Single-cell analysis of early progenitor cells that build coronary arteries

Abstract

Arteries and veins are specified by antagonistic transcriptional programs. However, during development and regeneration, new arteries can arise from pre-existing veins through a poorly understood process of cell fate conversion. Here, using single-cell RNA sequencing and mouse genetics, we show that vein cells of the developing heart undergo an early cell fate switch to create a pre-artery population that subsequently builds coronary arteries. Vein cells underwent a gradual and simultaneous switch from venous to arterial fate before a subset of cells crossed a transcriptional threshold into the pre-artery state. Before the onset of coronary blood flow, pre-artery cells appeared in the immature vessel plexus, expressed mature artery markers, and decreased cell cycling. The vein-specifying transcription factor COUP-TF2 (also known as NR2F2) prevented plexus cells from overcoming the pre-artery threshold by inducing cell cycle genes. Thus, vein-derived coronary arteries are built by pre-artery cells that can differentiate independently of blood flow upon the release of inhibition mediated by COUP-TF2 and cell cycle factors.

Extended Data Fig. 1. Single cell analysis of ApjCreER lineage labeled cells

(a) Comparison of rPCA and classical PCA at separation of subpopulations. PC scores were selected to best separate the Enpp+/Esam− population. Cells are colored by expression (log10 cpm, scaled to max per gene). N=352 cells. (b) Comparison of default and sum-of-60 modified PC scores. PC2 is the default PC score from rPCA; PC2.score is the modified sum-of-top-60 scores (expression is log10 cpm, scaled to max). Y-axis is the number of genes detected per cell (>1 count). N=426 cells. (c) Comparison of default and sum-of-top-60 scores. Scores were chosen that best separated the Vwf+ and Cxcr4 populations. N=426 cells. (d) Unique cell cycle signature on PC pos/neg biplots. PC1.pos (PC1.neg) is the sum of the top 30 genes by positive (negative) loading to PC1. Cells are colored by expression. Lower panel is same rPCA after removing the list of 202 cell cycle genes. Bolded numbers are the correlations between PC1.pos and PC1.neg. N=674 cells. (e) PC pos/neg biplot showing theoretical location of doublets expressing high levels of both gene sets. (f) Schematic of the pairwise discreteness test on a discrete (left) and continuous (right) pair of subpopulations. (g) FACS plots used to isolate GFP-positive cells (red box) from ApjCreER, Rosa^mTmG hearts at e12.5. (h) Upper panel: discreteness statistic generated by pairwise discreteness test as a function of number of intermediate cells (n_int) for simulated distributions. Lower panel: pairwise distributions of cell clusters in the dataset and the fraction of intermediate cells estimated by pairwise discreteness analysis. (i) rPCA plots and their accompanying gene expression patterns in the embryonic heart as reported by Euroexpress. In situ hybridization images show whole hearts in top panels while insets of specific areas are in lower panels with relative expression levels indicated. Expression levels in rPCA plots range from 0 (yellow) to 4 (brown) in log₁₀ cpm. Top panels: N=843 cells. (j) Summary of broadly defined cell populations as indicated by gene expression patterns. N=843 cells. (k) Example of manual clustering process. For i, n=732 cells; ii, n=531 cells; iii, n=415 cells; iv, n=284 cells; v, n=261 cells. (l) Comparison of pairwise discreteness test results for different numbers of genes per cell type signature (n).

Extended Data Fig. 2. Identification of a coronary progenitor niche within the SV

(a) Gene expression patterns identify cell types in rPCA plots of the venous valve-SV-coronary vessel (CV) continuum. Expression levels are log10 of counts per million reads (CPM) and range from 0 (yellow) to 4 (brown) as indicated. Left panel, n=843 cells, Right panels, n=732 cells. (b and c) Expression patterns in rPCA plots (b) and whole mount confocal immunofluorescence (c) of selected genes. For b, n=732 cells. (d) Overlaying gene expression patterns suggests that the SV has two distinct domains, the SVc (sinus venosus, coronary adjacent) and the SVv (sinus venosus, valve adjacent). (e) rPCA on the valve-SVv-SVc continuum identified specific markers of the SVv and SVc. Solid box, n=732 cells. Dotted box, n=415 cells. (f) In situ hybridization of SVv and SVc markers revealed complementary localization in vivo. (g) Color coding showing subpopulations that were used to calculate average expression levels. CV, coronary vessel; PC, principle component; SVc, sinus venosus-coronary; SVv, sinus venosus-valve. Scale bars: b, 200 μm, e, 30 μm.

Extended Data Fig. 3. Characterization of pre-artery cells

(a–d) rPCA plots of the e12.5 SVc-coronary vessel continuum. Each dot is an individual cell, and gene expression levels are indicated by the color spectrum as shown in Fig. 1d, which reflects log10 of counts per million. (a) Arterial genes highly enriched in the arterial areas of the plot. (b) Arterial genes significantly upregulated in, but not specific to, the arterial area of the plot. (c) Venous genes highly depleted in the arterial areas of the plot. (d) Venous genes downregulated, but not depleted, in the arterial area of the plot. For a–d, Bonferroni-adjusted p < 0.01; PCA plots, n=415 cells. Center and error bars are mean ± s.e.m. of log cpm expression values. (e) Genes expressed in adult coronary artery cells. Data is from the Tubula Muris consortium. N=445 cells. (f) Assignment of artery, capillary, and vein in adult coronary cells based on gene expression enrichment in e. N=445 cells. (g) Schematic for comparing e12.5 coronary cells to those along the adult artery-capillary-vein continuum. (h) Results of experiment schematized in g. The center line correspond to the median; the upper and lower hinges correspond to the first and third quartile, respectively; the whiskers extend to the largest value or to 1.5*IQR (inter-quartile range, or distance between quartiles), whichever is smaller. Pre-artery cells: n=20 cells. CV: n=277 cells. P = 6.2 x 10⁻¹³. Statistical test is two-tailed. A, artery; Art, arterial; CV, coronary vessel plexus; PC, principle component; SVc, sinus venosus-coronary; V, vein

Extended Data Fig. 4. Novel artery markers identified in scRNA-seq data

(a) e12.5, e14.5, and adult coronary cell rPCA plots with genes highly enriched or specific to the arterial area during development. Each dot is an individual cell, and gene expression levels are indicated by the color spectrum as shown in Fig. 1d, which reflects log10 of counts per million. Red/bolded genes are also enriched in adult artery cells. (b) Fluorescence in situ hybridization (RNAscope) for Slc45a4, which is expressed (arrowheads) in vessels positive for the arterial marker, Cx40, but not Cx40-negative capillaries (arrows). (c) Slc45a4 expression in pre-artery cells derived from the SV lineage (ApjCreER lineage labeled)(arrowheads). (d) Genes enriched in, but not specific to, arterial cells at e12.5 and 14.5. Genes in bold red are arterial specific in both the developing and adult heart. In both a and d: For PCA plots n=415 cells (upper panel, e12.5); n=347 cells (middle panel, e14.5); n=445 cells (lower panel, adult). For bar graphs E12.5, Art, n=20 cells; CV, n=277 cells; SV, n=118 cells; E14.5, Art, n=70 cells; CV, n=454 cells; SV, n=144 cells. Center and error bars are mean ± s.e.m. of log cpm expression values. Dots represent individual cells. Art, arterial; CV, coronary vessel plexus; SVc, sinus venosus-coronary; V, vein. Scale bars: 100 μm.

Extended Data Fig. 5. Additional whole mount immunofluorescence of marker genes

(a) Cx40 whole mount immunohistochemistry in late gestation hearts (e17.5). Cx40 is only expressed in cells lining large arteries and arterioles (overlapping blue and green signal). Low level, non-arterial signal is in myocardial cells. (b) rPCA plots from e12.5 and e14.5 with accompanying whole mount immunofluorescence in e13.5 hearts. VWF is enriched in the SV while Apln-nlacZ signal and DACH1 are present throughout the coronary plexus. CXCR4, CXCR7-GFP, and CXCL12-DsRed are enriched in the pre-artery and artery areas of rPCA plots and are as interspersed within the intramyocardial coronary plexus. N=415 cells (left panel, e12.5); n=347 cells (right panel, e14.5). Art, arterial; CV, coronary vessel plexus; SVc, sinus venosus-coronary. Scale bars: 100 μm.

Extended Data Fig. 6. Clustering and additional lineage analysis of pre-artery cells

(a) Clusters and relationships generated by rPCA and the pairwise discreteness test (left) and clusters generated by the Seurat pipeline (Louvain/SNN clustering, resolution = 2)(right). N=843 cells. (b) Violin plots show that arterial gene enrichment and venous gene de-enrichment is better with manual, iterative clustering, suggesting that this method leads to more precise populations. (c) Violin plots of cell cycle genes in the two CV plexus clusters generated by the indicated algorithms. Seurat clusters are more defined by cell cycle differences than iterative rPCA (iRPCA) clusters. For b and c, violin plots were made using Seurat VlnPlot. Each violin plot is one subtype and each dot corresponds to a cell. (d) Quantification of the SV and endocardium contributions to coronary arteries. Error bars are st dev. ApjCreER RCA: n=11 hearts. ApjCreER LCA: n=6 hearts. Nfatc1Cre RCA: n=5 hearts. Nfatc1Cre LCA: n=5 hearts. Centre is the mean. (e) Experimental design to lineage trace pre-artery cells. (f) Lineage labeling in e12.5 Cx40CreER, Rosa^tdtomato hearts induced with Tamoxifen at e11.5. (g) Arterial lineage labeling in hearts induced at e10.5. (h) Example of clones in Cx40CreER, Rosa^confetti heart at e15.5. Tamoxifen was administered at e12.5. Two groups of cells sharing the same fluorescent label (clones) are present, YFP-labeled (yellow circle) and nGFP labeled (green circle). Clone sizes are very small consistent with low proliferation rates in pre-arterial cells. (i) P8 heart lineage from Cx40CreER, Rosa^tdtomato animals dosed with Tamoxifen at e11.5. Heavy lineage labeling of the left coronary artery is shown (LCA). Arrowheads indicate branches of the right coronary artery. Myocardium (myo) of the left ventricle is also Cx40⁺ at e11.5, and is also lineage labeled. (j) Images from P8 Cx40CreER, Rosa^tdtomato hearts dosed with Tamoxifen at e11.5 or e16.5. Only the e11.5 dosage results in capillary labeling (arrows) resulting from reversion of pre-artery cells that differentiate during the burst of pre-artery specification between e12.5–14.5. Arrowheads point to arterial lineage labeling. (k). Postnatal lineage tracing in ApjCreER, Rosa^tdtomato or Cx40CreER, Rosa^tdtomato hearts where Tamoxifen was injected at P2. Tips of arteries are lineage labeled with ApjCreER, Rosa^tdtomato, but are depleted with Cx40CreER, Rosa^tdtomato label, indicating that artery tips can extend by incorporating capillary cells that differentiate into arterial endothelial cells. Unpaired two-tailed t test was used to calculate P values. For ApjCreER, N=78 artery tips at P2, n=41 artery tips at P6. p=4.4608 x 10^−19. For Cx40CreER, N=81 artery tips at P2, n=49 artery tips at P6. p=1.61705 x 10⁻¹⁵. Error bars are st dev. ****, p≤0.0001. Centre is mean. Scale bars: d, e, f, j, k, 100 μm; g, 50 μm.

Extended Data Fig. 7. A burst of pre-artery specification between e12.5–e14.5 specifies cells that build most of the embryonic left and right coronary arteries

(a) rPCA plots of the SVc-coronary vessel continuum show that Apj and Cx40 mark cells before and after pre-artery specification, respectively. N=415 cells. (b) Schematic of lineage tracing experiments. Black bars indicate Tamoxifen dosing, and arrows point to harvest date. (c–e) Right lateral views of early hearts show zones with heavy pre-artery specification. (f, g, and i) Low magnification of right lateral views (left most panels) at late embryonic stages show labeling in the main right coronary artery while right panels focus in on more distal branches of the right coronary artery. Table summarizes labeling and results. N= >3 hearts for each experiment. (h) Quantification of ApjCreER and Cx40CreER lineage labeling indicates that most of the embryonic coronary artery is formed by cells specified within the e12.5 to 14.5 time window. N=3 for each Cre. Error bars are st dev. Centre is mean. Tam, Tamoxifen. Scale bars, 100 μm.

Extended Data Fig. 8. Gene expression curves in e12.5 cells

(a) Expression of genes from the indicated categories along the SV-CV plexus-arterial differentiation continuum. The x-axis has individual cells organized as shown in Fig. 3a, and gene expression is plotted as LOESS curves. Raw data points are shown as dots. (b) Gene set enrichment analysis (GSEA) where cell cycle pathways are bolded. (c) Pre-artery cells (Cx40⁺Cxcr4⁺Apj⁻) segregate to the non-cycling quadrant of rPCA plots. Arrows indicate cell cycle progression. (d) Quantification of EdU labeling in pre-artery cells. Error bars are st dev. N=6 hearts. ***, p≤0.001. Unpaired two-tailed t test was used to calculate P value. Centre is the mean.

Extended Data Fig. 9. Effect of COUP-TF2 overexpression during coronary vessel development

(a) Schematic of transgenes used to study Coup-tf2 overexpression in coronary cells. (b and c) Recombination is not complete in the SV with e9.5 and 10.5 doses of Tamoxifen as show in whole mount confocal images (b) and quantification (c). Control GFP is visualized by direct fluorescence, and COUP-TF2^OE through immunostaining for the myc tag. For (c) ApjCreER, Rosa^mTmG, n=5 hearts. ApjCreER, Coup-tf2^OE, n=6 hearts. (d) Tamoxifen dosing at e11.5 and 12.5 fills capillaries with recombined cells, but still resulted in Coup-tf2^OE cells being excluded from arteries (A). (e) Induction of Coup-tf2^OE throughout vasculature shows that overexpressing cells can exist in arteries. (f) Quantification of ventricle coverage at e12.5. N=4 control hearts, n=7 COUP-TF2^OE hearts. ns, p>0.05. p=0.8868. (g) Whole mount confocal images of control and Coup-tf2^OE hearts at different stages of development. Coronary migration (dotted line) on the dorsal side of the ventricle (outlined with solid line) is similar in both genotypes. (h) High magnification of e12.5 Coup-tf2^OE heart shown in (g) highlights the positioning of transgenic cells at both the leading front and trailing cells. (i) COUP-TF2^OE cells can become part of the JAG-1-positive artery if induced after pre-artery specification with Cx40CreER. (j) Mosaic experiment where constitutive expression of the NOTCH intracellular domain (NICD) is induced at the same time as Coup-tf2^OE. This manipulation creates a vasculature containing three different transgene combinations: 1. NICD, 2. COUP-TF2^OE, or 3. NICD + COUP-TF2^OE (arrowheads). Those containing just the NICD (category 1) are the only transgenic cells that contribute to arterial vessels. (k) Quantification of the percentage of endothelial cells in capillaries and arteries (Art) with the three transgenic combinations. NICD-expressing cells preferred arteries while COUP-TF2^OE cells avoid arteries, the latter of which was not rescued by NICD. N=6 hearts. **, p≤0.01; ****, p≤0.0001. For NICD capillary versus artery, p=0.0070. For COUP-TF2^OE capillary versus artery, p=7.49224x10⁻⁰⁵. For COUP-TF2^OE + NICD capillary versus artery, p=8.07734x10⁻⁰⁵. (l) The CDK inhibitor, Flavopiridol, increased arterial specification (Cx40) in an SV sprouting assay. N=33 control explants, n=38 treated explants. ***, p≤0.001. (m) Immunostaining of endothelial sprouts (VE-cadherin⁺) migrating from SV/atria tissue explants with Cx40 showed the increased in this arterial marker (arrowheads) with Flavopiridol treatment. For all graphs: Error bars are st dev, A two-tailed unpaired t test was performed to determine P value, and centre is mean. A, artery; Cap, capillaries. Scale bars: b, 20 μm; d, e, g, h, i, and j, 100 μm; m, 25 μm

Extended Data Fig. 10. Gene expression curves in e14.5 control and Coup-tf2^OE cells

(a) FACS plots of the GFP-marked cells from control and Coup-tf2^OE hearts that were processed for scRNA-seq. (b) Criteria for identifying Coup-tf2^OE cells was >1 read of the flag and myc sequences included in the transgene. (c) Comparing the number of flag and myc reads in control and Coup-tf2^OE hearts confirms the specificity of this parameter for transgenic cells. Control: n=409 cells. COUP-TF2^OE: n=714 cells. The center line corresponds to the median; the upper and lower hinges correspond to the first and third quartile, respectively; the whiskers extend to the largest value or to 1.5*IQR (inter-quartile range, or distance between quartiles), whichever is smaller. (d) Expression of genes from the indicated categories along the vein-CV plexus-arterial axis. The x-axis has individual cells organized as shown in Fig. 5b. Lines are LOESS curves of gene expression and raw data points are shown as dots. Shaded region represents the 95% confidence interval of the LOESS curve. (e) Hypoplastic coronary vasculature with heterozygous deletion of Coup-tf2 in endothelial cells. Scale bars: 100 μm.

Figure 1. Identifying pre-artery cells using scRNA-Seq

Schematics of coronary artery development (a) and computational pipeline (b). (c) Relationship graph for ApjCreER-traced endothelial subtypes. (d) Pre-artery cells extend from the plexus in the SVc-coronary vessel (CV) continuum. Gene expression in brown. N=415 cells. (e) Heat map of venous and arterial genes. At, atria; SVv, Endo, endocardium; sinus venosus-valve; SVc, sinus venosus-coronary progenitors; SV, sinus venosus; ven, ventricle.

Figure 2. Pre-artery cells build coronary arteries

(a) Cx40 immunofluorescence in hearts to mark pre-artery cells (arrowheads). (b) Schematic of pre-artery cells during coronary development. (c) Cx40⁺ cells in sinus venosus (SV)- and endocardium-derived plexus. (d) Cx40CreER lineage labeling (e11.5 induction). N=7 hearts. (e and f) Pre-artery lineage labeling in arteries (arrowheads) and a subset of capillaries (arrows) at e15.5 (e) and P8 (f). (g) Cx40⁺ cells (arrowheads) in hearts that lack coronary blood flow. (h) Cx40⁺ vessels begin remodeling without blood flow. WT, n=7 hearts. Isl1 +/−, n=6 hearts. In d and h: error bars, st dev. LCA, left coronary artery; myo, myocardium; RCA, right coronary artery. Scale bars: d, 50 μm; a, g, h, and i, 100 μm. Graphs: centre is mean. P value: unpaired two-tailed t test. **, p≤0.01. ***, p≤0.001. ****, p≤0.0001.

Figure 3. The venous-to-arterial fate change is gradual and culminates in an expression threshold

(a) Coronary differentiation pathway (dashed arrow). (b–e) Gene expression along the differentiation pathway. Dotted lines, SVc expression levels; red shading, pre-artery cells. (f) Model based on known marker gene patterns. (g) Cell cycle genes decreased in pre-artery cells. CV, coronary vessel; SVc, sinus venosus-coronary progenitors.

Figure 4. COUP-TF2 specifically blocks pre-artery specification

(a and b) e15.5 hearts induced to express Coup-tf2^OE or Gfp before pre-artery specification. (c and d) e15.5 hearts induced to express Coup-tf2^OE after pre-artery specification. b, Coup-tf2^OE, n=11 hearts; Gfp, n=11 hearts. d, N=6 hearts. (e and f) Coup-tf2^OE induction in all endothelial cells before pre-artery specification. Control, n=12 hearts; Coup-tf2^OE, n=20 hearts. (g) Coup-tf2^OE induction in all endothelial cells after pre-artery specification. Control, n=16 hearts; Coup-tf2^OE, n=9 hearts. (h) Schematic displaying differentiation step blocked by COUP-TF2. A, artery; Cap, capillary; Endo, endocardium. Scale bars: 100 μm. Graphs: centre is mean, error bars are st dev. P value: unpaired two-tailed t test. ns, p>0.05. *, p≤ 0.05. **, p≤0.01. ****, p≤0.0001.

Figure 5. COUP-TF2 inhibits artery specification through cell cycle genes

(a) rPCA plots from e14.5 hearts (WT, n=347 cells; Coup-tf2^OE, n=321 cells). Red brackets, artery cells devoid of Coup-tf2 and Apj. (b) Coronary continuum based on gene expression patterns in a. N=347 cells. (c) Coup-tf2^OE cells do not populate the Coup-tf2⁻Apj⁻ artery population. WT, n=347 cells; Coup-tf2^OE, n=321 cells. (d) Progression towards artery is not generally affected by Coup-tf2^OE. WT, gray lines; Coup-tf2^OE, yellow lines. Red-shaded region: pre-artery cells. (e) Percentage of CV plexus cells in the indicated cell cycle phases. (f) Fold increase of control GFP or COUP-TF2^OE cells between e11.5 and e14.5. Control, n=8 hearts; Coup-tf2^OE, n=5 hearts. (g) EdU incorporation in coronary endothelial cells from Cdh5CreER; Coup-tf2 fl/+ hearts. E12.5: control, n=4 hearts; Coup-tf2 fl/+, n=2 hearts. E14.5: control, n=3 hearts; Coup-tf2 fl/+, n=3 hearts. (h) Cell cycle inhibition reverses the ability of Coup-tf2^OE to block artery formation, compare to Fig. 4f. Control, n=6 hearts; Coup-tf2^OE, n=6 hearts. p=0.4167. Graphs: centre is mean, error bars are st dev. P values: unpaired two-tailed t test.*, p≤ 0.05. **, p≤0.01. ***, p≤0.001.