Baseline characteristics of all and propensity score matched BD donor snRNAseq heart samples
We identified 20 BD donor left ventricular snRNAseq samples (healthy, N = 16 and LQTS, N = 4) in a study which fulfilled our criteria for further downstream analyses. Baseline characteristic of the samples was shown in Table 1 left. Before adjusting the variables, we found that age of LQTS were significantly higher compared to the other non-arrhythmic healthy hearts (average 64 vs 46 years, P = 0.037), whereas other baseline parameters including sex, BMI, hypertension, diabetes, chronic kidney disease and smoking history were similar. This indicates that aged donors itself may represent an increased risk for brain-dead related LQTS, but require further validating. Owing to the discrepancy of baseline characteristics between the group, we performed PSM with all the parameters to remove potential confounding factors prior to subsequent transcriptome analyses (Table 1). We applied PSM with a 1:1 sampling ratio, leaving 4 normal samples without arrhythmia and 4 LQTS samples with balanced baseline characteristics (SMD < 0.1 for all variables). Before and after PSM baseline characteristics comparing both groups were shown in Table 1.
Myocardial single-nuclei profiling reveals increased T cell composition in BD donor hearts with LQTS
Here we present single nuclei study of propensity score matched non-arrhythmic and LQTS BD donor left ventricular samples. Brief workflow of the current study as shown in Figure 1A. A total of 17075 and 13661 nuclei from 4 normal and 4 LQTS samples passed the quality control threshold to be included in our analysis (Figure 1B). By performing UMAP dimensionality reduction and KNN-clustering on the normalized integrated Seurat object, automated reference-based annotating of cells showed distinct left ventricular cell types: Level 1 predicted cell types (L1): Endothelial, Myeloid, NK/T, Pericyte, Fibroblast, Smooth Muscle, Cardiomyocyte, Lymphatic Endothelial, Neuronal, Mast, Adipocyte, B and Mesothelial (Figure 1C); Level 2 predicted cell types (L2): Endocardial, Venous Endothelial, Endothelial, Ariterial Endothelial, Monocyte/cDC, Macrophage, T, Pericyte, Fibroblast, Smooth Muscle, Ventricular Cardiomyocyte, NK, Capillary Endothelial, Lymphatic Endothelial, Neuronal, Mast, Atrial Cardiomyocyte, Adipocyte, B and Mesothelial. Each cluster highly expressed specific genes similar to be previously reported literature, to be specific, Endothelial (VWF), myeloid (C1QC), T (CD3E), Pericyte (KCNJ8), Fibroblast (DCN), Smooth Muscle (MYH11), Cardiomyocyte (RYR2), Lymphatic Endothelial (CCL21), Neuronal (NRXN1), Mast (KIT), Adipocyte (PLIN1), B (MZB1) and Mesothelial (HAS1) (Figure 1D). We subsequently compared cellular compositions between both conditions (Figure 1E). Interestingly, we found that immune cell compositions were increased in LQTS compared to healthy donors (Myeloid: P = 0.17, B cell: P = 0.19) and significant increase in Mast (P = 0.045) and NK/T (P = 0.029) composition.
Gene expression changes and cell-specific regulatory pathway enrichment show immune activation in LQTS
To assess the transcriptional changes with BD associated LQTS, we performed differential expression (DE) analysis using Seurat FindMarkers function with default non-parametric Wilcoxon rank sum test in each broad cell types. All total of 13640 significant DE genes were detected across all L1 cell types (Figure 2A). Up-regulated (average FC > 0.25, N = 8423) and down-regulated (average FC < -0.25, N = 5217) genes of the cell types were shown in Figure 2A. Interestingly, B emerged as the cell type with most DE genes (N = 4222) followed by Adipocyte (N = 3248), Lymphatic Endothelial (N = 1635), Mast (N = 1148), Neuronal (742), Myeloid (N = 586). NK/T (N = 497), Fibroblast (N = 442), Smooth muscle (N= 350), Endothelial (N = 333), Cardiomyocyte (N= 283) and Pericyte (N = 154), reflecting a significant difference occurring between the two conditions. Number of up and down-regulated genes were shown in Figure 2B. To further determine the biological relevance of cardiomyocytes and non-cardiomyocytes, we performed enrichment of all the cell-specific DEG using canonical KEGG pathways gene set (Figure 2C). In up-regulated genes, focal adhesion pathway was consistently enriched in several up-regulated DEG of cell types including B, Fibroblast, Myeloid and Neuronal (Figure 2C). Up-regulated genes of Myeloid was enriched in Th1 and Th2 cell differentiation pathway whereas both B and Myeloid were enriched in T cell receptor signaling pathway, suggesting activated innate and adaptive immune system in LQTS donor hearts (Figure 2C). Gene Set Enrichment Analysis (GSEA) of B cell DE genes revealed enrichment of Th1 cell differentiation pathway and neuron death in response to oxidative stress pathway (Figure 2D).
Cell receptor-ligand comparison analysis reveal significant increase in Neuronal signaling associated with inflammation and immune cell trafficking through adhesion molecules in LQTS
The determine whether cellular receptor-ligand communication differs between LQTS and healthy BD donors, we implemented CellChat multiple condition comparison analysis. In Figure 3A, it was shown that overall number of pathway interactions (1419 vs 2399) and their strength (51.047 vs 77.117) substantially increases in LQTS compared to normal donors. Cell-type specific differential interactions showed that non-immune cells including Fibroblast, Mesothelial, Pericyte and Smooth Muscle emerged as top differential communicating senders (Figure 3B) while Neuronal and Cardiomyocyte acts as top differential receiver in number of interactions. Evidently, Neuronal showed the highest received differentially increased in communication number and strength (Heatmap of Figure 3B, red indicates increased communication while blue indicates decreased in LQTS dataset), indicating the LQTS abnormality was indeed driven by neuronal changes. At the same time, significant differences were also observed in Cardiomyocyte to Lymphatic Endothelial signaling (Figure 3B), Next, we ranked all the significant pathways based on differences in the overall information flow within the inferred networks between LQTS and healthy donor hearts (Figure 3C). Interestingly, pathways associated with neuronal regulation of immunity such as Heparan Sulfate Proteoglycan (HSPG), neuronal growth factor (NGF), Contactin (CNTN), adhesion molecules including Junctional Adhesion Molecules (JAM), Intercellular Adhesion Molecule (ICAM), Platelet Endothelial Cell Adhesion Molecule (PECAM1), as well as immune cytokine IL6, LIGHT (TNFSF14), CXCL chemokines (CXCL) were only specific to LQTS. Furthermore, immune T lymphocyte signaling such as CD30, MHC-II and CD99 were also specific to LQTS samples. The overall, outgoing and incoming signaling patterns in normal and LQTS samples were shown in Figure 3 (DEF). Closer inspection of Fibroblast, Neuronal, Endothelial and Lymphatic Endothelial signaling changes of normal vs LQTS samples were shown in Figure 4A, showing significant increased LAMININ and adhesion (PECAM1 and NCAM) signaling strength and interaction within these cells. Closer inspection of the increased ligand-receptors between cell types with Neuronal as receiver as well as decreased ligand-receptors between cell types with Lymphatic Endothelial as receiver were shown in Figure 4B. Altogether, we hypothesize a that neuronal cell in LQTS modulates immune cell trafficking and inflammatory responses through regulating lymphatic and vascular endothelial adhesion molecules.
Integrated analysis of differentially expressed BD donor plasma proteome and Neuronal DE genes revealed integrin cell surface interactions as important regulatory pathway implicated in LQTS
A total of 463 proteins were identified in their study using label-free protein quantification using high-definition mass spectrometry. Using adjusted P value < 0.05 as significant, we identified 119 up-regulated and 124 down-regulated plasma proteins (Figure 5A). Through our single nuclei analysis, we hypothesized a major role of Neuronal in driving the molecular changes implicated in brain-dead related LQTS. Thus, we sought to explore the relationship between brain-dead plasma proteins and Neuronal in LQTS in order to determine potential cause and therapeutic targets. To accomplish this, we used an online multi-omics analysis platform, Omicsnet 2.033, which gave clear visualization of biological network visualization using DE BD donor plasma proteome and Neuronal gene expression with protein-protein interaction (PPI) network framework as shown in Figure 5B (Pink nodes as plasma protein and Blue nodes as Neuronal DE genes). ITGB1 (Integrin Subunit Beta 1) emerged as gene with highest connectivity (Figure 5B; increasing node size implicate higher connectivity). Through Reactome and KEGG database enrichment, relevant pathways were enriched. Integrin cell surface interactions, axon guidance, PDGF and NGF signaling, gap junction, apoptosis, as well as innate immune system and cytokine-cytokine receptor interaction pathways (Figure 5C). To validate the findings from bioinformatic analyses of human BD donor heart, we performed TBI mouse model to confirm the lymphocytic infiltration as well as relative expression of genes in donor hearts. In Figure 5D and 5E, total immune cell counts were increased at 6 and 24 hours compared with sham controls, but were not statistically significant, CD3+ T cells were significantly increased at 6h after experimental TBI and further infiltrated at 24 h, whereas B220+ B cells were increased only at 24 h after experimental TBI (Figure 5D and 5E). Western blot results of experimental TBI also showed that the expression of Itgb1 and CD44 proteins was significantly up-regulated at 6 hours (Figure 5F and 5F).
Integrin pathway activation significantly correlates with CD4 T cell composition in 166 BD donor hearts gene expression microarray
We determined that pathways associated with leukocyte trafficking including integrin cell surface interaction, focal adhesion and gap junction were activated in LQTS BD donor hearts, we tested whether the correlations of their enrichment score with CIBERSORT immune cell proportion score for the 22 leukocyte groups in the donor heart RNA sequencing samples (N = 166). Correlation heatmap in Figure 6A showed the significantly correlated pathways and immune proportion scores. Most evidently, we found that “Integrin cell surface interactions” was highly correlated with “T cells CD4 memory resting” (R = 0.57; P < 0.001), “B cells naïve (R = 0.27; P < 0.001)”, “Macrophages.M2” (R = 0.35; P < 0.001), “Allograft rejection” (R = 0.52; P < 0.001) and “Focal adhesion” (R = 0.33; P < 0.001) (Figure 6A and 6B).