Pilot studies were performed on discarded pig hearts obtained from the surgical research laboratory. IVS tissue was dissected free, minced and either cryopreserved or immediately used for nuclei preparation. Nuclei could easily be isolated from either fresh or frozen tissue and RNA integrity remained high after cryopreservation (Fig. 1). We chose to isolate nuclei from heart tissue rather than whole cells, to reduce the expression of stress response genes brought on by harsh enzymatic or mechanical treatment [9] and to reduce challenges in microfluidic cell isolation since cardiomyocytes are much larger than other cell types within the heart. We then proceeded to use our cryopreservation protocol on human samples. We initially performed pilot studies on human myectomy tissue from patients with hypertrophic cardiomyopathy. Surgical samples obtained from the operating room were subject to our processing and cryopreservation protocol. Thawed tissue was then processed for nuclear isolation and bulk RNA analysis, demonstrating high quality RNA (Fig. 2). Human heart nuclei were measured in cross section and the diameter was found to average 6.5 ± 1.8 µm. Four unused donor hearts from both males and females between the ages 23–57, were obtained from New England Donor Services (Fig. 3b). The hearts were prescreened for evidence of heart disease, and any donors with a history of diabetes, hypertension, hyperlipidemia, smoking, rheumatoid arthritis, or any other potential contraindication were excluded from this study. Each heart was cut into transverse sections, and the samples from the IVS were weighed and minced, and then cryopreserved. Nuclei were released from the cryopreserved tissue using gentle homogenization (Fig. 3a). Each cryopreserved sample produced approximately half a million nuclei suitable for use in the 10× Genomic Single Cell Gene Expression system or the Fluidigm C1 system.
After sequencing and initial data processing with Cell Ranger software [10], each sample dataset was processed further to remove called nuclei that were likely droplets with only ambient RNA, or droplets that contained two nuclei. The four datasets were combined using the Seurat Integration function [11]. The final combined dataset included 24,858 nuclei. Overall clustering of the integrated dataset revealed 23 cell populations within the IVS, and this was visualized using the dimensionality reduction algorithm uniform manifold approximation and projection (UMAP, Fig. 3c), where each dot represents a single nucleus and is colored by cluster identity. Splitting the integrated dataset into the individual datasets reveals that all 23 clusters are present in each dataset and no single sample is driving the clustering (Fig. 3d). Differentially expressed genes for each cluster were determined using the Seurat function “FindAllMarkers”, and the full list of differentially expressed genes for each cluster is listed in Additional file 1: Table 1. The top ten differentially expressed genes from each cluster were determined and their expression levels in each single nucleus in the dataset were represented in a heatmap with nuclei grouped together by cluster identity across the x-axis (Fig. 3e). Analysis of the top ten differentially expressed genes in each cluster was sufficient to indicate cluster similarity (e.g. clusters 2 and 3), likely representing similar cell types with different biological functions.
Cell identities were assigned to each cluster using known biomarkers of expected cell types, differentially expressed genes, gene ontology, and pathway analysis. Similar cell types were positioned close each other on the UMAP (Fig. 4a). The markers used to identify the different cell types are listed in the dot plot (Fig. 4c), and this plot illustrates that each cluster has a unique expression of those markers. Interestingly, we see five separate cardiomyocyte (CM) populations, revealing CM diversity (Fig. 4c, d). Pathway and gene ontology analysis indicate differences in oxidative phosphorylation, protein synthesis, while biomarker analysis reveals differences in expression of sarcomeric proteins such MYH7 and MYH7B. The fourth CM population shows an elevated metabolic phenotype compared to the other four CM populations.
Non-cardiomyocyte cells account for roughly two-thirds of the cells in the human IVS (Fig. 4b), with fibroblasts making up one-third of the total IVS cell population. There are six different fibroblast populations present with considerable differences in gene expression (Fig. 4c). Fibroblast 5 has the greatest expression of collagens, while Fibroblast 1 and 3 show little collagen expression. Ingenuity Pathway Analysis [14] for each fibroblast population illustrates the similarities in biological function for some of these fibroblast populations, such as 1 and 2, and 2 and 5 (Fig. 4e). Interestingly, populations 2 and 5 both exhibit increased activation of proinflammatory and other signaling pathways, indicating an important role in intercellular communication.
Other cell types identified include endothelial, lymphatic, smooth muscle and pericytes, neuronal, and several immune cell populations. Altogether, immune cells account for almost one-fifth of the cells in the adult human IVS (Fig. 4b). Several of the non-myocyte populations were assigned based on gene ontology and pathway analysis. Interestingly, one cell population showed very few upregulated genes, and gene ontology and pathway analysis were not informative. This cell population was labeled as stromal cells, and further analysis of this interesting cell population to understand the role it plays in the normal adult human heart is ongoing.