Abstract 4143528: HeartMap: An Integrated Cell Atlas of 2.6 million cells across 260 individuals

Background: Single-nucleus RNA sequencing (snRNA-seq) is a powerful tool to define the cell-type specific transcriptional signature in health and disease. Although it has been applied to study many forms of cardiovascular disease, few resources exist to compare results across studies and disease states. Methods: We identified 9 studies from the literature with > 10k nuclei, individuals > 18 years, and non-diseased controls. Existing data was downloaded, similarly re-processed, aligned to the GRCh38 reference transcriptome, and aggregated for quality control. Sixteen versions of data integration methods were benchmarked with the single-cell integration benchmarking (scIB) package. Clustering analysis was conducted in SCANPY using established methods to identify global cell types and distinct cell states within cell types. Cell type and cell state labels were assigned to clusters based on differential expression testing using limma-voom. Compositional analysis was conducted using scCODA. Results: HeartMap contains 2,600,078 cells across 36,601 genes from 9 studies, 260 individuals, 7 disease states and normal controls, and 8 anatomical regions. The median age was 54 years (IQR: 17) and 40% were female. Benchmarking selected single cell annotated variational inference as the optimal integration method. Global clustering of all nuclei revealed 14 major cell types. At the sample level, after controlling for differences between studies, the strongest sources of variation in transcription included disease status, patient sex, and anatomical region of sample. To demonstrate the utility of this resource, we identified 10 distinct fibroblast sub-populations among ~580,000 fibroblast cells. Two sub-populations of activated fibroblasts show POSTN+COL22A1+ and POSTN+TNC+ enrichment in primary (DCM/HCM) and ischemic (ICM/AMI/CAD) cardiomyopathies, respectively. Validation using RNA-scope and immunofluorescence staining revealed peri-vascular patterns of TNC+ fibroblasts and interstitial patterns of COL22A1+ fibroblasts in diseased tissues. Conclusions: HeartMap provides a unique resource to the cardiovascular research community and provides additional insights into the transcriptional diversity in health and disease.