| Traumatic spinal cord injury(SCI)is a life-shattering neurological disorder,and is increasingly recognized as global health priorities due to the preventability of most injuries,the expensive therapy and medical support care and the economic consequences deriving from loss in productivity.Traffic accidents,falls,and conflict violence are among the most common causes of injury-related SCI across the world.Mechanical damage to the spinal cord causes a series of inflammatory and pathological responses,leads to locomotor and autonomic dysfunctions,and even results in the dysregulation of multiple organ systems throughout the body.Often leading to varying degrees of disability,especially for young people,with a significant impact on their healthy life year.However,damage is often accompanied by the process of tissue self-repair,which can reduce the scope of damage and restore the damaged microenvironment.After spinal cord injury,there are often inflammatory reactions,reactive glial hyperplasia,and scar formation.However,compared with other diseases of the central nervous system,the repair efficiency of spinal cord injury is low,and the pathological changes caused by injury are difficult to reverse.Although varying degrees of therapeutic strategies have been tested,including experimental therapeutics to reduce secondary neuronal damage and enhance tissue repair and neuroplasticity,there are still no curative or regenerative treatments available to spinal injured individuals.At present,many studies have analyzed the changes of cell types and gene expression levels after spinal cord injury,and found that different cell subtypes(such as A1 reactive astrocytes and A2 reactive astrocytes)caused by injury play different functional roles in the repair process of spinal cord injury.Most of these studies,however,were based on bulk RNA sequencing and single cell sequencing(sc RNAseq),and due to the necessity of cellular dissociation prior to sequencing of individual tissues/cells,the spatial context for each sample is lost,thus limiting insight into the precise spatial organization of the cellular and molecular heterogeneity.However,this spatial information is crucial for understanding the structural heterogeneity of tissue,cell composition and the interaction between adjacent cells in tissue.Therefore,spatial transcriptomics came into being.Through combining microimaging and sequencing technology,spatial transcriptomics can obtain gene expression data while preserving the spatial location information of samples to the greatest extent,and this technology was rated as "technology of the year" by Nature Methods in 2020.Spatial transcriptomics has been applied to study the homeostasis/heterogeneity of tissue architectures in the brain,lung,breast,heart,liver,intestine,kidney,gastric,prostate,uterus,bladder,embryo,etc.These studies have broadened our understanding of tissue organization at unprecedented molecular resolution in biomedical research,especially in developmental biology,regenerative medicine,disease/tumor microenvironment.In this study,we explored the spatial heterogeneity of the wound architecture of injured spinal cord using spatial transcriptomics profiles at distinct time points(3,7,14 day post injury and 1,2 month post injury),aiming to systematically describe the dynamic characteristics of scar formation and the spatiotemporal atlas of cell heterogeneity after spinal cord injury.Specifically,this thesis focuses on the following four issues:(1)At present,spatial transcriptome studies are independent of each other,and the analysis methods used in each study are quite different,leading to poor comparability between the results of each study.Therefore,it is necessary to establish a unified spatial transcriptome databae to uniformly sort out and analyze these data,so that research results can be compared between different tissues and organs,and also within the same tissues and organs,so as to fully understand the spatial heterogeneity of specific tissues and organs.The first issue to be studied in this thesis is to establish a unified platform that includes storage,organization,analysis,comparison,and display.(2)For spatial transcriptomics,the identification of spatially variable genes is crucial to the study of spatial heterogeneity of tissue structure.However,the current methods for the identification of spatially variable genes only use the spatial location information and its corresponding gene expression information,ignoring the spatial heterogeneity information provided by tissue staining images.The second issue to be studied in this thesis is to establish a recognition method for spatially differentially expressed genes that integrates spatial heterogeneity information provided by H&E staining images.(3)At present,the spatial transcriptome sequencing data provided by the 10 x Visiim platform,each spatial location can contain 3-10 cells,involving multiple cell types.Simply defining it as a single cell type will affect the accuracy of understanding the spatial change characteristics of different types of cells after injury,thus further affecting the understanding of the relevant biological mechanisms after injury.Therefore,it is necessary to extract features from the gene expression information of all spatial locations in order to identify different biological functions,explore the underlying biological mechanisms,and provide clues for clinical treatment.However,there is currently no systematic study to elucidate the internal heterogeneity characteristics of spatial location points after spinal cord injury,which is the third issue to be studied in this thesis.(4)After traumatic spinal cord injury,the interaction between cells is crucial to the study of the microenvironment after injury.At present,the research on cell to cell communication after spinal cord injury is based on all cells in the entire tissue sample,and it is impossible to distinguish whether the cells come from the injury site or surrounding healthy tissues.Therefore,the accuracy of the results needs further verification.The spatial location information provided by spatial transcriptome technology makes it possible to conduct targeted and accurate research on the cell communication at the injury center,as well as the cell communication between the junction of damaged tissue and adjacent normal tissue.This precise targeting research is critical to the spread of damaged areas or the mechanism after tissue repair.The fourth issue to be studied in this thesis is to accurately target the communication between cells in specific spatial regions.In response to the above issues,the main work of the dissertation are as follows:(1)Construction of spatial transcriptomics database at single-cell resolution.To further understand the application of spatial transcriptomics in various tissues and organ systems,we searched the published literature on spatial transcriptomics from Pub Med and released studies from bio Rxiv.Finally,we obtained data from 43 studies,including 1082 sub-datasets,across 16 organ types and four species.After processing all the data,we provide systematic annotations of spatial transcriptomics,including(i)spatially patterned genes,(ii)spatially patterned pathways,(iii)gene regulatory networks,(iv)cell–cell interactions and(v)spatial transcriptomics deconvolution and interactions.In total,we detected 12 116,16 530,1476 and 4915 unique spatial pattern genes for human,mouse,chicken,and zebrafish,respectively.The number of pathways identified from human,mouse,chicken,and zebrafish was 7560,7650,7288 and 294,respectively.697 key regulons were identified.All the results were sorted and stored in the My SQL database,and provided to researchers in the form of web pages for retrieval by single gene,related article,tissue types and functional analysis.Users also can download all the results for further analysis.(2)Detect spatially patterned genes using mixed kernel functions in injured spinal cord tissue.Spatial pattern genes refer to genes with special distribution characteristics in space or highly variable genes in local regions.These genes with high spatial heterogeneity often reflect certain spatial structure characteristics,which can help to study the potential physiological functions of tissues.These spatially patterned genes can be used for downstream analysis,which is crucial to study the potential biological mechanism of pathological tissues.Gaussian kernel functions from different dimensions were proposed to identify genes with spatial pattern by combining spatial location information and tissue slice image information.Firstly,the appearance kernel function was constructed according to the spatial location information to describe the global spatial features,and then for the image information,the convolution neural network was used to extract the features of each spatial spot,and then the extracted features are constructed into the feature kernel function to describe the local spatial features.By combining the global and local spatial features,the spatial differential genes can be more accurately detected.Totally,1574,3257,2419,1387 and 2263 spatially patterned genes were detected for the 3,7,14 day post injury and 1,2 month post injury samples,respectively.These spatially patterned genes are mainly belonged to three types,which are distributed in the uninjured area,in the injured area,and in the whole area.These genes,which are located at the injured area,can promote the formation of synapses,signal transmission between axons,and have a positive role in the repair of injured nerves.(3)Extraction of injured related features and estimation of spatial distribution of different cell types.The spatial transcriptome sequencing data of this subject is based on the 10 x Visium technology platform.Each spatial spot can contain 3 to 10 cells,involving multiple cell types.Different cell types may involve different injury reactions and biological functions.Therefore,Non-negative Matrix Factorization(NMF)was adopted to deconvolute the spatial gene expression profiles,aiming to extract the different features from each spot.Combining the spatial changes of each feature at different time points after injury with its top ranked genes,the corresponding biological functions can be deduced.To further estimate the spatial distribution of each cell types,one public available sc RNA-seq data of injured spinal cord was downloaded,and adopted to estimate the expression signatures of each cell type of cells using the regularized negative binomial distribution model.Then these signatures were mapped into the spatial transcriptome to estimate the proportions of each cell type across the tissue architecture.Using these deconvoluted results,the characteristics of cell migration,aggregation,and the dynamic changes of tissue structure in the process of injury can be studied.Through non-negative matrix decomposition,we were able to identify the gene modules specifically distributed at the injury boundary and injury epicenter,and these characteristics are related to biological pathways such as EGFR,TGFb,JAK-STAT.(4)Cell-cell interactions based on the spatial location information.It is of great significance to study the cell-cell communication in specific regions(i.e.,the injury epicenter,the injury interface),which is crucial to understand the pathological progress of injury and the mechanism of scar formation.To study the spatial cross talk,the mouse ligand receptor pairs data was downloaded from public database.The mean value of all ligand-receptor pairs across the spots of prior selected region were calculated by averaging the ligand and receptor expression among each spot and its nearest neighbors,and then the mean expression of all these spots across the region was computed.For each ligand-receptor pair,this process was then repeated on randomized permutations of spot while keeping the total number of spots to generate a null distribution.P value was calculated according to the true average compared with the null distribution to test whether the pair was statistically significant.The results in this section described the spatial distribution of fibroblasts,astrocytes,microglia and macrophages and their subtypes,analyzed the highly expressed differential genes of each subtype and the biological enrichment pathways involved,and finally studied the spatial interaction of fibroblasts as ligand cells and receptor cells with other three cells,and find TGF-β,SPP1,MDK,CCL and other pathways interact between glial scar and fibrotic scar.In conclusion,this project first constructed a spatial transcriptomics database at single-cell-level,involving 4 species and 16 tissues and organs,and conducted several spatial related functional analyses,which can provide great help for researchers to study the spatial heterogeneity of these tissues.Then,we explored the spatial heterogeneity of the wound architecture of injured spinal cord using spatial transcriptomics profiles at distinct time points.We identified spatially patterned genes using our kernel functions model,and specifically distributed factors by using NMF.Integrating spatial gene expression with single-cell RNA data enabled us to study the dynamic changes of cell-type compositions,their colocalization and cross talk at increased resolution.We depicted the migration process of injury related cell types across the tissue architecture,identified different subpopulations of fibroblast at acute and chronic period which showed specific spatial distribution at the injury interface and lesion epicenter,and further explored their cross talk with other colocalized cell types.Overall,our data will provide an integrative molecular resource of injured spinal cord for the research communities,and bring new sight of the potential gene target for the treatment of spinal cord injury. |
PDF Full Text Request