Molecular markers of cardiac endocardial cushion development


  • Aaron D. Gitler,

    1. Department of Medicine, Cardiology Division, University of Pennsylvania Health System, 954 BRB II/III, Philadelphia, Pennsylvania
    2. Department of Cell and Developmental Biology, University of Pennsylvania Health System, Philadelphia, Pennsylvania
    Search for more papers by this author
    • A.D. Gitler and M.M. Lu contributed equally to this work.

  • Min Min Lu,

    1. Department of Medicine, Cardiology Division, University of Pennsylvania Health System, 954 BRB II/III, Philadelphia, Pennsylvania
    Search for more papers by this author
    • A.D. Gitler and M.M. Lu contributed equally to this work.

  • Yue Qin Jiang,

    1. Department of Medicine, Cardiology Division, University of Pennsylvania Health System, 954 BRB II/III, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Jonathan A. Epstein,

    Corresponding author
    1. Department of Medicine, Cardiology Division, University of Pennsylvania Health System, 954 BRB II/III, Philadelphia, Pennsylvania
    2. Department of Cell and Developmental Biology, University of Pennsylvania Health System, Philadelphia, Pennsylvania
    • 954 BRB II, 421 Curie Blvd., Philadelphia, PA 19104
    Search for more papers by this author
  • Peter J. Gruber

    1. The Cardiac Center, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author


Endocardial cushions are precursors of mature heart valves. They form within the looped heart tube as discrete swellings and develop into thin, pliable leaflets that prevent regurgitation of blood. The embryonic origins of cardiac valves include endothelial, myocardial, and neural crest cells. Recently, an increasing number of animal models derived from mutational screens, gene inactivation, and transgenic studies have identified specific molecules required for normal development of the cardiac valves, and critical molecular pathways are beginning to emerge. To further this process, we have sought to assemble a diverse set of molecular markers encompassing all stages of cardiac valve development. Here, we provide a detailed comparative gene expression analysis of thirteen endocardial cushion markers. We identify endocardial cushion expression of the transcription factor Fog1, and we demonstrate active Wnt/β-catenin signaling in developing endocardial cushions suggesting pathways that have not been previously appreciated to participate in cardiac valve formation. Developmental Dynamics 228:643–650, 2003. © 2003 Wiley-Liss, Inc.


Endocardial cushion formation begins at mouse embryonic day 9 (E9). At this stage, the heart consists of a simple tube composed of endothelial cells separated from myocardial cells by a thin layer of extracellular matrix called the cardiac jelly. In two distinct regions, the atrioventricular (AV) canal and the ventricular outflow tract (OT), this extracellular matrix thickens to form the first discernible endocardial cushions (Mjaatvedt and Markwald, 1989; Eisenberg and Markwald, 1995; Schroeder et al., 2003). Within the ensuing day, endocardial cells that line the AV and OT cushions respond to extrinsic molecular signals and transform into mesenchyme as they invade the underlying cardiac jelly. This process of endocardial–mesenchymal transformation has been extensively studied in explant assay systems and in vivo (Runyan and Markwald, 1983; Potts and Runyan, 1989; Brown et al., 1999). Molecular cascades involving transforming growth factor β (TGFβ; Brown et al., 1999), epidermal growth factor (EGF) (Chen et al., 2000), platelet-derived growth factor (PDGF) (Morrison-Graham et al., 1992), nuclear factor of activated T-cells (de la Pompa et al., 1998; Ranger et al., 1998), and Ras signaling (Lakkis and Epstein, 1998; Gitler et al., 2003) have all been implicated in this complex process.

Mesenchymal cells populating the endocardial cushions are also derived from migrating neural crest cells, which reach the OT cushions by E10.5 (Epstein et al., 2000; Gitler et al., 2003). However, neural crest contributes little or no cell mass to the AV cushions, a distinction that represents one of the major differences between the development of the AV and OT valves. Mesenchymal cells of both neural crest and endothelial origin proliferate, resulting in significant growth of the endocardial cushions between E11.5 and E13.5. Neighboring myocardial cells then invade the margins of the cellular endocardial cushions in a process known as “myocardialization” (van den Hoff et al., 1999). Thereafter, significant apoptosis is apparent and the cushions are remodeled into mature, thin valve leaflets (Poelmann et al., 2000; Zhao and Rivkees, 2000; Abdelwahid et al., 2002). These latter processes have received relatively little attention and less is known about molecular pathways regulating late valve formation. Importantly, the four primary endocardial cushions develop asymmetrically, interacting with each other as well as with adjacent tissues including the crest of the ventricular septum and the spine of the atrial septum. These interactions correlate with and may determine alignment of the right and left AV valves, atrial and ventricular septation, and the proper relationship of the great arteries to the ventricular chambers.

In summary, cardiac valve development involves formation of early endocardial cushion swellings, endocardial–mesenchymal transformation and invasion of neural crest-derived mesenchyme, myocardialization, proliferation, and late stages of remodeling that include apoptosis and maturation of valve leaflets. An ever-expanding number of single gene mutations in mice are being identified in association with abnormalities in one or more of these processes. In many cases, these abnormalities mimic human congenital cardiac disorders. In this report, we have attempted to assemble relevant expression data concerning the most informative examples and we include data implicating additional pathways that are likely to be relevant for the assembly of molecular cascades regulating the formation of cardiac valves.


We performed radioactive in situ hybridization on mouse embryos at E9.5, E11.5, E13.5, and E16.5 to examine the temporal and spatial expression pattern of 13 genes previously implicated in cardiac development (Table 1). The results of our analysis are summarized below, and images of all markers at each time point are available for download at

Table 1. Summary of Endocardial Cushion Genes Analyzed
Gene nameGene functionReference
Del1αVβ3 integrin receptor ligandHidai et al., 1998
ErbB3Tyrosine kinase receptor (EGFR family)Erickson et al., 1997
Fog1Zinc finger transcription factorTsang et al., 1997; Tsang et al., 1998
Fog2Zinc finger transcription factorSvensson et al., 2000; Tevosian et al., 2000
Foxc1Winged-helix transcription factorKume et al., 2001
Foxc2Winged-helix transcription factorKume et al., 2001
Msx1Homeobox transcription factorChan-Thomas et al., 1993
Msx2Homeobox transcription factorChan-Thomas et al., 1993; Kwang et al., 2002
NeuregulinSecreted ligand for ErbB receptorsMeyer and Birchmeier, 1995
PDGFRαAlpha subunit of PDGF receptorMorrison-Graham et al., 1992; Schatteman et al., 1995
Runx2Homeobox transcription factorD'Souza et al., 1999; Kruzynska-Frejtag et al., 2001
Smad6Intracellular TGFβ pathway inhibitorGalvin et al., 2000
Sox4HMG-box transcription factorYa et al., 1998

In the figures below, we present a subset of the molecular markers at each time point in the OT and AV region of the heart. Neuregulin is a secreted ligand that binds to EGF/ErbB receptors and is required for normal myocardial and endocardial cushion maturation (Meyer and Birchmeier, 1995). At E9.5 (Fig. 1), neuregulin expression is evident in endothelium of the atria, ventricles, and cushion regions (arrows, Fig. 1A,B). However, the gene encoding the neuregulin receptor ErbB3 is barely detectible at this time in the heart (Fig. 1C,D). By E11.5, it is strongly expressed by cushion mesenchyme (arrows, Fig. 2A,B). Likewise, expression of another growth factor receptor gene, PDGFRα, is not yet evident in the cushion regions at E9.5 (arrows, Fig. 1E,F), although by E11.5, it is strongly expressed by cushion mesenchyme (arrows, Fig. 2E,F) and remains active in this tissue through gestation (Figs. 3E,F, 4E,F). Mutations in both ErbB3 and PDGFRα result in endocardial cushion defects (Morrison-Graham et al., 1992; Erickson et al., 1997; Camenisch et al., 2002). Smad6, encoding an intracellular mediator of TGFβ signaling, which is required for endocardial cushion development (Galvin et al., 2000), is evident in cushion mesenchyme at E9.5 (arrows, Fig. 1G,H) and thereafter (Figs. 2G,H, 3G,H, 4G,H).

Figure 1.

A–P: Endocardial cushion gene expression at embryonic day 9.5. Radioactive in situ hybridization of transverse sections of E9.5 mouse embryos at the level of the cardiac outflow tract (OT) and the atrioventricular canal (AV). lv, left ventricle; a, atrium. Arrows point to the endocardial cushions, except in B where the arrow points to atrial endocardium.

Figure 2.

A–P: Endocardial cushion gene expression at embryonic day 11.5. Endothelial–mesenchymal transformation is under way and the endocardial cushions are populated by mesenchyme. Arrows in C–F indicate endocardial cushions in the outflow tract (OT) and atrioventricular canal (AV). Arrows in O,P indicate endothelial cells expressing Fog1 lining the endocardial cushions.

Figure 3.

A–P: Endocardial cushion gene expression at embryonic day 13.5. The endocardial cushions have begun to remodel such that discrete valve leaflets are now apparent (arrows). The leaflets are composed of a mesenchymal core surrounded by a single layer of endothelial cells. lv, left ventricle; rv, right ventricle.

Figure 4.

A–P: Endocardial cushion gene expression at embryonic day 16.5. The valve leaflets have thinned and appear as nearly mature valve structures (arrows). lv, left ventricle; rv, right ventricle.

Del1 (Hidai et al., 1998; Fig. 1I,J), Runx2 (Ducy et al., 1997; Fig. 1K,L), and Msx1 (Satokata and Maas, 1994; Fig. 1M,N) represent a group of genes that are first expressed by endocardial cells at E9.5; thereafter, these genes are also expressed by mesenchyme of the cushions after endothelial–mesenchymal transformation has taken place (Fig. 2I–L and data not shown). By E13.5, Del1 and Msx1 are quite specific for the forming valve regions that are undergoing remodeling into thin valve leaflets (arrows, Fig. 3I,J,M,N). Msx2 (not shown), a homoebox gene closely related to Msx1 (Satokata et al., 2000), is also expressed by cushion mesenchyme, suggesting potential functional redundancy. Runx2 (also called Osf2,Cbfa1, or Periostin) is strongly expressed by cushion mesenchyme and endothelium but is also expressed elsewhere in the heart (Figs. 2K,L, 3K,L, 4K,L). Runx2 plays an important role in epithelial to mesenchymal transformation during tooth formation and in the development of osteoblasts, processes in which TGFβ signaling has been implicated. Runx2 may be an downstream effector of TGFβ signaling (Lee et al., 2000). Runx2 knockout mice have bone and tooth defects, while cardiac valvular defects have not yet been reported (D'Souza et al., 1999), although further analysis and tissue-specific inactivation may be warranted.

Friend-of-Gata-1 (Fog1) is a transcriptional corepressor for Gata transcription factors, which have been implicated in endocardial cushion and cardiac development (Reiter et al., 1999; Garg et al., 2003). Fog1 knockout mice succumb to an early hematopoietic defect such that its role in valve development remains unknown (Tsang et al., 1997). Of interest, Fog1 is expressed during mid-gestation by both endocardial cells adjacent to the cushions, and by cushion mesenchyme, although cardiac expression of Fog1 has not been reported previously. Endocardial expression is apparent by E11.5 (arrows, Fig. 2O,P), whereas mesenchymal expression is not clearly evident until E13.5 (Fig. 3O,P) and subsequently declines (Fig. 4O,P). Endocardial cushion expression of Fog1 is stronger than that of Fog2 (not shown), despite that Fog2 knockout mice have AV canal defects (Svensson et al., 2000; Tevosian et al., 2000). A specific role for Fog1 during cushion remodeling may emerge from tissue-specific gene targeting or transgenic rescue studies.

FoxC2 is a forkhead transcription factor implicated in aortic arch remodeling. Of interest, FoxC2 expression is evident throughout cushion mesenchyme and endothelium at E11.5 (Fig. 2M,N) but then becomes asymmetrically restricted to the endothelium lining just one side of the maturing valve cusps. In both the OT (arrows, Figs. 3C, 4C) and AV (arrows, Figs. 3D, 4D) regions, FoxC2 is present on the downstream side of the valve cusps, regions that are exposed to low shear stress. FoxC1, a closely related transcription factor gene that displays partial functional redundancy with FoxC2 (Kume et al., 2001), does not show this asymmetric endothelial expression pattern (Fig. 4A,B).

Sox4 is one of the few examples of genes identified to date that display differential expression patterns in AV vs. OT cushions. At E13.5, Sox4 expression is relatively strong in the OT cushion mesenchyme but weak in the AV region (arrows, Fig. 3A,B). By E16.5, expression levels have declined, but stronger expression in the OT leaflets compared with the AV region remains (Fig. 4I,J). Sox4-deficient mice exhibit outflow tract defects consistent with this expression pattern (Schilham et al., 1996; Ya et al., 1998).

A series of recent studies have implicated Wnt signaling in cardiac development (Eisenberg and Eisenberg, 1999; Marvin et al., 2001; Schneider and Mercola, 2001; Tzahor and Lassar, 2001). However, it is unknown if Wnt signaling plays a role in endocardial cushion development. We sought to determine whether downstream effectors of Wnt signals were active during endocardial cushion development by analyzing β-galactosidase expression in TOPGAL mice. These reporter mice are engineered to express β-galactosidase under the control of a Lef/Tcf and β-catenin inducible promoter and function to indicate tissues in which active Wnt signaling is present (DasGupta and Fuchs, 1999). By E11.5, robust β-galactosidase activity is present in portions of both OT and AV endocardial cushions (arrows, Fig. 5A,B). The restricted pattern of expression within the cushion tissue suggests that only a subset of mesenchymal cells within the endocardial cushions are actively responding to Wnt signals. By E12.5, the pattern has broadened somewhat (Fig. 5C), although some mesenchyme remains unstained. These findings suggest that Wnt signals emerging from adjacent tissues are actively signaling to cushion mesenchyme during critical periods of cardiac valve formation. Future studies will be required to identify the nature and requirements for specific Wnt signals and downstream pathways.

Figure 5.

Wnt/β-catenin signaling during endocardial cushion development. β-Galactosidase activity in TOPGAL mice, which express lacZ in regions of active Wnt signaling, is shown in the region of the outflow tract at embryonic day (E) 11.5 (A) and the atrioventricular canal (B). C: β-Galactosidase expression at E12.5 in the atrioventricular endocardial cushions is also shown. Wnt-responsive cells are restricted to a portion of the endocardial cushions (arrows).

We have presented a comparative expression analysis of a series of molecular markers of endocardial cushion formation that demarcate critical tissues at all relevant stages of embryonic development. Discrete and robust expression of individual genes supports a large body of work that implicates multiple growth factor signaling cascades, including TGFβ, PDGF, and EGF pathways. Our data additionally suggest that Wnt signaling is likely to impact cardiac valve development. Within cushion tissue, molecular markers indicate that neither endocardium nor cushion mesenchyme is homogeneous, and expression within each domain is dynamic. We hope that the data provided here will inform and direct future investigation of cardiac valve development.


In Situ Probes

Probes for Del1 (Hidai et al., 1998), erbB3 (Meyer and Birchmeier, 1995), FoxC1 and FoxC2 (Kume et al., 2001), Msx1 and Msx2 (Chan-Thomas et al., 1993), Neuregulin (Meyer and Birchmeier, 1995), PDGFRα (Morrison-Graham et al., 1992), Runx2 (Kruzynska-Frejtag et al., 2001), and Sox4 (Ya et al., 1998) have been described previously. The Fog1 and Fog2 probes corresponded to 678-1138 bp and 404-856 bp of the mouse Fog1 and Fog2 cDNAs, respectively. We generated antisense probes by linearization with HindIII digestion, followed by transcription with T7 polymerase. To generate the Smad6 riboprobe, we linearized a mouse expressed sequence tag cDNA clone (GenBank accession no. BI659871) with EcoRI and transcribed with T3 RNA polymerase.

Radioactive In Situ Hybridization

In situ hybridization protocols (Wawersik and Epstein, 2000) are available at The 35S-labeled antisense riboprobes were synthesized with SP6, T7, or T3 RNA polymerase and 35S-UTP. Hybridization was carried out at 55°C overnight. Slides were dipped in Kodak NTB-2 emulsion, exposed for 5 to 7 days at 4°C, and developed and fixed in Kodak Dektol developer and fixer. Cell nuclei were counterstained with Hoechst 33258 (Sigma, St. Louis, MO) and mounted in Canada balsam/methyl salicylate. In situs were photographed on a Zeiss Axiophot 2 microscope and images were processed with Adobe Photoshop.

X-Gal Staining

TOPGAL transgenic mouse embryos expressing β-galactosidase were harvested into cold PBS and fixed for 2 hr in 2% PFA. Embryos were incubated in X-gal staining solution (5 mM K4Fe(CN)6, 5 mM K3Fe(CN)6, 2 mM MgCl2, 0.01% NP-40, 0.1% deoxycholate, 0.1% X-gal substrate [Roche Molecular, Indianapolis, IN] in PBS) at 37°C. Stained embryos were paraffin embedded, sectioned, and counterstained with eosin.


We thank Ed Morrisey for providing TOPGAL mice and for pointing out lacZ expression in endocardial cushions. J.A.E. received funding from the NIH and A.D.G. is supported by the Department of Cell and Developmental Biology predoctoral training grant from the National Institutes of Health.