Cross-Tissue Single-Nucleus RNA Sequencing Discovers Tissue-Resident Adipocytes Involved in Propanoate Metabolism in the Human Heart

Abstract

Background: Adipocytes are crucial regulators of cardiovascular health. However, not much is known about gene expression profiles of adipocytes residing in nonfat cardiovascular tissues, their genetic regulation, and contribution to coronary artery disease. Here, we investigated whether and how the gene expression profiles of adipocytes in the subcutaneous adipose tissue differ from adipocytes residing in the heart.

Methods: We used single-nucleus RNA-sequencing data sets of subcutaneous adipose tissue and heart and performed in-depth analysis of tissue-resident adipocytes and their cell-cell interactions.

Results: We first discovered tissue-specific features of tissue-resident adipocytes, identified functional pathways involved in their tissue specificity, and found genes with cell type-specific expression enrichment in tissue-resident adipocytes. By following up these results, we discovered the propanoate metabolism pathway as a novel distinct characteristic of the heart-resident adipocytes and found a significant enrichment of coronary artery disease genome-wide association study risk variants among the right atrium-specific adipocyte marker genes. Our cell-cell communication analysis identified 22 specific heart adipocyte-associated ligand-receptor pairs and signaling pathways, including THBS (thrombospondin) and EPHA (ephrin type-A), further supporting the distinct tissue-resident role of heart adipocytes. Our results also suggest chamber-level coordination of heart adipocyte expression profiles as we observed a consistently larger number of adipocyte-associated ligand-receptor interactions and functional pathways in the atriums than ventricles.

Conclusions: Overall, we introduce a new function and genetic link to coronary artery disease for the previously unexplored heart-resident adipocytes.

Keywords: adipocytes; cell communication; coronary artery disease; heart; subcutaneous fat.

Figure 1.. Cellular diversity in subcutaneous adipose tissue and each of the heart chamber.

A, Schematic overview of the single nucleus RNA-seq (snRNA-seq) analysis pipeline. Analysis of the subcutaneous adipose tissue (n=15) and heart (n=7) datasets included clustering of nuclei, followed by cell-type annotation and marker gene identification (see the Methods). B, Further analysis of adipocyte marker genes included identification of tissue-specific and tissue-shared adipocyte marker genes, followed by KEGG pathway enrichment analysis, transcription factor enrichment analysis, cis-expression quantitative trait locus (eQTL) analysis, gene set enrichment analysis for genetic disease associations, and cell-cell communication analysis between adipocytes and other cell-types. C, Uniform Manifold Approximation and Projection (UMAP) clustering of 37,865 nuclei displaying cellular diversity in human subcutaneous adipose tissue. D-G, UMAP visualization of 41,204, 42,867, 28,696, and 39,639 nuclei of right atrium (RA), left atrium (LA), right ventricle (RV), and left ventricle (LV), respectively, displaying cellular diversity in the heart. The black circle indicates the location of each chamber in the heart.

Figure 2.. Comparisons of adipocyte marker genes between subcutaneous adipose tissue and heart reveal shared and tissue-specific drivers of tissue-resident adipocytes.

A, Adipocyte marker genes that are specifically present in one of the heart chambers and absent in the subcutaneous adipose tissue, and vice versa, are considered tissue-specific adipocyte marker genes. B-E, Overlaps of adipocyte marker genes (left) between each of the heart chambers (right atrium (RA), left atrium (LA), right ventricle (RV), and left ventricle (LV)) and subcutaneous adipose tissue; and comparison of enriched pathways using unique adipocyte marker genes (right). P value is calculated using hypergeometric test and adjusted for multiple testing using BH method (FDR<0.05) (see Table S1–S5 for adipocyte marker gene lists in the subcutaneous adipose tissue, RA, LA, RV, and LV).

Figure 3.. Comparisons of adipocyte marker genes and functional pathways between the heart chambers reveal shared and chamber-specific features of heart adipocytes.

A, Venn diagrams showing the number of shared and chamber-specific adipocyte marker genes (AMGs) between the heart chambers (right atrium (RA), left atrium (LA), right ventricle (RV), and left ventricle (LV)). B, Dot plot showing enriched pathways in each heart chamber (RA, LA, RV, and LV) using all adipocyte marker genes. P value is calculated using the hypergeometric test and adjusted for multiple testing using the BH method (FDR<0.05).

Figure 4.. Module score and transcription factor (TF) enrichment analyses highlight the regulators of tissue-resident adipocytes.

Module scores analysis calculates the difference between the average expression level of genes within the module compared to aggregated expression of control gene sets. A, A schematic of the propanoate metabolism pathway with heart-specific adipocyte marker genes indicated by blue rectangles. B, Genes of the propanoate metabolism pathway among the tissue-specific adipocyte marker genes in the right atrium (RA), left atrium (LA), right ventricle (RV), and left ventricle (LV). C, Expression of propanoate metabolism pathway genes is enriched in the adipocyte cluster of LA using a module score analysis (see the Methods). D, Expression of propanoate metabolism pathway genes is less enriched in the adipocyte cluster of the subcutaneous adipose tissue using a module score analysis. E-F, Box plot of module score values by cell-types (adipocytes versus non-adipocytes) in LA (E) and subcutaneous adipose tissue (F). Each box extends from the 25^th to the 75^th percentile of the distribution of the module score values in every group. The horizontal line within each box denotes the median value. In the LA, the module scores of propanoate metabolism pathway genes are significantly higher (Wilcoxon rank sum p<2.2×10⁻³⁰⁸; a permutation p-value=1.00×10⁻⁶) in the adipocytes than non-adipocytes (Figure 4E), while in the subcutaneous adipose tissue the difference between the module scores in the adipocytes and non-adipocytes was smaller and did not pass the significance in the permutation analysis (Wilcoxon rank sum p=9.90×10⁻³⁸; permutation p-value=0.1272) (Figure 4F). G, TF enrichment analysis of the unique adipocyte marker genes (UAMG) in RA, LA, RV, LV, and subcutaneous adipose tissue identifies significantly enriched TFs using the Enrichr tool. Orange circle indicates heart adipocyte-specific TFs, which are enriched for all heart chambers. Yellow circle indicates adipose tissue adipocyte-specific TFs, which are enriched in at least three comparisons between subcutaneous adipose tissue and various heart chambers.

Figure 5.. Comparison of mean gene expression across the cell-types reveal adipocyte-specific expression of the propanoate metabolism pathway genes in the left (A) and right atrium (B) while including all available atrium samples.

The dot plots of the seven propanoate metabolism pathway genes (ACACA, BCKDHB, ECHDC1, EHHADH, HIBCH, LDHA, and PCCA) show their expression profiles across the cell-types in the left atrium (n=6 samples with the mean number of adipocytes=254 nuclei/sample (SD=189)) (A) and right atrium (n=5 samples with the mean number of adipocytes=48 nuclei/sample (SD=35)) (B). The size of the dot represents the percentage of cells where a gene is expressed within a cell-type while the color (blue is high) represents the average expression of each gene across all cells within a cell-type.

Figure 6. Cell-cell interaction analysis discovers significant heart-specific ligand-receptor interactions and signaling pathways where adipocytes represent either the sender or receiver cell-type.

. A, Circle plots showing the significant aggregated cell-cell communication networks among the cell-types in the right atrium (RA), left atrium (LA), right ventricle (RV), left ventricle (LV), and subcutaneous adipose tissue. The significant interactions (p<0.05) between two cell-types are identified using a one-sided permutation test. Lines in the circle plot indicate a significant cellular communication among the cell types, and the thickness of the lines corresponds to the relative intensity of intercellular communication. Each line is colored by the sender cell type. Plots show the strength of the interactions. Regarding the number of interactions, there are 59 adipocyte-associated ligand-receptors interactions in RA, 81 in LA, 18 in RV, and 31 in LV, respectively. B, Heatmap showing the presence (orange) of significant adipocyte-associated ligand-receptor (L-R) interactions in RA, LA, RV, LV, which are absent in the subcutaneous adipose tissue. Significant L-R pairs present in all heart chambers are further highlighted in violet color. Signaling pathways associated with these L-R interactions are shown in parenthesis. Tissue-specific adipocyte-associated L-R interactions, and their corresponding probability values, p-values, and signaling pathways are shown in detail in Table S11–S14. C-F, The adipocyte-associated significant L-R pairs that were specifically present in the four heart chambers were selected for further communication analysis mediated by each L-R pair: EFNA5-EPHA4 (C), EFNA5-EPHA3 (D), THBS1-CD36 (E), and IGF1-IGF1R (F). For each L-R pair, two chord diagrams are shown indicating the sender and receiver cell-types involved in RA and LA, respectively. Chords are colored by the sender cell-type. Signaling pathways associated with these L-R interactions are given in parenthesis.