Tissue-resident immune cells in health and disease: their heterogeneity and associated gene signatures
View/ Open
Date
31/07/2021Author
Patir, Anirudh
Metadata
Abstract
The immune system is comprised of numerous cell types, molecules and pathways whose primary purpose is to regulate tissue homeostasis and protect an individual from disease. Researchers have tried to identify and characterise components of the immune system and their interactions as modulating aspects of this system is a major goal of the pharmaceutical industry. This work has described the significant heterogeneity of immune cell types which are defined by their microenvironment, extending from their lineage commitment in specialised tissues, e.g. bone marrow and thymus, to their activation states in disease. A facet of this heterogeneity are tissues-resident immune cells (TRICs) which have tissue specific homeostatic functions, and differ from their tissue naïve counterparts at the transcriptomic and epigenetic level. Furthermore, these cells display unique activation states in disease, making them a target for tissue-specific therapies. Hence, in this thesis, I have sought to expand on the current knowledge of TRICs in health and disease, investigating their heterogeneity and how to define them using various computational approaches.
Initially in chapter two a single well defined TRIC population, microglia, the tissue-resident macrophage of the brain is investigated. Microglia are the dominant immune cell type of the brain and are strongly implicated in neurodegenerative disease. These cells exhibit great heterogeneity depending on the brain region they reside in, also influencing their activation states. Fifteen studies had previously sought to define the functional profile of microglia in humans and mice, but these ‘gene signatures’ showed poor agreement overall. To address this issue a core human microglia signature conserved across brain regions was derived. Accordingly, data derived from intact brain tissue and pooled cells derived transcriptomic data from various brain regions was collated. This included nine datasets across three resources, the Genotype-Tissue Expression (GTEx) project, the Allen Brain Atlas (ABA) and a study of central nervous system (CNS) cells from Zhang et al. From each dataset, a microglial signature was derived using gene coexpression network (GCN) analysis to capture genes sharing a common expression profile across samples and which likely represented the same biology. The final human microglia signature comprised of 249 genes which were present in three or more of the dataset-derived microglial signatures. This gene set was validated using various sources of evidence. The average expression of signature genes correlated with microglial numbers and was significantly higher in myeloid populations relative to other immune and CNS cell types. Furthermore, the proteins encoded by signature genes positively stained for microglia in different brain regions. The signature provides a means to understand the homeostatic state of these cells and a baseline against which their divergence in disease may be measured. Accordingly, the signature was used to analyse microglia in a transcriptomic dataset generated from post-mortem brain tissue of individuals of different ages and from Alzheimer’s patients in four regions of human brain. This helped untangle the qualitative (activation states) and quantitative (cell proportions) differences in microglia between conditions and brain regions. Microglial cell numbers correlated with neuroinflammation and tau pathology in a region-dependent manner. The activation state of these cells was characterised by the downregulation of homeostatic genes (CX3CR1 and P2RY12) and upregulation of TREM2-TYROBP pathway genes which have been implicated in Alzheimer’s disease through genome-wide association studies (GWAS).
In chapter three, the analysis of TRICs was expanded upon from microglia to other TRIC populations. Currently, several immune cell types have been defined by selected markers and cytokine/chemokine profiles in the context of different diseases and tissues. However, a comprehensive unbiased analysis of these phenotypes in the context of other immune cells is required to appreciate the breadth of cellular heterogeneity, revealing commonalities and further subdivisions of known cell types. Given the availability of transcriptomic atlases of immune- and tissue-derived cells, there is a considerable scope for further analyses of TRIC biology. Hence, publicly available transcriptomic datasets from mouse including that derived from pooled-cells of marker-defined cell types and that from unbiased single cell RNA-Seq (scRNA-Seq) were considered. The former was taken from the Immunological genome project (ImmGen) resource, which comprised of 128 combinations of marker-defined cell types from 26 tissues. The relationship between these cells was studied, as were the gene signatures associated with them. Comparing cell types based on their transcriptome showed the relative similarity between lymphoid cell types relative to the heterogeneous myeloid cell populations. Using GCN analysis, 157 gene modules associated with either cell lineages, cell types, cell subsets or TRICs were identified. Interestingly, it was difficult to distinguish certain marker-defined cell types from others, either suggesting that cell types could be defined by a few genes or current markers may encompass overlapping cell populations. As a complimentary unbiased approach, we also analysed immune subsets defined by the Tabula Muris Atlas, which included scRNA-Seq data derived from twelve tissues. Forty-three cell clusters were identified which were associated with forty-four gene clusters. The analyses highlighted gene clusters associated with the different cell lineages, cell types and TRICs, many of which significantly overlapped with those from the ImmGen GCN analysis. Some gene signatures were unique to TRICs or common across them, indicative of a tissue-dependent/independent biology. As expected, the greatest number, eleven signatures were associated with macrophages, eight of which agreed with cell types identified in the literature based on certain associated genes. To aid these analyses, novel approaches including annotating cells from a reference set of known cell types based on their transcriptomic profile; and capturing gene coexpression patterns using GCNs were developed, both of which are ongoing challenges unique to scRNA-Seq due to particular technical and biological variations.
Building upon the work described in chapter three, chapter four involved extending similar deconvolution analyses to human TRIC gene signatures from bulk tissue transcriptomic data taken from the GTEx resource. First, a set of reference immune signatures (RIS) was derived from a downsampled collection of all the 28 tissues of the GTEx, thus representing nine classifications of immune cell types. In the case of macrophages, monocytes and dendritic cells, combined gene signatures were identified, thus highlighting the resolution of bulk-transcriptomics for signature derivation. Finally, the RIS aided in deriving TRIC signatures individually from the 21 tissues of the GTEx considered for downstream analyses. As expected, signatures of macrophages-monocyte-dendritic cells were found in every tissue and across tissues 1,012 genes were associated with these cells, the highest number relative to other cell classifications. Interestingly, genes that were most commonly associated with a given cell type across tissues included many known markers for them, as found for macrophage-monocyte-dendritic cells, neutrophils, T cells, NK cells and B cells. Subsequently, each TRIC signature was compared with those derived from mouse in chapter three. Thirty-nine gene clusters overlapped between species and were associated with 12 TRICs. Seven TRIC populations and their associated genes were supported through literature.
In conclusion, this work has sought to examine the heterogeneity of TRICs, the transcriptomic signatures associated with them and the computational approaches to best derive them from tissue and cell level data. The work also shows the potential of using these TRIC signatures to explore disease states and the associated response of immune cells in this context.