Search

A cell’s function is defined by its intrinsic characteristics and its niche: the tissue microenvironment in which it dwells. Here, we combine single-cell and spatial transcriptomic data to discover cellular niches within eight regions of the human heart. We map cells to micro-anatomic locations and integrate knowledge-based and unsupervised structural annotations. For the first time, we profile the cells of the human cardiac conduction system, revealing their distinctive repertoire of ion channels, G-protein coupled receptors, and cell interactions using a customCellPhoneDB.orgmodule. We show that the sinoatrial node is compartmentalised, with a core of pacemaker cells, fibroblasts and glial cells supporting paracrine glutamatergic signalling. We introduce a druggable target prediction tool, drug2cell, which leverages single-cell profiles and drug-target interactions, providing unexpected mechanistic insights into the chronotropic effects of drugs, including GLP-1 analogues. In the epicardium, we show enrichment of both IgG+ and IgA+ plasma cells forming immune niches which may contribute to infection defence. We define a ventricular myocardial-stress niche enriched for activated fibroblasts and stressed cardiomyocytes, cell states that are expanded in cardiomyopathies. Overall, we provide new clarity to cardiac electro-anatomy and immunology, and our suite of computational approaches can be deployed to other tissues and organs.

INTRODUCTION Human heart failure is a highly morbid condition that affects 23 million individuals worldwide. It emerges in the setting of an array of different cardiovascular disorders, which has propelled the notion that diverse stimuli converge on a common final pathway. Consistent with this, initiating etiologies do not direct heart failure treatments, which are often inadequate and necessitate mechanical interventions and cardiac transplantation. The recent application of single-nucleus RNA sequencing (snRNAseq) transcriptional analyses to characterize the cellular composition and molecular states in the healthy adult human heart provides an emerging benchmark by which disease-related changes can be assessed. Moreover, the discovery of human pathogenic variants that cause dilated cardiomyopathy (DCM) and arrhythmogenic cardiomyopathy (ACM), disorders associated with high rates of heart failure, provides direct opportunities to evaluate whether genotype influences heart failure pathways. RATIONALE A systematic identification of shared and distinct molecules and pathways involved in heart failure is lacking, and knowledge of these fundamental data could propel the development of more effective treatments. To enable these discoveries, we performed snRNAseq of explanted ventricular tissues from 18 healthy donors and 61 heart failure patients. By focusing analyses on multiple samples with pathogenic variants in DCM genes (LMNA, RBM20, and TTN), ACM genes (PKP2), or pathogenic variant–negative (PV negative) samples, we characterized genotype-stratified and common heart failure responses. RESULTS From 881,081 nuclei isolated from left and right diseased and healthy ventricles, we identified 10 major cell types and 71 distinct transcriptional states. DCM and ACM tissues showed significant depletion of cardiomyocytes and increased endothelial and immune cells. Fibrosis was expanded in disease hearts, but, unexpectedly, fibroblasts were not increased, and instead showed altered transcriptional states that indicated activated remodeling of the extracellular matrix. Genotype-stratified analyses identified multiple transcriptional changes shared only among the hearts harboring pathogenic variants or distinctive for individual and subsets of DCM and ACM genotypes. We validated many of these by single-molecule fluorescent in situ hybridization. Through analyses of receptor and ligand expression across all cells, we observed changes in intercellular signaling and communications, such as increased endothelin signaling in LMNA hearts, tumor necrosis factor in PKP2 hearts, and others. We also identified specific cardiac cell lineages expressing genes with common polymorphisms that were identified in validated association studies of DCM. Because our findings indicated genotype-enriched transcripts and cell states, we harnessed machine learning to develop a graph attention network for the multinomial classification of genotypes. This network showed remarkably high prediction of the genotypes for each cardiac sample, thereby reinforcing our conclusion that genotypes activate very specific heart failure pathways. CONCLUSION snRNAseq of human ventricular samples illuminated cell types and states, molecular signals, and intercellular communications that characterize DCM and ACM. The cellular and molecular architectures that induce heart failure are both shared and distinct across genotypes. These data provide candidate therapeutic targets for future research and interventional opportunities to improve and personalize treatments for cardiomyopathies and heart failure.