cell communication inference and multiplex immunofluorescence spatial verification to construct the LVSI formation process as a three-step metastasis-promoting ecological model of "tumor epithelial invasion initiation - microenvironmental synergistic remodeling - vascular invasion".

[0013] This invention includes the following steps: a) Constructing a single-cell atlas of endometrial cancer under different lymphovascular space infiltration states Normal endometrial tissue, early low-grade endometrial carcinoma tissue with negative lymphovascular space infiltration (LVSI), and early low-grade endometrial carcinoma tissue with positive LVSI were collected. Single-cell suspensions were prepared, and single-cell RNA sequencing libraries were constructed and sequenced. After data quality control, integration, dimensionality reduction, and clustering, single-cell transcriptome atlases of endometrial and tumor tissues under different conditions were obtained.

[0014] b) Annotation of major cell types and tissue distribution preference analysis By screening for differentially expressed genes and combining them with classic cell type marker genes, the major cell types in the single-cell transcriptome atlas were manually annotated. Based on the odds ratio, the tissue distribution preferences of different cell populations in normal tissues, LVSI-negative tumors, and LVSI-positive tumors were evaluated, and differential abundance analysis was performed.

[0015] c) Identification and feature extraction of malignant epithelium associated with lymphovascular space infiltration Epithelial cells were extracted and re-clustered. Malignant epithelial cells were identified by inferCNV analysis. The functional module score was used to assess the invasion-related functional status of malignant epithelial cells. LVSI-related malignant epithelial subpopulations with pro-invasive characteristics such as partial epithelial-mesenchymal transition, hypoxia, glycolysis, and high expression of immunosuppression-related molecules were identified as basic data for the "tumor epithelial invasion initiation" stage in the three-step ecological model.

[0016] d) Reclustering analysis and functional status analysis of tumor microenvironment cell subsets T / NK cells, myeloid cells, fibroblasts, and endothelial cells were extracted separately and subjected to independent high-resolution re-clustering analysis to obtain refined subpopulations of each cell type. The functional status of cell clusters in different microenvironments was systematically evaluated using a functional scoring module, and transcription factor regulatory network analysis was performed on the fibroblast subpopulation.

[0017] e) Cell communication network deduction and screening of key ligand-receptor interactions Based on single-cell expression matrices and cell subpopulation annotation information, cell communication networks under different LVSI states are constructed, overexpressed ligand-receptor interactions are identified, inter-cell communication probabilities are calculated, and LVSI-related differential ligand-receptor interaction axes are screened through inter-group comparisons.

[0018] f) Spatial verification of multiplex immunofluorescence Paraffin-embedded tissue samples of endometrial cancer in different LVSI states were selected, serial sections were prepared, and multiple immunofluorescence staining and whole-section imaging were performed using multiple antibody combinations to obtain spatial colocalization information of key cell subpopulations and signaling molecules.

[0019] g) Construction of a three-tiered ecosystem model for promoting relocation By integrating the single-cell transcriptome atlas, malignant epithelial cell invasion characteristics, microenvironment cell subpopulation functional characteristics, differential ligand-receptor interaction axes, and spatial validation results obtained in steps a) to f), a three-step metastasis-promoting ecosystem model of lymphovascular space infiltration consisting of three consecutive stages: "tumor epithelial invasion initiation - microenvironment synergistic remodeling - vascular invasion".

[0020] Preferably, the LVSI positive group described in step a) is further divided into a focal LVSI group and a generalized LVSI group based on the number of affected blood vessels as determined by D2-40 / EPCAM and CD31 / EPCAM immunostaining, wherein generalized LVSI is defined as having ≥5 affected blood vessels.

[0021] Preferably, the early low-grade endometrioid endometrial cancer described in step a) is FIGO 2009 stage IA, grade G1-G2 endometrioid endometrial cancer.

[0022] Preferably, the classic cell type marker genes mentioned in step b) include: epithelial cell marker genes EPCAM, KRT8, KRT18; T cell marker genes CD3D, CD3E, CD3G; myeloid cell marker genes CD68, CSF1R; fibroblast marker genes COL1A1, COL1A2, DCN; endothelial cell marker genes PECAM1, VWF; B cell marker genes CD19, MS4A1; and plasma cell marker genes JCHAIN, IGKC.

[0023] Preferably, the LVSI-related malignant epithelial subgroup described in step c) is the TC4 subgroup, whose invasive characteristics include: partial epithelial-mesenchymal transition, hypoxia, glycolysis, high expression of immunosuppression-related molecules CD47 / CD274 / IDO1, and low expression of progesterone receptor co-genes.

[0024] Preferably, the refined subgroups obtained by the high-resolution re-clustering analysis in step d) include: T / NK cell subsets: Cycling_T (Cycling_MCM+CD8+T, Cycling_BUB+CD8+T and Cycling_CD4+T), ZNF683+SOX4+CD8+T, TNFRSF9+Treg; Myeloid cell subsets: SPP1+MMP9+ macrophages (SPP1+MMP9+Mac), APOE+CXCL9+ macrophages (APOE+CXCL9+Mac), LAMP3+ dendritic cells (cDC1_LAMP3_DC and cDC2_LAMP3_DC). Fibroblast subsets: iCAF, WNT5A+ myofibroblasts (WNT5A_mCAF), LRRC75A+ myofibroblasts (LRRC75A_mCAF); Endothelial cell subsets: Tip_ArtECs, Tip_CapECs, and Tip_VenECs (collectively referred to as Tip_ECs).

[0025] Preferably, the functional scoring module in step d) includes: TAM score, myeloid-derived suppressor cell score, M1 macrophage score, M2 macrophage score, activation score, phagocytosis score, migration score, endoplasmic reticulum stress score, and vascular function-related module.

[0026] Preferably, the transcription factor regulatory network analysis in step d) adopts the SCENIC method, including: filtering the original counting matrix, inferring the gene co-expression network using GENIE3, identifying transcription factor binding sites using RcisTarget, establishing transcription regulators, and quantifying single-cell transcription factor activity using AUCell scoring.

[0027] Preferably, the differential ligand-receptor interaction axes in step e) include: the SPP1-integrin axis, the LGALS9-immune checkpoint axis (involving P4HB / HAVCR2 / CD45 / CD44), the WNT5A-FZD / MCAM axis, and the collagen-integrin axis.

[0028] Preferably, the combination of multiple antibodies in step f) includes: Panel 1: Ki67, HIF1A, EPCAM, CD44, used to validate the spatial distribution of invasive tumor epithelial cell markers; Panel 2: CD68, PD-L1, EPCAM, and CD3 were used to verify the spatial relationships among macrophages, PD-L1 expression, tumor epithelial cells, and T cells. Panel 3: Ki67, PD-1, CD8, CD4, used to verify the spatial distribution of PD-1 positive T cells and proliferation-related T cells.

[0029] Preferably, the three-step model described in step g) is specifically configured as follows: The first step (initiation of tumor epithelial invasion): driven by the TC4 subset in a hypoxic microenvironment, which acquires initial invasive ability; The second step (co-remodeling of the microenvironment): Cell subpopulations such as Cycling_T, ZNF683+SOX4+CD8+T, SPP1+MMP9+Mac, WNT5A_mCAF, and LAMP3_DC form an immunosuppressive and angiogenic microenvironment through signaling axes such as LGALS9-immune checkpoint, SPP1-integrin, and WNT5A-FZD / MCAM. The third step (vascular invasion): Based on increased vascular permeability mediated by Tip_ECs and immunosuppression, invasive tumor cells cross the endothelial barrier and enter the vascular system.

[0030] Preferably, the method further includes outputting at least one of the following elements based on the constructed three-tier model: In the specific implementation, the key cellular components are shown in Table 1; the key cellular communication axes are shown in Table 2; and the functional scoring modules are shown in Table 3.

[0031] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention expands the description of lymphovascular space infiltration from a single static description of traditional postoperative pathological morphology to a dynamic, phased, multicellular ecological evolution process. By integrating single-cell RNA sequencing, cell communication network analysis, and multiplex immunofluorescence tissue in situ localization, the transcriptional characteristics and spatial distribution patterns of key cell subpopulations such as malignant epithelial cells, T / NK cells, myeloid cells, fibroblasts, and endothelial cells are systematically identified at cell type resolution, revealing the continuous evolution mechanism of LVSI formation from the perspective of multicellular synergistic interaction; (2) This invention constructs a three-step metastasis-promoting ecosystem model of “tumor epithelial invasion initiation - microenvironmental synergistic remodeling - vascular invasion” located at the forefront of tumor invasion, and clarifies the key cellular and molecular events in the invasion front region at different stages, providing a new theoretical framework for understanding the occurrence and development of LVSI; (3) This invention clarifies the key cell subpopulations (TC4, SPP1+MMP9+Mac, WNT5A_mCAF, Tip_ECs, Cycling_T) associated with LVSI, the differential ligand-receptor interaction axis (LGALS9-immune checkpoint, SPP1-integrin, WNT5A-FZD / MCAM and collagen-integrin, etc.), the combination of spatial validation markers (Panel 1, Panel 2, Panel 3), the functional module scoring system (p-EMT, hypoxia, glycolysis, immunosuppression, etc.), and the three-stage model structure, providing a theoretical basis and technical foundation for early risk assessment, auxiliary judgment, development of potential molecular detection markers, and screening of therapeutic targets for early low-grade endometrioid endometrial cancer LVSI. (4) This invention is the first to discover histological evidence of the spatial co-localization of PD-L1 positive tumor cells, PD-L1 positive macrophages and PD-1 positive T cells inside the vascular tumor thrombus of endometrial cancer, suggesting that the vascular tumor thrombus may be a micrometastatic unit with active immune regulation characteristics, providing a new theoretical basis for developing immune intervention strategies against vascular tumor thrombus. (5) This invention establishes a complete technical process from single-cell sequencing data processing, cell subpopulation analysis, functional scoring, communication network inference, spatial verification to three-step model integration, forming a systematic and reproducible LVSI mechanism modeling method, which solves the problem of the lack of standardized and systematic modeling process in the existing technology, and has significant advantages in the comprehensiveness of multi-cell collaborative analysis and the reliability of spatial verification. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall technical roadmap of the present invention. It demonstrates the complete technical process of tissue sampling, single-cell suspension preparation, library construction, sequencing, data analysis (cell annotation, malignant epithelial identification, microenvironment subpopulation analysis, functional scoring, transcription factor regulatory network analysis, cell communication inference), and multiplex immunofluorescence spatial validation.

[0033] Figure 2This chart presents the characteristics and invasion initiation functional status of LVSI-related tumor epithelial cell subsets; A: Heatmap showing the expression levels of representative marker genes for each epithelial cell subset (NC1, NC2, TC1-TC6); B: GO functional enrichment analysis bar chart showing significant enrichment pathways for each tumor epithelial cell subset; C: Bubble chart showing the expression of genes related to EMT, p-EMT, hypoxia, glycolysis, cytokines, immunosuppression, and progesterone receptor synergistic effects in different tumor epithelial cell subsets; D: Heatmap showing multiple functional scores of each tumor epithelial cell subset (p-EMT, hypoxia, glycolysis, inflammation / NF-κB, angiogenesis, immune escape, immunosuppression, progesterone response, etc.); E: Violin plot showing the ranking and comparison of p-EMT scores of each tumor epithelial cell subset; F: Correlation heatmap showing the correlation matrix between various functional scores of tumor epithelial cells. This figure illustrates the differences in invasion-related functional status among different tumor epithelial cell subsets. The LVSI-associated epithelial subset (TC4) shows enhanced p-EMT, hypoxia, glycolysis, cytokine and immunosuppression-related features, accompanied by changes in progesterone receptor synergistic features.

[0034] Figure 3 A: UMAP dimensionality reduction plot of T / NK cells, showing the spatial distribution of each T / NK subset; B: Stacked bar chart showing the proportional distribution of T / NK cell subsets under different LVSI states (normal endometrial tissue, LVSI negative, LVSI positive); C: Dot plot showing the distribution preference of each T / NK cell subset in normal endometrial tissue, LVSI negative and LVSI positive samples (based on odds ratio); D: UMAP plot showing the expression levels of T / NK cell exhaustion-related genes (PDCD1, HAVCR2, LAG3, TIGIT, etc.); E: Pseudo-temporal differentiation trajectory plot of CD8+ T cells, showing the evolutionary path from initial to terminal differentiation; F: Bubble plot showing the enrichment characteristics of STAT3 / STAT5 related pathways in ZNF683+SOX4+CD8+T and TNFRSF9+Treg subsets; G: Network diagram showing the comparison of neighborhood network analysis of T cells in LVSI positive and LVSI negative groups.

[0035] Figure 4The diagram shows the enrichment and functional status characteristics of monocyte-macrophage subsets related to LVSI; A: UMAP distribution map of monocyte-macrophage subsets; B: Stacked bar chart showing the proportion distribution of each monocyte-macrophage subset under different LVSI states; C: Box plot showing the characteristic functional scores of monocyte-macrophage under different LVSI states (TAM score, M1 score, M2 score, MDSC score, phagocytosis score, migration score, etc.); D: GO functional enrichment analysis bubble chart showing the significant enrichment pathway of the SPP1+MMP9+Mac subset.

[0036] Figure 5 A: UMAP distribution map of dendritic cell subsets; B: Stacked bar chart showing the proportion distribution of each dendritic cell subset under different LVSI states; C: Violin plot showing the expression levels of representative immunosuppressive genes (IDO1, LGALS9, PD-L1, etc.) of dendritic cells under different LVSI states; D: Box plot showing the activation score, migration score, and cell cycle score of different dendritic cell subsets; E: Bar chart showing the differential enrichment pathway analysis of LAMP3_DC subsets between the LVSI positive group and the LVSI negative group.

[0037] Figure 6 A: LVSI-related fibroblast subset enrichment and matrix remodeling characteristics; B: Bubble chart showing GO functional enrichment analysis results of iCAF, WNT5A_mCAF, and LRRC75A_mCAF subsets; C: Stacked bar chart showing the proportion distribution of each fibroblast subset under different LVSI states; D: GSEA enrichment analysis showing the significant enrichment of pathways (EMT, hypoxia, angiogenesis, etc.) of the WNT5A_mCAF subset compared to other fibroblast subsets; E: Heatmap showing the activity scores of key transcription factors (FOXO3, TWIST1, RUNX2, NFKB1, etc.) of different fibroblast subsets.

[0038] Figure 7A: LVSI-related Tip_ECs enrichment and pro-invasive endothelial function characteristics map; B: UMAP distribution map of endothelial cell subsets, labeled with Tip_ArtECs, Tip_CapECs, Tip_VenECs, etc.; C: Stacked bar chart showing the proportion distribution of each endothelial cell subset under different LVSI states; D: GSEA enrichment analysis map showing the significant enrichment pathways (angiogenesis, EMT, immune regulation, etc.) of Tip_ECs subsets compared with other endothelial cell subsets; E: Box plot showing the inflammatory response, angiogenesis, EMT, and vascular permeability scores of different endothelial cell subsets; F: Heatmap showing the scores of angiogenesis, hypoxia, inflammatory response, EMT, and related signaling pathways under different LVSI states.

[0039] Figure 8 Diagrams showing the reconstruction of LVSI-related cell communication networks and analysis of key interaction axes; A: Chord plot showing the interaction network between the main cell types of LVSI-negative (left) and LVSI-positive (right) endometrial carcinoma; B: Lollipop plot showing the differential ranking of interaction probabilities between LVSI-positive and LVSI-negative cell subsets; C: Heatmap showing the interaction probabilities between SPP1+MMP9+Mac and WNT5A_mCAF with endothelial cell subsets and T / NK cell subsets, respectively; D: Bubble plot showing the specific interactions between WNT5A_mCAF and SPP1+MMP9+Mac with LECs and Tip_ECs, respectively (SPP1-integrin axis, WNT5A-FZD / MCAM axis); E: Bubble plot showing the interaction between WNT5A_mCAF and SPP1+MMP9+Mac with T / NK cells (LGALS9-immune checkpoint axis); F: Bubble plot showing the interaction between LECs and Tip_ECs with T / NK cells.

[0040] Figure 9Multiplex immunofluorescence spatial validation maps for LVSI-related metastatic niches; A: Immunofluorescence images showing the spatial distribution characteristics of HIF1A, Ki67, and CD44 in invasive tumor epithelial cells (EPCAM+) in LVSI-negative, focal LVSI, and extensive LVSI stages; B: Immunofluorescence images showing the spatial distribution of the immune microenvironment in LVSI-negative, focal LVSI, and extensive LVSI stages, including the co-localization relationships of CD68+ macrophages, PD-L1 expression, EPCAM+ tumor cells, and CD3+ T cells; C: Immunofluorescence images showing the spatial distribution and phenotypic characteristics of T cells in LVSI-negative, focal LVSI, and extensive LVSI stages, including the co-expression of CD4, CD8, PD-1, and Ki67; D: Immunofluorescence images showing the spatial co-localization of PD-L1-positive tumor cells, PD-L1-positive macrophages, and PD-1-positive T cells within vascular tumor thrombi in extensive LVSI samples.

[0041] Figure 10 Schematic diagram of the three-step metastatic ecosystem model of LVSI in endometrial cancer; A: The diagram shows the local inflammation and epithelial-dominated state in the LVSI-negative stage, corresponding to the early stage of the first step; B: The diagram shows the stage of invasion initiation and microenvironmental remodeling, where TC4-like malignant epithelial cells, SPP1+MMP9+ macrophages, WNT5A_mCAF, proliferative exhausted T cells, and Tip_ECs form a pro-invasive niche through multiple ligand-receptor axes (SPP1-integrin, WNT5A-FZD / MCAM, LGALS9-immune checkpoint, etc.), corresponding to the late stage of the first step and the second step; C: The diagram shows the stage of LVSI-positive and vascular tumor embolization, where PD-L1-positive tumor cells, PD-L1-positive macrophages, and PD-1-positive T cells co-localize within the tumor embolus, constituting an immunosuppressive vascular invasion microenvironment, corresponding to the third step. Detailed Implementation

[0042] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are only for explaining the present invention and are not intended to limit the scope of the invention in any way. Any equivalent substitutions or modifications made based on the concept of the present invention are within the protection scope of the present invention. Unless otherwise specified, the technical means used in the following embodiments are all conventional means well known to those skilled in the art.

[0043] Table 1 Key cellular components of the three-step model of lymphovascular infiltration

[0044] Table 2 Key Cell Communication Axis

[0045] Table 3. Functional scoring modules related to lymphovascular infiltration (some representative modules)

[0046] Example 1: Construction of single-cell transcriptome atlases related to lymphovascular space infiltration.

[0047] This embodiment follows the appendix. Figure 1 The overall technical approach shown is to construct single-cell transcriptome atlases of endometrial cancer under different lymphovascular space invasion (LVSI) states.

[0048] Sample Collection and Grouping: Normal endometrial tissue, LVSI-negative early low-grade endometrial carcinoma tissue, and LVSI-positive early low-grade endometrial carcinoma tissue were collected. All samples were from patients undergoing initial surgical treatment who had not received radiotherapy, chemotherapy, or hormone therapy prior to surgery. Normal endometrial tissue was obtained from patients who underwent hysterectomy for uterine fibroids and whose postoperative pathology confirmed no endometrial lesions; tumor tissue was obtained from patients with FIGO 2009 stage IA, G1 to G2 grade endometrial carcinoma. LVSI status was independently interpreted by two senior pathologists based on HE staining, D2-40 / EPCAM, and CD31 / EPCAM immunostaining. Extensive LVSI was defined as ≥5 affected blood vessels. Based on the interpretation results, tumor samples were divided into LVSI-negative, focal LVSI, and extensive LVSI groups.

[0049] Single-cell suspension preparation and sequencing: Fresh tissue was collected, dissociated, red blood cells were removed, cell viability was assessed, and cell concentration was adjusted. A single-cell RNA sequencing library was constructed and high-throughput sequencing was performed. Sequencing data underwent Fastp preprocessing, GRCh38 reference genome alignment, Seurat quality control (removal of cells with <200, >7500 genes, or mitochondrial UMI ratio >15%), Harmony batch effect integration, dimensionality reduction, clustering, and UMAP visualization to obtain single-cell transcriptome atlases of endometrial and tumor tissues under different LVSI states.

[0050] Results: As attached Figure 1 As shown, the main cell types in this atlas include epithelial cells, T / NK cells, myeloid cells, fibroblasts and myocytes, endothelial cells, B cells, and mast cells, providing basic data for subsequent identification of malignant epithelial cells, analysis of tumor microenvironment subpopulations, and construction of cell communication networks.

[0051] Example 2: Identification of the initiation characteristics of malignant epithelial invasion associated with lymphovascular space infiltration.

[0052] In this embodiment, epithelial cells were extracted from single-cell transcriptome maps to identify LVSI-related malignant epithelial subpopulations and their invasion initiation characteristics.

[0053] Epithelial cell re-clustering and malignancy identification: Epithelial cells were extracted and re-clustered, and combined with normal epithelial cells as a reference, tumor-associated malignant epithelial cells were identified by inferring CNV copy number variation. Epithelial cells were divided into normal epithelial subsets NC1 and NC2 and tumor epithelial subsets TC1, TC2, TC3, TC4, TC5, and TC6.

[0054] Functional scoring and feature analysis: Differential expression analysis, cell entropy analysis, pseudo-temporal analysis and functional module scoring were performed on malignant epithelial cells to calculate the scores of gene sets related to partial epithelial-mesenchymal transition (p-EMT), epithelial-mesenchymal transition (EMT), hypoxia, glycolysis, inflammatory response (NF-κB), angiogenesis, immunosuppression and progesterone response.

[0055] Results: As attached Figure 2 As shown, the LVSI-associated malignant epithelial subset TC4 exhibits pro-invasive characteristics, including high expression of p-EMT, hypoxia, glycolysis, inflammation, and immunosuppression-related molecules (CD47, CD274, IDO1), and low expression of progesterone receptor co-genes. (See attached image.) Figure 2 A heatmap shows representative genes from each epithelial cell subset; (attached) Figure 2 BGO analysis showed that the TC4 subset was enriched in hypoxia, EMT, and immune escape pathways; Appendix Figure 2 C-bubble plots show high expression of genes related to p-EMT, hypoxia, glycolysis, immunosuppression, and progesterone response in the TC4 subset; (See attached image) Figure 2 The D-function score heatmap showed that the TC4 subset had the highest scores in p-EMT, hypoxia, glycolysis, and immunosuppression; (see attached image) Figure 2 The E-violin diagram shows that the p-EMT score of the TC4 subgroup is significantly higher than that of other subgroups; (attached) Figure 2 The F-correlation heatmap reveals the synergistic relationships among invasion-related functional modules. These results serve as the basis for constructing the "tumor epithelial invasion initiation" stage in the three-step metastasis-promoting ecosystem model.

[0056] Example 3: Identification of T / NK cell subsets associated with lymphovascular infiltration.

[0057] In this embodiment, T / NK cells were extracted from single-cell transcriptome maps to identify LVSI-related T cell remodeling features.

[0058] T / NK cell re-clustering and subpopulation annotation: T / NK cells were extracted and subjected to independent high-resolution re-clustering analysis. Subpopulation annotation was performed by combining differentially expressed genes, classic T / NK cell marker genes, and previous single-cell annotation systems. The characteristics of LVSI-related T / NK cell remodeling were determined through cell cycle scoring, exhaustion-related gene expression analysis, pseudo-temporal analysis, and neighborhood network analysis.

[0059] Results: As attached Figure 3 As shown, the enriched or expanded T cell subsets in LVSI-positive tumors include Cycling T cells, ZNF683+SOX4+CD8+ T cells, and TNFRSF9+Treg cells. (See attached image) Figure 3 A shows the UMAP clustering of T / NK cells; (attached) Figure 3 B shows the distribution of T / NK subsets under different LVSI states, indicating that the proportions of Cycling_T and ZNF683+SOX4+CD8+T are significantly increased in the LVSI-positive group; Appendix Figure 3 C. Distribution preference analysis based on odds ratios showed that the above subgroups were enriched in LVSI positive samples; Appendix Figure 3 D shows the expression of exhaustion-related genes in specific T cell subsets; Appendix Figure 3 The simulated time-series differentiation trajectory of E CD8+ T cells indicates that ZNF683+SOX4+CD8+ T cells are in the terminal differentiation stage; Appendix Figure 3 F shows enrichment of STAT3 / STAT5-related pathways in the ZNF683+SOX4+CD8+T and TNFRSF9+Treg subsets; Appendix Figure 3 G-neighborhood network analysis showed that Cycling_T and ZNF683+SOX4+CD8+T cell-related neighborhoods significantly expanded in the LVSI-positive group. These results serve as evidence for the immune cell composition in the "microenvironment synergistic remodeling" phase of the three-step metastasis-promoting ecosystem model.

[0060] Example 4: Identification of myeloid cell subsets associated with infiltration of the lymphovascular space.

[0061] In this embodiment, myeloid cells were extracted from single-cell transcriptome maps, and LVSI-associated monocyte-macrophage and dendritic cell subsets were identified.

[0062] Myeloid cell re-clustering and functional scoring: High-resolution re-clustering analysis was performed on myeloid cells to identify monocyte-macrophage and dendritic cell subsets by combining differentially expressed genes and classic myeloid cell marker genes. The AddModuleScore function was used to score the functional characteristics of tumor microenvironment cell subsets, including TAM score, MDSC score, M1 macrophage score, M2 macrophage score, phagocytosis score, activation score, and migration score. Simultaneously, miloR was used for differential abundance analysis, and GSEA, GSVA, GO, and KEGG were used to analyze the functional enrichment characteristics of different cell subsets.

[0063] Results (monocytes-macrophages): As attached. Figure 4 As shown, LVSI-related enriched subsets in monocytes and macrophages include SPP1+MMP9+Mac and APOE+CXCL9+Mac. (See attached image) Figure 4 A shows the UMAP distribution map of monocyte-macrophage subsets; (attached) Figure 4 B shows the changes in the proportions of each subgroup under different LVSI states; SPP1+MMP9+Mac was significantly enriched in the LVSI-positive group; Appendix Figure 4 The C-function scoring chart shows that SPP1+MMP9+Mac has high scores in matrix remodeling and angiogenesis; (attached) Figure 4 DGO functional enrichment analysis showed that SPP1+MMP9+Mac was enriched in collagen metabolism, fibroblast proliferation, and angiogenesis pathways. APOE+CXCL9+Mac exhibited high antigen presentation-related characteristics.

[0064] Results (dendritic cells): As attached Figure 5 As shown, cDC1_LAMP3_DC, cDC2_LAMP3_DC, and pDC are mainly distributed in LVSI-positive tumor samples. (See attached image) Figure 5 A shows the UMAP distribution map of dendritic cell subsets; (attached) Figure 5 B shows a significantly increased proportion of the LAMP3_DC-related subset in the LVSI-positive group; Appendix Figure 5 C shows high expression of immunosuppressive genes (such as IDO1, LGALS9, PD-L1) in LAMP3_DC; Appendix Figure 5 D-scans showed that LAMP3_DC exhibited high activation, high migration, and moderate cell cycle score; (See attached image) Figure 5 Differential enrichment pathway analysis showed that LAMP3_DC was enriched in pathways related to inflammation, antigen presentation, and immunosuppression. These results serve as evidence for myeloid cells in the "microenvironment synergistic remodeling" stage of the three-step metastasis-promoting ecosystem model.

[0065] Example 5: Identification of fibroblast subsets and transcriptional regulatory characteristics associated with infiltration into the lymphovascular space.

[0066] In this embodiment, fibroblasts were extracted from single-cell transcriptome maps, and LVSI-related fibroblast subsets and their transcription factor regulatory networks were identified.

[0067] Fibroblast Reclustering and Subpopulation Annotation: High-resolution reclustering analysis was performed on fibroblasts, and subpopulation annotation was conducted by combining differentially expressed genes, classic fibroblast marker genes, and functional enrichment results. The transcription factor regulatory network of fibroblast subpopulations was analyzed using SCENIC.

[0068] Results: As attached Figure 6 As shown, the fibroblast subsets enriched in LVSI-positive samples include iCAF, WNT5A_mCAF, LRRC75A_mCAF, and dCAF. (See attached image) Figure 6 A shows the UMAP distribution map of fibroblast subsets; (attached) Figure 6 BGO functional enrichment analysis showed that iCAF was associated with angiogenesis, leukocyte migration, and TGF-β receptor signaling; WNT5A_mCAF was associated with EMT, hypoxia, inflammatory response, TGF-β stimulation response, collagen metabolism, and angiogenesis; and LRRC75A_mCAF was associated with lymphocyte differentiation inhibition and immunosuppressive inflammatory programs. Figure 6 C shows the proportions of different fibroblast subsets under different LVSI states; WNT5A_mCAF and LRRC75A_mCAF were significantly enriched in the LVSI-positive group; Appendix Figure 6 D GSEA analysis showed that WNT5A_mCAF was enriched in EMT, hypoxia, and angiogenesis pathways compared to other fibroblast subsets; (See attached image) Figure 6 The E transcription factor activity heatmap revealed differences in the regulatory activities of key transcription factors (such as FOXO3, TWIST1, RUNX2, and NFKB1) among different fibroblast subsets. These results serve as the basis for matrix remodeling in the "microenvironment synergistic remodeling" stage of the three-step metastasis-promoting ecosystem model.

[0069] Example 6: Identification of endothelial cell subsets associated with infiltration into the lymphovascular space.

[0070] This embodiment extracts endothelial cells from single-cell transcriptome maps and identifies LVSI-related endothelial cell subsets and their functional characteristics.

[0071] Endothelial cell re-clustering and functional scoring: High-resolution re-clustering analysis was performed on extracted endothelial cells, and subpopulation annotation was performed by combining differentially expressed genes, classic endothelial cell marker genes, and functional scoring results. The AddModuleScore was used to assess functional modules such as angiogenesis, inflammatory response, EMT, and vascular permeability.

[0072] Results: As attached Figure 7 As shown, LVSI-related endothelial cell subsets include Tip_ECs-related subsets such as Tip_ArtECs, Tip_CapECs, and Tip_VenECs. (See attached image) Figure 7 A shows the UMAP distribution map of endothelial cell subsets; (attached) Figure 7 B shows the proportions of different endothelial subsets under different LVSI states; Tip_ECs were significantly enriched in the LVSI-positive group; Appendix Figure 7 C-level differential enrichment pathway analysis showed that Tip_ArtECs, Tip_CapECs, and Tip_VenECs were all enriched in pathways related to angiogenesis, endothelial migration, and basement membrane remodeling; (See attached...) Figure 7 DGSEA analysis showed that Tip_ECs were enriched in angiogenesis, EMT, and immune regulation pathways compared to other endothelial subsets; Appendix Figure 7 The E-function scoring chart showed that Tip_ECs had the highest scores in angiogenesis, EMT, and vascular permeability; (See attached image) Figure 7 The F-thermogram shows changes in angiogenesis, hypoxia, inflammatory response, EMT, and related signaling pathway scores under different LVSI states, with all of these scores significantly elevated in the LVSI-positive group. These results serve as the basis for endothelial cells in the "vascular invasion" stage of the three-step metastasis-promoting ecosystem model.

[0073] Example 7: Identification of multicellular communication networks associated with lymphovascular infiltration.

[0074] This embodiment constructs cell communication networks under different LVSI states based on single-cell expression matrices and cell subpopulation annotation information, and identifies key ligand-receptor interaction axes.

[0075] Cell communication analysis: Cell communication networks were constructed using the CellChat R package. CellChat objects were created using a normalized expression matrix and cell metadata, and overexpressed ligand-receptor interactions were identified based on the human CellChatDB database. Intercellular communication probabilities were calculated and filtered, pathway-level analysis and communication network aggregation were performed, and key sending and receiving cells were identified through centrality analysis. CellChat objects were created for both LVSI-positive and LVSI-negative groups, and the two groups were then merged and compared to identify altered cell interactions under different LVSI states.

[0076] Results: As attached Figure 8 As shown, candidate communication axes were obtained: SPP1-integrin axis, LGALS9-immune checkpoint axis, WNT5A-FZD / MCAM axis, and collagen-integrin axis. (See attached image) Figure 8A-string diagrams show the differences in interactions between major cell types of endometrial carcinoma in LVSI-negative (left) and LVSI-positive (right) endometrial-like uterine linings; the number and intensity of communications were significantly increased in the LVSI-positive group. (See attached diagram.) Figure 8 The lollipop plot illustrates the difference in interaction probabilities between LVSI-positive and LVSI-negative cell subsets, showing that communication among SPP1+MMP9+Mac, WNT5A_mCAF, and Tip_ECs is significantly enhanced in the LVSI-positive group; (See attached image) Figure 8 The heatmap shows the interaction probabilities between SPP1+MMP9+Mac and WNT5A_mCAF and endothelial cell subsets and T / NK cell subsets, respectively; (See attached image) Figure 8 Bubble diagrams show the SPP1-integrin and WNT5A-FZD / MCAM interactions between WNT5A_mCAF and SPP1+MMP9+Mac with LECs and Tip_ECs, respectively; (See attached diagram) Figure 8 E bubble plot shows the LGALS9-immune checkpoint interaction between WNT5A_mCAF and SPP1+MMP9+Mac cells and T / NK cells; (See attached image) Figure 8 The F-bubble plot shows the immune checkpoint interactions between LECs and Tip_ECs and T / NK cells. These results serve as evidence for the communication network of multicellular synergistic mechanisms in the three-step metastasis-promoting ecosystem model.

[0077] Example 8: Spatial verification using multiplex immunofluorescence.

[0078] In this embodiment, paraffin-embedded tissue samples of endometrial carcinoma with different LVSI states were selected for multiplex immunofluorescence spatial verification.

[0079] Sample processing and multiplex immunofluorescence staining: Spatial validation was performed on paraffin-embedded tissue samples of endometrial carcinoma from different LVSI states. After dewaxing, hydration, antigen retrieval, antibody incubation, signal amplification, nuclear staining, and whole-section scanning, tissue sections were subjected to multiplex immunofluorescence detection using three antibody panels. Panel 1: EPCAM, HIF1A, Ki67, CD44, to detect tumor epithelial cell proliferation and hypoxia status; Panel 2: CD68, PD-L1, EPCAM, CD3, detection of macrophage, PD-L1 expression, spatial relationship between tumor epithelium and T cells; Panel 3: Ki67, PD-1, CD8, CD4, to detect T cell proliferation and PD-1 expression.

[0080] Results: As attached Figure 9 As shown, attached Figure 9A shows the spatial distribution of invasive tumor epithelial cells in LVSI-negative, focal LVSI, and extensive LVSI stages: In LVSI-negative cases, the HIF1A-highly expressing tumor epithelial area is confined to the glandular lumen; in focal LVSI, the HIF1A-highly expressing epithelial cells extend towards the stromal invasion front; in extensive LVSI, a large number of HIF1A / Ki67 double-positive tumor cells are visible at the invasion front. (Attached) Figure 9 B shows the spatial distribution of the immune microenvironment: PD-L1-positive macrophages are localized in the epithelial stroma and glandular lumens when LVSI is negative, but extend to the peritumoral stroma when LVSI is positive. (See attached image) Figure 9 C shows the spatial distribution and phenotype of T cells: In LVSI-negative / focally positive samples, T cells are mainly distributed in the tumor stroma and have low PD-1 expression; in extensive LVSI samples, a large number of CD4+ / CD8+ T cells are aggregated around the invading tumor epithelium, with significantly upregulated PD-1 expression, and some cells simultaneously express Ki67 and PD-1. (See attached image) Figure 9 D-scans revealed the spatial co-localization of PD-L1-positive tumor cells, PD-L1-positive macrophages, and PD-1-positive T cells within vascular tumor thrombi in extensive LVSI samples. These spatial validation results confirmed the findings of single-cell transcriptomics and cell communication analyses, providing histological evidence for LVSI-related metastatic niches.

[0081] Example 9: Construction of a three-tiered lymphatic vascular interstitial infiltration-promoted metastasis ecosystem model.

[0082] This embodiment integrates the results of all the above embodiments to construct the LVSI three-tiered transfer-promoting ecosystem model.

[0083] Model Integration: The LVSI three-step metastasis-promoting ecosystem model was constructed by integrating the single-cell transcriptome atlas (Example 1), the malignant epithelial cell status analysis results (Example 2), the T / NK cell, myeloid cell, fibroblast and endothelial cell subset analysis and functional scoring results (Examples 3-6), the CellChat cell communication network (Example 7), and the multiplex immunofluorescence spatial validation results (Example 8).

[0084] Results: As attached Figure 10 As shown, the model consists of three consecutive stages: First stage (initiation phase of tumor epithelial invasion): As attached Figure 10 As shown in AB, the TC4 subset with p-EMT characteristics acquires initial invasive ability driven by the hypoxic microenvironment and begins to remodel the surrounding microenvironment by secreting chemokines and cytokines. This stage corresponds to LVSI-negative and early focal LVSI.

[0085] Second stage (microenvironmental collaborative reshaping stage): As attached Figure 10As shown in B, based on cell subsets such as Cycling_T, ZNF683+SOX4+CD8+T, SPP1+MMP9+Mac, APOE+CXCL9+Mac, LAMP3_DC, iCAF, WNT5A_mCAF, and LRRC75A_mCAF, and their immunosuppressive, inflammatory, matrix remodeling, and angiogenesis characteristics, a complex communication network is established with immune cells and endothelial cells through multiple signaling axes such as LGALS9-immune checkpoint, SPP1-integrin, and WNT5A-FZD / MCAM, gradually constructing an immunosuppressive and angiogenesis-promoting microenvironment ecosystem.

[0086] Third stage (vascular invasion stage): As attached Figure 10 As shown in Figure C, based on Tip-ECs-mediated increased vascular permeability, basement membrane remodeling, and immunosuppression, highly invasive tumor cells cross the endothelial barrier and enter the vascular system, completing LVSI. The spatial co-localization of PD-L1-positive tumor cells, PD-L1-positive macrophages, and PD-1-positive T cells within the vascular tumor thrombus suggests that the tumor thrombus is not only a passive invasion result but also a micrometastatic unit with active immune regulation characteristics.

[0087] Ultimately, a metastasis-promoting ecosystem model for LVSI was formed, consisting of "tumor epithelial invasion initiation - microenvironmental synergistic remodeling - vascular invasion," providing a technical foundation for early risk assessment, auxiliary judgment, development of potential molecular biomarkers, and screening of therapeutic targets for LVSI.

[0088] The above embodiments 1-9 fully describe the specific implementation of the present invention, and are combined with the appendix. Figure 1-10 The invention has been described in detail above. Those skilled in the art can reproduce the invention based on the above description without any inventive effort. The above embodiments describe in detail the implementation methods and verification results of the present invention. Those skilled in the art, based on the teachings of the above embodiments, can reproduce the present invention and obtain the corresponding technical effects without creative effort. The scope of protection of the present invention is defined by the claims; the embodiments are for explanation and illustration only.

Claims

1. A method for constructing a three-step metastatic ecological model of endometrial cancer lymphovascular invasion, characterized in that, Includes the following steps: a) Obtain normal endometrial tissue, early low-grade endometrial carcinoma tissue with negative lymphovascular space invasion, and early low-grade endometrial carcinoma tissue with positive lymphovascular space invasion, perform single-cell RNA sequencing, and construct a single-cell transcriptome atlas. b) Perform cell type annotation and differential abundance analysis on the single-cell transcriptome atlas to obtain the cell composition characteristics under different lymphovascular space infiltration states; c) Extract epithelial cells for re-clustering and copy number variation analysis to identify malignant epithelial cell subsets, and identify malignant epithelial cell subsets with invasive initiation characteristics in the lymphovascular space through functional module scoring. d) T / NK cells, myeloid cells, fibroblasts and endothelial cells were extracted and subjected to high-resolution re-clustering, functional module scoring and transcription factor regulatory network analysis to obtain the functional status and regulatory characteristics of each cell subpopulation during the infiltration of the lymphovascular space. e) Based on the annotation information of cell subpopulations under different lymphovascular space infiltration states, construct a cell communication network and identify differential ligand-receptor interaction axes. f) Select endometrial cancer tissue samples with different lymphovascular space infiltration states, perform multiple immunofluorescence staining and spatial imaging with multiple antibody combinations, and obtain spatial colocalization information of key cells and signaling molecules. g) Integrate the single-cell transcriptome atlas, malignant epithelial cell invasion characteristics, microenvironment cell subpopulation functional characteristics, differential ligand-receptor interaction axes, and spatial colocalization information obtained in steps a)-f) to construct a three-step metastasis-promoting ecosystem model of lymphovascular space infiltration consisting of three consecutive stages: "tumor epithelial invasion initiation - microenvironmental synergistic remodeling - vascular invasion".

2. The method according to claim 1, characterized in that, The positive lymphovascular space infiltration group described in step a) is further divided into a focal lymphovascular space infiltration group and a widespread lymphovascular space infiltration group based on the number of affected vessels as determined by HE staining and D2-40 / EPCAM and CD31 / EPCAM immunostaining. The widespread lymphovascular space infiltration is defined as having ≥5 affected vessels.

3. The method according to claim 1, characterized in that, The invasion initiation features described in step c) include: partial epithelial-mesenchymal transition, hypoxia, glycolysis, high expression of immunosuppression-related molecules, and low expression of progesterone receptor co-genes; the immunosuppression-related molecules are selected from one or more of CD47, CD274, and IDO1.

4. The method according to claim 1, characterized in that, The high-resolution re-clustering and functional module scoring mentioned in step d) include: i) Subgroups of T / NK cells were further divided into Cycling_T cells, ZNF683+SOX4+CD8+T cells and TNFRSF9+Treg cells, and their exhaustion and proliferative functions were evaluated. ii) Subgroups of myeloid cells were further subdivided into SPP1+MMP9+ macrophages, APOE+CXCL9+ macrophages and LAMP3+ dendritic cells, and their matrix remodeling, angiogenesis and immunosuppressive functions were evaluated. iii) Subgroups of fibroblasts were further subdivided into iCAF, WNT5A+ myofibroblasts and LRRC75A+ myofibroblasts, and their matrix remodeling, collagen metabolism and immunosuppressive inflammatory functions were evaluated. iv) Subgroup the endothelial cell population into Tip_ECs and assess their angiogenesis, endothelial migration, vascular permeability and basement membrane remodeling function.

5. The method according to claim 1, characterized in that, The differential ligand-receptor interaction axes mentioned in step e) include: the SPP1-integrin axis, the LGALS9-immune checkpoint axis, the WNT5A-FZD / MCAM axis, and the collagen-integrin axis; wherein the LGALS9-immune checkpoint axis involves P4HB, HAVCR2, CD45, and CD44 molecules.

6. The method according to claim 1, characterized in that, The multiple antibody combinations mentioned in step f) include: Group 1: Ki67, HIF1A, EPCAM, and CD44, used to validate the spatial distribution of markers of invasive tumor epithelial cells; The second group consisted of CD68, PD-L1, EPCAM, and CD3, used to verify the spatial relationships among macrophages, PD-L1 expression, tumor epithelial cells, and T cells. The third group consists of Ki67, PD-1, CD8, and CD4, used to verify the spatial distribution of PD-1 positive T cells and proliferation-related T cells.

7. The method according to claim 1, characterized in that, The three-step ladder model mentioned in step g) is specifically as follows: The first step, the tumor epithelial invasion initiation stage: driven by the TC4 subset with some epithelial-mesenchymal transition characteristics in a hypoxic microenvironment, it acquires initial invasive ability; The second stage, the microenvironment synergistic remodeling stage: Cycling_T, ZNF683+SOX4+CD8+T, SPP1+MMP9+Mac, WNT5A_mCAF and LAMP3_DC cell subsets, through the LGALS9-immune checkpoint, SPP1-integrin and WNT5A-FZD / MCAM signaling axis, form an immunosuppressive and angiogenic microenvironment; The third stage, the vascular invasion stage: Based on the increased vascular permeability mediated by Tip_ECs and immunosuppression, invasive tumor cells cross the endothelial barrier and enter the vascular system.

8. The method according to any one of claims 1-7, characterized in that, It also includes step h): based on the constructed three-step metastasis-promoting ecosystem model, screening or evaluating a combination of molecular biomarkers for early risk prediction, molecular subtyping, auxiliary diagnosis, or therapeutic targets of endometrial cancer lymphovascular space invasion; the combination of molecular biomarkers includes a combination of nine prognostic genes: HMGA1, MGST1, NHP2, PRDX6, RBIS, TCEA1, TPD52, TUFT1, and YBX1.

9. A three-step ecological model for promoting metastasis in endometrial cancer by infiltration into the lymphovascular space, characterized in that, It is constructed by the method described in any one of claims 1-7.

10. A system for constructing a three-step metastatic ecological model of endometrial cancer lymphovascular invasion, characterized in that, include: A single-cell sequencing module for performing step a) as described in claim 1; The data processing and analysis module is used to perform steps b) to e) and g) as described in claim 1; The spatial verification module is used to perform step f) as described in claim 1. The data processing and analysis module includes a cell subpopulation analysis unit, a functional scoring unit, a cell communication inference unit, and a model integration and construction unit.

Images