Single-cell transcriptomics reveal distinctive patterns of fibroblast activation in heart failure with preserved ejection fraction

Abstract

Inflammation, fibrosis and metabolic stress critically promote heart failure with preserved ejection fraction (HFpEF). Exposure to high-fat diet and nitric oxide synthase inhibitor N[w]-nitro-l-arginine methyl ester (L-NAME) recapitulate features of HFpEF in mice. To identify disease-specific traits during adverse remodeling, we profiled interstitial cells in early murine HFpEF using single-cell RNAseq (scRNAseq). Diastolic dysfunction and perivascular fibrosis were accompanied by an activation of cardiac fibroblast and macrophage subsets. Integration of fibroblasts from HFpEF with two murine models for heart failure with reduced ejection fraction (HFrEF) identified a catalog of conserved fibroblast phenotypes across mouse models. Moreover, HFpEF-specific characteristics included induced metabolic, hypoxic and inflammatory transcription factors and pathways, including enhanced expression of Angiopoietin-like 4 (Angptl4) next to basement membrane compounds, such as collagen IV (Col4a1). Fibroblast activation was further dissected into transcriptional and compositional shifts and thereby highly responsive cell states for each HF model were identified. In contrast to HFrEF, where myofibroblast and matrifibrocyte activation were crucial features, we found that these cell states played a subsidiary role in early HFpEF. These disease-specific fibroblast signatures were corroborated in human myocardial bulk transcriptomes. Furthermore, we identified a potential cross-talk between macrophages and fibroblasts via SPP1 and TNFɑ with estimated fibroblast target genes including Col4a1 and Angptl4. Treatment with recombinant ANGPTL4 ameliorated the murine HFpEF phenotype and diastolic dysfunction by reducing collagen IV deposition from fibroblasts in vivo and in vitro. In line, ANGPTL4, was elevated in plasma samples of HFpEF patients and particularly high levels associated with a preserved global-longitudinal strain. Taken together, our study provides a comprehensive characterization of molecular fibroblast activation patterns in murine HFpEF, as well as the identification of Angiopoietin-like 4 as central mechanistic regulator with protective effects.

Keywords: Angiopoietin-like 4; Fibroblast activation; Heart failure with preserved ejection fraction (HFpEF); Immune activation; Single-cell RNA sequencing.

Fig. 1

Study model and cell type assignment. A Murine HFpEF model characterization by ratio of heart weight to tibia length (HW/TL) and echocardiographic hallmarks (E/E’, global-longitudinal strain and LVEF), purple data points represent the animals used for single-cell RNA sequencing (scRNAseq). Statistical analysis performed by one-way ANOVA, bar graphs indicate mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001. ns = deemed not significant (p > 0.05), LVEF = left ventricular ejection fraction, w weeks. B Schematic summary of experimental setup for scRNAseq experiments using mice after 7 weeks of HFpEF or control diet. Created with BioRender.com. C UMAP embeddings of normalized scRNAseq data after processing and filtering. D Marker gene expression for cell type assignment. E Cell type composition of main cell types as mean percentage per group, compared between HFpEF and control mice. *p < 0.05, p values were calculated via label permutation. F Cosine distance ratios of highly variable genes between pseudobulked cell type profiles. Median between group distance is divided by median within group distance. G Representative Picrosirius-Red stainings of interstitial fibrotic fibers (arrowheads) and perivascular fibrosis (arrows) from control and different stages of HFpEF heart sections. Imaging performed in 594 nm (Picrosirius-Red) and 488 nm (autofluorescence) channels. White scale bars in the right bottom corner correspond to 100 μm

Fig. 2

Integrated atlas of cardiac fibroblasts from different disease models. A Schematic of the integrated murine HFpEF and HFrEF (AngII and MI) fibroblast studies. B + C UMAP embeddings of integrated fibroblasts, colored by disease (HF, Heart Failure) vs. control (B), study (C). D Overview of top cell state marker expression of integrated fibroblast states. E UMAP embeddings, showing the integrated fibroblast atlas colored by cell clusters, i.e. the integrated fibroblast states (IFS). Labels indicate possible fibroblast differentiations based on functional characterization. F Estimated pathway activities with PROGENy based on effect size (avg log2 fold change) of footprint genes in integrated fibroblast states. *PROGENy z score > 2. G Overrepresentation analysis of extracellular matrix related gene sets with markers of integrated fibroblast states. Hypergeometric test with Benjamini–Hochberg correction, *q < 0.05, **q < 0.01, ***q < 0.001

Fig. 3

Comparison and interpretation of fibroblast disease signatures from different heart failure models. A Comparing intersections of upregulated genes in different heart failure (HF) models. B Intersection quantification via Jaccard index. C Comparison of direction of regulation between studies. Pearson correlation was calculated between log fold change vectors of signature genes in pairwise comparisons. Each study comparison was based on the upregulated genes from the study on the x-axis. **p < 0.01. D Heatmaps of gene set overrepresentation in study specific fibroblast disease signatures. Hypergeometric test with Benjamini–Hochberg correction, *q < 0.01, **q < 0.001, ***q < 0.0001. E Estimated pathway activities with PROGENy based on effect size (log fold change) of footprint genes compared between HF models. F Expression values of selected fibrosis and inflammatory genes in individual fibroblasts in HFpEF (purple) and control (orange) mice. All genes were significantly upregulated (Wilcoxon test, adj. p value < 0.05). G Immunofluorescence images of collagen IV (red) and DAPI (blue) staining of left ventricular heart sections. Lower panels show magnifications of the areas marked by white boxes. White arrows indicate capillaries or larger blood vessels. Scale bars in the right bottom corner indicate 50 μm length. H Immunohistological staining of Angptl4 protein in left ventricular heart sections

Fig. 4

Decomposing Fibroblast disease signatures. A Schematic of different expression patterns in regard to cell states that could yield an upregulation of a disease signature. Compositional shifts by expanding cell number are distinguished from transcriptional shifts via uniform (state independent) or non-uniform (state dependent) upregulation of disease signatures. B Composition change of integrated fibroblast states (IFS) between control and heart failure per study. p values calculated via label permutation, *p < 0.05, **p < 0.01. C Overrepresentation analysis of disease-specific fibroblast signatures (x-axis) and top 100 IFS markers (y-axis). Hypergeometric test, *p < 0.05. D Gene set scores of study specific signatures (x-axis) were used to calculate the area under the receiver operator curve (AUROC, y-axis) between control and diseased cells within each IFS (color). E HFpEF signature expression dependent on IFS category by calculating the explained variance (eta² values) of gene-wise ANOVAs. Violin plots display normalized expression values of three genes with lowest (lower panel) and highest (upper panel) variance explained by cell state. F) Quantification of differences in state-dependent regulation of disease signatures across heart failure models. The ratio of the explained variance by IFS and disease class was calculated for each HF model and its disease signature. Wilcoxon test p values are shown. G) Explained variance (eta² values) by IFS on x-axis and explained variance by disease class (gene ~ disease class) on y-axis. Violet dots are part of the disease signature. H The ratio of explained variance by state and disease class of selected genes that were upregulated in all HF models. I Corroboration of murine fibroblast signatures in human myocardial samples. Human HFpEF and HFrEF studies were curated and top differentially upregulated genes were selected (y-axis). Gene set overlaps with fibroblast disease signatures from different study models (left panel) or fibroblast state marker (right panel) (hypergeometric test). AngII =angiotensin II model, HFpEF heart failure with preserved ejection fraction, MI myocardial infarction. q value = Benjamini–Hochberg-corrected p value, *q < 0.05, **q < 0.01, ***q < 0.001

Fig. 5

Macrophage engagement in HFpEF. A UMAP embedding of integrated and clustered macrophages from control and HFpEF mice. B Representative flow cytometry plots of Ly6C^high/low monocytes and macrophages (MΦ) (top row), monocyte-derived/resident MΦ (bottom row) in HFpEF vs control mice. Cells were gated on CD45⁺Lin⁺CD11b⁺ cardiac cells. C Quantification of flow cytometry results. Statistical analysis using t test, bar graphs indicate mean ± SD, n = 6/group *p < 0.05, ns not significant. D Ligand-Receptor network based on LIANA, receptors in fibroblasts shown in red, blue depicts ligands from macrophages. Node size visualizes effect size of upregulation in HFpEF mice, edge width visualizes HFpEF specificity (see methods). E Pearson correlation of top predicted ligands in HFpEF (from D) in NicheNet (left panel). Top NicheNet ligands and their regulatory potential with fibroblast target genes (right panel)

Fig. 6

Angptl4 improves diastolic dysfunction by reducing collagen IV deposition in vivo and in vitro. A In vivo study design comparing recombinant murine Angptl4 peptide (rANGPTL4, 200 ng in 50 µl NaCl) vs. NaCl control (0.9%, 50 µl) administration every second day i.p. for 5 weeks starting after 5 weeks of dietary induction. Murine HFpEF induction by 0.5 g/L L-NAME and 60% high-fat diet for 10 weeks in total. Echocardiography (echo) captured cardiac systolic and diastolic function at baseline, after 5 (prior to i.p. injection start) and 10 weeks. Created with BioRender. B Time course of diastolic function determined by E/E’ (PW Doppler velocity across the mitral valve (E) and peak tissue Doppler at the mitral valve annulus (E’) during early diastole). n = 10/10/9/9. C Comparison of the experimental groups after 10 weeks. n = 10/10/9/9. One-way ANOVA with Tukey correction for multiple comparison or Kruskal–Wallis test according to normality determined by Shapiro–Wilk test. p values < 0.09 shown above bars, p < 0.05 defined as statistically significant. HW/TL heart weight/ tibia length, LA left atrium, LVEF left ventricular ejection fraction, LWDd left ventricular lateral wall diameter in end-diastole from short axis views D Comparison of E/E’ after 10 weeks in HFpEF mice either treated with control or rANGPTL4. n = 9/9, unpaired t test, p value shown above bar. E Experimental design (left panel) of stimulating human ventricular cardiac fibroblasts (cFBs) in vitro with 2 µg/ml human recombinant ANGPTL4 or PBS for 24 h. Resulting mRNA levels (right panel) were determined by qPCR and values indicate fold change relative to control mean. n = 12/12, unpaired t test or Mann–Whitney test according to normality determined by Shapiro–Wilk test, p values shown above bars. Created with BioRender. F Representative immunofluorescence stainings of collagen IV 1:200 (pink) and DAPI (blue). Whole heart long-axis cryo-sections depicted in the top row with white boxes indicating magnifications shown in bottom row and respective scale bars. G Quantification of F by scanning whole heart sections using a slide scanner and semi-automated analysis of the whole left ventricular (LV) tissue by normalizing the collagen IV positive LV area to total LV area using QuPath. One-way ANOVA, p values < 0.05 shown above bars

Fig. 7

Plasma ANGPTL4 is increased in HFpEF patients. A Circulating levels of ANGPTL4 in human plasma samples of HFpEF and age-matched controls measured by sandwich ELISA. n = 19/20, Mann–Whitney U test, *p < 0.05. B) ANGPTL4 plasma levels in relation to NYHA functional class of all recruited patients. ANOVA, p value < 0.05, n = 10/21/3 in baseline and n = 11/18/5 in 12 months (12 M) follow-up. C Correlation of clinical parameters to ANGPTL4 circulating levels in all patients (control and HFpEF) and D as subanalysis only in HFpEF patients using simple linear regression. p-val indicates uncorrected p value. hs high sensitivity, LA left atrial, MFU months follow-up, SVES supraventricular extrasystoles. Plots in (A, B) display mean ± SD