The heart endocardium is derived from vascular endothelial progenitors

Heart development as a model for the study of organogenesis has attracted substantial attention owing to the complexity of the molecular and cellular developmental programs that can be linked to numerous congenital heart defects (Buckingham et al., 2005; Kirby, 2002; Olson and Schneider, 2003; Srivastava and Olson, 2000). During vertebrate gastrulation, cardiac progenitors located in the anterior portion of the primitive streak migrate anterior-laterally to form bilateral heart fields, which are also known as the cardiac crescent (Garcia-Martinez and Schoenwolf, 1993; Wei and Mikawa, 2000). As the embryo folds ventrally, the two heart fields coalesce into a linear heart tube composed of myocardial and endocardial cells.

The generation of multiple cell types in the heart can be achieved through the recruitment of a set of non-overlapping embryonic precursors of distinct origins, or via the progressive diversification of multipotent cardiovascular progenitors; conceivably, these developmental strategies for establishing multiple cell lineages within the heart are not mutually exclusive (Laugwitz et al., 2008). The second heart field (SHF) contains progenitor cells, initially located outside the heart within the pharyngeal mesoderm, that migrate into the heart and give rise to the endocardium, myocardium and smooth muscle lineages (Buckingham et al., 2005). Whether the SHF encompasses multipotent cardiovascular progenitors or distinct lineage-restricted progenitor populations is a fundamental question in embryology.

The origins of the endocardium are controversial (Harris and Black, 2010; Vincent and Buckingham, 2010). Previous lineage studies in chick (Cohen-Gould and Mikawa, 1996; Lough and Sugi, 2000; Wei and Mikawa, 2000) demonstrated that myocardial and endocardial progenitors derive from separate subpopulations within precardiac mesoderm that diverge prior to, or at, the primitive streak stage. In support of these findings, recent studies in zebrafish suggest that endocardial cells are derived from the hematopoietic and vascular lineages, and then migrate to populate the developing heart tube (Bussmann et al., 2007; Schoenebeck et al., 2007). Lineage studies in avian embryos demonstrated that a population of cells within the anterior pharyngeal mesoderm could contribute to outflow tract endocardium but not to myocardium (Noden, 1991). Together, these studies support a model in which part of the endocardium is derived from mesodermal cells that do not have myocardial potential.

In contrast to the idea that endocardial and myocardial lineages have already diverged post-gastrulation, several recent studies have suggested the existence of a common myocardial-endocardial progenitor at cardiac crescent stages, or even later. These studies, primarily performed in vitro, provide evidence that endocardial, myocardial and smooth muscle cells arise from multipotent cardiovascular progenitors that co-express Islet1 (ISL1), FLK1 (also known as VEGFR2 and KDR) and NKX2.5 (Kattman et al., 2006; Moretti et al., 2006; Wu et al., 2006). In response to temporal and positional cues from the microenvironment, these progenitors undergo progressive lineage diversification and differentiation, similar to that seen in the hematopoietic system (Laugwitz et al., 2008). Obviously, the existence of multipotent cardiovascular progenitor cells that contribute to all major heart lineages has clinical implications and provides the foundation for cardiac regenerative medicine.

At early stages of development, endocardial cells are distinguished from myocardium by the expression of endothelial markers, such as FLK1, CD31 (PECAM1) and VE-cadherin (cadherin 5) (Baldwin, 1996; Drake and Fleming, 2000). Although endocardial and endothelial cells share substantial molecular and functional characteristics, a recent study demonstrated molecular and lineage distinctions between these two cell types and suggested that the endocardium and myocardium derive from a common precursor, distinct from endothelial cells (Misfeldt et al., 2009). These previous in vitro studies, and lineage studies of SHF progenitors in mice (Cai et al., 2003; Moretti et al., 2006; Verzi et al., 2005), have given rise to the idea that the endocardium derives from multipotent progenitors within the first and second heart fields (Harris and Black, 2010; Vincent and Buckingham, 2010).

In this study, we have addressed a longstanding question regarding the origin(s) of the endocardium in both avian and mouse models. In DiI mapping studies in chick embryos, cells medial to the cardiac crescent at Hamburger and Hamilton stage (St.) 7 gave rise to endocardial, but not myocardial, cells, demonstrating a divergence of endocardial and myocardial progenitors at this stage. Time-lapse movies of endothelial (QH1-positive) cells in quail embryos demonstrated that endothelial cells medial to the cardiac crescent migrated upwards into the arterial pole of the heart, contributing to the endocardial population. These data are consistent with there being an endocardial progenitor population within the SHF that is distinct from myocardial progenitors. To investigate the ability of endothelial progenitors to contribute to either endocardium or myocardium we transplanted endothelial cells from St. 8 quail embryos into the cardiac region of chick host embryos. The results of these studies demonstrated that endothelial cells were able to contribute only to endocardium, not myocardium, demonstrating endothelial cell fate restriction at this stage.

In mouse, the Isl1 lineage contributes to the SHF-derived myocardium (e.g. outflow tract, right ventricle and atria). We found that only a restricted subset of endocardial cells was derived from this lineage; we suggest that the rest of the endocardium in these regions derives from the Isl1^– endothelial lineage. Finally, genetic ablation of Flk1 in mouse in the Mesp1 and Tie2 (Tek – Mouse Genome Informatics) lineages diminished endothelial and endocardial cells. Ablation of Flk1 in the Isl1 lineage altered the proportion of Isl1⁺ and Isl1^– lineage-derived endocardial cells in Isl1Cre;Flk1 mutant embryos. Taken together, our findings indicate that endocardial cells derive, at least in part, from vascular endothelial cells. These cells give rise to endocardium, but not myocardium, following a migratory pathway comparable to that described for SHF cells. Hence, the SHF contains distinct myocardial and endocardial progenitor populations.

Fertilized chick/quail eggs were incubated at 38°C under 80% humidity; embryos were staged according to Hamburger and Hamilton (Hamburger and Hamilton, 1951).

Live quail embryos at St. 7 were permeabilized, blocked, then incubated with QH1 primary antibody (DSHB) followed by Cy3-conjugated anti-mouse IgG1 secondary antibody (1:100; Jackson ImmunoResearch). Embryos were then placed in a humidified temperature-controlled chamber under a Nikon 90i fluorescence microscope. Images were acquired using ImageProPlus (Media Cybernetics) and assembled using Photoshop CS (Adobe).

Quail or chick grafts of distinct tissues were dissected and inserted into an incision created in the chick or quail host, respectively. For the mouse-chick transplantations, YFP-expressing mouse endothelial cells isolated by FACS were transferred to an agar plate for 16 hours to form aggregates, which were then selected for implantation into chick embryos. All host embryos were incubated for an additional 24 hours.

Whole-mount in situ hybridization was performed using digoxigenin-labeled antisense riboprobes synthesized from the cDNA (see Table S1 in the supplementary material). Images were obtained using a Leica MZ16FA stereomicroscope attached to a digital camera (DC300F, Leica Microsystems).

Fate-mapping experiments were performed on St. 7-8 embryos. DiI (D282, Molecular Probes) at 5 mg/ml (ethanol) was subsequently diluted 1:2 with tetraglycol and was pressure-injected into cultured chick embryos using a micromanipulator.

Conditional knockout embryos were generated using either Mesp1Cre (Saga et al., 1996), Tie2Cre (Koni et al., 2001) or Isl1Cre (Yang et al., 2006) mice, crossed with the Flk1 conditional mice supplied by Genentech. Cre activity was detected using the Rosa26R reporter (Jackson ImmunoResearch). Endothelial cells for the transplantation studies were isolated by FACS from VE-cadherinCre (Alva et al., 2006) crossed with the Rosa26YFP reporter (Srinivas et al., 2001). As a control, neuronal YFP-expressing cells were isolated from YFP 2.2 (Feng et al., 2000) mouse embryos.

Cryosections were blocked with 5% horse serum and then incubated with the following antibodies: QCPN, QH1, anti-PECAM1, anti-ISL1, MF20, anti-tropomyosin (DSHB), anti-NKX2.5 (Santa-Cruz), anti-smooth muscle actin (SMA) (Sigma), anti-β-galactosidase (β-gal) (Cappel) and anti-FLK1 (gift of Philip Thorpe, University of Texas Southwestern Medical Center). Secondary antibodies were Cy5- and Cy2-conjugated anti-rabbit and anti-mouse; Cy3-conjugated anti-rat, anti-hamster, anti-mouse, anti-goat (1:100; Jackson ImmunoResearch).

Following the immunostaining procedure, high-magnification images were taken and processed using Photoshop CS. Single cells were counted manually according to their DAPI staining and were identified according to the expression of the indicated markers in the green or red channels. FLK1 and ISL1 double-positive cells were quantified relative to the total number of ISL1-expressing mesodermal cells. To determine endocardial lineage composition, quantification of β-gal and PECAM1 double-positive cells was performed relative to the total number of PECAM1-expressing cells. The results are presented as the percentage averaged from four different images for each developmental stage.

Chick embryos were placed in New-Culture (Nathan et al., 2008), grown to St. 3, and placed above a platinum cathode. The pCAGG-GFP construct (0.03 μg/μl) (Nathan et al., 2008) was injected between the blastoderm and the vitelline membrane using a glass capillary. Next, the anode was placed above the embryo and two pulses of 6V for 25 mseconds and 500 mseconds were performed using an ECM830 electroporator (BTX). Embryos were analyzed under the Nikon 90i fluorescence upright microscope.

To examine the origins of endocardial cells in chick embryos, we first mapped the expression patterns of cardiac and endothelial markers by in situ hybridization at cardiac crescent stages (St. 7-8). At these stages, NKX2.5, ISL1 and GATA4 are expressed in cardiac progenitors within the lateral splanchnic mesoderm, labeling the cardiac crescent (Fig. 1A-C,G-I) (Nathan et al., 2008; Tirosh-Finkel et al., 2006). By contrast, the endothelial markers VEGFR2, TAL1 and LMO2 are broadly expressed as a punctuate sheet, extending from the midline and spreading throughout the anterior mesoderm (Fig. 1D-F,J-L).

We next traced the fate of cells along the medial-lateral axis of the chick embryo to determine their contribution to the endocardium and myocardium. We began by labeling cardiac crescent cells with the lipophilic carbocyanine dye DiI. DiI-labeled cells could be seen in both the endocardium and the myocardium within the heart tube (Fig. 2A-C′, n=6/6). These findings in the chick confirm that the cardiac crescent at St. 7 (corresponding to the first heart field) contains both endocardial and myocardial progenitors. Injection of DiI into the medial mesoderm resulted in labeling of the cranial paraxial mesoderm, a subregion within the pharyngeal mesoderm (Tzahor and Evans, 2011) (Fig. 2D-F′, n=4/5). Subsets of these labeled cells lined the dorsal aorta, suggesting that some endothelial cells had been labeled (Fig. 2F′). Lastly, we injected DiI adjacent to the cardiac crescent, from its medial side; subsequently, labeled cells were found in the endocardium, but not in the myocardium, and were also detected in the pharyngeal mesoderm (Fig. 2G-I′, n=5/7). A whole-mount view of a representative embryo and serial transverse sections of its heart are shown in Fig. S1 in the supplementary material.

The fate-mapping data could suggest that cells with endothelial but not myocardial potential contribute to the endocardium. To prove that this is indeed the case, we developed a wide-field time-lapse microscopy assay (see Materials and methods) to track the migration of endothelial cells in live quail embryos (Fig. 3 and see Movie 1 in the supplementary material). This analysis enabled us to follow dynamic endothelial cell (stained with QH1 antibody, QH1^pos) movements from the cardiac crescent to linear heart tube stages. By examining the movies, and still images taken from them, we could identify the endothelial cell path (Fig. 3, green circle): endothelial cells adjacent to the inner cardiac crescent initially move towards the head, and then bifurcate medially as they enter the heart tube endocardium via the outflow tract (Fig. 3M,N). These findings further support the concept that the endocardium is formed from endothelial cells that migrate into the linear heart tube, using distinct patterns of movement. The pattern of movement of the QH1^pos endothelial cells into the outflow tract is reminiscent of the movement of myocardial progenitors from the second/anterior heart field. Hence, we conclude that we have identified an endocardial progenitor population within the second/anterior heart field that is distinct from myocardial progenitors.

We next investigated whether distinct sources of endothelial cells are capable of forming endocardium when incorporated into the cardiac environment. Accordingly, we dissected distinct quail tissues containing endothelial cells and transplanted them into the developing chick heart (Fig. 4). First, a small tissue sample from the vitelline vessel of an E4 quail embryo was transplanted to replace the cardiac outflow tract of a St. 10 chick embryo (Fig. 4A-D). At St. 12, quail-derived endothelial cells, identified by QCPN antibody staining, could be detected in the endocardium of the host chick (Fig. 4C,D; n=3/4). We then transplanted a somite, which contains skeletal muscle as well as endothelial progenitors (Ema et al., 2006; Kardon et al., 2002), from a St. 8 quail donor into the cardiac crescent of a St. 8 chick embryo (Fig. 4E-H). Somite-derived cells from the quail integrated with the endocardium, 24 hours later (Fig. 4G,H; n=2/2). In both of these transplantation assays, quail cells (stained with QCPN, red) were only found within the endocardium of the chick embryo (Fig. 4A-H). Endocardial cells derived from the transplant were positive for the quail endothelial marker QH1 (data not shown).

Next, we transplanted a small section of the cardiac crescent from a St. 8 chick embryo adjacent to the vitelline vessel of the quail embryo (Fig. 4I). Quail-derived endothelial cells (QH1^pos) migrated into the heart field explant and formed an endocardial-like structure (Fig. 4J,L; n=4/4). By contrast, these endothelial cells did not integrate into an explant of trunk lateral mesoderm (Fig. 4K,L).

We then considered whether mouse endothelial cells could contribute to the chick endocardium in a similar manner. To this end, we FACS purified endothelial cells obtained from E14.5 embryos derived from VE-cadherinCre (Alva et al., 2006) crossed with Rosa26YFP (Srinivas et al., 2001) mice (Fig. 4M-P). Strikingly, mouse VE-cadherin⁺ embryonic endothelial cells implanted into the cardiac crescent of a chick embryo were detected in endocardial, but not myocardial, cells (Fig. 4O,P; n=4/4). In control experiments, in which FACS-sorting YFP-expressing neuronal cells were similarly transplanted into the cardiac crescent of chick, we observed no YFP expression in endocardial cells (data not shown; n=5/5).

Collectively, our findings in avian embryos suggest that cardiac and endothelial progenitors from distinct, non-overlapping sources give rise to myocardial and endocardial cells, respectively. Furthermore, we identified a field adjacent to the cardiac crescent that contributes to the heart endocardium. We tracked the migratory path of single endothelial cells into the outflow tract, where they contribute to endocardial cells. Lastly, we provide evidence that any endothelial cells transplanted into the cardiac environment can form endocardial cells, supporting the idea that the endocardium derives from endothelial cells.

ISL1 is a major player in the development of pharyngeal mesoderm/SHF cells during embryogenesis (Evans et al., 2010; Rochais et al., 2009; Srivastava, 2006; Tzahor and Evans, 2011; Vincent and Buckingham, 2010). This transcription factor is expressed in undifferentiated progenitors of the two heart fields, although it is only required for the normal development of the SHF (Cai et al., 2003; Prall et al., 2007; Tirosh-Finkel et al., 2006). Our data suggest that endocardial cells could be derived from vascular endothelial cells; however, we cannot exclude the possible existence of multipotent mesoderm progenitors located outside the cardiac crescent. To gain deeper insight into these issues, we performed immunostaining with QH1 and for ISL1 in St. 8 quail embryos (Fig. 5A) and for PECAM1 and ISL1 in sections of E8.0 mouse embryos (Fig. 5B). Few endothelial cells (QH1^pos and PECAM1^pos; to clarify our terminology, ^pos refers protein expression, whereas ⁺ indicates a lineage) were detected between the endoderm and the mesoderm, along the medial-lateral (quail) or dorsal-ventral (mouse) axes (Fig. 5A-C). We did not observe co-expression of QH1/PECAM1 and ISL1.

We next sought to address the existence of multipotent cardiovascular progenitors (reviewed by Laugwitz et al., 2008) expressing both FLK1 and ISL1 in E7.5-11.5 mouse embryos (Fig. 5E-G′). FLK1 expression within ISL1^pos mesoderm cells gradually increased, reaching a peak at E9.5; we could not detect such cells at E11.5 (Fig. 5E-G′, quantified in 5D). Three subpopulations of mesoderm cells could be detected: ISL1^pos, FLK1^pos and ISL1 FLK1 double positive at E9.5 (Fig. 5F-G′). The aortic sac, a dilation just distal to the outflow tract, serves as the primordial vasculature from which the aortic arches arise. Endothelial cells within the aortic sac lie in a continuum with the endocardium in the anterior pole of the heart, with both expressing FLK1 and PECAM1 (Fig. 5G-H). It appears that endothelial cells do not express ISL1 at cardiac crescent stages, although we do see rare ISL1^pos cells in the endocardium at E9.5 (Fig. 5H). Our gene expression data, coupled with anatomical examinations, is in line with the idea that endocardial cells are derived from FLK1^pos ISL1^neg cells, rather than from the double-positive population. Importantly, we cannot rule out the possibility that some FLK1^pos ISL1^pos cells give rise to endocardial cells.

We next re-evaluated the lineage origin of the endocardium in the mouse, using Mesp1Cre (Saga et al., 2000) and Isl1Cre (Yang et al., 2006) mouse lines. MESP1 is broadly expressed in cardiac and craniofacial mesoderm cells at the anterior primitive streak (Saga et al., 2000). Within the heart, the Mesp1 lineage marks all cardiac cells of mesodermal origin (see Fig. S2 in the supplementary material) (Saga et al., 1999).

The Isl1Cre line crossed with the Rosa26 conditional reporter marks broad regions of the heart and the mesoderm core of the pharyngeal arches (Harel et al., 2009; Nathan et al., 2008); notably, most of the left ventricle is not marked by the Isl1 lineage (Fig. 6A-C) (see also Cai et al., 2003; Moretti et al., 2006; Sun et al., 2007). Difficulties in distinguishing between myocardial cells and the thin endocardial lining led us to examine Isl1Cre-mediated lacZ expression in the endocardium in greater detail using anti-β-gal and PECAM1 double staining (Fig. 6D-F′; quantified in 6G). Examination of β-gal-expressing and PECAM1-expressing cells in the outflow tract, right ventricle and atria revealed that ∼50-60% of endocardial cells in these regions expressed both markers. Therefore, endocardial cells are derived from two lineages: Isl1⁺ and Isl1^–. These findings suggest that the endocardium is more heterogeneous than previously suggested.

The expression of FLK1 in early mesodermal cells marks progenitors with a broad lineage potential, although it is thought that this gene is only necessary for the formation of endothelial and hematopoietic lineages (Ema et al., 2006; Motoike et al., 2003; Shalaby et al., 1995). To further investigate the endothelial origins of the endocardium, and the necessity of FLK1 for endocardium formation in the mouse, we conditionally ablated Flk1 in both cardiac and endothelial progenitors. To this end, we first employed an Flk1 conditional allele crossed with the Mesp1Cre mouse. Mesp1Cre;Flk1 null conditional mutants die between E9.5 and E10.5, one day later than the non-conditional Flk1 knockout embryos, enabling us to obtain clearer insight into the cardiac phenotypes of these mutants (Fig. 7). At E9.5, the mutant embryos displayed cardio-craniofacial defects, as compared with their control littermates (Fig. 7A,B). As expected, these mutants were devoid of endothelial cells in the head and heart regions, in agreement with the expression pattern of Mesp1 lineage progenitors (Fig. 7C,D). Immunostaining of sectioned E9.5 heart tubes revealed normal myocardial gene expression; NKX2.5, SMA, myosin heavy chain (MHC) and tropomyosin were expressed throughout the entire myocardium (Fig. 7G-J). By contrast, we observed complete loss of the endocardium in the hearts of Mesp1Cre;Flk1 null conditional mutants (Fig. 7K-N).

Because the Mesp1 lineage gives rise to a broad range of mesodermal cells, including cardiac and endothelial progenitors, we examined the function of FLK1 in endothelial cells by deleting it using Tie2Cre mice (Koni et al., 2001). Although this Cre is less efficient than Mesp1Cre, the morphology of the Tie2Cre;Flk1 null mutants was comparable to that of the Mesp1Cre;Flk1 null mutants (Fig. 8A,D). In addition, the number of endothelial and endocardial cells (PECAM1^pos) was significantly reduced (Fig. 8B,C,E,F). Taken together, our findings in the mouse, based on ablation experiments of Flk1 in both the Mesp1⁺ and Tie2⁺ lineages, confirm that FLK1 is required for the formation of endothelial and endocardial cells. The loss of endothelial cells that accompanied the loss of the endocardium in these mutants further suggests that the endocardium derives from endothelial cells in the mouse. In both Mesp1Cre and Tie2Cre Flk1 null mutants, heart looping was abnormal (Figs 7 and 8): mutant embryos displayed a shortened outflow tract and hypoplastic right ventricle and pharyngeal arches, similar to classical knockouts of SHF regulators (Buckingham et al., 2005; Cai et al., 2003). These findings in both mutants suggest that endothelial cells play non-cell-autonomous roles in the incorporation of pharyngeal mesoderm/SHF cells into the heart, presumably by affecting their migration and/or survival.

To elucidate the specific requirement for FLK1 in Isl1⁺ cardiac progenitors, we crossed the Flk1 conditional allele with the Isl1Cre mouse line (Yang et al., 2006). Isl1Cre;Flk1 null conditional mutants died at ∼E14. At E9.5, the Isl1Cre;Flk1 null conditional mutants had no apparent cardio-craniofacial defects and the endocardium was properly formed, as compared with control littermates (Fig. 9A-H). Likewise, expression of Pecam1 at E9.5 was comparable in control and mutant embryos (Fig. 9A,E). In fact, it appears that there are more endocardial cells (PECAM1^pos) in sections of these mutants, as compared with controls, at E9.5-13 (Fig. 9A-H).

Owing to the observed heterogeneity in the origins of endocardial cells, we examined the proportion of Isl1⁺ and Isl1^– lineage-derived endocardial cells in controls and Flk1 null conditional mutants (Fig. 9I-K). Although Isl1Cre labels endothelial cells (Fig. 6) (Moretti et al., 2006; Sun et al., 2007), many endothelial cells are Isl1^– in terms of their lineage (Figs 5 and 6). Strikingly, quantification of the proportion of Isl1⁺ (β-gal^pos) and Isl1^– (β-gal^neg) lineage-derived endocardial cells (PECAM1^pos) revealed a decrease in the Isl1⁺ population and an increase in the Isl1^– population (Fig. 9I-K). These data suggest that ablation of Flk1 in the pharyngeal mesoderm [Isl1 lineage (Tzahor and Evans, 2011)] affected endocardial cell composition in the Isl1⁺ and Isl1^– lineages, and further point toward the vascular endothelial origin of the heart endocardium. In addition, we demonstrated that FLK1 plays key roles in the formation of the endocardium in Mesp1⁺ and Tie2⁺ (but not Isl1⁺) cells. Notably, FLK1 is required for cardiogenesis in this lineage, as mutant embryos die at ∼E14.5; the molecular pathology and underlying mechanisms of this are currently under investigation.

In this study, we addressed a controversial question in heart development concerning the origins of the endocardium (Harris and Black, 2010; Lough and Sugi, 2000; Vincent and Buckingham, 2010). We provide multiple experimental findings based on gene expression, fate mapping, time-lapse movies and transplantation assays in the avian system demonstrating that during the cardiac crescent stages, endocardial cells derive from a vascular endothelial population that is distinct from myocardial progenitors. Furthermore, lineage tracing and conditional ablation of Flk1 in cardiovascular progenitors in the mouse demonstrated that FLK1 is required in Mesp1⁺ and Tie2⁺, but not Isl1⁺, lineages, for formation of the endocardium in vivo. Ablation of Flk1 in the Isl1⁺ progenitors was compensated for by Isl1^– endocardial cells, which are likely to be derived from endothelial cells. Taken together, we demonstrated in both avian and mouse models that the heart endocardium derives from, and is continuous with, lineage-restricted vascular endothelial cells (Fig. 10A).

We suggest that pharyngeal mesoderm cells within the second/anterior heart field harbor distinct myocardial and endocardial progenitor populations; the Mesp1 lineage marks pan-mesodermal cardiovascular progenitors. These cells stand at the top of the hierarchy, upstream of the ISL1 and TIE2 progenitor populations (Fig. 10B). Mesp1 lineage-derived cells give rise to both myocardial and endocardial progenitors (as well as to other mesodermal derivatives). Myocardial cells are also derived from the Isl1 lineage (Fig. 10B, green), whereas endocardial cells originate from the Tie2 endothelial lineage (Fig. 10B, red; data not shown). However, ∼50% of the endocardium derives from the ISL1 lineage. Because Isl1 lineage-derived cells mark some endothelial cells (Fig. 6), we suspect that the Isl1⁺ endocardial cells are also derivatives of vascular endothelial cells.

In vitro studies in chick and mouse have provided evidence for a bipotential progenitor for myocardial and endocardial cells (Hutson et al., 2010; Kattman et al., 2006; Moretti et al., 2006). This observation, combined with Cre recombinase studies of SHF lineages (Cai et al., 2003; Moretti et al., 2006; Verzi et al., 2005), suggest that bipotential ISL1 FLK1 progenitors within the SHF give rise to both myocardial and endocardial lineages in vivo. In the mouse, we observed the existence of FLK1 and ISL1 co-expressing pharyngeal mesoderm/SHF cells at ∼E7.5-9.5 (see also Moretti et al., 2006), although they seem to be distinct from the FLK1^pos and PECAM1^pos endothelial/endocardial populations (Fig. 5). The fact that ∼50% of endocardial cells within outflow tract, right ventricle and atria derive from an ISL1 lineage does not necessarily imply that these cells are derivatives of a bipotential progenitor of the SHF. Instead, our DiI labeling and in vivo imaging data suggest that these cells might originate from committed vascular endothelial progenitors within, or adjacent to, the SHF. Because FLK1 expression is not absolutely required for the formation of the endocardium in the Isl1 lineage, we suggest that progressive lineage diversification of cultured pluripotent embryonic stem cells into distinct heart progenitor populations might not follow the same developmental paths that occur in vivo.

Activation of Cre recombinase in Mesp1Cre mice precedes its activation in Isl1Cre mice, a finding which could suggest that FLK1 expression is required for the endocardium only within an early developmental window. Thus, FLK1 is required for the formation, but not for the maintenance, of endocardial cells. The absence of endothelial cells in Mesp1Cre;Flk1 null and Tie2Cre;Flk1 null, as compared with Isl1Cre;Flk1 null, mutants (Figs 7, 8, 9) could account for the complete loss of endocardial cells.

Finally, using novel in vivo imaging technology we observed that cells with endothelial identity adjacent to the cardiac crescent migrate inwards to give rise to the endocardium. These observations demonstrate that, in vivo, myocardial and endocardial lineages have already diverged by the time that ISL1 is first expressed in cardiac progenitors. Along these same lines, we performed single-cell labeling experiments in the anterior primitive streak in the chick and identified separate populations of endocardial and myocardial progenitors (see Fig. S3 in the supplementary material), consistent with previous studies (Cohen-Gould and Mikawa, 1996; Wei and Mikawa, 2000). In summary, this work provides strong evidence that the endocardium is derived, at least in part, from vascular endothelial cells that have diverged from myocardial lineages in SHF progenitors.

We thank Sean Wu, Karina Yaniv and Margaret Buckingham for insightful discussions. E.T. was the incumbent of the Gertrude and Philip Nollman Career Development Chair.