Defining the earliest step of cardiovascular lineage segregation by single-cell RNA-seq

See allHide authors and affiliations

Science  09 Mar 2018:
Vol. 359, Issue 6380, pp. 1177-1181
DOI: 10.1126/science.aao4174

Committing the heart

The heart is a complex organ composed of multiple cell types such as cardiomyocytes and endothelial cells. Cardiovascular cells arise from Mesp1-expressing progenitor cells. Lescroart et al. performed single-cell RNA-sequencing analysis of mouse wild-type and Mesp1-deficient cardiovascular progenitor cells at early gastrulation (see the Perspective by Kelly and Sperling). When Mesp1 was eliminated, embryonic cells remained pluripotent and could not differentiate into cardiovascular progenitors. During gastrulation, the different Mesp1 progenitors rapidly became committed to a particular cell fate and heart region. Notch1 expression marked the earliest step of cardiovascular lineage segregation.

Science, this issue p. 1177; see also p. 1098


Mouse heart development arises from Mesp1-expressing cardiovascular progenitors (CPs) that are specified during gastrulation. The molecular processes that control early regional and lineage segregation of CPs have been unclear. We performed single-cell RNA sequencing of wild-type and Mesp1-null CPs in mice. We showed that populations of Mesp1 CPs are molecularly distinct and span the continuum between epiblast and later mesodermal cells, including hematopoietic progenitors. Single-cell transcriptome analysis of Mesp1-deficient CPs showed that Mesp1 is required for the exit from the pluripotent state and the induction of the cardiovascular gene expression program. We identified distinct populations of Mesp1 CPs that correspond to progenitors committed to different cell lineages and regions of the heart, identifying the molecular features associated with early lineage restriction and regional segregation of the heart at the early stage of mouse gastrulation.

The mammalian heart is composed of different regions (ventricles, atria, and great vessels) and cell types, including cardiomyocytes (CMs), endocardial cells (ECs), smooth muscle cells (SMCs), and epicardial cells (EPs) (1). Heart development begins at gastrulation, during which CPs leave the primitive streak (PS) and migrate toward the antero-lateral pole of the embryo (2). From embryonic day 6.25 (E6.25) to E7.25, Mesp1 marks the early CPs within the PS, whereas a day later, Mesp1 is expressed in the somites (2, 3). Mesp1+ cells give rise to all heart cells, ECs of the aorta and brain, some muscles of the head and neck, as well as to few somitic derivatives and liver cells from its later expression (2, 46). Temporally inducible Mesp1 lineage tracing shows that at E6.5, Mesp1+ cells mark left ventricle (LV) progenitors, whereas the right ventricle, atria, outflow, and inflow tracts and head muscles arise from Mesp1+ cells at E7.25, which correspond respectively to the first and second heart fields (FHF and SHF, respectively). No somitic or liver derivatives were labeled at these early time points (4, 5, 7). In addition, most of the Mesp1 CPs differentiate into either CMs or ECs, suggesting that lineage segregation occurs early during gastrulation (4, 5). It remains unknown whether molecular heterogeneity between E6.5 and E7.25 Mesp1+ CPs reflects stochasticity in gene expression, transcriptional priming, or early lineage and regional segregation.

To investigate the molecular and cellular basis of the earliest stages of CP specification and diversification, we performed single-cell RNA sequencing (scRNA-seq) of Mesp1 CPs at E6.75 and E7.25. To this end, Mesp1-rtTA/tetO-H2B–GFP (green fluorescent protein) mice were treated with doxycycline at different time points after fertilization in order to label only early Mesp1-expressing cells and no somitic derivatives (Fig. 1A), embryos were dissociated into single cells, and H2B-GFP–positive CPs were isolated by means of fluorescence-activated cell sorting (FACS) (fig. S1). A total of 172 and 341 Mesp1 CPs at E6.75 and E7.25, respectively, were sequenced and analyzed further after passing through a stringent quality control pipeline (supplementary materials, materials and methods). We recently reported single-cell transcriptomes for E6.5 epiblast cells, as well as E7.25/7.5 Flk1-expressing progenitors (8). Visualization by use of dimensionality reduction techniques allowed us to order the cells along developmental progression and assign a time stamp to each cell, demonstrating that the Mesp1+ CPs at E6.75 and E7.25 likely represent a continuum of differentiation (Fig. 1, B and C, and fig. S2) (8).

Fig. 1 scRNA-seq of Mesp1+ CPs fills the gap between E6.5 epiblast cells and E7.5 mesodermal cells.

(A) Scheme of the experimental strategy used for isolating Mesp1-expressing CPs in vivo. Scale bar, 200 μm. (B) SPRING plot of 892 cells showing Mesp1 CPs at E6.75 and E7.25 and the published epiblast cells (E6.5_Scia) and E7.5 Flk1+ progenitors (E7.5_Scia) with read count of Mesp1 > 0. (C) SPRING plot colored by the inferred pseudotime time for all 892 cells.

To determine the role of Mesp1 in regulating the cardiovascular differentiation program and the heterogeneity of early CPs, we performed scRNA-seq of FACS-isolated Mesp1-expressing cells in Mesp1 knockout (KO) context (fig. S3) (2). We sequenced transcriptomes of 85 single Mesp1-null cells isolated at E6.75, before the appearance of the developmental defect associated with Mesp1 deficiency (Fig. 2A). Pseudotime analysis revealed that Mesp1 KO cells presented a developmental block, being stuck in the gene expression program of epiblast cells (Fig. 2B). Principal components analysis showed that principal component 2 captured expression differences between wild-type (WT) and Mesp1 KO cells (fig. S4), with 206 down-regulated and 136 up-regulated genes (table S1). We found a highly significant overlap for genes differentially expressed between WT and Mesp1 KO cells in vivo and genes that are down- or up-regulated after Mesp1-induced gain of function in embryonic stem cells (ESCs) in vitro (9), many of which are direct Mesp1 target genes (Fig. 2C, fig. S5, and table S1). Several well-known regulators of pluripotency—including Nanog, Eras, Pou5f1/Oct4 (10), and markers of the epiblast, including E-Cadherin/Cdh1, Epcam, Cldn6, and Cldn7—were up-regulated in single Mesp1 KO cells (Fig. 2, D to F, and table S1), which is consistent with the defect of exiting the pluripotent epiblast stage. By contrast, the genes down-regulated in Mesp1 KO cells were greatly enriched for Mesp1 target genes controlling epithelial-mesenchymal transition (EMT) (Snai1 and Zeb2), migration (Rasgrp3), and cardiovascular commitment (Etv2, Hand1, Myl7, Gata4, Flk1, and Pdgfra) (Fig. 2, F and G, and fig. S5) (9, 11). Pdgfra/Flk1–expressing cells that mark Mesp1 CPs in human and mouse ESC differentiation in vitro and during mouse gastrulation in vivo (5, 12, 13) were much reduced in Mesp1 KO cells, supporting the absence of CP specification (fig. S6).

Fig. 2 Mesp1 controls the exit from pluripotency, EMT, and cardiovascular specification.

(A) SPRING plot of all 892 cells, including Mesp1 KO cells colored by cell types. (B) Pseudotime time distribution for WT and Mesp1 KO cells at E6.75. (C) Comparison of the genes differentially expressed in scRNA-seq experiments between control and Mesp1 KO cells and the genes regulated by Mesp1 gain of function (GOF) in ESCs. The 58 genes in agreement with the scRNA-seq experiment with false discovery rates of <0.1 were highlighted in red. The significance of the overlap was calculated by means of hypergeometric test using the phyper function in R. (D) Gene ontology enrichment for genes down-regulated in Mesp1 KO cells. (E to G) Violin plots showing the mean and variance difference between WT and Mesp1 KO cells of (E) genes regulating pluripotency (Nanog and Eras), (F) EMT (Cdh1 and Snai1), and (G) cardiovascular fate (Gata4, Etv2, and Myl7).

SPRING analysis, which allows visualizing high-dimensional single-cell expression data (14), of WT Mesp1-expressing cells at E6.75 and E7.25 identified five distinct destination cell types (DCTs) protruding from a core of intermingled cells (Fig. 3, A and B). All cells present within the DCTs came from E7.25 embryos, which is consistent with cell fate diversification of Mesp1-derived lineages during the late stages of gastrulation. To further define the nature of the five DCTs, we identified genes with specifically elevated expression in each of the five groups (Fig. 3, B and C; fig. S7; and table S2). This analysis identified both known and previously unrecognized genes associated with cardiovascular development. DCT1 was enriched among others in Sox7, Etv2, and Tal1 transcripts, which are markers of the endothelial or endocardial lineage (15, 16). DCT2 was marked by the expression of Hand1, Bmp4, Tnnc1, Tbx3, Hand2, Tbx20, Gata4, Myl4, and Mef2c, which are well-known CM markers (Fig. 3C, fig. S7, and table S2) (7, 17). Bmp4 promotes CM differentiation (18). Moreover, Hand1 lineage tracing showed that Hand1-expressing cells contribute to the LV and to the myocardial and epicardial lineages, with no contribution to the endocardium (17). These data suggest that DCT2 corresponds to CPs committed to the CM lineage. In situ hybridization of Sox7 (EC marker) and Bmp4 (CM marker) on Mesp1-rtTA/tetO-H2B-GFP embryos showed that these two markers did not colocalize in Mesp1-H2B-GFP–expressing cells at E7.25, which is consistent with the notion that DCT1 and DCT2 mark two distinct Mesp1 populations committed to EC and CM differentiation (Fig. 3D). Unsupervised hierarchical clustering of DCT1 and DCT2 markers with CM and EC markers identified by means of scRNA-seq of more mature mouse hearts at E10.5 (19) further showed that DCT1 and DCT2 cells clustered respectively with EC and CM lineage (Fig. 3E).

Fig. 3 Mesp1 single-cell analysis identifies different progenitors committed to different fates and heart regions.

(A) SPRING analysis of the 807 WT Mesp1-H2B-GFP+ cells at E6.75 and E7.25 and Mesp1+ Scia cells. (B) The five end points revealed by means of SPRING analysis were considered as five distinct cell types (DCT1-5). (C) Expression of key genes specific for DCT1-4. (D) RNA-FISH of Sox7 and Bmp4 on sections of Mesp1-rtTA/tetO-H2B-GFP embryos at E7.25. (E) Heatmap of DCT1 and DCT2 end point cells based on unsupervised clustering of the expression of CM and EC marker genes identified at E10.5 (24) combined with newly identified genes enriched in DCT1 or DCT2. (F and G). RNA-FISH of Foxc2 and Wnt2b on sections of Mesp1-rtTA/tetO-H2B-GFP embryos at E7.25. Higher magnification is found in (G). Scale bar, 50 μm.

DCT3 and DCT4 are enriched in genes expressed and regulating SHF development (Tbx1, Foxc2, Hoxb1, and Hoxa1) (Fig. 3C, fig. S7, and table S2) (2022). Lineage tracing experiments have previously shown that Tbx1 (DCT3) marks anterior SHF progenitors, whereas Hoxb1 and Hoxa1 (DCT4) mark posterior SHF (20, 22), suggesting that DCT3 and DCT4 correspond to Mesp1 CPs committed to the anterior and posterior SHFs. RNA–fluorescence in situ hybridization (FISH) experiments further showed that Foxc2 (DCT3) and Wnt2b (DCT4) were largely nonoverlapping, with Wnt2b localized closer to the PS, whereas Foxc2-expressing cells were found more anterolaterally at E7.25 (Fig. 3, F and G). A subset of SHF progenitors, called cranio-pharyngeal progenitors, express Tcf21 and contribute to the formation of some head muscles (23). RNA-FISH showed that Tcf21 is preferentially expressed in a subset of DCT3 Mesp1 H2B-GFP expressing Foxc2, whereas little overlap was observed between DCT4 progenitors expressing Wnt2b and Tcf21 (fig. S8). Consistent with the notion that DCT3 marks cranio-pharyngeal progenitors, the “pharyngeal” cluster found in our previous Flk1+ scRNA-seq (8) are closest to DCT3 (fig. S7). DCT5 expressed endoderm markers such as Sox17 and Foxa2 and may have no relation with cardiac development (fig. S7 and table S2).

On the basis of the cumulative evidence that suggested our scRNA-seq captures the developmental progression from epiblast to early cardiovascular lineage segregation, we next investigated the expression of genes with known biological function, which revealed specific expression of cardiovascular, mesodermal, and other genes regulating signaling pathways across the various DCT populations (fig. S9). Analysis of genes dynamically expressed during the trajectory toward DCT1 and DCT2 revealed the existence of gene clusters peaking in expression at sequential points along the pseudotemporal ordering. These gene clusters showed enrichment for gene ontology categories associated with developmental progression in which the predicted trajectory provided consistency between the pseudotime and the real developmental time (fig. S10 and table S3). Moreover, cells coexpressing genes enriched in both DCT3 and DCT4 presented early pseudotime values than either DCT3 or DCT4 cells, which is consistent with the presence of immature progenitors that undergo multilineage priming before making cell fate decision into either DCT3 or DCT4 lineages (fig. S11).

To more clearly differentiate between the putative differentiation paths to the DCT1/EC and DCT2/CM progenitors, we determined the genes involved in embryonic development that are specifically up- or down-regulated in a given DCT (fig. S12). Of particular interest, DCT2 cells showed reduced Notch1 expression (Fig. 4A). Different studies have shown the importance of Notch1 in the latter stages of cardiovascular development, in regulating endocardium, valve formation, trabeculation, and myocardium compaction (24). However, a role for Notch in the early steps of cardiovascular lineage commitment during gastrulation has not been previously described. RNA-FISH and immunostaining showed that Notch1 was indeed expressed and active in a subset of Mesp1-H2B-GFP+ cells (Fig. 4B and fig. S13). To determine whether Notch1 expression in DCT1 cluster marks Mesp1+ CPs committed to the EC fate, we induced lineage tracing by administrating tamoxifen to Notch1-CreERT2/Rosa-tdTomato mice at E6.5 (25) and assessed the fate of marked cells at E12.5 (Fig. 4, C to G). tdTomato+ cells in the ventricles were almost exclusively found in the EC (83.9 ± 3.0%), with minor contribution to the CM (6.0 ± 1.4%) and to the EPs (10.2 ± 2.1%) (Fig. 4H), which is consistent with the notion that DCT1 marks early Mesp1/Notch1 double-positive CPs committed to the EC fate. In the atria, Notch1-CreERT2 marked preferentially the ECs (76.9% ± 2.9), although a smaller contribution to the CM lineage was also observed (16.6% ± 2.7) (Fig. 4I), which is consistent with a lower expression of Notch1 in DCT2 cluster. In addition to marking ECs of the heart, Notch1-CreERT2, similar to Mesp1-Cre, also marked, at the early stage of gastrulation, ECs of the aorta, intersomitic, and brain vessels (fig. S14).

Fig. 4 Notch1 marks Mesp1 progenitors committed to the endocardial fate.

(A) Violin plot of Notch1 expression in DCT1 and DCT2 cells. (B) Notch1 RNA-FISH on a section of an E7.25 Mesp1-rtTA/tetO-H2B-GFP embryo. Scale bar, 50 μm. (C) Experimental strategy used for tracing Notch1-expressing cells at E6.5. (D to G) Confocal analysis of immunostaining for [(D) to (F)] endoglin (EC marker) and (G) cTNT (CM marker) of Notch1-CreERT2/Rosa-tdTomato heart sections at E12.5. (D) Lower magnification. Scale bar, 500 μm. Higher magnifications of the (E) LV and (G) RA showed that most tdTomato+ cells are ECs, although rare CMs are also marked (asterisks). Scale bars, 200 μm. (H and I) Percentage of tdTomato+ cells in ECs, CMs, and EPs in (H) the ventricles (7804 cells counted; n = 6 embryos from three different litters) and (I) the atria (4819 cells counted; n = 7 embryos from three different litters).

Altogether, our single-cell profiling of early CPs shows that Mesp1 CPs segregate rapidly from the epiblast into distinct cardiovascular lineages. The analysis of Mesp1 KO cells showed that Mesp1 is required to exit the pluripotent state of the epiblast and promotes EMT, migration, and cardiovascular specification in vivo. Consistent with the early regional and lineage segregation found with clonal analysis (4, 5), our scRNA-seq demonstrates that Mesp1 CPs are also molecularly heterogeneous, as previously suggested by scRNA-seq during in vitro ESC differentiation (26), and identifies temporally and spatially distinct Mesp1 subpopulations that likely correspond to CPs committed to the different cardiovascular lineages and regions of the heart at the early stages of gastrulation (fig. S15). Future studies will be required to determine whether the early cardiovascular lineage segregation uncovered here is also occurring for the other mesodermal and endodermal cells and whether defects in the early commitment steps are associated with congenital diseases and heart malformations. Last, our results will be important to design new strategies to direct the differentiation of ESC into a specific cardiovascular lineage.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Tables S1 to S4

References (2750)

References and Notes

Acknowledgments: We thank the Light Microscopy Facility for help with confocal imaging and F. Hamey for help with SPRING analysis. F.L. has been supported by the Fonds de la Recherche Scientifique (FNRS), the European Molecular Biology Organization long-term fellowship, and the Leducq Fondation. Work in the Gottgens laboratory is supported by grants from the Wellcome, Bloodwise, Cancer Research UK, National Institute of Diabetes and Digestive and Kidney Diseases, and core support grants by the Wellcome to the Wellcome–Medical Research Council Cambridge Stem Cell Institute. C.B. is an investigator of WELBIO. Work in C.B.’s laboratory was supported by the FNRS, the Université Libre de Bruxelles Foundation, the European Research Council, the Bettencourt Schueller Foundation (C.B. and F.L.), and the Leducq Fondation as part of the network “22q11.2 deletion syndrome: Novel approaches to understand cardiopharyngeal pathogenesis.”
View Abstract

Stay Connected to Science

Navigate This Article