PDGFRα demarcates the cardiogenic clonogenic Sca1+ stem/progenitor cell in adult murine myocardium

Abstract

Cardiac progenitor/stem cells in adult hearts represent an attractive therapeutic target for heart regeneration, though (inter)-relationships among reported cells remain obscure. Using single-cell qRT-PCR and clonal analyses, here we define four subpopulations of cardiac progenitor/stem cells in adult mouse myocardium all sharing stem cell antigen-1 (Sca1), based on side population (SP) phenotype, PECAM-1 (CD31) and platelet-derived growth factor receptor-α (PDGFRα) expression. SP status predicts clonogenicity and cardiogenic gene expression (Gata4/6, Hand2 and Tbx5/20), properties segregating more specifically to PDGFRα(+) cells. Clonal progeny of single Sca1(+) SP cells show cardiomyocyte, endothelial and smooth muscle lineage potential after cardiac grafting, augmenting cardiac function although durable engraftment is rare. PDGFRα(-) cells are characterized by Kdr/Flk1, Cdh5, CD31 and lack of clonogenicity. PDGFRα(+)/CD31(-) cells derive from cells formerly expressing Mesp1, Nkx2-5, Isl1, Gata5 and Wt1, distinct from PDGFRα(-)/CD31(+) cells (Gata5 low; Flk1 and Tie2 high). Thus, PDGFRα demarcates the clonogenic cardiogenic Sca1(+) stem/progenitor cell.

Figure 1. Single-cell profiles map cardiogenic gene expression to Sca1⁺ SP cells co-expressing Pdgfra.

(a) Left, flow sorting of fresh cardiac Lin⁻ Sca1⁺ cells after immunomagnetic enrichment for Sca1. Center, further purification of Lin⁻ Sca1⁺ cells for the SP phenotype by Hoechst 33324 staining±ABC transporter inhibitors: FTC, fumitremorgin C; Res, reserpine; Ver, verapamil. Right, bar graph, mean±s.e.m.; n=4; *P≤0.0001. (b) Single-cell qRT–PCR profiles of fresh total Sca1⁺, SP and non-SP cells, compared with cardiomyocytes (CMC). The heatmap illustrates expression as −ΔCt values (blue, low or absent; red, high) and hierarchical clustering reveals the co-expression of functionally related genes in the populations indicated. Highlighted are: yellow, the Sca1 gene Ly6a; blue, Kdr and Cdh5, enriched in Sca1⁺ and non-SP cells; red, cardiogenic transcription factors, enriched in SP cells and CMC; green, Pdgfra and Tcf21, enriched in SP cells; violet, CMC genes. Sca1⁺, n=23; non-SP, n=44; SP, n=43; CMC, n=18. For the full set of 44 genes, see Supplementary Fig. 1. (c) Density plots of expression (−ΔCt) for selected genes in SP (light red) versus non-SP (light blue) cell populations. Genes are ordered according to increasing P-values. Those with a significantly divergent prevalence of expression between SP and non-SP cells are indicated by an asterisk. (d) PCA of the single-cell expression profiles. (Top) PC2 (19% variability) separates SP from non-SP sample scores, whereas PC3 (9% variability) establishes a distinct separation between SP/non-SP/Sca1⁺ cells and CMC. Non-SP and unfractionated Sca1⁺ cells cluster together in the PC projection. (Bottom) Gene loadings contributing to each PC indicate that a small subset of genes explain the cross-group variability captured by PC2 and PC3. Cdh5 and Kdr are predominantly associated with non-SP and unfractionated Sca1⁺ cells, while Pdgfra and Tcf21 are correlated with SP cells (as given by PC2). Differences between CMCs and the remaining samples are strongly reflected in PC3, with cardiac structural genes (Myh6 and Myl2) clustered consistently.

Figure 2. The Sca1⁺ SP phenotype identifies clonogenic, self-renewing cardiac cells.

(a) Bar graph showing the % of adult cardiac clone-forming cells from Lin⁻ Sca1⁺ SP and non-SP fractions, after single-cell deposition. n=7; *P≤0.02. (b) Long-term growth of cloned cardiac SP cells. Clones shown are those used for grafting in Figs 3e and 4; Supplementary Fig. 4. (c) Lack of increased senescence-associated β-galatosidase (SA-β-gal) in four independent clones at increased passage number. *P≤0.02; **P≤0.0005. (d) Persistent Sca1 expression (left) and enrichment for the SP phenotype (right) in cardiac SP clones after long-term culture. Results shown here (clone 16 at passage 29) were confirmed in four additional clones. (e,f) Enrichment for 2° clone formation in cloned cardiac SP cells. (e) Serial bright-field images showing generation of 2° clone by a single cell from clone 3, passage 14. Scale bar, 100 μm. (f) 2° clone formation by six independent clones at passage 13–14. *P≤0.005. Data are shown as the mean±s.e.m. between independent experiments.

Figure 3. Fidelity of cloned cardiac SP cells to freshly isolated SP cells

(a) Pearson correlation plot, each cell of the heatmap showing the sample pair's correlation coefficient (r). The clustering algorithm separates highly correlated samples (red: r close to 1) from weakly correlated ones (blue: r close to 0). The 20 independent lines of cloned cardiac SP cells show strong within-group correlation, moderate correlation to MSCs and ESCs, and weak correlation to other samples including heart. All clones were analysed at <30 passages. BM, bone marrow; CM, neonatal cardiomyocytes; eH, embryonic day 10 heart; ESC, undifferentiated AB2.2 ESCs; H, adult heart (encompassing whole heart, atria, and each ventricle); K, kidney; L, liver; MSC, PDGRα⁺ bone marrow MSCs; Sk, skeletal muscle; Sp, spleen. (b) Molecular signature of cloned cardiac SP cells. Bar graph, mean±s.e.m. for all 40 genes' expression in 20 independent clones. Genes are colour coded based on functional association or tissue specificity. For comparison with reference samples, see Supplementary Fig. 3. (c) Density plot showing the prevalence for expression of key cardiac transcription factors in the 20 clones. See Fig. 1d. The clustering algorithm separates genes with heterogeneous (multimodal) expression from those with homogeneous low or high expression. (d) Heatmap showing scaled expression of the four most heterogeneous transcription factors in the 20 clones (Gata4, Mef2c, Tbx5, Hand2 and Nkx2-5). Expression within each sample is scaled to Z-score, 0 indicating mean expression of the four genes. Red indicates higher expression than the mean, while blue indicates lower expression. The two-dimensional hierarchical clustering algorithm orders the clones and genes based on co-expression profiles. Asterisks denote four clones taken forward to more detailed studies. (e) Western blots for cardiac transcription factors in the clonal lines. Cytoplasmic (C) and nuclear (N) fractions were analysed using glyceraldehyde 3-phosphate dehydrogenase and histone H1 to authenticate the fractions and β-actin as a loading control. The bands for TBX20 correspond to isoforms with or without the C-terminal 145 a.a. extension.

Figure 4. Cloned cardiac SP cells show tri-lineage potential after cardiac grafting.

Clones were transduced with mOrange and delivered by intramural injection into mice subjected to sham operation (control) or coronary artery ligation (infarction). (a) Co-expression of cardiac troponin I (cTnI, green) and sarcomeric α-actin (sarc actin, violet) in striated grafted cells at 12 weeks. (b) Bar graph showing the proportion of mOrange⁺ cells expressing sarcomeric proteins at 2 or 12 weeks. (c,d) Cardiomyocyte isolation was performed on hearts injected with mOrange⁺ cloned cardiac SP cells at the time of infarction 12–14 weeks earlier. mOrange⁺ rod-shaped cardiomyocyte-like cells were detected in the preparations. (c) Representative cells from three independent hearts are shown. Scale bar, 100 μm. (d) Representative cells co-stained for mOrange⁺ and sarcomeric α-actin. Control, secondary antibody only. Scale bar, 20 μm. (e) Induction of von Willebrand factor (vWF, green) and localization of mOrange⁺ cells in vessels at 12 weeks. (f) Bar graph showing the proportion of mOrange⁺ cells expressing vWF at 2 weeks. (g) Induction of SM-myosin heavy chain (SM-MyHC, green) and localization of mOrange⁺ cells in vessels at 12 weeks. (h) Bar graph showing the proportion of mOrange⁺ cells expressing vWF at 2 weeks. At least 200 cells were scored for each clone for each condition, using at least three sections ≥80 μm apart containing mOrange⁺ cells. Data for each clone and time point are the mean±s.e.m. for 3–4 control hearts and 5–7 infarcted ones. Immunohistochemistry is shown here for two clones at 12 weeks and for all four clones at 2 weeks in Supplementary Fig. 4.

Figure 5. Intramyocardial delivery of cloned cardiac SP cells improves cardiac structure and function.

Views by cine-MRI are shown at end-diastole and end-systole, along with late Gadolinium (Gd)-enhanced MRI (LGE) for infarct size. (a) Representative images of mouse hearts at 1 day and 12 weeks after infarction (MI) and grafting of clones. The red bar highlights end-systolic and end-diastolic diameters. (b) Serial measurements showing improved ejection fraction and diminished infarct size at 12 weeks in infarcted hearts that received cloned cardiac SP cells. Bar graphs present the mean±s.e.m. (sham+vehicle, n=4; sham+SP cells, n=10; MI+vehicle, n=8; MI+SP cells, n=20). *P<0.05 compared with the corresponding uninfarcted control hearts; ^†P<0.05 compared with infarction plus the cell-free vehicle control injection.

Figure 6. Sca1⁺ SP cells derive from Mesp1-, Nkx2-5-, Isl1-, Gata5- and Wt1-fated cells.

(a) Cre drivers were crossed with the Ai14 R26R-tdTom line. Left, longitudinal sections of the heart at low magnification. Right, ventricular myocardium and brain at high magnification. Orange, Cre-dependent expression of tdTom; blue, 4′,6-diamidino-2-phenylindole. Scale bar, 1 mm (left), 25 μm (right). (b,c) Quantitation of cardiac SP and non-SP cells derived from the Cre⁺ ancestors shown. (b) Left, flow cytometry showing no interference of Cre with SP cell number. Right, tdTom in the SP and non-SP fractions of Lin⁻/Sca1⁺ cells. (c) Bar graph, % tdTom⁺ cells; n=3–8; *P≤0.05; **P≤0.001. Data are shown as the mean±s.e.m. for independent experiments.

Figure 7. cGata5 and Wt1 contribute specifically to PDGFRα⁺/CD31⁻ cells.

(a) Flow cytometry illustrating the four-way strategy to dissect fate mapping in cardiac Lin⁻/Sca1⁺ cells by SP staining plus PDGFRα and CD31. Representative results are shown at the right for strong induction of tdTom in all four populations by antecedent expression of Nkx2-5-Cre. (b) Precursors expressing cGata5 and Wt1 give rise to the PDGFRα⁺/CD31⁻ cells. Little or no difference was seen between SP and non-SP cells having the same PDGFRα/CD31 phenotype. Data are the mean±s.e.m. for tdTom induction by each Cre line in the four populations; n=2–7 excepting n =1 for Tie2-Cre; *P<0.05; **P<0.01; ***P<0.001. (c) Schematic representation of the origin of cardiac Sca1⁺ cells. The developmental stages and transitions shown are detailed in the text.

Figure 8. Precise co-segregation of the cardiogenic signature to PDGFRα⁺/CD31⁻ SP and non-SP cells.

(a) Left, FACS density plots illustrating the four-way strategy to flow-sort Lin⁻/Sca1⁺ cells by SP staining plus PDGFRα and CD31. Right, bar graph of the subpopulations' prevalence. Mean±s.e.m.; n=5; *P<0.0001. (b) Bar graph indicating the % of single cells generating clones. Mean±s.e.m.; n=4–5; *P<0.0001. (c) Single-cell qRT–PCR profiles, showing co-expression of selected genes in Pdgfrα⁺/CD31⁻ versus Pdgfrα⁻/CD31⁺ cells from the SP and non-SP fractions. The heatmap shows expression as −ΔCt values (blue, low; red, high). Genes are ordered based on the variance of mean expression levels of the four groups. Highlighted genes: green, Cdh5 and Kdr, enriched in PDGFRα⁻/CD31⁺ cells; light red, Pdgfra, Tcf21, Tbx20, Gata4 and Hand2, in PDGFRα⁺/CD31⁻ cells; yellow: Ly6a, in all four subpopulations. n=60–70 for each. (d) Left, gating strategy to identify mutually exclusive PDGFRα⁺ and CD31⁺ cells within the Sca1⁺ population. Right, bar graph showing the % of cells from the indicated populations. Mean±s.e.m.; n=5; *P<0.0001. (e) Density plots show −ΔCt values of the indicated genes in single PDGFRα⁺/CD31⁻ (light red) and PDGFRα⁻/CD31⁺ (light blue) cells from the three sample groups (SP, non-SP and Sca1⁺). Genes are ranked based on the loadings extracted from PC2, which reflect the across-sample variability. (f) PCA of the single-cell expression profiles from the six populations in panel e. (Top) PC1 (43% of variability) captures within-group variance. PC2 (20% of variability) corresponds to between-group variance, mainly separating cells according to PDGFRα/CD31 status, irrespective of their isolation as SP, non-SP or total Sca1⁺ cells. (Bottom) Gene loadings associated with PC1 and PC2. Genes associated with PC1 are unrelated to cell class: consistently absent/low expression across all samples at the left (Nkx2-5) and high expression at the right (Ly6a). PC2 resolves the genes expressed in different cell populations: Cdh5 and Kdr, prevalent in PDGFRα⁻/CD31⁺ cells, at the top, with Pdgfra, Tcf21 and cardiac transcription factors, prevalent in PDGFRα⁺/CD31⁻ cells, at the bottom.