3.1 ScRNA-seq analyses of human hearts
To characterise non-cardiomyocytes in human hearts, we first generated a scRNA-seq dataset by integrating two groups of human heart samples (Control: normal heart; HF: Heart failure) and obtained single-cell transcriptomes for 22,537 cells after strict quality control. We defined nine major cell types: fibroblasts (DCN+, PDGFRA+), endothelial cells (CDH5+), cardiomyocytes (ACTA1+), T/NK cells (CD3G+, NKG7+), macrophages/monocytes/dendrtic cells (CD68+), B cells (CD79A+), neuronal cells (PLP1+), neutrophils (S100A8+), and smooth muscle cells (ACTA2+), based on their respective molecular signatures (Fig. 1a, Supplementary Figure S1a-c). Furthermore, as shown in Fig. 1b, we found that fibroblasts were the largest group of the non-cardiomyocyte cell types. The proportion of fibroblasts was obviously higher in the heart failure group (56.49%) than in the control group (42%).
Given the above results and the vital role of fibroblasts in fibrosis [3, 6], in the subsequent experiments, fibroblasts from the two groups (10,424 cells) were extracted and partitioned into distinct subtypes. The dispersion of the fibroblast clusters in the Uniform Manifold Approximation and Projection (UMAP) plots indicated a high degree of fibroblast heterogeneity (Supplementary Figure S1d); fibroblasts from both groups were partitioned into 13 distinct subtypes (Fig. 1c, Supplementary Figure S2a-c). Moreover, we found that the expression levels of fibroblast genes (POSTN, COLLAGEN 1A1, COLLAGEN 1A2, THBS4, VIMENTIN, and FN1) were significantly higher in the heart failure group than the control group (Supplementary Figure S2d). Additionally, as shown in the volcano plot in Supplementary Figure S2e, expression of the active fibroblast marker POSTN was increased in the heart failure group compared with the control group (Supplementary Figure S2e).
Surprisingly, cluster 3 showed elevated expression levels of multiple progenitor cell markers, including CD34, PI16, CD248 and NT5E, which indicate multi-lineage differentiation [24, 28–29] (Fig. 1c). In addition, gene ontology (GO) enrichment analysis showed that vasculature development and differentiation were highly enriched in cluster 3, indicating that this sub-cluster of fibroblasts may exhibit specific functions during cardiac homeostasis and after injury (Fig. 1d). Therefore, we investigated the potential developmental relationships among these subclusters of fibroblasts. Pseudotime(s) were visualised using principal curves representing the trajectories of fibroblast differentiation across the steady-state atlas, with cluster 3 set as the root (Fig. 1e). Fibroblasts were ordered along a trajectory, and cells in different states were identified (Fig. 1e-f). Using the Monocle R package [30–31], we defined the different cellular states along the trajectory. Interestingly, different subgroups of fibroblasts arose and were ordered from a certain point; a split view of the trajectory analysis of each cluster is shown in Supplementary Figure S1e. Further trajectory analysis clearly showed that the expression of specific MSC markers (including CD34, ENG, NT5E, and PI16) significantly changed with the state, corresponding to the pseudotime trajectory (Fig. 1g-j).
We further measured the expression of these biomarkers at the RNA level using quantitative PCR (Fig. 1k, primer sequences used for qPCR including human and mice are listed in Supplementary Table S1) and found that the expression levels of VIMENTIN, DDR2, POSTN, PDGFRA, COLLAGEN 1A, COLLAGEN 3A, and FN1 were significantly higher in the heart failure group than in the control group. In addition, immunofluorescence staining showed increased expression of VIMENTIN, DDR2, and PDGFRA in the heart failure group compared with the control group (Fig. 1l, Supplementary Figure S3a-b).
3.2 ScRNA-seq analyses of murine hearts.
To delineate the heterogeneity of mouse fibroblasts during the development of heart failure, we collected scRNA-seq datasets of total non-cardiomyocytes from a public repository (GSE198833). After applying strict quality control filters, we obtained single-cell transcriptomes for 12,157 cells by integrating two groups (sham and TAC), and graph-based clustering was performed to classify these cells according to their gene expression profiles using UMAP plots (Supplementary Figure S4a). We defined 7 major cell types: fibroblasts (Dcn+, Pdgfra+, Ddr2+), lymphocytes (Cd3g+, Cd79a+), neutrophils (S100a8+), endothelial cells (Cdh5+, Pecam1+), neuronal cells (Plp1+), smooth muscle cells (Acta2+, Myh11+) and Mac_Mo_DC cells (Cd68+, Adgre1+), based on established canonical markers (Supplementary Figure S4a-d). Compared with the sham group, we found that the proportion of fibroblasts increased following TAC (sham: 53.74%; TAC: 79.99%) (Fig. 2a). We distinguished these fibroblasts according to characteristic biomarkers above and displayed the distribution of each subpopulation. Based on transcriptome characteristics, fibroblasts were further analysed to investigate their heterogeneous features, and 10 subclusters were identified between the sham and TAC groups (Fig. 2b, Supplementary Figure S4e and S5a). The dispersion of the fibroblast clusters in the UMAP plots indicated a high degree of fibroblast heterogeneity (Fig. 2b). Similar to the previous human results, we also found a group of fibroblast subsets with high expression of multiple progenitor cell markers, including Ly6a, Cd34, Cd248, Thy1 and Pi16 (Fig. 2c). These fibroblast subpopulations mentioned above showed an elevated capacity for extracellular matrix organisation, extracellular structure organisation and ossification, which are all involved in fibrosis (Fig. 2d).
In addition, GO enrichment analysis showed that extracellular structure organisation, positive regulation of blood vessel development, and vasculogenesis were highly enriched in cluster 7 (Supplementary Figure S5b), indicating that this subcluster of fibroblasts may perform specific functions during heart failure. We then conducted pseudotime trajectory analyses for each subcluster of fibroblasts, as for human samples, visualising the pseudotime(s) using principal curves representing trajectories of fibroblast differentiation across a steady-state atlas with cluster 7 set as the root. Fibroblasts were ordered along a trajectory, and cells in different states were identified (Fig. 2e, Supplementary Figure S5d). Surprisingly, further trajectory analysis clearly showed that the expression of specific MSC markers, including Cd34, Ly6a, and Pi16, significantly changed with the state, corresponding to the pseudotime trajectory (Fig. 2f-h, Supplementary Figure S5c and S5e).
Combining the human and mouse data, we identified fibroblast subsets in the heart with high expression levels of MSC markers and a progenitor cell-like phenotype, which can transdifferentiate into multiple cell types, especially fibroblasts, and perform specific functions in cardiac fibrosis. Furthermore, we performed TAC procedures in C57BL/6 mice to establish a heart failure model; mice were analysed after five weeks (Fig. 2i). Immunostaining of Vimentin, Ddr2, Postn and Pdgfra revealed a significant increase in fibroblasts during the development of murine heart failure (Fig. 2j, Supplementary Figure S6a-c). Given the high proportion and heterogeneity of fibroblasts in both human and murine hearts as revealed by scRNA-seq analyses and the comprehensive differentiation profile of MSC markers such as Sca1 shown by trajectory analyses, we next explored the role of Sca1+ cells in heart failure.
3.3 Sca1+ cells can differentiate towards fibroblasts in vivo
As shown in Supplementary Figure S6d-e, by analysing the correlation between human and mouse single-cell datasets, we found that the cellular composition of mouse and human hearts was very similar, as was the high expression of MSC markers (CD34, PI16, CD248) in fibroblast subsets. Thus, in order to further investigate the specific fibroblast subset expressing MSC markers in injury-induced cardiac fibrosis and the relationship between this fibroblast subset and other fibroblasts in vivo, we used an inducible genetic stem/progenitor cell lineage tracing mouse model (Sca1-CreERT2;Rosa26-tdTomato) (Fig. 3a). After genotyping (primer sequences used for genotyping in this study are listed in Supplementary Table S2), mice were treated with five doses of tamoxifen (0.15 mg/g body weight) over the course of one week to induce tdTomato (tdT) labelling of Sca1+ cells. We performed TAC surgery approximately two weeks after the last tamoxifen dose avoiding the potential side effect of tamoxifen (Fig. 3b), and five weeks postoperatively echocardiography was performed on each mouse to comprehensively evaluate cardiac function. Echocardiography results demonstrated poor average cardiac function in mice that underwent TAC surgery (Fig. 3c), as demonstrated by a lower EF (P = 0.0001) and left ventricular FS (P = 0.0001) than the sham group (Fig. 3d-e). In addition, histology confirmed that the volume of the mouse heart and the thickness of the myocardial wall were significantly greater in mice from the TAC group than in those from the sham group (Fig. 3f). TdT+ positive cells increased in TAC heart samples, and the expression of characteristic fibroblast markers such as VIMENTIN, DDR2, POSTN, and PDGFRA also significantly increased compared with those from the sham group. There was also greater co-staining of tdT+ with these fibroblast markers in TAC heart samples than in samples from the sham group (Fig. 3i-j, Supplementary Figure S7a-b), which was statistically significant (Fig. 3g-h, Supplementary Figure S7c-d).
Ablation of target cells by inducing intracellular expression of DTA is a well-established mouse model [32]. To further clarify the role of Sca1+ cells in myocardial remodelling in response to TAC injury, we used the DTA system to deplete Sca1+ cells to determine their role in myocardial fibrosis. We generated Sca1-CreERT2; R26-eGFP-DTA mice which were then pulsed with tamoxifen to induce recombination for DTA expression in Sca1+ cells. TAC was performed in Sca1-CreERT2; R26-eGFP-DTA mice treated with tamoxifen. As shown in Fig. 3k, depletion of Sca1+ cells significantly affected heart function, as Sca1-CreERT2;R26-DTA mice had better EF and FS than Sca1-CreERT2;R26-tdTomato mice (Fig. 3l-m). We also observed a reduction in the expression of fibroblast markers along with tdT+ in the cardiac cells (Fig. 3p-q, Supplementary Figure S7e-f), which was statistically significant (Fig. 3n-o, Supplementary Figure S7g-h). This suggests that Sca1+ cells have the ability to differentiate into fibroblasts in vivo during cardiac fibrosis, and that depletion of Sca1+ cells not only affects their transformation, but also the degree of cardiac fibrosis.
3.4 Sca1 + cells which differentiate into fibroblasts to accelerate fibrosis after injury have a non-bone marrow source
We investigated the exact source of these special clusters of fibroblasts and whether they originated from BM or a non-BM source. Therefore, we generated two types of chimeric mice using a bone marrow transplantation (BMT) model. One was BMTSca1→WT, where bone marrow cells from Sca1-CreERT2;R26-tdTomato mice were transferred to irradiated wild-type C57BL/6J mice via a tail vein injection. The other was BMTWT→Sca1, which reversed this combination (Fig. 4a). After a recovery period of approximately two weeks, all chimeric mice were treated with tamoxifen by oral gavage and subjected to myocardial remodelling by transverse aortic constriction (TAC). Successful bone marrow reconstruction was indirectly confirmed by a control group of recipient mice all died within one week of irradiation without BMT.
After 5 weeks, heart samples were collected from all chimeric mice and immunostaining was performed to observe the extent of myocardial fibrosis. In the BMTSca1→WT chimeric mouse model, almost no tdT+ cells were detected, let alone co-expressing cells among the abundant cardiac fibroblasts. However, in the BMTWT→Sca1 chimeric mouse model, a large proportion of tdT+ cells co-stained with fibroblast markers such as VIMENTIN, DDR2, POSTN, and PDGFRA (Fig. 4b-e). This suggests that this cluster of cells participates in fibroblast regeneration. Thus, our data demonstrate that the crucial group of Sca1+ fibroblasts that respond to myocardial fibrosis is derived from a non-BM source.
3.5 Sca1+ cells can differentiate towards fibroblasts in vitro
We then investigated whether Sca1+ cells could differentiate into fibroblasts in vitro. We harvested hearts from Sca1-CreERT2;R26-tdTomato mice that had completed tamoxifen treatment, sorted the Sca1+ cells, and cultured them in complete stem cell culture medium. Next, we treated them with connective tissue growth factor (CTGF), which is sufficient to differentiate MSCs into fibroblasts [33]. Immunofluorescence staining showed that tdT-positive cells co-stained for Vimentin, Ddr2, Pdgfra, and Postn, indicating that Sca1+ cells can differentiate into fibroblasts in vitro (Fig. 5a-d, Supplementary Figure S8a-d). Furthermore, quantitative PCR analysis showed that the expression levels of characteristic fibroblast markers significantly increased after treatment with CTGF or transforming growth factor-β for 3, and 5 days in vitro (Fig. 5f, Supplementary Figure S8f), which was consistent with the results of western blotting (Fig. 5e, 5g, Supplementary Figure S8e and S8g). Collectively, these results confirmed that Sca1+ cells have the potential to differentiate into fibroblasts at the molecular level.
3.6 Role of Wnt4 activation in the differentiation of Sca1+ cells
As a result of the above data, we explored in detail how Sca1+ cells differentiate into fibroblasts. Therefore, we performed bulk RNA-seq analyses on control and CTGF-treated cardiac fibroblasts from Sca1+ lineage-tracing mice. Using differential gene expression profiling (Fig. 6a, Supplementary Figure S9a), we examined genes related to the Wnt signalling pathway and identified Wnt4, which has recently been associated with the development and differentiation of many cell types [19]. We then knocked down Wnt4 in Sca1+ cardiac fibroblasts using small interfering RNA (siRNA). The primer sequences used are listed in Supplementary Table S3 and the verification results are shown in Supplementary Figure S9c. Knocking down Wnt4 led to a significant reduction in the expression of fibroblast markers such as VIMENTIN, DDR2, POSTN, and PDGFRA (Fig. 6b-d). Considering the expression of PDGFRA following FB-inducing factors treatment (Fig. 5e-g, Fig. 6c-d, Supplementary Figure S8e-g), we suggest that Sca1+ cells can differentiate into PDGFRA-expressing cells involved in fibrosis, possibly via Wnt4.
Tentatively, we next knocked down Pdgfra in Sca1+ cardiac fibroblasts using siRNA, and cultured these cells with or without CTGF, and added the groups of Pdgfra deficient by siRNA. Surprisingly, bulk RNA-seq analyses revealed an updated series of Wnt pathway-related differentially expressed genes, including Wnt4 (Fig. 6e). The genes that were differentially expressed among the four groups are shown in Supplementary Figure S9b. The verification results, including those of other primary differentially expressed genes are shown in Supplementary Figure S9d. The Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses, which revealed a tendency for cardiac remodelling, are shown in Fig. 6f. Furthermore, RNA expression levels of the fibroblast markers Ddr2, Collagen I, Collagen3a, Postn, Cd248, and Fn1 were significantly reduced when Pdgfra was knocked down (Fig. 6g). To confirm this result, overexpression of Pdgfra using plasma was performed, the primer sequences are shown in Supplementary Figure S9g. The protein expression levels of the fibroblast markers VIMENTIN, POSTN, and PDGFRA were downregulated or upregulated, in accordance with the inhibition or overexpression of Pdgfra (Fig. 6h-i), the statistical results were demonstrated in Fig. 6j-k, which were to a great extent consistent with our hypothesis. In addition to western blotting, we used Cell Counting Kit-8 and a lactate dehydrogenase (LDH) kit to observe the effect on Sca1+ cellular activity in vitro. These demonstrated that higher expression of Pdgfra impaired cellular activity and caused increased LDH release (Supplementary Figure S9e-f).
We also investigated WNT4 expression in human samples and found that it was significantly increased in patients with heart failure (Supplementary Figure S9h-j, P = 0.0001 in Supplementary Figure S9h, P = 0.003 in Supplementary Figure S9j), which was consistent with our findings in mice. We therefore hypothesised that Pdgfra plays a crucial role in the differentiation of Sca1+ cells into fibroblasts via Wnt4.
3.7 Sca1+ cells play a role in myocardial fibrosis significantly related to PDGFRα
To test this hypothesis, we generated a new genetic lineage tracing system, Sca1-CreERT2 Pdgfraflox/flox; Rosa26-tdTomato mice, to observe the effect of conditional Pdgfra knockdown on myocardial fibrosis in vivo. We followed the same experimental design as previously described (Fig. 7a) and collected heart samples and blood from mice after evaluating their cardiac function using echocardiography. The serum isolated from blood samples was used to test the brain natriuretic peptide (BNP) concentration in peripheral circulating blood and found that BNP levels decreased significantly in Sca1-CreERT2 Pdgfraflox/flox;R26-tdTomato mice compared with Sca1-CreERT2;R26-tdTomato mice after TAC surgery (P = 0.0024) (Fig. 7b). The results of echocardiography showed that Sca1-CreERT2 Pdgfraflox/flox;R26-tdTomato mice had a significant improvement in left ventricular ejection fraction (EF) (P = 0.002) and left ventricular fractional shortening (FS) (P = 0.0001) compared with Sca1-CreERT2;R26-tdTomato mice after TAC surgery (Fig. 7c-e). In agreement with previous research, we observed a sharp reduction in the ratio of tdT+ cells, as well as significantly lower expression of fibroblast markers such as VIMENTIN, DDR2, POSTN, and PDGFRA in conditional Pdgfra knock down mice compared with Sca1-CreERT2;R26-tdTomato mice, as evidenced by immunofluorescence staining (Fig. 7f-g, Supplementary Figure S10a-f) and confirmed by histology (Fig. 7h). Based on these findings, we confirmed that Sca1+ cells can transform into fibroblasts and affect the severity of fibrosis through the Wnt4-Pdgfra pathway.