Heterogeneity of ECs between LA and LV. By analyzing single-cell RNA sequencing data from healthy cardiac cells obtained from the left ventricle and left atrium in the first dataset, we identified 11 distinct clusters (Fig. 2a). Feature plots revealed that PECAM1 and VHF, well-known EC biomarkers, were highly expressed in clusters 0, 5, and 10 (Fig. 2b). We zoomed in all ECs only and distinguished two relatively distinct clusters of ECs originating from the left ventricle and atrium (Fig. 2c), and we found 621 DEGs between these two chambers of the heart. The top 10 most significant DEGs' expression levels are represented in the heatmap (Fig. 2d).
To identify potential disease-causing genes, we compared gene expression levels between healthy controls and patients with dilated cardiomyopathy (DCM) in the second dataset, which yielded 146 differentially expressed genes (DEGs) associated with the disease. Additionally, we analyzed the third dataset to compare gene expression levels between week 1 and week 11 in mice with transverse aortic constriction (TAC), identifying 145 DEGs. These three datasets had seven overlapped genes: B2M, FN1, SAT1, MYL2, MYH6, MYL3, and TNNC1 (Fig. 3a). To explore protein-protein interactions between these seven genes, we constructed a network using STRING and found that MYL2, MYH6, TNNC1, and MYL3 were highly associated with each other (Fig. 3b). SAT1, an enzyme that catalyzes the acetylation of polyamines, showed no clear association with the other six genes.
We then analyzed the expression levels of the seven overlapped genes in the left ventricle and left atrium (Fig. 3c). MYH6, SAT1, FN1, and B2M were significantly upregulated in the left atrium, while MYL2, MYL3, and TNNC1 showed significantly higher expression levels in the left ventricle. These genes are involved in cardiac functions such as muscle filament sliding (p-value = 7.78E-12), regulation of the force of heart contraction (p-value = 2.99E-07), ventricular cardiac muscle tissue morphogenesis (p-value = 1.71E-11), and cardiac muscle contraction (p-value = 4.13E-08, Fig. 3d).
Predicting heart disease by using overlapping genes.To investigate the impact of the 7 overlapped DEGs on disease development, a linear regression model was trained using the expression matrix of these 7 genes in the second dataset, which contained approximately 1,100 cells from 2 healthy samples and 4 DCM samples. To validate the model, an external independent dataset of 1,700 cells from 2 healthy samples and 2 CAD samples, a related disease that also leads to heart failure, was used. Each cell is treated as a sample in the prediction model. The prediction model achieved excellent results, with 83.08% accuracy in the DCM dataset (Fig. 4a) and 85.67% accuracy in the independent CAD dataset (Fig. 4b), using the 7 overlapped genes as the basis. These results provide further evidence of the significant contribution of these 7 genes' expression levels to heart failure development and suggest a potential avenue for early prediction of heart failure.
Characterizing potential disease-causing genes in the development of the fetal heart. To understand if 7 potential disease-causing genes contribute to the development of heart disease by affecting the development of human’s heart during fetal stage, the expression levels of these genes were analyzed during the fetal heart development. The data was collected over 7 different time periods and compared between adjacent days to generate 6 lists of DEGs. Among these lists, 65, 21, 42, 42, 82, and 45 significant DEGs were found between days 90 and 94, 94 and 110, 110 and 113, 113 and 115, 115 and 120, and 120 and 122, respectively. Two genes, B2M and FN1, were found to be common in both the previous list of 7 DEGs and these DEGs. B2M plays a key role in antigen presentation, processing, inflammation, the complement cascade, and stress response, and previous research suggests positive associations of higher B2M levels with cardiovascular disease outcomes. FN1, on the other hand, is involved in various cell processes, such as embryogenesis, wound healing, and blood coagulation. Interestingly, these two genes exhibit a similar expression patterns in the different time stage, indicates a potential joint action. Figure 5 shows the average expression levels of these two genes over time.