A human cell atlas of the pressure-induced hypertrophic heart

Figures

• 

  Fig. 1. Dysregulated genes in cardiomyocytes from hypertrophied hearts. a, Representative uniform manifold approximation and projection (UMAP) plot after snRNA-seq and integration of data from cardiac tissues from the septum of 14 control hearts and five patients with AS and hypertrophy. A total of 88,536 nuclei were pooled. Nineteen cell clusters were identified: CM, EC, FB, PC, SMC, LC, MC and NLC. Three clusters had shared gene expression profiles (CM–EC, CM–PC and CM–FB), and one cluster was of low quality (lowQC-CM). b, Violin plots for the MYH isoforms MYH6 and MYH7 shown as number of unique molecular identifiers (nUMI). c, The cardiomyocyte disease score was calculated by the expression of ten established cardiomyocyte disease and stress markers per cell (NPPA, NPPB, MYH7, MYH7B, XIRP2, CMYA5, ANKRD1, TNNI3, ACTA1 and PFKP). d, Unbiased reclustering of the four cardiomyocyte clusters, shown as a UMAP plot, revealed six subclusters of cardiomyocytes (CM0–CM5). e, GO term analysis for the genes upregulated in cluster 0 versus those in all other clusters. f, Quantitative analysis of the distribution of cardiomyocytes in the subclusters from d respective to their origin in percent. g,…

• 

  Fig. 2. Disturbed intercellular crosstalk in hypertrophied hearts. a, Circos plots visualizing ligand–receptor interactions. Interactions between the different cell types were analyzed using CellPhoneDB to analyze cells from healthy hearts (left) and patients with AS (right). Incoming signals are shown in red. Outgoing signals are shown in the colors assigned to the respective cell type. The thickness of the lines indicates the number of interactions. b, Total ligand–receptor interactions in hearts of healthy individuals versus patients with AS. c, Total ligand–receptor interactions in cardiomyocytes of hearts of healthy individuals versus those with AS. d, Cardiomyocyte ligand–receptor interactions with other cardiac cell types. Shown are incoming signals (number of interactions) with a ligand expressed by any other cell type and the respective receptor expressed by cardiomyocytes. e,f, Representative cardiomyocyte ligand–receptor interaction scores calculated with CellPhoneDB are shown. e, Downregulated pathways from cardiomyocyte interactions with endothelial cells in hearts of patients with AS (left, of 32 total significant interactions) and upregulated pathways from those in hearts…

• 

  Fig. 3. EPHB1 is dysregulated in cardiomyocytes of hypertrophied hearts. a, Mean expression (log2-transformed and normalized UMI counts) of Eph receptors in all cardiomyocytes. b, Heatmap of Eph receptors expressed (log2-transformed and normalized UMI counts) in all cardiomyocytes of healthy individuals versus patients with AS. Colors indicate expression levels in each nucleus. c, Differential expression of Eph receptors in cardiomyocytes between hearts of healthy individuals versus patients with AS, ranked by differential expression and calculated by subtracting mean expression of genes (normalized UMI counts; values from individuals with hypertrophy minus those from healthy individuals). d, FeaturePlot for expression of EPHB1 (log2-transformed and normalized UMI counts). Color reflects expression levels in each nucleus. e, EPHB1 expression (log2-transformed and normalized UMI counts) according to cell type. f, Violin plot for EPHB1 expression in cardiomyocytes (log2-transformed and normalized UMI counts). g, Violin plot for EPHB1 expression in cardiomyocytes from patients with AS versus healthy individuals shown for individual patients (log2-transformed and normalized UMI counts). h,…

• 

  Fig. 4. The EFNB2–EPHB1 ligand–receptor interaction has anti-hypertrophic effects in cardiomyocytes. a, FeaturePlot of EFNB2 expression in the human cardiac snRNA-seq dataset (log2-transformed and normalized UMI counts). Color reflects expression levels in each nucleus. b, Mean EFNB2 expression (log2-transformed and normalized UMI counts) per cell type. c, Phosphorylated (p) EPHB1 protein levels in human cardiomyocytes treated with recombinant (rec)-EFNB2 (15 min) compared to those in untreated cardiomyocytes (control). Left, quantification as fold change versus control (n = 3; normalized to α-tubulin-1B (α-TUB1B). Right, representative western blot. α-tubulin-1B served as the loading control. Blots were processed in parallel; the loading control was run on the same blot. d, Left, representative immunofluorescence images of the hypertrophy assay with primary neonatal rat cardiomyocytes (PE, recombinant EFNB2, 72 h). Blue, DAPI; red, sarcomeric α-actinin. Right, quantification of the hypertrophy assay only counting individual cardiomyocytes. The cardiomyocyte cellular area is displayed in µm2 (n = 4). e, Representative immunofluorescence images of the co-culture model with primary…

• 

  Extended Data Fig. 1. snRNA-seq dataset quality control (1). (a) Representative uniform manifold approximation and projection (UMAP) plots before (left panel) and after (right panel) harmonization of the two datasets (Healthy, n=14, vs. aortic valve stenosis=AS, n=5). (b) UMAP plot showing the local inverse Simpson’s index (LISI) integration score per nucleus, indicated by a color code (a score of 2 represents the ideal integration of the datasets). (c) Density plot for the distribution of the LISI integration scores per nucleus. (d) Violin plot for the LISI integration score per cell-cluster. (e) Violin plots of numbers of genes expressed per nucleus (top) and for the total amount of transcripts expressed per nucleus (bottom) in each cell-cluster. (f) Left: box plots for the ratio of the number of genes vs. the LISI score per cluster (visualized as median and 25th and 75th percentiles with whiskers indicating maximal and minimal values.). Right: bar graphs for the ratio of the number of genes vs. the LISI score per cluster given as mean ± s.e.m. The red lines indicate the threshold value calculated using the identify_outlier() function, where mean ratios above Q3+1.5 × IQR or below Q1-1.5 × IQR were…

• 

  Extended Data Fig. 2. snRNA-seq dataset quality control (2). (a) Box plots for the distribution of annotated cell types according to condition; CM=cardiomyocytes; PC=pericytes; FB=fibroblasts; EC=endothelial cells; MC= monocytes; SMC=smooth muscle cells; LC=leukocytes; NLC=neuronal-like cells. Red= Healthy (n=14); blue= AS (n=5); visualized as median and 25th and 75th percentiles with whiskers indicating maximal and minimal values. (b) Bar plot for the number of singlets (green) and doublets (violet) per cluster identified by the bioinformatic tool scDblFinder. (c) Cell-type specific marker expression for all annotated clusters from panel (a). Markers shown are a combination of well-known cell-type markers and genes identified using the unbiased Seurat function ‘FindAllMarkers’. Colors of the dots reflect the fold change (natural log) of the respective gene and size reflects the percentage of nuclei expressing the marker. (d) FeaturePlots for the expression of TNNT2, MYH7 (cardiac markers), CDH5, PECAM1 (endothelial cell markers), PDGFRA and LAMB1 (fibroblast markers) in the human cardiac snRNA-seq dataset (log2 transformed and normalized UMI counts). Color represents the expression.

• 

  Extended Data Fig. 3. Disturbed gene expression signatures in hypertrophic cardiomyocytes. (a) Heatmap for the top 30 down-regulated genes in hypertrophied (AS) vs. healthy cardiomyocytes (log2 transformed and normalized UMI counts). Color reflects the expression. (b) Heatmap for the top 30 up-regulated genes in hypertrophied (AS) vs. healthy cardiomyocytes (log2 transformed and normalized UMI counts). Color represents the expression. (c) Gene ontology (GO) term analysis for the 2775 down-regulated genes in hypertrophied vs. healthy cardiomyocytes performed with Metascape. (d) Gene ontology (GO) term analysis for the 527 significantly up-regulated genes in hypertrophied vs. healthy cardiomyocytes performed with Metascape.

• 

  Extended Data Fig. 4. Cardiomyocyte sub-cluster specific signatures and gene ontology terms. (a) Heatmap for the top ten uniquely up-regulated genes (natural logarithm of the fold change) per cardiomyocyte sub-cluster. (b-e) Gene ontology (GO) term analysis for the genes up-regulated in the distinct cardiomyocyte sub-clusters vs. all other cardiomyocytes performed with Metascape. Adjusted p-values based on Bonferroni correction using all genes of the dataset to compare the expression in violin plots were calculated with the Seurat function ‘FindAllMarkers’ using ‘bimod’ as the statistical test (a).

• 

  Extended Data Fig. 5. Cardiomyocyte sub-cluster 0 specific signatures. (a) Cardiomyocyte representation in cluster CM0 as percentage of all cardiomyocytes shown for each individual patient (Healthy n=14; AS n=5). (b) Violin plots for the genes ERBB4, TBX20 and GATA4 in cardiomyocytes of cluster 0 only (log2 transformed and normalized UMI counts). (c) Violin plots for the genes ADRA1A, NLGN1 and FGF12 in cardiomyocytes of cluster 0 only (log2 transformed and normalized UMI counts). Data are shown as mean ± s.e.m. (a). Normal distribution was assessed using the Kolmogorov-Smirnov test (a). Statistical analysis to compare two Gaussian distributed groups was performed using the unpaired, two-sided Student’s t-test (a). t=3,013, 17 degrees of freedom (a). Adjusted p-values based on Bonferroni correction using all genes of the dataset to compare the expression in violin plots were calculated with the Seurat function ‘FindAllMarkers’ using ‘bimod’ as the statistical test (b-c). Source data

• 

  Extended Data Fig. 6. VEGFA-VEGFB switch in hypertrophied cardiomyocytes. (a) Violin plots for the genes VEGFA and VEGFB in cardiomyocytes of cluster 0 only (log2 transformed and normalized UMI counts). (b) Mean expression of VEGFA and VEGFB in healthy (n=14) and hypertrophied (AS, n=5) hearts (log2 transformed and normalized UMI counts). (c) Vegfa and Vegfb mRNA levels for untreated vs. phenylephrine (PE, 200μM, 72h) treated primary neonatal rat cardiomyocytes identified by qRT-PCR (houskeeping gene: Hprt, n=7, displayed as fold change to untreated cardiomyocytes). (d) Left: representative immunofluorescence images of cardiac cryosections from non-hypertrophied hearts and hearts from patients with hypertrophic cardiomyopathy. Blue: DAPI, red: VEGF-B. Scale bar indicates 25 μm. Right: quantification for histological assessment of VEGF-B protein expression in n=4 non-hypertrophied vs. n=5 hypertrophied hearts. Adjusted p-values based on Bonferroni correction using all genes of the dataset to compare the expression in violin plots were calculated with the Seurat function ‘FindAllMarkers’ using ‘bimod’ as the statistical test (a). Data are shown as mean ± s.e.m. (b-d). Normal distribution was assessed…

• 

  Extended Data Fig. 7. Cross-talk between cardiomyocytes and other cardiac cells in hypertrophy. (a) Cardiomyocyte (=CM) ligand-receptor interactions with other cardiac cell-types in the healthy (blue) or aortic valve stenosis (=AS) heart (red). Shown are outgoing signals (number of interactions) from CMs to the respective interacting gene expressed by any other cell type. PC=pericytes; FB=fibroblasts; EC=endothelial cells; MC=monocytes; SMC= smooth muscle cells; LC=leukocytes; NLC=neuronal-like cells. (b) Circos plots visualizing the number of ligand receptor interactions analyzed using CellChat between the different cell types in cells from healthy hearts (left panel) and aortic valve stenosis (=AS) patients (right panel). Cardiomyocyte incoming signals are shown in red, outgoing signals are shown in blue. (c) Left: Total number of ligand-receptor interactions in hearts of healthy vs. AS patients assessed by using CellChat. Right: Total strength of ligand-receptor interactions in hearts of healthy vs. AS patients assessed with CellChat. (d) Dot plots for the number and strength of incoming and outgoing interactions for all cell types in healthy (left) vs. AS patients (right) assessed with CellChat.…

• 

  Extended Data Fig. 8. Regulation of the potential ephrin-B2 receptor EPHA4. (a) FeaturePlot for the expression of EPHA4 in snRNA-seq dataset (log2 transformed and normalized UMI counts). Color refers to the expression. (b) Linear regression for EPHB1 protein expression in cardiomyocytes (see Fig. 4i) and maximal wall thickness in mm (see Supplementary Data 3), shown for individual patients (Healthy n=14; AS n=5). The confidence bands reflect the standard error of the beta coefficients. (c) Violin plot for EPHA4 in hypertrophied (AS) vs. healthy cardiomyocytes (log2 transformed and normalized UMI counts). (d) Box plot showing expression of EPHA4 in cardiomyocytes in individual patients of respective groups (n=14 healthy, n=5 AS) (log2 transformed and normalized UMI counts; visualized as median and 25th and 75th percentiles with whiskers indicating maximal and minimal values). (e) Violin plot for EFNB2 in hypertrophied (AS) vs. healthy cardiomyocytes (log2 transformed and normalized UMI counts). (f) Phosphorylated EPHA4 protein in human cardiomyocytes treated with recombinant ephrin-B2 (10μg/ml, 15min), compared to non-treated CMs (=Ctr). Left: Quantification displayed as fold change vs Ctr (n=3;…