Tissue and cell-type identity lie at the core of human physiology and disease. Understanding the genetic underpinnings of complex tissues and individual cell lineages is crucial for developing improved diagnostics and therapeutics. We present genome-wide functional interaction networks for 144 human tissues and cell types developed using a data-driven Bayesian methodology that integrates thousands of diverse experiments spanning tissue and disease states. Tissue-specific networks predict lineage-specific responses to perturbation, reveal genes’ changing functional roles across tissues, and illuminate disease-disease relationships. We introduce NetWAS, which combines genes with nominally significant GWAS p-values and tissue-specific networks to identify disease-gene associations more accurately than GWAS alone. Our webserver, GIANT, provides an interface to human tissue networks through multi-gene queries, network visualization, analysis tools including NetWAS, and downloadable networks. GIANT enables systematic exploration of the landscape of interacting genes that shape specialized cellular functions across more than one hundred human tissues and cell types.

Skeletal muscle is a heterogeneous tissue comprised of muscle fiber and mononuclear cell types that, in addition to movement, influences immunity, metabolism and cognition. We investigated the gene expression patterns of skeletal muscle cells using RNA-seq of subtype-pooled single human muscle fibers and single cell RNA-seq of mononuclear cells from human vastus lateralis, mouse quadriceps, and mouse diaphragm. We identified 11 human skeletal muscle mononuclear cell types, including two fibro-adipogenic progenitor (FAP) cell subtypes. The human FBN1+ FAP cell subtype is novel and a corresponding FBN1+ FAP cell type was also found in single cell RNA-seq analysis in mouse. Transcriptome exercise studies using bulk tissue analysis do not resolve changes in individual cell-type proportion or gene expression. The cell-type gene signatures provide the means to use computational methods to identify cell-type level changes in bulk studies. As an example, we analyzed public transcriptome data from an exercise training study and revealed significant changes in specific mononuclear cell-type proportions related to age, sex, acute exercise and training. Our single-cell expression map of skeletal muscle cell types will further the understanding of the diverse effects of exercise and the pathophysiology of muscle disease.Skeletal muscle is a complex heterogeneous tissue consisting of multinucleated muscle fibers, immune cells, endothelial cells, muscle stem cells (satellite cells), non-myogenic mesenchymal progenitors (e.g., fibro-adipogenic progenitors, or FAPs), and other mononuclear cells 1 . To improve the understanding of skeletal muscle cell types and their transcriptional signatures, we studied human and mouse skeletal muscle mononuclear cells by single-cell RNA-sequencing and single human muscle fiber subtypes by RNA-seq.The majority of skeletal muscle is composed of the multinucleated fibers that facilitate movement. These muscle fibers include several fiber types of differing metabolic and functional properties 2-4 . While slow-twitch (or Type I) muscle fibers possess high oxidative capacity, fast-twitch (or Type II) muscle fibers have a high glycolytic capacity and are capable of supplying more power than Type I fibers 2-4 . Fiber-type composition differs across individuals and can change by as much as 10-30% during exercise training regimens 5-7 . Furthermore, the transcriptomic response to physical activity is different in each fiber-type as each fiber-type responds differently to different modes of exercise 8,9 . Crucially, muscle fibers secrete myokines, which both act locally within muscle tissue as well as influence other organs and tissues via hormone-like signaling 10 . Myokines may be responsible for the immune-, metabolism-, and cognition-related benefits of physical activity, as well as the chronic diseases that are caused by lack of physical activity (insulin resistance, cardiovascular disease, etc.) 10 .Besides multinucleated fibers, skeletal muscle contains many mononuclear cells, such as immune cells,...

*These authors contributed equally to this work.In inflammatory central nervous system conditions such as multiple sclerosis, breakdown of the blood-brain barrier is a key event in lesion pathogenesis, predisposing to oedema, excitotoxicity, and ingress of plasma proteins and inflammatory cells. Recently, we showed that reactive astrocytes drive blood-brain barrier opening, via production of vascular endothelial growth factor A (VEGFA). Here, we now identify thymidine phosphorylase (TYMP; previously known as endothelial cell growth factor 1, ECGF1) as a second key astrocyte-derived permeability factor, which interacts with VEGFA to induce blood-brain barrier disruption. The two are co-induced NFB1-dependently in human astrocytes by the cytokine interleukin 1 beta (IL1B), and inactivation of Vegfa in vivo potentiates TYMP induction. In human central nervous system microvascular endothelial cells, VEGFA and the TYMP product 2-deoxy-D-ribose cooperatively repress tight junction proteins, driving permeability. Notably, this response represents part of a wider pattern of endothelial plasticity: 2-deoxy-D-ribose and VEGFA produce transcriptional programs encompassing angiogenic and permeability genes, and together regulate a third unique cohort. Functionally, each promotes proliferation and viability, and they cooperatively drive motility and angiogenesis. Importantly, introduction of either into mouse cortex promotes blood-brain barrier breakdown, and together they induce severe barrier disruption. In the multiple sclerosis model experimental autoimmune encephalitis, TYMP and VEGFA co-localize to reactive astrocytes, and correlate with blood-brain barrier permeability. Critically, blockade of either reduces neurologic deficit, blood-brain barrier disruption and pathology, and inhibiting both in combination enhances tissue preservation. Suggesting importance in human disease, TYMP and VEGFA both localize to reactive astrocytes in multiple sclerosis lesion samples. Collectively, these data identify TYMP as an astrocyte-derived permeability factor, and suggest TYMP and VEGFA together promote blood-brain barrier breakdown.

In this study, we consider a clinically relevant situation where neither accurate proportion estimates nor pure cell expression is of direct interest, but where we are rather interested in detecting and interpreting relevant differential expression in mixture samples. We develop a method, Cell-type COmputational Differential Estimation (CellCODE), that addresses the specific statistical question directly, without requiring a physical model for mixture components. Our approach is based on latent variable analysis and is computationally transparent; it requires no additional experimental data, yet outperforms existing methods that use independent proportion measurements. CellCODE has few parameters that are robust and easy to interpret. The method can be used to track changes in proportion, improve power to detect differential expression and assign the differentially expressed genes to the correct cell type.