Characterization of p16-positive stromal cells in age-related cardiac disorders

Abstract

The heart undergoes structural alterations, including fibrosis and cardiomyocyte hypertrophy with age. These alterations are accompanied by functional decline, resulting in heart failure. Although cardiac stromal cells such as fibroblasts and macrophages are known to play a role in age-related cardiac changes, specific subsets of these cells may have a more pronounced effect on the development of these alterations. In this study, we analyzed p16, a marker of senescence,-positive (p16+) cells in cardiac fibrosis by single-cell RNA sequencing of aged p16-Tom mice, in which p16+ cells can be labelled in the presence of tamoxifen. We found that TGF-β signalling were significantly enriched in the transcriptome of p16+ fibroblasts compared with p16− fibroblasts. Besides, BMP4 was upregulated in p16+ fibroblasts. This activation potentially promoted the expression of collagen genes such as Col4a1 and Col5a3. Selective elimination of p16+ fibroblasts ameliorated cardiac fibrosis in aged p16–Col1a2–LRTD mice, with levels comparable to those observed in young mice. These findings suggest that p16+ fibroblasts play a critical role in age-associated cardiac fibrosis. We further demonstrate the potential use of transcriptomic signatures of p16+ fibroblasts to identify human fibroblast subsets causing age-related cardiac diseases such as dilated cardiomyopathy.

Graphical Abstract

Graphical Abstract.

In recent years, advances in medical technology and improvements in nutritional status have substantially extended human life expectancy, thereby enabling the realization of a long-lived society. Consequently, ‘aging’, which is characterized by a decline in physiological function, has emerged as a significant social concern. Aging is a phenomenon observed across virtually all organs and is associated with an increased risk of various age-related diseases. In heart, considerable number of cardiovascular diseases, including heart failure, are classified as age-related disorders (1). Even in the absence of overt disease, the aging heart exhibits functional and structural changes, such as impaired diastolic and systolic function, myocardial hypertrophy, fibrosis and reduced regenerative capacity (2).

Recent studies have suggested that the accumulation of senescent cells with age may drive systemic aging. When exposed to various stresses, cells express cell cycle inhibitors such as p16^INK4a and undergo irreversible cell cycle arrest, a process referred to as cellular senescence. Although senescent cells accumulate in relatively small numbers within tissues, they can impair overall organ function through the secretion of inflammatory and other bioactive molecules, collectively termed the senescence-associated secretory phenotype (SASP). The genetic or pharmacological elimination of senescent cells has been reported to ameliorate aging phenotypes in multiple organs, including the heart (3, 4). Previous studies have demonstrated that cardiac aging is influenced by stromal cells, including fibroblasts and macrophages. These cells contribute to cardiac aging through extracellular matrix remodelling and the secretion of inflammatory mediators (5–7). Notably, these phenotypic features overlapped with characteristic of senescent cells, suggesting a potential role for senescent cells in age-related cardiac remodelling. However, senescent cell populations are heterogeneous and exhibit organ-specific properties, and thus, there have been few studies that directly label and isolate senescent cells in the heart. Consequently, the cellular origins of cardiac senescent cells and the mechanisms by which they contribute to cardiac aging, including fibrosis, remain poorly understood. We previously established a transgenic mouse model (p16-Cre^ERT2/Rosa26-CAG-lsl-tdTomato, p16-Tom mice) that enables in vivo labelling of senescent cells using p16 gene expression as a marker (8). In this study, we employed single-cell RNA sequencing (scRNA-seq) of the heart using p16-Tom mice to directly characterize gene expression patterns of p16-positive (p16+) cells and to elucidate the contribution of senescent cells to cardiac aging.

Results

p16-positive stromal cells accumulate in aged murine heart

It has been reported that p16+ cells accumulate in the heart with aging (9). However, further research is necessary to determine the specific cell types present and their respective proportions. We investigated the accumulation of p16+ cells in the heart with age using p16-Tom mice (Fig. 1A). At 3 months of age, p16High senescent cells were rarely detected; however, a progressive accumulation of senescent cells was observed with aging (Fig. 1B). Based on cellular morphology, most p16+ cells manifested as non-cardiomyocytes. To further characterize these non-cardiomyocyte senescent cells, scRNA-seq was performed on 24-month-old aged p16-Tom mouse hearts (Fig. 1C and D). The cells were then classified into 19 clusters, and UMAP clustering revealed no significant structural disparities between p16-negative (p16−) and p16+. The expression level of tdTomato gene was markedly higher in p16+ cells across all clusters, thereby validating the efficacy of the fluorescence-based sorting (Fig. 1E). The largest cell population was constituted by fibroblasts, followed by macrophages and endothelial cells (Fig. 1F and G). Notably, the proportion of macrophages increased in p16+ cells compared with p16− cells.

Fig. 1.

Senescent stromal cells accumulate in mouse heart with age. (A) Immunostaining of p16High cells in mouse heart from indicated ages of p16-Tom mice. The tamoxifen was administrated two weeks before sample collection, and tdTomato-positive cells (p16High cells) were stained by using anti-RFP. Scale bar = 200 μm. (B) Quantification of p16High cell area (RFP-positive area) from the sections described in figure A. (N = 3 for each age) (C) The gating strategy of the tdTomato-positive cells from 24-month-old aged p16-Tom mouse. (D) UMAP showing cell clusters derived from one 24-month-old aged p16-Tom male mouse. There were 8551 p16-negative and 3700 p16-positive cells obtained from this dataset. (E) Violin plot showing the expression level of tdTomato in p16- cells and p16+ cells of indicated clusters. (F) Feature plots showing the expression patterns of marker genes of fibroblasts (Pdgfra), endothelial cells (Pecam1) and macrophages (Adgre1). (G) Cell type composition of p16- cells and p16+ cells.

p16+ fibroblasts showed high TGF-β signalling response and collagen gene expressions

We then focused on fibroblasts, the most abundant p16+ cell type, and a re-clustering fibroblast was performed exclusively (Fig. 2A). The fibroblasts were subdivided into nine clusters, and while the overall UMAP structure was largely preserved, clusters C4 and C7 were markedly reduced in p16+ cells (Fig. 2B). The clusters C0, C5 and C6, which exhibited a high proportion of p16+ cells, were designated as TomHigh clusters. In contrast, the remaining clusters were designated as TomLow clusters for the subsequent analysis. A differentially expressed gene (DEG) analysis was conducted between TomHigh and TomLow fibroblasts, which identified 114 upregulated and 162 downregulated genes in TomHigh fibroblasts (adj. P < 0.05, Log2FC > 0.3) (Fig. 2C). Among DEGs, we found several SASP factors, including Ccl2, Cxcl1, Cxcl14 and Cxcl16 (10, 11), were upregulated in the TomHigh compared with TomLow fibroblast clusters (Fig. 2D). Gene set enrichment analysis (GSEA) revealed a significant enrichment of the interferon alpha response pathway in TomHigh fibroblasts (Fig. 2E), suggesting enhanced interferon-mediated inflammatory responses in senescent fibroblasts. Gene ontology (GO) analysis further indicated upregulation of TGF-β–related pathways and extracellular matrix (ECM) organization in TomHigh fibroblasts (Fig. 3A). Given that TGF-β is a known promoter of tissue fibrosis (12, 13). Therefore, the activation of these pathways may contribute to the development of age-associated cardiac fibrosis. However, pathways related to ECM organization were also identified among downregulated pathways in TomHigh fibroblasts, suggesting potential alterations in ECM composition during ECM remodelling. Expression analysis of representative fibrosis-related genes revealed comparable expressions of major fibrillar collagen genes (Col1a1, Col1a2 and Col3a1) (14) across clusters, whereas Col4a1 and Col5a3 were preferentially expressed in TomHigh clusters C0, C5 and C6 (Fig. 3B). Additionally, Smoc2 and Crispld2, which have been previously demonstrated to be highly expressed in p16+ bladder fibroblasts (15), were also enriched in TomHigh clusters. The inflammatory chemokine Ccl2 exhibited particularly high expression in cluster C5, further supporting enhanced inflammatory activity.

Fig. 2.

Cluster proportions were changed in senescent fibroblasts. (A) UMAP showing clustering results of 5564 p16-negative and 2189 p16-positive fibroblasts obtained from aged p16-Tom mouse. (B) Population of each cluster in p16− and p16+ fibroblasts shown in figure A. (C) Volcano plot showing DEGs comparing TomHigh and TomLow fibroblast clusters. All DEGs were identified by adj. P < 0.05, Log2FC > 0.3. (D) Violin plots showing differentially expressed SASP factors in TomHigh and TomLow fibroblast clusters. (E) The GSEA plot showing the enrichment of interferon alpha response hallmark between TomHigh and TomLow fibroblast clusters. GSEA term was identified by adj. P < 0.05 using the B-H method.

Fig. 3.

TGF-β pathway was activated in senescent fibroblasts. (A) Dot plots visualizing GO terms enriched in up-regulated DEGs (above) and down-regulated DEGs (below) in TomHigh comparing with TomLow fibroblast clusters. (B) Violin plot showing expression levels of indicated genes in fibroblast clusters.

p16+ macrophages changed their chemokine expression patterns

We then examined characteristics of macrophages, whose proportion was increased among p16+ cells. A re-clustering of macrophages alone was performed (Fig. 4A). The macrophages were then subdivided into 5 clusters, with minimal changes in overall UMAP structure. According to the expression of Mrc1, an M2 macrophage marker (16), most of the macrophages were M2 macrophages (Fig. 4B). The cluster C0, which was enriched in p16+ cells, was designated as TomHigh cluster, while the cluster C1 and C4 were designated as TomLow clusters (Fig. 4C). DEG analysis was conducted between TomHigh and TomLow macrophages, which identified 303 upregulated and 339 downregulated genes in TomHigh macrophages (adj. P < 0.05, Log2FC > 0.3) (Fig. 4D). GSEA revealed a significant enrichment of the protein secretion pathway in TomHigh macrophages (Fig. 4E), suggesting an enhanced secretion of inflammatory mediators by p16+ macrophages. GO analysis revealed enrichment of immune cell chemotaxis-related pathways in TomHigh macrophages (Fig. 5A). Interestingly, chemotaxis-related terms were also detected among the downregulated pathways. We found that TomHigh and TomLow macrophages had different preferential patterns on these chemotaxis-related cytokine expressions (Fig. 5B). These findings suggest that p16+ macrophages in aged mouse hearts undergo substantial changes in the profile of secreted inflammatory factors.

Fig. 4.

Cluster proportions were changed in senescent macrophages. (A) UMAP showing clustering results of 931 p16-negative and 828 p16-positive macrophages obtained from aged p16-Tom mouse. (B) Feature plots showing Mrc1 expression in macrophages from aged p16-Tom mouse. (C) Population of each cluster in p16- and p16+ macrophages shown in figure A. (D) Volcano plot showing DEGs comparing TomHigh and TomLow macrophages clusters. All DEGs were identified by adj. P-value<0.05, Log2FC > 0.3. (E) The GSEA plot showing the enrichment of protein secretion hallmark between TomHigh and TomLow macrophage clusters. GSEA term was identified by adj. P < 0.05 using the B-H method.

Fig. 5.

Chemokine expression patterns were changed in senescent macrophages. (A) Dot plots visualizing GO terms enriched in up-regulated DEGs (above) and down-regulated DEGs (below) in TomHigh comparing with TomLow macrophages clusters. (B) Heatmap showing expression levels of up-regulated cytokines and down-regulated cytokines in TomHigh macrophages. DEGs were identified by adj. P < 0.05 and Log2FC > 0.3.

Based on these findings, we hypothesized that p16+ fibroblasts play a pivotal role in development of fibrosis and inflammation through the enhanced secretion of specific factors, and macrophages were the mediators of senescent fibroblasts. To investigate intercellular communication, CellChat analysis was performed among four groups: TomHigh and TomLow fibroblasts, and TomHigh and TomLow macrophages (Fig. 6A). This analysis revealed that the ANGPTL, BMP, CX3C and PTN signalling pathways were predominantly activated by TomHigh fibroblasts. A detailed examination of the expression patterns revealed that Angptl1, Angptl2, Itga1, Bmp4, Cx3cl1 and Ptn were particularly enriched in TomHigh fibroblasts (Fig. 6B). Bmp4, a member of the TGF-β superfamily, has been demonstrated to activate both Smad-dependent and Smad-independent signalling pathways (17). Previous studies reported its involvement in cardiac inflammation and fibrosis associated with myocarditis or cardiac hypertrophy (18, 19). This suggests the possibility of a critical role for Bmp4 in age-related cardiac fibrosis and enhanced TGF-β–related signalling in TomHigh fibroblasts. Given that there was no significant difference in TGF-β receptor signalling between TomHigh and TomLow fibroblasts. Therefore, Bmp4 may represent a dominant driver of TGF-β–related pathway activation in the aged heart. On the other hand, it has been mentioned that Bmp4 promotes M2 polarization of macrophages, which also explains the M2 macrophage infiltration in aged hearts (20). Additionally, Cx3cl1 is a chemokine that promotes macrophage migration and may play a role in age-associated macrophage infiltration into cardiac tissue (21).

Fig. 6.

BMP secretion from senescent fibroblasts was enhanced. (A) Heatmap showing signalling patterns between fibroblasts and macrophages calculated by using CellChat package. Secreted signals (left panel) and received signals (right panel) were quantified and visualized by using relative strength. (B) Violin plot showing expression levels of indicated genes involved in TomHigh fibroblasts specific outgoing signalling patterns. The ligand and receptor genes listed were referenced from the CellChat database.

Cardiac fibroblasts of DCM patients showed similar gene signatures of murine p16+ fibroblasts

Finally, to directly assess the contribution of senescent fibroblasts to cardiac fibrosis and macrophage infiltration, we conducted p16-Cre^ERT2/Col1a2-Dre^ERT2/Rosa26-CAG-LSL-RSR-tdTomato-2A-DTR (p16-Col1a2-LRTD) mice to specifically eliminate p16+ fibroblasts. Aged mice (24-month-old) exhibited significantly increased cardiac fibrosis compared with young mice (3-month-old); however, selective elimination of p16+ fibroblasts in aged mice markedly reduced fibrosis to levels comparable to those observed in young mice (Fig. 7A). Similarly, ablation of p16+ fibroblasts significantly reduced macrophage infiltration, as compared with littermate aged mice (Fig. 7B). To assess the translational relevance of these findings in the context of human disease, we employed the p16+ fibroblast gene signature derived from mouse hearts to human scRNA-seq data from patients with dilated cardiomyopathy (DCM) using AUCell analysis (22, 23). The fibroblasts extracted from the human dataset were classified into ten clusters, with clusters C2, C4 and C7 predominantly derived from healthy donors, and clusters C0, C1, C3, C5, C6 and C8 derived from DCM patients (Fig. 7C and D). AUCell analysis using genes upregulated in TomHigh fibroblasts revealed the lowest AUCell scores in donor-derived cluster C7. In contrast, cluster C5, which was derived from DCM, exhibited significantly higher AUCell scores. These results suggest that fibroblast populations in DCM hearts, particularly clusters C3 and C5, exhibit molecular characteristics similar to those of senescent fibroblasts compared with fibroblasts from healthy donors.

Fig. 7.

Elimination of senescent fibroblasts ameliorated age-associated cardiac changes. (A) Sirius Red staining of hearts from 3mo WT, 24mo p16^Ink4a-CreERT2/Rosa26-CAG-lsl-rsr-tdTomato-2A-DTR (p16-LRTD, littermate control), and 24mo p16^Ink4a-CreERT2/Col1a2-DreERT2/Rosa26-CAG-lsl-rsr-tdTomato-2A-DTR (p16-Col1a2-LRTD) mice. The representative images were shown in the left panels. All the aged mice were i.p. injected with tamoxifen (TAM, 80 mg/kg BW) every 2 days for two weeks, then with diphtheria toxin (DT, 25 μg/kg BW) every 2 days for three doses. The mice were sacrificed 2 weeks after the final dose of DT (N = 4). Scale bar = 100 μm. *P < 0.05, **P < 0.01, one-way ANOVA followed by Tukey’s test was used. (B) Immunostaining results of macrophages by using anti-F4/80 from the samples described in figure A. The representative images were shown in the left panels. (N = 4). Scale bar = 100 μm. (C and D) UMAP showing (C) clustering results and (D) sample sources of 12,595 and 37,128 fibroblasts obtained from two healthy donors and five DCM patients, respectively (23). (E) Violin plot showing AUCell scores of each cluster calculated with up-regulated DEGs in TomHigh fibroblasts. DEGs were identified by adj. P < 0.05 and Log2FC > 0.3.

Discussion

In age-related cardiac diseases such as cardiac fibrosis, inflammation, and heart failure, it has been reported that not only cardiomyocytes, but also stromal cells including fibroblasts, macrophages and endothelial cells play important roles (24–26). In particular, fibroblasts have been shown to promote extracellular matrix remodelling associated with aging and diseases. Additionally, they are also known to induce DNA damage accumulation in cardiomyocytes via TGF-β signalling (13). The present study focused on p16+ cells, a marker of senescent cells. We found increased expression of collagen genes such as Col4a1 and Col5a3 in p16+ fibroblasts. Moreover, the elimination of p16+ fibroblasts markedly improved age-associated cardiac fibrosis, with levels comparable to those observed in young mice. These findings suggest that p16+ fibroblasts play a central role in extracellular matrix remodelling in the aged heart. In addition, the TGF-β signalling pathway, which is known to induce upregulation of collagen gene expression, was enhanced in p16+ fibroblasts. This finding suggests a possible role of TGF-β signalling in the increased collagen expression. However, CellChat analysis revealed that secretion of the TGF-β family was more active in macrophages than in fibroblasts, and there was little difference in TGF-β–mediated signal reception between TomHigh and TomLow fibroblasts. In contrast, the BMP family, which belongs to the TGF-β superfamily and activates both Smad-dependent and Smad-independent pathways similar to TGF-β (17), was actively secreted by TomHigh fibroblasts and strongly received by fibroblasts. Notably, Bmp4 was highly expressed exclusively in TomHigh fibroblasts, suggesting that autocrine BMP4 signalling in p16+ cardiac fibroblasts potentiates TGF-β signalling and collagen gene expression. In addition to Bmp4, expressions of Angptl2, Smoc2 and Crispld2 were also elevated in TomHigh fibroblasts (15, 27) These genes have been reported as being upregulated in senescent cells, supporting the classification of TomHigh fibroblasts as a population enriched in p16+ cells.

Although, based on Mrc1 expression levels, we considered that macrophages in the aged heart were M2-like, previous studies have also indicated that M2-like macrophages could still express pro-inflammatory cytokines in an environment-dependent manner, particularly chemotaxis-related cytokines (28, 29). ScRNA-seq of macrophages revealed enhanced secretion of distinct pro-inflammatory cytokines in TomHigh and TomLow populations, making it difficult to determine the relationship between p16+ macrophages and macrophage infiltration associated with heart aging. However, given that senescent fibroblasts secrete BMP4, which promotes macrophage polarization toward the M2 phenotype (20), as well as CX3CL1, which induces macrophage migration, and that the elimination of p16+ fibroblasts substantially ameliorated macrophage infiltration, p16+ fibroblasts likely contribute significantly to age-associated macrophage infiltration. However, chemokines such as CCL secreted by TomHigh macrophages may partially contribute to macrophage recruitment.

Recently, several studies have attempted to extract senescent cells from bulk RNA-seq data using single-cell transcriptomic data (30–32). While these approaches have expanded our understanding by identifying senescent cells based on comprehensive gene expression patterns, they do not always fully align with analysis using p16. One potential explanation for these observations is that p16 expression levels are frequently low and may not be detected by RNA-seq even when the genes is expressed (30). In this study, we calculated AUC scores based on single-cell data obtained from p16-Tom mice. This approach allows for the assessment of p16 expression independent of RNA-seq detection limits. Using this approach, we identified fibroblasts with characteristics analogous to those of murine p16+ fibroblasts, which were specifically present in DCM. Despite the limited scope of the data from a single mouse, the results suggest that a subset of fibroblasts causing human DCM may share characteristics with murine p16+ fibroblasts.

In conclusion, the present study suggests that p16+ fibroblasts in the murine heart promote fibrosis by activating the TGF-β signalling pathway potentially via autocrine BMP4 secretion, which in turn leads to enhanced collagen gene expression. Furthermore, the role of p16+ fibroblasts in age-associated macrophage infiltration appears exceed that of macrophages themselves. Additionally, our findings demonstrate the potential of utilizing data from p16+ fibroblasts in the mouse heart to make inferences about cellular properties in human heart disease. These results suggest that p16+ fibroblasts play a significant role in cardiac aging, particularly fibrosis, in aged mice. Consequently, they may serve as a crucial therapeutic target for future anti-aging interventions.

Methods

Mouse experiment

Mice were housed in a temperature (23–25 °C), and humidity-controlled colony room, maintained on a 12-h light/dark cycle (08:00 to 20:00 light on), with standard food (CA-1, CLEA Japan), and water provided ad libitum. All animals were handled following the Guidelines for Animal Experiments of the Institute of Medical Science, the University of Tokyo, and the Institutional Laboratory Animal Care. p16-LRTD mice were generated by crossing p16^Ink4a-CreERT2 mice with Rosa26-CAG-lsl-rsr-tdTomato-2A-DTR mice. p16-Col1a2-LRTD mice were generated by crossing p16^Ink4a-CreERT2 mice with Rosa26-CAG-lsl-rsr-tdTomato-2A-DTR mice and Col1a2-2A-DreERT2-WPRE-pA mice. All p16-Col1a2-LRTD mice and littermate control, p16-LRTD mice, were i.p. injected with tamoxifen (TAM, 80 mg/kg BW) every two days for 2 weeks, then with diphtheria toxin (DT, 25 μg/kg BW) every 2 days for three doses at the designated time point. The mice were sacrificed 2 weeks after the final dose of DT.

scRNA-Seq library preparation

The mouse was sacrificed and systematically perfused with PBS to remove the blood cells. Heart was collected, minced, and incubated for 20 min in 50 μg/mL Liberase TH, 10 mM HEPES, 2 kunitz/mL DNase in RPMI at 37 °C and passed through 18G needles and 21G needles with a 20-min incubation in between. Cells were filtered through a 70 μm cell strainer and resuspended into RBC lysis buffer for 3 min at room temperature and washed with wash buffer. Cells were treated with an FcR blocking reagent (Miltenyi Biotec) on ice for 10 min, stained with Zombie Violet™ Fixable Viability Kit (BioLegend), washed, and sorted using FACS AriaIII (BD Biosciences).

scRNA-Seq analysis

For the scRNA-Seq library construction, we followed the manufacturer's instructions of Chromium Next GEM Single Cell 3′ Kit v3.1 (10X Genomics) and Chromium Controller (10× Genomics). The single-cell suspensions of Zombie Violet−/tdTomato− and Zombie Violet−/tdTomato+ cells were sorted from the aged heart lysates with 10,000 cells and 8000 cells, respectively. Library sequencing was performed on the DNBSEQ-G400RS (MGI Tech) with 150 bp paired-end reads. The Cell Ranger package (version 7.0.1) was used to process unique molecular identifiers (UMIs) and barcodes and align the transcripts to an mm10 mouse reference genome. After obtaining the feature-barcode matrix, we utilized Seurat package (v4.3.0) (33) in R language (v4.3.1) to process the quality control, clustering, dimension reduction, cell type annotation and identification of differentially expressed genes (DEGs). The thresholds of quality control included 1000 < nFeatures<8000, 5000 < nCounts<40,000 and mitochondrial counts ratio < 15%. Following quality control filtering, all single-cell transcriptomes were selected for further normalization, log transformation, dimensional reduction, and clustering. The DEGs between TomHigh and TomLow clusters were identified by using the FindMarkers function. The two-sided Wilcoxon rank-sum test followed by B-H method was conducted to calculate the adjusted P-values for each identified DEG with Log2FC > 0.3, adjusted P-values<0.05. Gene ontology (GO) analysis to identify the enriched terms of biological processes by packages clusterProfiler (v4.16.0) (34). The significance of all terms was identified by adjusted P-value < 0.05 using B-H method. AUCell analysis was performed to estimate the enrichment of DEGs in TomHigh fibroblasts for human data using AUCell (v1.24.0) (22). The cell–cell interactions analysis was performed by using CellChat (v2.1.2) (35) package. TomHigh or TomLow of fibroblasts or macrophages were transformed into the CellChat object, then followed the default analysis pipeline to process the ligand-receptor determination. The ‘Secreted Signalling’ reference ligand-receptor datasets were conducted in the inference.

Histological analysis

For histochemical and immunohistochemical staining, heart tissues were fixed in paraffin, sectioned, and stained with H&E or anti-F4/80 (abcam, ab6640) or anti-RFP (Rockland, 600-401-379) or Sirius Red. The stained tissue sections were observed with an Olympus BX51 microscope (Olympus) under bright-field illumination. To quantify the macrophage infiltration and fibrosis area in the heart, we acquired five images from each section and calculated the number of F4/80-positive cells and the amount of red-coloured area within each image.

Statistical analysis

For animal experiments, all the mice were randomly assigned to each group and independently followed the same age-dependent schedule in each experimental design. The sample sizes were not predetermined by pilot studies. Blind designs were employed in the analysis of fibrosis and macrophage infiltration. One-way analysis of variance (ANOVA) and a post hoc Tukey’s test was used for statistical analysis of the results obtained from histological experiment. Comparison between cluster C5 and C7 in human heart datasets was made by an unpaired two-tailed Student's t-test.