Mapping the cardiac vascular niche in heart failure

Abstract

The cardiac vascular and perivascular niche are of major importance in homeostasis and during disease, but we lack a complete understanding of its cellular heterogeneity and alteration in response to injury as a major driver of heart failure. Using combined genetic fate tracing with confocal imaging and single-cell RNA sequencing of this niche in homeostasis and during heart failure, we unravel cell type specific transcriptomic changes in fibroblast, endothelial, pericyte and vascular smooth muscle cell subtypes. We characterize a specific fibroblast subpopulation that exists during homeostasis, acquires Thbs4 expression and expands after injury driving cardiac fibrosis, and identify the transcription factor TEAD1 as a regulator of fibroblast activation. Endothelial cells display a proliferative response after injury, which is not sustained in later remodeling, together with transcriptional changes related to hypoxia, angiogenesis, and migration. Collectively, our data provides an extensive resource of transcriptomic changes in the vascular niche in hypertrophic cardiac remodeling.

Fig. 1. Fate tracing of the vascular niche after TAC induced injury.

a Experimental timeline of genetic labeling of specific cell types using tamoxifen, 3-week washout period, surgery time point (transverse aortic constriction (TAC) or sham) and duration until harvest. The TAC procedure is illustrated. b Overview of genetic Cre driver mouse lines and animal numbers per group. c Heart weight to tibia length ratio between TAC and sham groups. Each dot represents an individual mouse (mean ± SD; number of independent mice per group is displayed in b; *p < 0.05; unpaired t-test, two-sided). Source data are provided as a Source Data file. d Representative pictures of picro-sirius red staining for TAC (14 and 28 days) and sham hearts, one per genotype (scale bar 50 µm). e After extraction murine hearts were cut transversely to separate the upper part for imaging and lower part from basal region to apical area for dissociation. Prior to scRNA-seq samples were pooled per group, time point and genotype. Single-cell suspensions were DAPI stained, followed by FACS sorting for DAPI⁻ and tdTom⁺ cells. Fourteen individual single-cell cDNA libraries were generated using the 10x genomic platform. f Representative stainings visualizing localization of fate traced cell types (tdTomato, red) in TAC and sham (scale bars 50 µm). The Figure was partly generated using Servier Medical Art, provided by Servier, licensed under a Creative Commons Attribution 3.0 unported license.

Fig. 2. Single-cell sequencing of fate traced vascular niche cells.

a UMAP representation of 77,602 cells based on the integration of all 14 individual scRNA-seq libraries. Dot colors indicate the original fate traced genotype of each cell. Major cell type annotations are included. b Dot plot of top five expressed marker genes per cluster. Dot size refers to proportion of cells expressing the gene per cluster. c UMAP embedding as in a, split up according to lineage tracing genotype (sham and TAC conditions combined per genotype). d Bar graph showing contribution of the different lineage tracing genotypes to the major cell type clusters. e UMAP embedding labeling cells according to underlying condition (sham/TAC). f UMAP plots of selected genes used in the induced lineage tracing models, depicting their expression pattern. Scaled gene expression is indicated by color.

Fig. 3. Subclustering of fibroblasts and subsequent analysis of ECM expression.

a UMAP embedding showing extracellular matrix (ECM) score in the overall integrated dataset. b UMAP of subsequent subclustering of 54905 fibroblasts, consisting of PdgfrβCreER, Gli1CreER, Col1a1CreER derived cells. Cluster annotation: fibroblast 1 (Fib1), Atf3 fibroblasts (Atf3-Fib), fibroblast 2 (Fib2), fibroblast 3 (Fib3), ECM fibroblast (ECM-Fib), interferon fibroblast (IntFib). c Normalized contribution of each fate traced genotype to the corresponding cluster. d Dot plot of top five expressed marker genes per cluster. Dot size refers to proportion of cells expressing the gene per cluster. e Top three representative gene sets enriched (hypergeometric test, one-sided) in marker genes per fibroblast subtype. Dot size refers to overlap of tested genes and gene set (precision). Full list in Supplementary Data 1. f Normalized proportion of fibroblasts per subcluster from sham and TAC (*: false discovery rate (FDR) <0.05 and absolute log2 fold change >0.58, see also Supplementary Fig. 4e). g Violin plot of ECM scores per condition and fibroblast cluster (all fibroblast Cre drivers combined). Integrated boxplots show center line as median, box limits as upper and lower quartiles. All clusters show significant difference between TAC and sham, ECM-Fib show significant differences to all other subtypes (*p < 0.001, two-sided Wilcoxon rank sum test, unpaired, Bonferroni adjusted p value). h Dot plot of expression scores for collagen subtypes per fibroblast cluster separated by condition and time point. Dot size refers to proportion of cells expressing the gene set per cluster. i Heatmap of top 100 non-ECM related genes correlating in gene expression to the ECM score of all fibroblasts. Fibroblasts are sorted by ECM score on the x-axis from low-to-high score. Genes are listed on the y-axis, color indicates gene expression level. j Top 20 genes sets enriched (hypergeometric test, one-sided) in the top 100 genes correlating to ECM score from i. Dot size refers to overlap of test genes and gene set (precision). For details on statistics and reproducibility, see “Methods”.

Fig. 4. Mapping transcriptomic changes in fibroblast subtypes of highest ECM producing fibroblast, ECM-Fib.

a Hierarchical clustered gene expression heatmap of differentially expressed genes (DEG) analyzed per cluster and condition, combined with gene set enrichment analysis. DEG were identified using MAST (only upregulated genes with adjusted p value <0.01 and logFC > 0.3). Top five enriched gene sets (hypergeometric test, one-sided) per gene cluster on the right. Dot size refers to overlap of tested genes and gene set (precision). b DEG of ECM-Fib comparing sham and TAC 14 days visualized in a volcano plot, displaying cluster specific (red) and non-specific (blue) DEG. c Violin plot comparing Thbs4 expression across all fibroblast subtypes per condition (*: adjusted p value <0.01, differential gene expression analysis by MAST). d Confocal XY scans of entire cross sections of PdgfrβCreER;tdTomato hearts stained for THBS4 (scale bars 500 µm, inserts 50 µm). e Signal pathway activity prediction of fibroblast subtypes based on pathway responsive genes (PROGENy). Color indicates relative predicted activity per pathway. f Transcription factor (TF) activity prediction based on TF regulons (DoRothEA) for fibroblast subtypes. Color indicates relative predicted activity per TF. g Correlation of TF activity scores to ECM score. Top 10 highest correlating TF displayed. Color indicates relative predicted activity per TF. Fibroblasts sorted by ECM score from low to high on the x-axis. h Violin plot of predicted TEAD1 transcription factor activity across fibroblast subtypes (all conditions combined). Integrated boxplots show center line as median, box limits as upper and lower quartiles. ECM-Fib shows significant differences to all other subtypes (*p < 0.001, two-sided Wilcoxon rank sum test, unpaired, Bonferroni adjusted p value). i Schematic overview of cardiac fibroblast cell line generation. j Bar graphs of relative gene expression of Tead1, Acta2 and Col1a1 measured by RT-qPCR. Data points represent normalized expression by 2^–∆∆Ct method (mean ± SD, n = 3 independent experiments per group, *p value < 0.05; **p value < 0.01; ***p value < 0.001; ns not significant; one-way ANOVA with Tukey’s post hoc). Source data are provided as a Source Data file. For details on statistics and reproducibility, see “Methods”. The Figure was partly generated using Servier Medical Art, provided by Servier, licensed under a Creative Commons Attribution 3.0 unported license.

Fig. 5. Analysis of potential fibroblast differentiation trajectories.

a Velocities derived from the dynamical model for subclustered fibroblasts (TAC day 28) visualized as streamlines in a UMAP embedding. b UMAP embedding of fibroblast, separated per condition with calculated latent time (from scVelo). c Supervised trajectory from Fib1 to ECM-Fib estimated with Monocle 3. Color indicates calculated pseudotime. d Density plots for cell density per pseudotime of different groups of fibroblasts. Upper panel, Fib1 and ECM-Fib density per pseudotime, lower panel density of all fibroblasts separated by condition. e Heatmap of genes with a changing expression pattern over the calculated pseudotime in c and associated pathway activities. The heatmap was clustered into five gene clusters by hierarchical clustering and the gene clusters were sorted according to pseudotime. Enriched pathways were mapped with EnrichR using Bioplanet pathways database as a resource. Dot color indicates significance (−log10(p value)), dot size refers to overlap of tested genes and gene set (count). f Smoothened gene expression for Cald1, Mical2, Palld, Col4a2 and Col3a1 over pseudotime (x-axis) separated by condition: Sham, TAC 14 days and TAC 28 days (gray error band: 95% confidence interval). g Ligand receptor interactions based on CrossTalkeR between selected subclusters defined in this study separate for each time point. Color indicates weight of interaction. h Bar graph showing changes in interactions, comparing sham and TAC (left TAC 14 days, right TAC 28 days). Increase in interactions indicated in red, decrease in blue. For details on statistics and reproducibility, see “Methods”.

Fig. 6. Exploration of TAC induced effects on cardiac EC.

a UMAP of 14,595 subclustered Cdh5CreER;tdTomato fate traced endothelial cells (EC) after integration from all conditions. Cluster annotation: Capillary (CapEC), capillary artery (CapA-EC), capillary vein (CapV-EC), stressed (StrEC), angiogenic (AngEC), interferon (IntEC), artery (ArtEC), lymphatic (LymEC), cycling (CyclEC), DNA replicating (RepEC) and endocardial (EndoEC). b Left, gene expression heatmap of 20 marker genes per EC atlas subtype (x-axis). Right, Cdh5 lineage derived dataset annotation and subcluster proportion. c Top five representative gene sets enriched (hypergeometric test, one-sided) in marker genes per EC subtype. Dot size refers to overlap of tested genes and gene set (precision). Full list in Supplementary Data 1. d Top five expressed marker genes per cluster. Dot size refers to proportion of cells expressing the gene per cluster. e Violin plot of Npr3 and Vwf expression in EC subtypes. f UMAP visualization of Npr3 expression. g Normalized proportion of EC per subcluster (*: false discovery rate (FDR) <0.05 and absolute log2 fold change >0.58, see also Supplementary Fig. 8d). h Cardiac Ki-67 quantification, values are expressed as percentage Ki-67⁺-tdTom⁺ of all tdTom⁺ cells (mean ± SD; n = 3 sham, n = 3 TAC 14 days, n = 4 TAC 28 days, independent replicates; *p value <0.05; one-way ANOVA). Source data are provided as a Source Data file. Right, representative images (scale bar 50 µm). i TF activity prediction based on TF regulons (DoRothEA) for selected EC subtypes. j Signal pathway activity prediction of selected EC subtypes based on pathway responsive genes (PROGENy). k Gene expression heatmap of hierarchical clustered, differentially expressed genes (DEG) combined with gene set enrichment analysis (GSEA). DEG were identified using MAST (only upregulated genes with adjusted p value <0.01 and logFC >0.3). Right, top five enriched gene sets (hypergeometric test, one-sided) per gene cluster. Dot size refers to overlap of tested genes and gene set (precision). l UMAP visualization of positive cell migration regulation scores. m GSEA of changing interactions between sham and TAC 28 days involving ArtEC. Color indicates up/down significance odds. For details on statistics and reproducibility, see “Methods”.

Fig. 7. Mural cell heterogeneity and changes in TAC induced cardiac remodeling.

a UMAP of 7309 subclustered mural cells after integration of PdgfrβCreER, Myh11CreER, Ng2CreER derived cells from all conditions. Cluster annotation: pericyte 1 (Peri1), pericyte 2 (Peri2), vascular smooth muscle cell (VSMC 1), VSMC 2, stressed VSMC (StrVSMC), stressed pericyte (StrPeri), interferon pericyte (IntPeri) and Schwann cells (Sw). b Dot plot of top five expressed marker genes per cluster. Dot size refers to proportion of cells expressing the gene per cluster. c UMAP embedding showing gene expression of selected marker genes: Myh11, Acta2, Notch3, Kcnj8, Abcc9, Colec11. d Top two representative gene sets enriched (hypergeometric test, one-sided) in marker genes per mural subtype. Dot size refers to overlap of tested genes and gene set (precision). Full list in Supplementary Data 1. e Normalized contribution of each fate traced genotype to the corresponding cluster. f Normalized proportion of mural cells per subcluster from sham and TAC condition samples (*: false discovery rate (FDR) <0.05 and absolute log2 fold change >0.58, see also Supplementary Fig. 10f). g Confocal immunofluorescence image of Myh11CreER tagged TAC heart stained for THBS4 (scale bar 50 µm). h UMAP embedding with ECM score per cell. i Violin plot of summarized ECM scores per condition and subcluster (all three fate traced fibroblast genotypes combined). Integrated boxplots show center line as median, box limits as upper and lower quartiles. Peri1 and Peri2 show a significant difference between sham and TAC (*p < 0.001, two-sided Wilcoxon rank sum test, unpaired, Bonferroni adjusted p value). j Top significant enriched gene sets associated with either shared or unique differentially expressed genes (DEG) per subtype. Dot size refers to overlap of tested genes and gene set (precision). k Signal pathway activity prediction of selected EC subtypes based on pathway responsive genes (PROGENy). l Gene set enrichment analysis (GSEA) of changing interaction involving Peri1 and Peri2. Color indicates up- or down significance odds. For details on statistics and reproducibility, see “Methods”.