A lung adenocarcinoma prognosis molecular screening method based on single cell spatial transcriptome

Abstract

The application discloses a lung adenocarcinoma prognosis molecular screening method based on single cell spatial transcriptome, which integrates and analyzes single cell sequencing and high-throughput sequencing data of lung adenocarcinoma histological modes lepidic and solid by a bioinformatics method, obtains prognosis molecules NECTIN2 and PVR, and verifies the molecules by using multiplex immunofluorescence experiments and spatial transcriptome data. Specifically, the single cell sequencing data of patients in the two histological modes are subjected to cytoplasm control, standardization, dimension reduction, clustering, cell type annotation; cell communication analysis is performed on tumor cells and immune cells, and significant co-inhibitory receptor ligand pairs TIGIT_NECTIN2 and TIGIT_PVR are obtained; the prognosis of tumor cell related molecules NECTIN2 and PVR is analyzed by using high-throughput sequencing data; and the expression difference and spatial distribution of the above molecules in different modes are verified by multiplex immunofluorescence experiments and spatial transcriptome data of patients in the two histological modes.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of biomedical information technology, in particular to a lung adenocarcinoma prognosis molecular screening method based on single-cell spatial transcriptome. BACKGROUND

[0002] Lung cancer has the highest incidence of all malignant tumors, causing more than 100,000 deaths each year. Among them, lung adenocarcinoma is the most common histological subtype, accounting for 40% of lung cancer. In lung adenocarcinoma, the development and prognosis of the disease are related to the appearance of morphologically different tumor regions, known as histological patterns. Common histological patterns of lung adenocarcinoma include lepidic, papillary, acinar, and solid. As the histological pattern progresses, the invasiveness of the tumor becomes stronger and stronger, and the heterogeneity becomes higher and higher. Lung adenocarcinoma has a complex tumor microenvironment, containing infiltrating immune cells, stromal cells, chemokines, and extracellular matrix components. Tumor cells dynamically interact with the peripheral environment, forming a hypoxic, pH-lowering, inflammation-increasing, and immunosuppressive environment, promoting the progression of lung adenocarcinoma and affecting the patient's response to drug therapy.

[0003] Traditional high-throughput bulk RNA-seq obtains the average level of cell mixtures, and the signal is usually dominated by cell clones with growth advantages, masking the differences in cell type-specific expression patterns, which hinders the exploration of deep molecular mechanisms of tumors and intratumoral heterogeneity. Single-cell RNA sequencing (scRNA-seq) not only provides a comprehensive and unbiased picture of the diversity of cells within lung adenocarcinoma tumor tissue, but also allows analysis of gene expression in individual cells, providing a clear understanding of intratumoral and intertumoral heterogeneity mechanisms and ligand-receptor signaling cell-cell interactions. This provides insights into the key drivers of tumor progression and biological mechanisms, and expands the thinking for finding effective treatment targets for patients with different patterns of lung adenocarcinoma. SUMMARY

[0004] The present application focuses on the histological patterns of lung adenocarcinoma, and the main technical contribution of the present application is: based on single-cell transcriptome, genes related to the prognosis of patients with two histological patterns are obtained, and the expression of prognostic molecules and spatial localization are verified through high-throughput sequencing, multiplex immunofluorescence experiments, and spatial transcriptome data.

[0005] The present application specifically provides:

[0006] The lung adenocarcinoma prognosis molecular screening method based on single-cell spatial transcriptome is revealed by integrating single-cell and high-throughput transcriptome through bioinformatics methods, and the steps include:

[0007] (1) Cell quality control, standardization, dimension reduction, clustering, cell type annotation on single-cell sequencing data of two histological patterns of lung adenocarcinoma patients;

[0008] (2) Cell communication analysis on tumor cells and immune cells, obtaining significant co-inhibitory receptor-ligand pairs: TIGIT_NECTIN2 and TIGIT_PVR;

[0009] (3) Analysis of the prognostic situation of tumor cell-related molecules NECTIN2 and PVR using high-throughput sequencing data;

[0010] (4) Verification of the expression differences and spatial distribution of the above molecules in different patterns through multiple immunofluorescence experiments and spatial transcriptome data of two histological pattern patients.

[0011] In step (1):

[0012] The two histological patterns of lung adenocarcinoma include: lepidic (squamous) and solid (solid);

[0013] Cell quality control is to filter cells with less than 200 genes, less than 10 cells expressing genes, and mitochondrial gene content higher than 30%; The standardization of the single-cell data is standardized by the LogNormalize method in the Seurat package; The dimension reduction is to select the top 2000 highly variable genes using the FindVariableFeatures function in the Seurat package, and reduce the dimension of the highly variable genes to several principal components through principal component analysis; The clustering is to cluster cells according to 30 principal components, and the resolution is controlled at 0.8 by selecting the FindNeighbors and FindClusters functions in the Seurat package; The cell type annotation is to calculate the differential marker genes of each cell cluster using the FindAllMarkers function in the Seurat package to perform cell annotation.

[0014] In step (2):

[0015] The cell communication is analyzed using the Python package CellPhoneDB to analyze the interaction between different cell types. For the genes expressed by the cell population, the percentage of cells expressing the gene and the average gene expression are calculated, and if the gene is expressed in 10% or less of the cells in the population, it is removed; The interaction between cells and cells is inferred by 1000 permutation tests on gene expression levels, and then an adjacency matrix is generated for all cell-cell interactions; The relative expression level of the co-inhibitory ligand or receptor is represented by z-score, and ggplot2 package is used for visualization.

[0016] In step (3):

[0017] The high-throughput sequencing data analysis is survival analysis using gene expression and survival data of lung adenocarcinoma patients in Cancer and Tumor Gene Atlas (TCGA);

[0018] The survival analysis is statistical analysis of survival data of patients by dividing the subjects into high and low groups according to the median using the survival package, drawing a Kaplan-Meier survival curve using the surivminer package, and calculating the P value using the Logrank test.

[0019] In step (4):

[0020] The multiplex immunofluorescence experiment and spatial transcriptome data verification are fluorescence staining and spatial transcriptome sequencing of pathological tissues of patients with two histological patterns of lung adenocarcinoma collected by clinical surgery, and calculation of the expression difference and spatial distribution of tumor-related molecules NECTIN2 and PVR in different histological patterns;

[0021] The method for calculating the expression difference of tumor-related molecules NECTIN2 and PVR in different histological patterns is to process the histological images and FASTQ files using the 10x SpaceRanger (https: / / www.10xgenomics.com / support) pipeline, and to align the sequencing data with the human reference genome (Ensemble Genome GRCh38). The generated output results are imported into Seurat for data preprocessing, including normalization, dimension reduction, clustering and data integration;

[0022] The normalization is performed using the SCTransform method in the Seurat package; the dimension reduction is performed using the RunPCA function in the Seurat package for principal component analysis; the clustering is performed according to 30 principal components, and the resolution is controlled at 0.8 by selecting the FindNeighbors and FindClusters functions in the Seurat package; and the data integration is performed using the FindTransferAnchors function in Seurat.

[0023] The lung adenocarcinoma prognosis molecular screening method based on single-cell spatial transcriptome constructed by the present application can obtain gene expression information and cell spatial position information at the single-cell level at the same time compared with the average level of traditional research on cell mixtures of tumor and cancer-adjacent tissues, thereby revealing the correlation and functional characteristics between prognosis-related molecules and tissue structures. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1: Schematic diagram of the flow of the lung adenocarcinoma prognosis molecular screening method based on single cell spatial transcriptome.

[0025] Figure 2 : Schematic diagram of the microenvironment cell composition of the single cell transcriptome of the lung adenocarcinoma patient of two histological patterns of the embodiment.

[0026] Figure 3a : Schematic diagram of the cell communication analysis between tumor cells and immune cells of the lung adenocarcinoma patient of the lepidic pattern of the embodiment.

[0027] Figure 3b : Schematic diagram of the cell communication analysis between tumor cells and immune cells of the lung adenocarcinoma patient of the solid pattern of the embodiment.

[0028] Figure 4a : Schematic diagram of the prognosis of the prognosis molecule NECTIN2 in the TCGA lung adenocarcinoma patient of the embodiment.

[0029] Figure 4b : Schematic diagram of the prognosis of the prognosis molecule PVR in the TCGA lung adenocarcinoma patient of the embodiment.

[0030] Figure 5a : Schematic diagram of the expression of the prognosis molecule NECTIN2 in the spatial transcriptome sample of the lung adenocarcinoma patient of the lepidic pattern of the embodiment.

[0031] Figure 5b : Schematic diagram of the expression of the prognosis molecule NECTIN2 in the spatial transcriptome sample of the lung adenocarcinoma patient of the solid pattern of the embodiment.

[0032] Figure 5c : Schematic diagram of the expression of the prognosis molecule PVR in the spatial transcriptome sample of the lung adenocarcinoma patient of the lepidic pattern of the embodiment.

[0033] Figure 5d : Schematic diagram of the expression of the prognosis molecule PVR in the spatial transcriptome sample of the lung adenocarcinoma patient of the solid pattern of the embodiment. DETAILED DESCRIPTION

[0034] The present application provides a lung adenocarcinoma prognosis molecular screening method based on single cell spatial transcriptome. The present application integrates and analyzes single cell sequencing and high-throughput sequencing data of lung adenocarcinoma patients of two histological patterns (lepidic and solid) by bioinformatics methods, obtains prognosis molecules NECTIN2 and PVR, and verifies them by using multiplex immunofluorescence experiments and spatial transcriptome data.

[0035] Reference Figure 1 , the steps of the present application include:

[0036] 1) Single-cell sequencing data of two histological pattern patients are sequentially subjected to cell quality control, standardization, dimensionality reduction, clustering and cell type annotation;

[0037] 2) Cell communication analysis is performed on tumor cells and immune cells, and significant co-inhibitory receptor ligand pairs TIGIT_NECTIN2 and TIGIT_PVR are obtained;

[0038] 3) High-throughput sequencing data is used to analyze the prognosis of tumor cell-related molecules NECTIN2 and PVR;

[0039] 4) The expression difference and spatial distribution of the above molecules in different patterns are verified by multiplex immunofluorescence experiments and spatial transcriptome data of two histological pattern patients.

[0040] Next, the technical solutions of the present application are further described in conjunction with the examples.

[0041] The lung adenocarcinoma prognostic molecular screening method based on single-cell spatial transcriptome of the present embodiment comprises the following steps:

[0042] (1) The source of the two histological pattern data involved in the present embodiment is two histological patterns, i.e. lepidic (scaly) and solid (solid)

[0043] Obtain single-cell transcriptome data of 3 lung adenocarcinoma patients with two histological patterns from https: / / codeocean.com / capsule / 8321305 / tree / v1;

[0044] Download 539 lung adenocarcinoma gene transcriptome expression data from TCGA database (https: / / portal.gdc.cancer.gov / );

[0045] Obtain pathological samples required for multiplex immunofluorescence experiments and pathological samples required for spatial transcriptome experiments of 4 lung adenocarcinoma patients with two histological patterns from clinical surgical operations.

[0046] (2) Sample processing of multiplex immunofluorescence experiment

[0047] For the pathological samples obtained in step (1) for multiplex immunofluorescence experiments, further, the localization of TIGIT, NECTIN2 and PVR molecules in the pathological tissues of two histological pattern patients is detected, comprising the following steps:

[0048] First, the paraffin sections of the pathological tissues of the lung adenocarcinoma patients with two histological patterns are subjected to paraffin removal and hydration treatment;

[0049] Then, antigen retrieval was performed; primary antibody mix was added at a ratio of 1:1000: TIGIT (Abeam, Ab243903) + NECTIN2 (Abeam, Ab269721) or TIGIT (Abeam, Ab243903) + PVR (Ab307687), placed in a humidified chamber and incubated overnight at 4°C;

[0050] Then, two fluorescent secondary antibodies were added: Proteintech, SA00003-2 + CoraLite594-conjugated Goat Anti-Mouse IgG (H+L), Proteintech, SA00013-3 and Fluorescein (FITC)-conjugated Affinipure Goat Anti-Rabbit IgG (H+L), Proteintech, SA00003-2 + Rhodamine (TRITC)-conjugated Goat Anti-Rat IgG (H+L), Proteintech, SA00007-7, placed in a humidified chamber at room temperature, incubated in the dark for 2 hours;

[0051] Then, counterstaining was performed with DAPI and incubated in the dark for 15 minutes, and finally mounted with an antifluorescence quencher and stored at 37°C.

[0052] (3) Spatial transcriptome sample processing

[0053] For the pathological samples obtained in step (1) for spatial transcriptome experiments, further, the two histological patterns of pathological tissue were fixed with formalin and subsequently encapsulated in paraffin-embedded tissue blocks;

[0054] Then, the samples were sectioned and stained with hematoxylin and eosin (H&E) for imaging using an Aperio GT450 scanner at a resolution of 40x (equivalent to 0.25 microns / pixel);

[0055] After the tissue was unmounted, the transcriptome probes were transferred from the original glass slides to the capture area on the Visium slide with dimensions of 11 mm x 11 mm using the Visium CytAssist device;

[0056] After mRNA penetration, a comprehensive transcriptome analysis was achieved by poly(A) capture and probe hybridization; the resulting libraries were sequenced using the Illumina Novaseq 6000, using paired-end sequencing with a read length of 150 base pairs.

[0057] (4) Single-cell transcriptome sequencing data quality control and cell filtering

[0058] Since low-quality cells, such as low-expression cells, contaminated cells, double cells, etc., will have uncontrollable effects on the experiment, it is necessary to filter low-quality cells according to the number of genes, mitochondrial proportion, etc. Specifically, filter cells with less than 200 genes, less than 10 genes expressed in cells, and mitochondrial gene content higher than 30%, to reduce the adverse effects of low-quality cells on the experiment.

[0059] (5) Single-cell transcriptome sequencing data standardization, dimensionality reduction, clustering

[0060] a. The gene expression level in single-cell transcriptome sequencing data usually has large variability and dynamic range. In order to eliminate this variability and make the comparison between different samples more reliable, it is necessary to standardize the data. Specifically, standardization is performed by the LogNormalize method in the Seurat package, that is, the expression value of each gene is divided by the expression value of all genes in the whole cell, multiplied by 10000, and logarithmically transformed;

[0061] b. Single-cell transcriptome sequencing data usually has high dimensionality, that is, each cell may express thousands of genes. In order to reduce the dimensionality and visualize the data, dimensionality reduction is needed. The dimensionality reduction method can map high-dimensional data to a lower-dimensional space, so as to better observe the similarity and difference between cells. Specifically, the first 2000 highly variable genes are selected using the FindVariableFeatures function in the Seurat package, and principal component analysis is performed by the RunPCA function in the Seurat package to reduce the dimensionality of highly variable genes to several principal components;

[0062] c. Clustering is the process of grouping similar cells into the same category or cluster. Through clustering analysis, cell clusters with similar gene expression patterns can be identified, revealing the existence of different cell types, subtypes or states. Clustering can help identify cell types, discover cell subpopulations, and explore the transcriptional dynamics during cell development. Specifically, the first 30 principal components (dims = 1:30) obtained by dimensionality reduction are clustered using the FindNeighbors and FindClusters functions in the Seurat package, with a resolution of 0.8 (resolution = 0.8).

[0063] (6) Single-cell transcriptome sequencing cell type annotation

[0064] Use the FindAllMarkers function in the Seurat package to calculate the differential marker genes of each cell cluster to perform cell annotation and show the microenvironment cell composition of the single-cell transcriptional atlas of patients with two histological patterns of lung adenocarcinoma, referenceFigure 2 Specifically,

[0065] The parameters are set as logfc.threshold=0.25, min.pct=0.1, i.e. filter out genes with differential expression less than 0.25 and similarity less than 0.1. The final obtained marker genes for each cell type are:

[0066] CD8 T (CD8A, CD8B), CD4 T (IL7R), tumor cell (EPCAM, KRT19), regulatory T cell (FOXP3, IL2RA), macrophage (CD68, C1QA), plasma cell (IGHG1, JCHAIN), B cell (MS4A1, CD19), NK cell (NKG7, GNLY), monocyte (S100A8, S100A9), fibrocyte (LUM, PDGFRA), mast cell (TPSAB1, TPSB2), endothelial cell (VWF, CLDN5), plasmacytoid dendritic cell (LILRA4, IL3RA).

[0067] (7) Cell communication analysis

[0068] The gene expression matrix of each cell in the single-cell transcriptome data is extracted, and the CellPhoneDB algorithm of the Python package is used to analyze the interaction between tumor cells and immune cells, referring to Figure 3a 、 Figure 3b Specifically,

[0069] First, for each cell in each cell cluster, the interaction probability between it and other cells is calculated, and this probability is calculated by comparing the expression levels of paired receptor and ligand genes in the cell;

[0070] Next, statistical significance analysis is performed to screen significant receptor-ligand pairs with P<0.05 to determine which paired receptor-ligand interactions are significant in the cell cluster. This analysis is based on information in the paired receptor-ligand database and a statistical model of single-cell data.

[0071] (8) Survival analysis

[0072] The RNA-seq data of the TCGA-LUAD (lung adenocarcinoma) project STAR process and the clinical data (survival status and overall survival OS) were downloaded and sorted from the TCGA database (https: / / portal.gdc.cancer.gov / );

[0073] Proportional hazards assumption test was performed using survival package and survival regression was fitted, P value was calculated using Logrank test, results were visualized using survminer package and ggplot2 package, refer to Figure 4a , Figure 4b .

[0074] (9) Spatial transcriptome data processing and calculation of molecular expression

[0075] The SCTransform algorithm in Seurat package was used to normalize the gene expression matrix of spatial transcriptome data. Principal component analysis was performed using the RunPCA function, and the first 30 principal components were selected for cell clustering with a resolution of 0.8. The FindTransferAnchors function in the Seurat package was used for single-cell joint spatial transcriptome analysis of the two histological patterns of lung adenocarcinoma. Specifically, the following steps were included:

[0076] First, the gene expression matrix and corresponding cell type labels in the single-cell transcriptome data were extracted as feature representations;

[0077] Then, the feature representations of the single-cell transcriptome data were mapped to the expression image of the spatial transcriptome data. This process can be seen as mapping low-dimensional features back to high-dimensional space to reconstruct the expression image of the spatial transcriptome data;

[0078] Finally, the probability of all cell types in each spot on the image was obtained, and the cell type with the highest probability was selected as the annotation result of the spot.

[0079] The expression visualization of the two molecules NECTIN2 and PVR in the spatial transcriptome image was completed by the SpatialFeaturePlot function in the Seurat package, refer to Figure 5a- Figure 5d .

Claims

1. A lung adenocarcinoma prognosis molecular screening method based on single cell spatial transcriptome, characterized by the steps of The method comprises the following steps: 1) cell quality control is performed on single-cell transcriptome sequencing data of patients with two histological patterns of lung adenocarcinoma, and then the screened single-cell transcriptome sequencing data is standardized, dimensionally reduced and clustered; finally, cell type annotation is performed; 2) cell communication analysis is performed on tumor cells and immune cells to obtain significant co-inhibitory receptor-ligand pairs: TIGIT_NECTIN2 and TIGIT_PVR; 3) high-throughput sequencing data analysis is used to analyze the prognosis of tumor cell-related molecules NECTIN2 and PVR; 4) the expression difference and spatial distribution of the above molecules NECTIN2 and PVR in different patterns are verified through multiple immunofluorescence experiments and spatial transcriptome data of patients with two histological patterns of lung adenocarcinoma; In step 1), the two histological patterns of lung adenocarcinoma include: lepidic and solid; 1.1) cell quality control is performed according to the number of genes, mitochondrial proportion to filter low-quality cells; 1.2) standardization is to divide the expression value of each gene by the expression value of all genes in the whole cell, multiply by 10,000, and take the logarithmic transformation; 1.3) dimension reduction is to map high-dimensional data to a lower-dimensional space; 1.4) clustering is to group similar cells into the same category or cluster; 1.5) cell type annotation is to annotate cells by calculating the differential marker genes of each cell cluster; In step 2), the gene expression matrix of each cell in the single-cell transcriptome sequencing data is extracted, and the interaction between tumor cells and immune cells is analyzed, specifically: 2.1) for each cell in the cell cluster, the interaction probability between the cell and other cells is calculated; this probability is calculated by comparing the expression levels of paired receptor and ligand genes in the cell; 2.2) statistical significance analysis is performed to screen significant receptor-ligand pairs and determine which paired receptor-ligand interactions are significant in the cell cluster; In step 3), high-throughput sequencing data analysis is performed using gene expression and survival data of TCGA lung adenocarcinoma patients in cancer and tumor gene atlas; the steps of survival analysis include: 3.1) the research object is divided into high and low two groups according to the median, and the survival data of the patients is statistically analyzed; 3.2) Kaplan-Meier survival curve is drawn; 3.3) Logrank test is used to calculate P value; In step 4), the multiple immunofluorescence experiments and spatial transcriptome data verification are performed by collecting pathological tissues of patients with two histological patterns of lung adenocarcinoma through clinical surgery to perform fluorescence staining and spatial transcriptome sequencing, and calculating the expression difference and spatial distribution of tumor-related molecules NECTIN2 and PVR in different histological patterns.

2. The lung adenocarcinoma prognosis molecular screening method based on single-cell spatial transcriptome according to claim 1, wherein 1.1) cell quality control is to filter cells with less than 200 genes, genes expressed in less than 10 cells, and mitochondrial gene content higher than 30%; 1.2) Standardization is performed by the LogNormalize method in the Seurat package; 1.3) Dimension reduction is performed by selecting the top 2000 highly variable genes using the FindVariableFeatures function in the Seurat package, and reducing the dimension of the highly variable genes to several principal components by principal component analysis; 1.4) Clustering is performed by clustering cells according to 30 principal components, by selecting the FindNeighbors and FindClusters functions in the Seurat package, and controlling the resolution at 0.8; 1.5) Cell type annotation is performed by calculating the differential marker genes of each cell cluster using the FindAllMarkers function in the Seurat package.

3. The single-cell spatial transcriptomic-based lung adenocarcinoma prognostic molecular screening method according to claim 2, characterized in that, In step 2), cell communication analysis between tumor cells and immune cells is performed using the Python package CellPhoneDB to analyze the interaction between cells; for the genes expressed by the cell population, the percentage of cells expressing the gene and the average gene expression value are calculated, and if the gene is expressed in 10% or less of the cells in the population, it is removed; the interaction between cells is inferred by 1000 permutation tests on gene expression levels, and then an adjacency matrix is generated for all cell-cell interactions; the relative expression level of co-inhibitory ligands or receptors is represented by z-score; visualization is performed using the ggplot2 package.

4. The single-cell spatial transcriptomic-based prognostic molecular screening method for lung adenocarcinoma according to claim 3, characterized in that, In step 4), the method for calculating the expression difference of tumor-related molecules NECTIN2 and PVR in different histological patterns is to use the 10x SpaceRanger pipeline to process histological images and FASTQ files, and align the sequencing data with the human reference genome Ensemble Genome GRCh38, and the generated output results are imported into the Seurat package for data preprocessing.

5. The single-cell spatial transcriptomic-based prognostic molecular screening method for lung adenocarcinoma according to claim 2, characterized in that, The cell annotation is as follows: the parameter settings are logfc.threshold=0.25, min.pct=0.1, i.e. filtering out genes with a differential expression level less than 0.25 and a similarity less than 0.1; the final marker genes for each cell type are: CD8T (CD8A, CD8B), CD4T (IL7R), tumor cells (EPCAM, KRT19), regulatory T cells (FOXP3, IL2RA), macrophages (CD68, C1QA), plasma cells (IGHG1, JCHAIN), B cells (MS4A1, CD19), NK cells (NKG7, GNLY), monocytes (S100A8, S100A9), fibrocytes (LUM, PDGFRA), mast cells (TPSAB1, TPSB2), endothelial cells (VWF, CLDN5), and plasmacytoid dendritic cells (LILRA4, IL3RA).

6. The single-cell spatial transcriptomic-based prognostic molecular screening method for lung adenocarcinoma according to claim 4, characterized in that, Data preprocessing is performed in the Seurat package, including normalization, dimension reduction, clustering, and data integration; Normalization is normalized using the SCTransform method in the Seurat package; Dimensionality reduction is principal component analysis using the RunPCA function in the Seurat package; clustering is cell clustering according to 30 principal components, and the method selects the FindNeighbors and FindClusters functions in the Seurat package, with a resolution of 0.8; data integration is integrated using the FindTransferAnchors function in Seurat.

Images