Communication between integrin receptors facilitates epicardial cell adhesion and matrix organization


  • So Hyun Pae,

    1. Northwestern University, Feinberg School of Medicine, Department of Pediatrics, Neonatology Research Laboratory, Chicago, Illinois
    Search for more papers by this author
  • Danijela Dokic,

    1. Northwestern University, Feinberg School of Medicine, Department of Pediatrics, Neonatology Research Laboratory, Chicago, Illinois
    Search for more papers by this author
  • Robert W. Dettman

    Corresponding author
    1. Northwestern University, Feinberg School of Medicine, Department of Pediatrics, Neonatology Research Laboratory, Chicago, Illinois
    2. Northwestern University, Feinberg Cardiovascular Research Institute, Chicago, Illinois
    3. Children's Memorial Research Center, Developmental Biology Core, Chicago, Illinois
    • 303 E. Chicago Ave., Neonatology, Ward 12-191, Chicago, IL 60611
    Search for more papers by this author


Formation of the epicardium requires interactions between α4β1 integrin, and the extracellular matrix. We investigated the role of other integrins expressed by epicardial cells. We detected transcripts for α5, α8, αv, β1, β3, and β5 integrins in the chick proepicardial organ (PE). We demonstrate that α5β1, α8β1, and αvβ3 integrins are expressed by chick epicardial mesothelial cells (EMCs). Migration of EMCs in vitro was reduced by RGD-containing peptides. Using adenoviruses expressing an antisense to chick α4 (AdGFPα4AS), full-length (Adhα4V5), and C-terminal deleted α4 (Adhα4ΔCV5), we found that EMCs were less able to adhere to vitronectin and fibronectin120 indicating that α4β1 plays a role in regulating EMC adhesion to ligands of α5β1, α8β1, and αvβ3. In Adhα4ΔCV5-infected EMCs, α5β1 was diminished in fibrillar adhesions and new FN matrix assembly was abnormal. We propose that cooperation between α4β1 and RGD integrins is important for EMC adhesion and subepicardial matrix formation. Developmental Dynamics 237:962–978, 2008. © 2008 Wiley-Liss, Inc.


Formation of the visceral pericardium or epicardium is critical to normal heart development (Manner et al.,2001; Reese et al.,2002). The epicardium forms the third layer of the early heart tube as mesothelial cells from the proepicardial organ migrate cranially and superficially to cover the myocardium (Manasek,1969; Ho and Shimada,1978; Manner,1992; Viragh et al.,1993). In avian, amphibian and rat embryos, EMCs migrate over the myocardium as a mesothelial sheet contiguous with the proepicardium (Hiruma and Hirakow,1989; Fransen and Lemanski,1990; Nesbitt et al.,2006). In fish and mice, EMCs migrate from aggregates of proepicardial cells that detach from the proepicardium (PE) and float across the pericardial cavity to attach to the myocardial surface (Komiyama et al.,1987; Kuhn and Liebherr,1988; Munoz-Chapuli et al.,1996). In either case, once attached to the heart PE cells migrate as mesothelial groups with individual cells joined by adherens and tight junctions (Wada et al.,2003; Compton et al.,2006; Dokic and Dettman,2006) as well as desmosomes and gap junctions (Li et al.,2002).

Superficial epicardium performs several essential roles in cardiac development. By signaling to the underlying myocardium, it stimulates myocardial cell proliferation in the compact zone (Sucov et al.,1994; Chen et al.,2002; Pennisi et al.,2003; Stuckmann et al.,2003). Signals received from the myocardium also influence extracellular matrix formation in between the epicardium and myocardium (Jenkins et al.,2005). Superficial epicardial mesothelial cells are a population of multipotent cells that undergo epithelial-mesenchymal transformation (EMT) to invade deeper aspects of the developing heart (Dettman et al.,1998; Gittenberger-de Groot et al.,1998; Manner,1999). Invading epicardially-derived cells move throughout the heart and become coronary vascular endothelial, smooth muscle, and pericyte cells, as well as cardiac fibroblasts and atrioventricular cushion cells (Dettman et al.,1998; Perez-Pomares et al.,2002a,b).

The developmental potential of the PE and epicardium (Perez-Pomares et al.,1997,1998; Gittenberger-de Groot et al.,1998) has been extensively analyzed but little is known about molecular pathways that control epicardial development. Our hypothesis is that interactions between EMCs and their extracellular environment regulate their development. This hypothesis is supported by several studies. In avian hearts, the PE grows cranially from the septum transversum by Hamburger and Hamilton stage (HH) 16 of development to attach to the inner curvature of the heart (Manner et al.,2001; Nahirney et al.,2003). Even before PE cells interact with the myocardium they interact with extracellular matrix in the pericardial coelom (Nahirney et al.,2003). After PE cells attach to myocardium and transition to EMCs, superficial migration requires cell-matrix interactions. This was demonstrated by the targeted deletion of α4 integrin (α4) in the mouse (Yang et al.,1995; Sengbusch et al.,2002). With β1 integrin, α4 forms a receptor for fibronectin (FN) and vascular cell adhesion molecule (VCAM-1) and in the developing heart α4β1 is exclusively expressed in the epicardium. Loss of function of α4 resulted in cardiac defects including pericardial effusion, failure of proepicardial aggregate attachment to the myocardium, and absence of coronary vessels (Yang et al.,1995; Sengbusch et al.,2002). Similarly, in hearts deficient for the myocardially-expressed ligand VCAM-1, the epicardium was missing and coronary vascular development failed to occur (Kwee et al.,1995). These comparable phenotypes showed that epicardial α4 integrin is essential for EMC migration and adhesion of the epicardium to the myocardium.

While loss of α4 and VCAM-1 in mouse heart development prevents epicardial formation and adhesion, interaction of the epicardial mesothelium with the myocardium is likely to be complex and involve other cell–cell and cell–matrix interactions. In VCAM-1-deficient hearts, FN is still expressed and could still mediate adhesion to α4β1. This is supported by the observation of small patches of epicardial cells on the atrial surface of VCAM-1-deficient hearts (Kwee et al.,1995). In addition, while α4 null proepicardial cells cannot attach to VCAM-1, they were capable of binding to FN in vitro (Sengbusch et al.,2002). These observations suggest that in the mouse proepicardial aggregate, adhesion to the heart is not exclusively mediated by α4β1 integrin.

Given that loss of α4 produces profound alterations to normal epicardial development and that EMCs adhere to FN in the absence of α4β1 expression, we set out to define the role of other integrins in epicardial development. We focused on the class of integrins that bind to the RGD core sequence in matrix proteins like FN, vitronectin (VN), and tenascins (TN), which have been shown to be expressed in subepicardial extracellular matrix (Burch et al.,1995; Bouchey et al.,1996). Here we report that EMCs express transcripts for four RGD binding integrins in addition to α4β1. These integrins are α5β1, α8β1, αvβ3, and αvβ5. In cultured EMCs, we found that α5β1 is found in cell-substratum attachment sites resembling focal complexes and fibrillar adhesions, sites implicated in cells to catalyze FN fibrillogenesis (Pankov et al.,2000; Clark et al.,2005). We found expression of αvβ3 in cell surface puncta resembling focal complexes and we identified α8 integrin in chick EMCs in arrays that resembled cytoplasmic actin fibers. RGD peptides changed the migratory properties of chick EMCs implicating a role for these integrins in EMC migration. We tested if α4 can regulate the function of RGD binding integrins in EMCs. We found that by altering α4 levels or function in chick EMCs, we modified adhesion to VN and a variant of FN that cannot bind α4. In addition, expression of a C-terminal deleted variant of human α4 reduced α5β1 localization in focal complexes and fibrillar adhesions as well as impaired new FN fibrillogenesis. The observations presented here support a cooperative role for α4, α5 and αv integrins in epicardial cell adhesion, migration and FN matrix deposition.


Proepicardial Expression of RGD Binding Integrins

Based on the observation that α4 null PEs adhered to FN in vitro (Sengbusch et al.,2002), we predicted that PE cells express additional integrin receptors. Since in embryos proepicardial cell migration is likely to involve FN (Nahirney et al.,2003), VN (Bouchey et al.,1996), and TN (Burch et al.,1995), we focused on integrins that bind to the core RGD sequence common to these molecules. Several integrins have been identified as primarily binding to RGD sequences in FN, VN, TN, and in other ECM proteins (Ruoslahti,1996; Hynes,2002). While forming various dimeric receptors, the integrins that form receptors for the RGD core sequence are α5, α8, αv, β1, β3, β5, β6, and β8. Sequences are available for all chick integrin orthologs except for α5 integrin. Using gene-specific primers that identified alternate exons in integrin transcripts, we did reverse transcriptase PCR (RT PCR). Total RNA was isolated from PEs dissected from HH16 chick embryos. Attached PEs were not used for RNA isolation. We were able to amplify transcripts for α8, αv, α4, β1, β3, and β5 but not β6 or β8 (Fig. 1A). While α4, α5, and α8 pair with β1, αv pairs with either β3 or β5 integrins. We observed that the band for β5 was less abundant than that of β3 suggesting that αvβ5 contributes less to the αv integrin pool. Together, these observations indicated that besides the known expression of α4 and β1, epicardial PE cells express transcripts for αv, α5, α8, β3, and β5 integrins.

Figure 1.

Amplification of integrin transcripts in chick proepicardial organs. Amplification products from total RNA isolated from chick PEs after 35 cycles. Isoforms are indicated above each lane. Markers sizes are indicated at the left. Expected band sizes were: α8: 343 bp, αv: 672 bp, α4: 802 bp, β1: 402 bp, β3: 476 bp, β5: 380 bp, β6: 310 bp, β8: 460 bp, GAPDH: 579 bp.

Localization of α5, α8, and αv Integrins in Epicardial Cells

To demonstrate protein expression of integrins on the surface of chick EMCs, we performed immunostaining on monolayers of cells grown on collagen-coated coverslips (Fig. 2). EMCs cultured in this manner continue to express markers of undifferentiated EMCs such as cytokeratins and Wilm's tumor factor (WT1) (Dettman et al.,1998; Dokic and Dettman,2006). When imaging, we focused on two portions of the monolayer (Fig. 2A). These were at the leading edge of the monolayer (e.g., the blue box in Fig. 2A) and in the middle of the monolayer (e.g., the green box in Fig. 2A). Migrating cells at the leading edge were imaged to determine localization of integrins at the front of cells near lamella (e.g., the orange box in Fig. 2B) and at the rear of cells (e.g., the red box in Fig. 2B). Cells in the middle of the monolayer were imaged to determine if the organization of integrins in cell-matrix adhesions differs in this portion of the mesothelium. In these experiments, we incubated antibodies with living monolayers in culture medium prior to fixation. This allowed us to determine if antibodies were interacting with cell surface integrins.

Figure 2.

Accumulation of α5, αv, and α8 integrins in migrating chick EMCs. A,B: Phase contrast images of EMCs that migrated from a HH24 heart to the surface of a collagen-coated coverslip. The heart is visible at top left (A). The green box in A indicates the “middle” region of the monolayer and the blue box in A indicates the “edge” of the monolayer. The orange and red boxes in B surround regions of a cell at the front of the migrating mesothelium. The orange box is over the lamellum and the red box surrounds both the leading and trailing edge of a cell at the migrating front. Thus, cells in C, D, I, J, O, P are typical of the red box; cells in E, F, K, L, Q, R are typical of the orange box; cells in G, H, M, N, S, T are typical of the green box. Positions of cells are indicated at the bottom of the figure as edge, front of edge, or middle. C–T: Pairs of images with a phase contrast image on the left and a fluorescent image on the right. In C–H, living EMCs as in A were incubated with U1α for 1 hr, washed, fixed, and incubated with goat anti-mouse Alexa 488. In I–N, living EMCs were incubated with LM609 for 1 hr, washed, fixed, and incubated with goat anti-mouse Alexa 488. In O–T, EMCs were fixed, permeabilized with 0.1% (v/v) triton detergent and then incubated with α8-cyto and visualized with goat anti-rabbit Alexa 488. Magnification bars = 500 μm (A); 20 μm (B, G, H, M–P, S, T); 10 μm (C, D, I, J, Q, R); 5 μm (E,F,K,L).

To detect α5β1 receptors on the surface of cultured chick EMCs, we utilized a monoclonal antibody raised against the chick α5 subunit. This antibody (U1α) recognizes the extracellular domain of α5 when it is in its high affinity conformation for FN (Boettiger et al.,1995). In primary chick EMC cultures, we observed that U1α reacted with EMCs and that it stained specific sub-cellular structures (Fig. 2C–H). In cells at the edge of monolayers, we observed that U1α reacted with punctate structures at the front of cells near lamella and in longer fibrillar structures at the rear of cells (Fig. 2D, F). In middle regions of monolayers, U1α stained primarily fibrillar structures throughout the cell (Fig. 2H). Remarkably, during the 1-hr incubation of monolayers with U1α, all of the cells in the monolayer were labeled to an equal degree, indicating that throughout, structures containing α5β1 are dynamically organized. We observed similar localization of α5β1 in EMCs fixed before incubation with U1α although staining was slightly less intense and less organized due to fixation (see Fig. 6).

Figure 6.

Accumulation of α5, α8, and αv integrins in EMCs knocked down for α4 integrin. A,B: Diagrammatic representation of an EMC monolayer with infected cells (green). The box in B represents a typical microscopic field analyzed. CN: Confocal microscopy of EMCs infected with AdGFPlacZ (C, D, G, H, K, L) and AdGFPα4AS (E, F, I, J, M, N) for three days. Infected cells were fixed and stained with U1α (C–F), α8-cyto (G–J), and LM609 (K–N). Magnification bars = 20 μm.

The organization of α5β1 in chick EMCs was reminiscent of human α5β1 in human foreskin fibroblasts (HFFs) where it was shown that α5-integrins are arranged into focal adhesions, which anchor stress fibers and fibrillar adhesions, which bind FN fibrils aligned in parallel to actin bundles (Zamir et al.,1999). α5β1 is first organized into focal adhesions and then is translocated along stress fibers into fibrillar adhesions (Pankov et al.,2000). If U1α is labeling adhesion complexes similar to those described in HFFs, then we predicted that in shorter times of incubation of cells with U1α we would observe labeling primarily in puncta and not in fibrillar structures. We treated monolayers with U1α for 10 min and then washed off unbound antibody. Coverslips were fixed after 15-min intervals up to 1 hr. Here we observed that after 10 min (time 0), U1α labeled mostly puncta (see Supplemental Fig. 1A, E, which can be viewed at and that fibrillar structures became predominant by 45 min (Suppl. Fig. 1C, G). From these experiments, we concluded that chick EMCs express high-affinity α5β1 receptors and they are dynamically organized on their surface in puncta resembling focal complexes and in fibers similar to fibrillar adhesions.

We used the monoclonal antibody LM609 (Cheresh,1987) to determine if αvβ3 receptors are expressed on the surface of chick EMCs (Fig. 2I–N). As with U1α, we incubated EMCs with LM609 (10 μg/ml) for 1 hr prior to fixation. Here we observed staining at the edge of monolayers in puncta similar to those described with U1α staining (Fig. 2J, L). In the lamella of these cells, there were discernable puncta that were even smaller than those detected underneath the cell body and the trailing edge. These smaller puncta were detected also in filopodia extending in front of the lamella (Fig. 2K,L arrowheads). In addition, puncta were particularly apparent toward the trailing portion of cells at the edge. However, unlike α5β1, αvβ3 did not appear in fibrillar structures (Fig. 2L, N). In cells in the middle of the monolayer, αvβ3 was detected uniformly throughout the cell in small puncta. Thus, αvβ3 is expressed on the surface of EMCs and while it appears to change its organization within the monolayer, it does not appear in fibrillar adhesions.

To confirm the presence of α8 protein in chick EMCs, we used a polyclonal antiserum raised against the C-terminus of human α8-integrin: α8-cyto (Bossy et al.,1991). In these experiments, cells were cultured on coverslips as above. However, since α8-cyto recognizes the C-terminus, we could not use it to demonstrate surface expression of α8, so EMCs were fixed and permeabilized prior to incubation with α8-cyto. EMCs were stained by this antiserum throughout the cytoplasm primarily in a fibrillar organization (Fig. 2P, R, T). However, these fibrillar structures filled the entire cell regardless of whether cells were on the edge or middle of the monolayer and they were reminiscent of actin fiber arrays and not fibrillar adhesions described in other cell-types or as detected in EMCs by U1α. In contrast to α5β1 and αvβ3, we did not observe staining with α8-cyto in punctate structures. While we could not demonstrate surface expression of α8 with this antibody reagent, staining with α8-cyto confirmed the accumulation of α8 in EMCs. Thus, integrin protein expression in cultured EMCs is in agreement with the mRNA data in PEs in Figure 1. Furthermore, each of the integrins we detected appeared in specific sub-cellular locales suggesting that each receptor performs different roles in EMCs.

We tested if α4 integrins co-localized with α5β1, α8β1, or αvβ3 in chick EMCs. We tested several commercially available antibodies to human and mouse α4 on sections of HH28 embryos and identified one that reacted with the chick epicardium (Fig. 3C). This was a polyclonal antiserum raised against an N-terminal epitope of human α44N-19). We tested this antiserum on living chick EMCs and found that it labeled the surface of chick EMCs (Fig. 3B). When cells were incubated with α4N-19 and either U1α or LM609 (Fig. 3A, B, G–I) simultaneously, we observed that staining for α4 was in areas where αv puncta were most abundant and in areas where α5 fibrillar adhesions were appearing. However, α4 was not overtly localized to either puncta, fibrillar adhesions or along actin fibers. Thus, while α5β1, α8β1, and αvβ3 accumulate in distinct sub-cellular locations, α4 is found in broader areas that overlap with these localized structures.

Figure 3.

Comparison of α5, α8, and αv localization to α4, focal adhesion and actin filament localization. Confocal micrographs of chick EMCs stained with U1α (A, D–F), N-19 (B, C, H, I), LM609 (G, I–L), α8-cyto (M–R), Texas Red-X phalloidin (N, O), and anti-β-catenin (P–R). A, B: EMCs at the edge of the monolayer stained for α5β1 (green) and α4 (red). C: Coronal section of an HH28 chick heart stained with N-19 showing epicardial (Epi) staining (arrow). Myocardium is indicated (Myo). DF: Images of an EMC at the edge of the monolayer expressing GFP-paxillin and stained for α5β1 (red). Note the trail of α5β1 integrin behind the GFP-paxillin expressing cell (arrows). E, F: Higher magnification of the cell in D. α5β1 staining is predominant in the trailing portion of the cell (E) and is distinct from GFP-paxillin accumulation in focal adhesions throughout the cell (F). GI: Images of an EMC at the edge of the monolayer stained for αv (green, G) and α4 (red, H). Merged image is shown in I. JL: Images of an EMC expressing GFP-paxillin and stained for αvβ3 (red). K, L: Higher magnification of the cell in J. Punctate αvβ3 staining (K) is distinct from GFP-paxillin accumulation in focal adhesions throughout the cell (L). MO: Images of an EMC at the edge of the monolayer stained for α8 (green) and actin (red). At lower magnification α8 (M) appears to coincide with actin stress fibers (N). At higher magnification, α8 appears as green speckles on the red actin fibers (O). PR: Images of EMCs in the middle of the monolayer stained for β-catenin (red) and α8 (green). Arrowheads point to α8 staining reminiscent of cortical actin filaments. Magnification bars = 6 μm (O); 10 μm (E, F, K, L–N); 20 μm (A–D, G–J); 33 μm (Q, R); 50 μM (P).

We performed localization studies in conjunction with markers for other cytoplasmic structures to more carefully define the sites in which each of these integrins accumulates. To determine if the sites where α5β1 and αvβ3 accumulate are focal adhesions, we transfected chick EMCs with a plasmid expressing human paxillin fused to green fluorescent protein (GFP). GFP-paxillin was demonstrated to be an excellent marker for classical focal adhesions but did not localize in fibrillar adhesions in HFFs (Zamir et al.,2000). We observed that GFP-paxillin was organized into focal adhesions (Fig. 3D–F, J–L). U1α or LM609 did not distinctly label these structures indicating that α5 and αv are not primarily organized into focal adhesions in migrating EMCs. However, on occasion there was minor accumulation of α5β1 and αvβ3 in puncta in the vicinity of GFP-paxillin containing focal adhesions. We made a similar observation in EMCs stained with LM609 and a polyclonal antibody to chick vinculin. Here, we occasionally observed αvβ3 puncta near vinculin-containing structures (Suppl. Fig. 2).

To determine if α8 fibrillar staining was similar to actin filaments, we co-stained with Texas-Red X phalloidin and performed confocal microscopy (Fig. 3M–O). At lower magnifications, α8 appeared to coincide with actin stress fibers (Figs. 2P, R, T, 3M, N). However, at higher magnifications this staining appeared as speckles that decorated the actin filaments (Fig. 3O). Cortical actin filaments are adjacent to cell junctions but do not overlap with junctional complexes. To mark intercellular junctions, we stained monolayers with β-catenin, a protein component of the adherens junction (Fig. 3P–R). This allowed us to distinguish α8 in association with cortical actin next to adherens junctions in EMCs (Fig. 3P–R, arrowheads). Thus, α8 appears in EMCs in association with stress fibers and cortical actin filaments. Together, the observations presented in Figures 2 and 3 confirmed our initial observations with RT-PCR and demonstrated that EMCs express at least three RGD binding integrin receptors in addition to α4β1: α5β1, α8β1, and αvβ3.

α5β1 Accumulates on the Basolateral Surface of the Epicardium and in Coronary Vascular Endothelium

We investigated if integrins expressed by cultured EMCs were present in the developing heart and if they were limited to epicardium. To do this, we stained sections of chick hearts with antibodies to α5, α8 and αv. When HH20 hearts were stained with U1α, we were surprised to see that it was not expressed throughout the PE (Fig. 4A). While all cells reacted with pan anti-cytokeratin, only cells in the center of villous PE protrusions reacted with U1α (Fig. 4A, B). PE cells on the heart, but separated from myocardium by mesenchymal space, did not react with U1α, whereas a small number of subepicardial mesenchymal cells did (Fig. 4C). On the ventricular myocardium, ventral to the PE, we observed epicardium that did not have distinctive staining with U1α (Fig. 4D). These epicardial cells were loosely organized and were not flattened. However, in regions even further from the PE, where epicardial cells were flattened, there was a clear basolateral, staining pattern (Fig. 4F). This bright staining formed a distinct stripe underneath the epicardium (Fig. 4D–F). U1α stained other cells in the HH19 heart; in particular the myocardium and endocardium reacted with this antibody. Myocardial staining was discernable but not as intense as subepicardial and endocardial staining (Fig. 4C–E). Staining was present in the endocardium (Fig. 4D). We observed that in coronary vessels of hearts older than HH19, U1α reacted with the vascular endothelium. This was apparent until late embryogenesis into stage HH40 where U1α reacted with endothelial cells of both coronary arteries and veins (Fig. 4G, H). Like staining in epicardium closely associated with myocardium, α5 appeared to be basolateral to vascular endothelial cells (Fig. 4H). Thus, α5β1 is expressed throughout the heart during development, in endocardium, myocardium, and epicardium as well as in coronary vascular endothelium.

Figure 4.

Localization of α5, α8, and αv in the developing heart. A–E: Sagittal sections of an HH20 chick embryo showing stained with anti-cytokeratin (red), U1α (green) and DAPI (blue). A: Dorsal-caudal portion of the pericardial coelom showing the PE (arrowheads). B: Magnification of area shown in (A) showing staining for α5β1 in cells in the center of PE villi. C: More ventral view of the same heart showing staining for α5β1 in myocardium (Myo) and in a mesenchymal cell in the subepicardium (arrowhead). D: Lower magnification and more ventral view of epicardium on the myocardium. α5β1 staining is in the myocardium and in epicardium, but only where cells have begun to flatten on the heart (arrow). E: Ventral view of same heart over the left ventricle where a distinct stripe of α5β1 staining is observed between the myocardium and epicardium (arrows). F: Higher magnification of panel (E) showing α5β1-staining basolateral to epicardial cells (arrowheads). G: Coronal section of an HH40 heart showing α5β1-staining in a coronary artery and vein (asterisks). Staining is apparent in the vascular endothelium (VEC) and in perivascular cells but not in coronary vascular smooth muscle (VSMC). H: Higher magnification of artery in G showing α5β1-staining basolateral to vascular endothelial cells. Arrowheads point to cells in section where cytoplasmic stain surrounds the nucleus. I, J: Sagittal section of an HH19 chick heart showing accumulation of α8 (green) myosin heavy chain (red) and nuclei (blue). α8 appears in both the epicardium (Epi) and myocardium (Myo). K, L: Sagittal section of an HH19 chick showing accumulation of αv (green) myosin heavy chain (red) and DAPI (blue). J, L: Identical images of I and K shown without the green channel to highlight myocardial staining. I: Magnification bars = (A, H, I–L) 50 μm; (B,C,E,G) 20 μm; (D) 100 μm; (F) 7 μm; (H) 16 μm.

Using α8-cyto and an antiserum to the cytoplasmic tail of human αvv-cyto), we observed that α8 and αv integrins are expressed in the developing heart (Fig. 4I–L). Antisera to α8 and αv reacted with myocardium as well as epicardium and staining was more uniform than was observed for α5. In HH19 embryos, staining with α8-cyto was brighter in the heart than in extracardiac regions (data not shown). Staining was observed in the epicardium, although it was not as bright as in the myocardium (Fig. 4I). αv-cyto reacted with both myocardium and epicardium in HH19 hearts (Fig. 4K). Together these experiments confirm the presence of α5, α8, and αv integrins in the developing heart and support the hypothesis that they play a role in EMC migration on or adhesion to the developing myocardium.

EMCs Utilize RGD Binding Integrins During Migration

EMCs spontaneously migrate to the surface of tissue culture plates from both explanted proepicardial organs and HH24 hearts (Fig. 2A), so we used this as an assay to test the role of integrins during EMC migration. Several integrins interact with the cell-binding domain of FN, which contains the RGD core sequence. Peptides that compete for binding of integrins to the RGD core sequence are sufficient to interfere with adhesion of integrins to FN to block cell migration (Ruoslahti,1996). α4β1 binds to distinct domains of FN including the CS-1 and HepII domains and its interaction with FN can similarly be inhibited by small peptides. To test the roles of integrins in EMC migration, we explanted HH24 hearts onto uncoated tissue culture plastic in the presence of blocking peptides to the cell-binding domain (GRGDSP) and CS1-domain of FN (EILDVPST). Since plates are charged but not coated, EMCs must organize new extracellular matrix during migration. In preliminary experiments, we observed that there were small differences between monolayers produced by individual hearts under the same conditions of migration. Most importantly, there was variability in the distance that the monolayer migrated from a given region of the heart. Thus, in order to quantify the migration of EMCs from HH24 hearts, we measured the area of the dish covered by each monolayer. When migration is greater, the area of the monolayer will be larger and when migration is reduced, the area covered by EMCs will be smaller. The applied method is summarized in Figure 5A–D.

Figure 5.

Reduced migration of chick EMC in the presence of blocking peptides. A: Schematic summary of the cell migration assay used in this study. B: Formula showing the calculation used to determine the area over which EMCs migrated. C: Representative phase contrast image of a monolayer generated from an explanted heart. D: Representative image of a hematoxylin-stained monolayer imaged for quantification. E: Bar graph showing the results of EMC migration under identical conditions with the addition of 1, 10, 50, and 100 μM GRGDSP (RGD), GRADSP (RAD), EILDVPST (LDV), and EILEVPST (LEV) peptides. Treatment and concentration is indicated under the X-axis. Sample size is indicated at the base of each bar in the graph. Error bars represent standard error of the mean. Significance was determined using a paired Student's t-test. P values in (E) are *P < 0.05, **P < 0.01, †P < 0.05, and ††P < 0.05.

We tested four concentrations of GRGDSP and observed that this peptide reduced EMC migration from HH24 hearts in a dose-dependent manner (Fig. 5E, light gray bars). This decrease became statistically significant at 10 μM and was maximal at 100 μM where it decreased from 3.41 mm2 (untreated) to 2.21 mm2. This represented an overall decrease in area of 35%. EMC migration assays in the presence of a control peptide (GRADSP) were not significant (Fig. 5E, white bars). The CS1 peptide (EILDVPST) was able to decrease migration of EMCs with statistical significance at concentrations of 50 and 100 μM (Fig. 5, black bars). The maximum effect on migration with CS1 was at 50 μM yielding a decrease from 3.41 mm2 (untreated) to 2.59 mm2 or a 24% decrease in area. The average effect observed at 100 μM was almost the same as at 50 μM suggesting that the maximal affect may have been achieved. The control peptide for CS1 (EILEVPST) did not have a statistically significant effect on migration at any concentration tested (Fig. 5E, dark gray bars). While we have not ruled out a minor additive component due to cell proliferation, our results indicated that in addition to the known requirement for α4β1 in mouse EMC migration, integrins that bind ECM proteins with the core RGD sequence play a role in EMC migration.

α4β1 Influences Chick EMC Adhesion to VN and FN120

Having established that EMCs express integrins other than α4β1 and that these integrins affected their migratory behavior, we wanted to know whether loss of α4 is sufficient to alter the expression of other integrins in EMCs. To test this, we utilized an adenovirus (AdGFPα4AS) that expresses a 1.3-Kb antisense transcript to chick α4 (Dettman et al.,2003). To determine if loss of α4 resulted in altered expression of α5, α8, or αv integrins, we cultured chick EMCs from HH24 hearts on collagen-coated coverslips, removed the hearts and then infected the cells with AdGFPα4AS for three days. Monolayers were cultured in low-serum medium to limit spontaneous EMT. Monolayers were then incubated with U1α, LM609, and fixed. To determine if α8 expression was altered in AdGFPα4AS-infected cells, monolayers were fixed and stained with the α8-cyto antiserum. Control monolayers were infected with AdGFPlacZ. We observed that AdGFPα4AS-infected EMCs all continued to express α5, α8, and αv integrins in the patterns we had observed in uninfected cells (Fig. 6). The observation that integrins α5, α8, and αv are still expressed in epicardial cells infected with AdGFPα4AS indicates that downregulation of α4 does not affect the expression of these integrins. If this follows in the mouse, it would support the hypothesis that impaired adhesion or migration of EMCs in the absence of α4 is not due to concomitant loss of α5, α8, and αv but may be due to an alteration to the ability of these integrins to participate in EMC adhesion.

One manner in which α4 could alter PE adhesion in the presence of other integrins is by changing the ability of these integrins to bind to their ligands. This could occur either through the presence of α4 on the surface of EMCs or by signaling through α4β1. To test this, we constructed two recombinant replication incompetent adenoviruses expressing variants of human α4 (Fig. 7A). One was designed to express full-length human α4 fused to a C-terminal V5 epitope tag (Adhα4V5). The other was designed to express a C-terminal truncated version of human α4 replacing the cytoplasmic domain of α4 with a V5 epitope tag (Adhα4ΔCV5). In both adenoviruses, a 17–amino acid linker sequence (Fig. 7A, underlined sequence) was placed in between the C-terminal amino acid of α4 and the V5 sequence (Fig. 7A, bold sequence). We confirmed co-expression of α4 and V5 in human 293FT cells after infection with either virus (Suppl. Fig. 3). In addition, infection with Adhα4V5 or Adhα4ΔCV5 increased 293FT cell adhesion to culture dishes coated with soluble VCAM-1 indicating that the expressed α4 variants form active adhesion receptors (Suppl. Fig. 3).

Figure 7.

Altered adhesion of EMCs to VN and FN120 after α4 modification. A: Graphic representation and single letter amino acids for the intracellular domain of human α4 encoded by the adenoviruses Adhα4V5 and Adhα4ΔCV5. Underlined amino acids are the spacer peptide included in the AdCMV/V5/Dest vector. Bold amino acids represent the V5-epitope tag. Extracellular, transmembrane (PM) and intracellular domains of the α4 variants are indicated. B,C: Cell adhesion assays of chick EMCs adhered to VN (B) or FN120 (C). The Y-axis of both graphs represents the number of infected cells bound per mm2. The viruses used to infect cells are indicated underneath the X-axis. Bars = standard error. Sample size is indicated at the bottom of each bar. P values are *P < 0.001 and **P < 0.002 in B and *P < 0.02 and **P < 0.02 in C.

To determine if downregulation of α4 had an effect on EMC adhesion to putative extracellular matrix ligands of α5β1, α8β1, and αvβ3, we performed cell adhesion assays (Fig. 7B,C). Adhesion to VN was done to assay if alteration of α4 affected αvβ3 integrin mediated adhesion. FN120 is an alpha chymotryptic fragment of FN that acts as a substrate for integrins that recognize the RGD domain, but lacks the CS1 and HepII domains recognized by α4β1. Thus, we used FN120 as an independent test of EMC adhesion to integrin ligands not mediated by α4β1. For example, FN120 can interact with α5β1 (Rabinowich et al.,1996) and α8β1 (Muller et al.,1995). Here, we observed that downregulation of α4 (by infection with AdlacZα4AS for 3 days) altered the ability of chick EMCs to adhere to both VN (Fig. 7B) and FN120 (Fig. 7C). Although both experiments were significantly different from the AdGFPlacZ controls, the effect on adhesion to FN120 was smaller (Fig. 7C). AdGFPlacZ-infected cells adhered to FN120 at a density of 49.9 cells per mm2 and AdlacZα4AS-infected cells adhered to FN120 at a density of 43.2 cells per mm2. AdGFPlacZ-infected EMCs adhered at a density of 47.9 cells per mm2 and AdGFPα4AS-infected EMCs adhered at a density of 13.3 cells per mm2. Thus, the effect of AdGFPα4AS was about 3.5-fold greater on VN (Fig. 7B). This much larger decrease in adhesion supports the hypothesis that α4 expression in EMCs is important for αv integrin-mediated adhesion to VN.

Deletion of the cytoplasmic tail of α4 has been used as a strategy to test α4 signaling in several studies (Bittner et al.,1998; Hsia et al.,2005). In these studies, truncated α4 is capable of forming cell surface–expressed, dimeric receptors that mediate cell adhesion but not signaling since α4 cannot interact with cytoplasmic adapter proteins like paxillin. Chick EMCs were infected with Adhα4ΔCV5 and tested in cell adhesion assays to VN and FN120. We observed that expression of the C-terminal deleted form of α4 had a dominant effect on binding to VN but not FN120 (Fig. 7B,C). EMCs infected with Adhα4ΔCV5 adhered to VN-coated wells at a density of 25.6 cells per mm2. Compared to cells infected with AdGFPlacZ, this was a two-fold decrease in cell adhesion. Cells infected with Adhα4V5 adhered at a density of 45.5 cells per mm2, but this was not statistically different than the control. These observations support the hypothesis that expression of a C-terminal deleted variant of human α4 in chick EMCs significantly alters their ability to adhere to VN. However, overexpression of full-length human α4 does not significantly alter binding to VN. Infection with Adhα4ΔCV5 had no effect on EMC binding to FN120, with infected cells adhering at a density of 49.6 cells per mm2 relative to 49.9 cells per mm2 for AdGFPlacZ-infected cells. Surprisingly, the virus that had the largest effect on EMC binding to FN120 was Adhα4V5, which reduced adhesion to 38.7 cells per mm2. This effect was statistically significant and was not observed in Adhα4ΔCV5-infected cells. Thus, by carrying out EMC adhesion assays, we observed that binding to FN120 was sensitive to altering levels of α4 and binding to VN was sensitive to loss of α4 or expression of C-terminal deleted human α4 but not overexpression of human α4.

α4β1 Influences Organization of EMC Fibrillar Adhesions and FN Matrix

While infection with AdGFPα4AS did not alter expression of α5, α8, and αv, it was possible that expression of either full-length or C-terminal deleted human α4 modified the expression of these integrins in chick EMCs. To test this, we made primary cultures of chick EMCs and infected them with Adhα4V5 or Adhα4ΔCV5 for two days. Cells were incubated with either U1α or LM609, fixed and stained with anti-V5 to detect infected cells. When staining with α8-cyto, cells were fixed prior to incubation with antibodies. We observed that expression of α8 and αv integrins was unchanged either when cells were infected with Adhα4V5 or Adhα4ΔCV5 (Fig. 8E–L). In addition, combined with the AdGFPlacZ results (Fig. 6), we did not detect any changes to α8 or αv protein expression or localization caused by adenoviral infection alone. In contrast, localization of α5 in fibrillar adhesions was reduced in EMCs infected with Adhα4ΔCV5 (Fig. 8B) but not in EMCs infected with Adhα4V5 (Fig. 8D). This result suggested that the C-terminus of α4 plays an important role in the assembly of fibrillar adhesions in chick EMCs.

Figure 8.

Decreased α5β1 accumulation in EMC fibrillar adhesions after Adhα4ΔCV5infection. Confocal microscopy of EMCs cultured on coverslips and infected with Adhα4ΔCV5 (A, B, E, F, I, J) or Adhα4V5 (C, D, G, H, K, L) for 2 days. AD, IL: EMCs were incubated with U1α (A–D) or LM609 (I–L) for 1 hr, fixed and stained with polyclonal anti-V5 and Alexa conjugated secondary antibodies. EH: EMCs were fixed permeabilized and stained with monoclonal anti-V5 and α8-cyto. Proteins stained in cells are indicated in the lower left-hand corner of each panel. Arrows in A, B, E–J point to V5-positive cells. Typically, the polyclonal anti-V5 antiserum (A, C, I, K) yielded higher background than the monoclonal anti-V5 antibody (E, G) but positive cells were distinct during imaging at lower magnifications. Dotted line in B outlines the V5 pattern for the infected cell in A. Scale bars = 20 μm.

Given that expression of C-terminal deleted human α4 reduced U1α staining in fibrillar adhesions in chick EMCs and the known role of α5β1 integrin in fibronectin fibrillogenesis, we postulated that these cells would not efficiently incorporate FN into an extracellular matrix. To determine if α4 loss-of-function or gain-of-function adenoviruses changed the FN matrix organized by chick EMCs, we infected EMC monolayers with AdGFPlacZ or AdGFPα4AS for 3 days or Adhα4V5 or Adhα4ΔCV5 for 2 days and stained them for FN using a pan-FN antiserum (Fig. 9A–D, I–L). In all cases, we did not observe any changes to the FN matrix produced on the basolateral surface of the cultured EMCs. However, it was possible that the FN fibers we observed were deposited prior to any alteration to α4 caused by adenoviral infection. In order to observe new FN fibers produced after virus infection, we performed a FN fibrillogenesis assay (Wu et al.,1993). In this assay, soluble human FN dimers are added to the medium and allowed to incorporate into the FN matrix of the chick cells. Cells are then fixed and stained with a monoclonal antibody (IST-4) that exclusively recognizes human FN. If human cellular FN incorporates into the chick matrix, then IST-4 will recognize FN fibers. If it does not, then the antibody will yield diffuse non-specific staining. In pilot experiments with uninfected monolayers, we observed that human FN was readily incorporated into a chick EMC matrix after an 18-hr incubation. However, when human FN was not added to the medium, IST-4 failed to stain chick fibronectin secreted by the EMCs (data not shown). In monolayers infected with AdGFPlacZ, human FN was observed in the matrix underneath infected cells (Fig. 9E,F). This was also true for EMCs infected with AdGFPα4AS (Fig. 9G,H). However, cells infected with Adhα4ΔCV5 and Adhα4V5 were not able to incorporate human FN into the sub-adjacent matrix nearly as well as in controls (Fig. 9N, P). Thus, while the FN matrix was not altered, new incorporation of FN into this matrix was impaired by expression of either the full-length or C-terminal deleted variants of human α4. Based on our observations of Adhα4ΔCV5-infected EMCs (Figs. 8 and 9), we conclude that α4 plays a role through its C-terminus in regulating the incorporation of α5 into fibrillar adhesions. This also affects EMCs by diminishing their ability to incorporate soluble FN into extracellular matrix.

Figure 9.

Reduced FN matrix incorporation in EMCs infected with Adhα4ΔCV5and Adhα4V5. Confocal microscopy of EMCs cultured on coverslips and infected with AdGFPlacZ (A, B, E, F), or AdGFPα4AS (C, D, G, H) for 3 days or Adhα4ΔCV5 (I, J, M, N) or Adhα4V5 (K, L, O, P) for 2 days. AD, IL: Cells were untreated. EH, MP: Cells were treated with human plasma FN (50 μg/ml) for 18 hr. Cells were fixed and either stained with pan anti-FN (A–D, I–L) or mAb Ist-4, which detects human FN (E–H, M–P). Proteins stained in cells are indicated in the lower left-hand corner of each panel. Dotted lines in F, H, N, P outline infected cells. Scale bars = 20 μm.


In both chick and mouse, the epicardium forms from cells that originate in the PE and migrate over the surface of the heart. Integrin α4β1 performs several key roles during this process. It mediates the attachment of PE cells to the myocardium and subsequent superficial migration (Yang et al.,1995; Sengbusch et al.,2002). Additionally, α4 helps maintain or strengthen the junctions that hold these cells together as groups by promoting pathways that sustain intercellular adhesion (Dokic and Dettman,2006). In this study, we have presented evidence that α4β1 communicates with other integrins expressed by EMCs to mediate adhesion and FN polymerization. These new insights indicate that α4β1 does not simply function to mediate PE cell adhesion to the heart, as suggested by targeted deletion studies in the mouse. Rather, it performs several important regulatory roles in EMCs to mediate cell adhesion, migration, and to preserve the mesothelial phenotype of epicardial cells. We propose that α4 is a signaling receptor that senses the attachment of the PE to the heart by interacting with ligands expressed by the myocardium. After touching the heart, pathways are activated in most PE cells, in part through α4,which promote an adhesive, migratory, and mesothelial phenotype. These events are critical for the transition of the PE to the epicardium.

Chick Epicardial Mesothelial Cells Organize Distinct Cell Adhesion Complexes In Vitro

Our initial hypothesis was that the PE and migrating EMCs expressed integrins that interacted with matrix proteins expressed in the PE and subepicardial matrix (Burch et al.,1995; Bouchey et al.,1996; Nahirney et al.,2003). These included FN, VN, and TN, all proteins that utilize RGD core sequences to interact with their integrin receptors. Consistent with this hypothesis, we detected integrin transcripts for α8, αv, β1, β3 and β5 in PE RNA and confirmed expression of three integrin receptors (α5β1, α8β1 and αvβ3) in cultured chick EMCs and in sections of developing chick hearts. In cultured EMCs, α5β1 and αvβ3 accumulated in cell adhesion complexes similar to that reported in HFFs (Katz et al.,2000; Zamir et al.,2000). Like HFFs, we observed α5β1 in long fibrillar adhesions in chick EMCs. However, we never observed α5β1 or αvβ3 in larger structures resembling focal adhesions. Rather these integrins were found in small puncta that were rarely associated with GFP-paxillin-containing structures. Since chick EMCs are squamous epithelial cells, they may organize different cell-adhesion complexes than HFFs. Focal complexes are defined as small, dot-like adhesions enriched in activated αvβ3 that form at the edges of lamellipodia and transition into focal adhesions (Zamir and Geiger,2001). We observed similar dot-like structures in the lamella of EMCs. However, αvβ3 remained in dot-like structures towards the rear of cells (at the edge of the monolayer) and were uniformly distributed underneath cells in the middle and back of the monolayer. We did not observe their transition into larger, GFP-paxillin-containing focal adhesions. Based on this, we propose that αvβ3 adhesions in EMCs are a stable form of focal complexes. In addition, based on antibody chasing experiments we believe that in chick EMCs, fibrillar adhesions are dynamically organized just as in HFFs. However, we believe that EMC fibrillar adhesions are organized in association with focal complexes and not focal adhesions, as was observed in HFFs.

While none of the integrins we detected in EMCs were abundantly organized into focal adhesions, all were found to overlap actin filaments. Focal complexes containing α5β1 and αvβ3 were associated with stress fibers (data not shown) and α4β1 was broadly expressed on the surface of EMCs in areas of EMCs that accumulate stress fibers. α8β1 was not localized in either focal complexes or fibrillar adhesions, rather it was found along fibrous cytoplasmic arrays similar to and overlapping with both actin stress fibers and cortical actin filaments. Our appreciation of the importance of regulating the actin cytoskeleton in epicardium is increasing. For example, we have observed that cortical actin arrays are important for maintaining the mesothelial state of epicardium (Dokic and Dettman,2006). Also, epicardium robustly expresses the adapter protein LMP4, a protein that couples T-box transcription factors such as Tbx5 to the actin cytoskeleton (Camarata et al.,2006; Bimber et al.,2007). Tbx5 expression is important for epicardial migration (Hatcher et al.,2004), so regulation of Tbx5 function may be linked to the dynamics of actin polymerization in EMCs. It will be important to investigate the links between the extracellular matrix, integrins, and the actin cytoskeleton in regulating genes that control epicardial formation and coronary vascular development.

α4β1 Cooperates With RGD Binding Integrins During Epicardial Cell–Matrix Adhesion

Our results support the hypothesis that α4β1 plays a role in regulating α5β1-, α8β1-, or αvβ3-mediated adhesion in epicardial cells. Thus, chick EMCs express RGD-binding integrins and if this holds true in the mouse, the question remains, why can't these integrins support adhesion in the absence of α4? One explanation is that loss of α4 expression results in loss of expression of other integrins in EMCs. However, we found that chick EMCs knocked down for α4 (by AdGFPα4AS infection) continued to express α5β1, α8β1, and αvβ3. An alternate explanation is that α4 exerts its influence on epicardial cell adhesion, in part, through other integrins expressed on EMCs. Consistent with this idea, we found that chick EMCs were reduced in their ability to bind to either VN or FN120 when we changed the levels of α4 or expressed variants of human α4. Reduced adhesion to VN and FN120 in cells infected with AdGFPα4AS is consistent with the hypothesis that EMCs require α4 for adhesion to the heart even in the presence of other integrins. Decreased attachment of EMCs was greater for VN than it was for FN120 suggesting that α4 has a greater ability to affect αvβ3 than it does α5β1 and α8β1. However, the manner in which the viruses decreased adhesion was different for VN and FN120. AdGFPα4AS and Adhα4ΔCV5 decreased attachment to VN whereas Adhα4V5 did not. These observations are consistent with the idea that the presence of wild-type α4 is important for binding of EMCs to VN. AdGFPα4AS and Adhα4V5 decreased attachment to FN120 suggesting that adhesion via α5β1 and α8β1 is sensitive to levels of α4. Increasing or decreasing α4 could alter the balance of receptors on the surface of cells by competing for β1 subunits. However, since Adhα4ΔCV5 did not alter binding to FN120 and AdGFPα4AS did not affect α5 incorporation into fibrillar adhesions, there may also be other modes of action. This could include the modulation of RhoA signaling by α4β1 (Moyano et al.,2003; Dokic and Dettman,2006). Increased expression of α4 from Adhα4V5 could antagonize RhoA activation and this could alter the ability of α5β1 or α8β1 to mediate adhesion to FN.

Reduced EMC adhesion to VN or FN120 in response to altered α4 suggests that there is functional cooperation between integrin receptors on EMCs. Functional cooperation (or “cross-talk”) between integrin receptors is a mechanism that allows cells to finely tune their response to extracellular matrix (Ginsberg et al.,2005). Cross-talk occurs when one integrin responds to ligand and this alters the ability of other integrins to interact with or respond to their ligands (Schwartz et al.,1995). We believe that our observations support a role for α4 in modifying the function of other integrins in EMCs. If α4 plays a role in regulating other integrins in EMCs, then, by this paradigm, signaling through α4 could change the affinity of epicardial α5β1, α8β1, or αvβ3 for their ligands. In fact, α4 is known to do this in T-lymphocytes where binding of α4β1 to VCAM-1 or FN increased αLβ2 adhesion to ICAM-1, promoting transendothelial migration (Chan et al.,2000). Similarly, α4β1 ligation was shown to activate β2 integrins in leukocytes (May et al.,2000). We have also found that pretreatment of rat EMCs with soluble VCAM-1 increases their adhesion to FN (Pae and Dettman, unpublished observation). The results presented here are consistent with a mechanism of EMC cell adhesion that requires α4β1 for maintaining or increasing the affinity of α5β1, α8β1, or αvβ3 for their ligands. Thus, an attractive hypothesis is that PE cell attachment and migration on the heart may be analogous to T-lymphocyte adhesion and migration through activated endothelium via α4β1.

α4β1 Influences Fibrillar Adhesion Formation and FN Matrix Formation

Little is known about the requirements for the formation of fibrillar adhesions. The current model is that α5β1 is presented on the cell surface in a high-affinity conformation available for binding and organizing soluble FN and organizes into focal adhesions (Zamir et al.,2000). Another protein, tensin, bound to actin filaments exerts force upon α5β1 and pulls it into longer fibrillar adhesions generating strong extracellular matrix contacts. This simultaneously changes the localization of α5β1 and generates a force outside of the cell on folded FN dimers. Dimers are pulled apart, exposing intramolecular binding sites. By uncovering these sites in FN, tensin stimulates FN fibrillogenesis (Pankov et al.,2000). Our demonstration of α5β1 integrin in focal complexes and fibrillar adhesions of cultured, migrating EMCs is similar to this model and is consistent with the hypothesis that α5β1 participates in FN polymerization in the epicardium.

We have observed that EMCs infected with Adhα4ΔCV5 were impaired in their ability to bind to U1α and to incorporate human plasma FN into matrix. This suggests that in EMCs, α4β1 plays a part in the formation of fibrillar adhesions. This could occur for several reasons. One possibility is that α4β1 could modulate the affinity of α5β1 for FN, placing it into a low-affinity conformation. α5β1 exists at different affinity states for its soluble ligand FN in a number of cell-types (Faull et al.,1993). U1α recognizes an activated form of α5β1 and, upon binding, maintains this receptor in a high-affinity state (Boettiger et al.,1995). So if expression of α4ΔCV5 changes the affinity state of α5β1, infected EMCs would not bind to U1α with the same intensity. Another possibility is that expression of α4ΔCV5 reduces the amount of α5 incorporated into fibrillar adhesions. This could also result in a change in affinity of α5β1 for soluble FN. In this case, α5β1 would be less efficient in its ability to bind to FN dimers and incorporate them into focal complexes. Therefore, tensin would translocate less α5β1 into fibrillar adhesions. This idea is supported by our observation that in EMCs infected with Adhα4ΔCV5, U1α was detected at much lower levels in either focal complexes or fibrillar adhesions. Another possibility is that α4 plays a direct role in FN fibrillogenesis in cells that express it. While the RGD binding integrins are thought to play the major role in FN fibrillogenesis, it has been demonstrated that under some circumstances α4 alone can stimulate an alternative assembly pathway (Sechler et al.,2000). We don't think this is likely since in EMCs α5β1 is expressed in fibrillar adhesions while α4β1 is not expressed in specific adhesion complexes.

Integrins, Adhesion Complexes, and Heart Development

Epicardial formation starts with the protrusion of the PE and interaction with the heart. α4 is one of the first true markers for the PE and the gene encoding α4 (Itga4) has recently been identified as a locus regulated by Wilms-tumor factor (WT1) (Kirschner et al.,2006). WT1 is expressed in both the PE and epicardial cells as they are migrating, as well as in older epicardium and epicardially derived mesenchymal cells (Perez-Pomares et al.,2002b). It is possible that WT1 triggers the expression of α4 in the PE and this renders PE cells capable of adhering to the heart. Furthermore, Sengbusch and colleagues (2002) observed that in α4 null mice, the ability of the PE to release aggregates from villous protrusions was impaired. This indicates that through α4, WT1 sets into motion a series of events that allows the PE to transition to epicardium. We propose that α4 is also critical for this transition because it potentially modulates the activation state of the other integrins expressed in migrating EMCs enabling them to adhere and migrate on myocardium and organize the subepicardial extracellular matrix. Matrix deposition is likely to be critical for the PE to epicardial transition and may be exquisitely regulated. For example, subepicardial matrix deposition can be disregulated as it is in RxRα null mice where the PE to epicardial transition is delayed, which is in part due to excessive buildup of FN in subepicardial matrix (Jenkins et al.,2005). Thus, α4, α5, α8, and αv expression in superficial epicardium during these embryonic stages likely contributes to a complex set of events that governs the transition of PE to epicardium.

It is likely that α4 and RGD-binding integrins play an important role in the maturation of the epicardium. α4 expression persists in epicardium through the later stages of development (our unpublished observations) where it promotes intercellular adhesion in superficial EMCs (Dokic and Dettman,2006). In addition, the subepicardial matrix contains VN and FN and it is, therefore, logical that the epicardium expresses receptors for these matrix proteins. In regions of the heart where epicardium is directly adjacent to the myocardium, α5 is expressed in a basolateral stripe that probably represents strong adhesive connections. Supporting this was our observation that fibrillar adhesions were abundant in the middle of the monolayer and at the trailing edge of migrating EMCs where we frequently saw trails of membrane blebs containing α5 and FN behind migrating EMCs at the edge of the monolayer. These blebs form because it is easier to tear the cell membrane off than break the tight association between α5 and FN (Kirfel et al.,2004).

Finally, it is tempting to speculate that in epicardium, α8 functions as a tenascin receptor. Tenascins are a family of ECM proteins that have been detected in the subepicardial ECM, and while the individual members of the family vary between birds and mammals, these proteins have been proposed to play a role in epicardial or smooth muscle development (Burch et al.,1995). The α8 subunit binds exclusively with β1 and can bind to several RGD-containing ligands. These include FN, VN, TN-C, and nephrectin. α8 has been deleted in the mouse and this led to a variable phenotype that primarily included defects in kidney morphogenesis. However, no heart defects have been described (Muller et al.,1997). The signaling properties and subcellular localization of α8 are currently not well understood so it will be interesting to see what roles α8 plays in epicardial adhesion and coronary vascular development.


Chicken Eggs

Fertile White Leghorn chicken (Gallus gallus domesticus) eggs were obtained from Phil's Fresh Eggs (Forreston, IL) and were incubated in a humidified bird hatching incubator at 37°C. Embryos were staged according to the method of Hamburger and Hamilton (Hamburger and Hamilton,1951).

Antibodies and Fluorescent Stains

The monoclonal antibody to chick α5-integrin (U1α, Boettiger et al.,1995) was a generous gift from Dr. Ruth Chiquet-Ehrismann. Monoclonal anti-αvβ3 (LM609) and anti-α4 (HP2/1) were obtained from Chemicon International. Polyclonal anti-mouse α4-integrin (Santa Cruz Biotechnology, N-19 and Y-18) reacted with sectioned chick hearts specifically in epicardium and was used at a final concentration of 2 μg/ml. Monoclonal anti-myosin heavy chain (MF20) was obtained from the Developmental Studies Hybridoma Bank. Monoclonal anti-human FN (IST-4) and polyclonal pan anti-FN were obtained from Sigma-Aldrich. Polyclonal anti-human α8-integrin cytoplasmic tail (α8-cyto) was a generous gift from Dr. Louis Reichardt. Polyclonal anti-human αv-integrin cytoplasmic tail (αv-cyto) was from Chemicon International (no. AB1930). Monoclonal anti-V5 was from Invitrogen and polyclonal anti-V5 was from GeneTex (no. 30564). Of these antibodies, three yielded some background staining. These were α8-cyto, αv-cyto, and polyclonal anti-V5. Secondary antibodies used in this study were: goat anti-rabbit IgG, goat anti-mouse IgG, donkey anti-goat IgG, and donkey anti-mouse IgG coupled to either Alexa 488 or Alexa 568 (Invitrogen, Molecular Probes). Texas Red-X phalloidin, and 4′,6 diamidino-2-phenylindole, dihydrochloride (DAPI) were from Invitrogen, Molecular Probes.

Peptides and Soluble Proteins

All peptides were purchased from American Peptide and suspended in sterile water and diluted to the concentrations indicated in M199 containing antibiotics and fetal bovine serum (1% v/v). To competitively inhibit binding of integrins to the core RGD binding domain of FN, VN, and other ECM proteins, we used the GRGDSP peptide. The peptide GRADSP was used as the control. To competitively inhibit binding of α4β1 to the V25 (CS-1) domain of FN, we used EILDVPST. The peptide EILEVPST was used as the control. The following proteins were used in cell adhesion and FN fibrillogenesis assays: human vitronectin (Promega), human serum FN (BD Biosciences), purified human FN alpha-chymotryptic fragment 120K (Chemicon), and recombinant human soluble VCAM-1 (R&D Systems).

Reverse Transcriptase PCR

Total RNA was isolated from 10 HH16 embryonic chick proepicardial organs or epicardial cells cultured from 10 HH24 hearts using the Cells-to-cDNA II RTPCR kit (Ambion). PCR was done using Platinum Taq (Invitrogen) according to the manufacturer's instructions. Amplification was carried out with the following profile: 94°C 1 min 30 sec; 55°C 30 sec; 72°C 1 min 30 sec; 94°C 30 sec. Varying numbers of cycles were used up to 35 cycles. Primers were synthesized by IDT DNA. Primers were designed so that forward and reverse primers were in alternate exons of the transcript. This allowed us to distinguish between amplification of RNA and genomic DNA. The following primers were used: α4 forward: 5′-atggtaaccgtagctgtacct; α4 reverse: 5′-agtcatccttgttcccacttg-3′; αv forward: 5′-cctcaaccagttcatccctgctaa-3′; αv reverse: taaggccactgcagagtcatcatc-3′; α8 forward: 5′-cgcaaattctgacaggcactgaag-3′; α8 reverse: 5′-gtcacacaggtctggctctgtaa-3′; β1 forward: 5′-tccacggatgctggatttcact-3′; β1 reverse: 5′-ttcctgcgtgtcattcactcca-3′; β3 forward: 5′-tggaaactcctcatcaccatccac-3′;β3 reverse: 5′-acagctcacagccttgagttctac-3′; β5 forward: 5′- aaatgtgcgtggtgctccaaag-3′; β5 reverse: 5′-ttcgcatctcttcagccagct-3′; β6 forward: 5′-tcatgtgcaagggacctgttct-3′; β6 reverse: 5′-tgtttcttcatgtcctggtcgc-3′; β8 forward: 5′- accctatatctgcattcacccagg-3′; β8 reverse: ccaatggaactggcttccttgaac-3′; GAPDH forward: acgccatcactatcttccag-3′; GAPDH reverse: 5′-cagccttcactaccctcttg-3′.

Plasmids and Viruses

Adhα4V5 and Adhα4ΔCV5 were constructed using the Gateway system (Invitrogen). A plasmid containing human α4 cDNA was obtained from ATCC (cat no. 95494). The coding sequences were amplified with primers that removed either the stop codon (Adhα4V5) or 30 codons encoding the C-terminus of hα4 (Adhα4ΔCV5). The forward primer was 5′-caccatgttccccaccgagagcgc-3′, the reverse primer for Adhα4V5 was 5′-atcatcattgcttttactgtt-3′, and the reverse primer for Adhα4ΔCV5 was 5′-gccagccttccacataacata-3′. PCR products were generated with pfu Ultra Taq polymerase (Stratagene) and cloned into the entry vector, pENTR/D/Topo. Individual clones were isolated and sequenced. Entry clones were recombined with pAd/CMV/V5/Dest using LR clonase (Invitrogen) and tested by PCR and restriction mapping with Bst XI and Pac I (New England Biolabs). Through recombination with pAd/CMV/V5/Dest, the cDNAs were placed under the control of a CMV promoter and a V5 epitope tag was fused to the C-terminus of each expressed protein. Pac I digested adenoviral clones were ethanol precipitated and 4 μg of DNA was transfected into human embryonic kidney 293FT cells using lipofectamine 2000 (Invitrogen) on 6-cm dishes. Viral supernatants were generated and amplified in 293FT cells until sufficient viral particles were generated to carry out CsCl purification. Here, we followed the method described in the AdEasy manual from Q-Biogene. Construction of AdGFPlacZ and AdGFPα4AS were previously described (Dettman et al.,2003).

The GFP-paxillin fusion expression plasmid was a gift from Dr. Teng-Leong Chew. We used Lipofectamine 2000 (Invitrogen) to transfect HH24 hearts. Hearts were excised and placed into serum and antibiotic-free Opti-MEM medium. We combined 10 μl of lipofectamine 2000 with 4 μg of plasmid DNA and 0.5 ml of Optimem and incubated this mixture for 30 min at room temperature. Hearts were placed onto collagen coated glass coverslips in 30-mm dishes and we added the DNA/lipofectamine/Opti-MEM mix to the dishes. Hearts were incubated overnight until cells had migrated from the heart to the surface of the coverslip. Hearts were removed and coverslips were fixed prior to incubation with antibodies.


Embryos or hearts (tissues) were dissected in Dulbecco's PBS (PBS) and fixed in formaldehyde (4% v/v from paraformaldehyde, EM Sciences) at room temperature for 60 min. Samples were washed in PBS and then embedded in 1.5% (w/v) agarose and 5% (w/v) sucrose in plastic molds. Agarose blocks were trimmed and placed in 20% sucrose (w/v) overnight and then 30% sucrose (w/v) until the blocks sunk to the bottom of the tube. Blocks were then covered with OCT and frozen. Sections (10 μm) were cut on a Leica CM1850 cryostat and allowed to air dry. Slides were placed and washed three times in PBS before transferring into PBT (PBS, 0.1% v/v Tween-20, 0.1% w/v bovine serum albumin) for 1 hr. For α8-cyto and αv-cyto, we used antigen retrieval. Sections were warmed to room temperature and washed in PBS. Sections on slides were then incubated in sodium citrate buffer (10 mM, pH 6.0) with Triton X-100 (0.25% v/v) at 96°C for 20 min in a slide mailer. The mailer containing the slides and buffer was then placed at room temperature until the buffer reached 25°C. Sections were washed in PBS before further manipulation. Sections on slides were incubated in primary antibody diluted in PBT overnight at 4°C. After three 5-min washes in PBS, tissues were incubated in secondary antibody diluted in PBT for 2 hr at room temperature. Staining was observed using a Zeiss LSM 510 UV Meta confocal inverted microscope.

Mesothelial Sheet Migration

Chick eggs were incubated in humidified egg incubators at 37°C. Embryos were staged according to Hamburger and Hamilton (1951). Hearts were removed from HH24 embryos in PBS and transferred to serum-free M199. To allow residual blood to be pumped from hearts, hearts were incubated for 2 hr at 37°C in 5% CO2 before transferring to uncoated tissue culture dishes. Hearts were positioned on plates with 500 μl M199 supplemented with heat-inactivated fetal bovine serum (1% v/v) such that medium was just covering the surface. Peptides and antibodies were added to this medium as indicated. Mesothelial migration was allowed to proceed for 18 hr at 37°C in 5% CO2 after which dishes were photographed with a Photometrics Cool Snap digital camera mounted on a Nikon TE300 eclipse inverted microscope using the 4× objective and phase contrast optics. Hearts were removed by washing and monolayers were fixed in formaldehyde (4% v/v) diluted in PBS for 10 min at room temperature. Dishes were washed twice in PBS and monolayers were stained with hematoxylin QS (Vector labs) for 30 min. Dishes were washed twice with PBS and entire monolayers were photographed using a Nikon SMZ1000 stereomicroscope at 1× magnification. The area of monolayers was calculated by tracing the outlines of the outer edge of the monolayer and the inner edge of the monolayer left by the “footprint” of the removed heart. The number of pixels found in the areas of both outlines was calculated using Metamorph (Molecular Devices). The area of the smaller inner outline was subtracted from the area of the larger outline to determine the number of pixels within the monolayer area. Then the number of pixels was converted into square micrometers after determining the number of pixels per millimeter at the magnification used (13.9 pixels/mm). Data is expressed as the mean of areas obtained under each condition tested. Statistical significance was calculated using a Student's t-test and error bars were calculated as the standard error of the mean.

Fibronectin Fibrillogenesis Assays

Chick EMCs were cultured on glass coverslips coated with rat-tail collagen I (BD Biosciences, no. 354089) in M199 medium supplemented with antibiotics and 1% fetal bovine serum. After 24 hr in culture, the medium was replaced with M199 containing human serum FN diluted to 50 μg/ml (w/v) and cultures were incubated overnight at 37°C in 5% CO2. Cells were fixed in formaldehyde (4% v/v from paraformaldehyde, EM Sciences) and washed in PBS and PBT. Cells were incubated with either anti-panFN or anti-human FN overnight at 4°C (clone IST-4, Sigma, 1:100 dilution in PBT), washed in PBS and then with goat anti-mouse Alexa-Fluor 488 (green) or 568 (red, Molecular Probes). Stained cells or tissues were imaged on the Zeiss LSM 510 confocal microscope as above.

Cell Adhesion Assays

Primary cultures of chick EMCs were generated from 20–30 explanted HH24 hearts in M199 supplemented with FBS (10%v/v) and antibiotics. Hearts were removed after one day and cells were cultured for 2 days. Adenovirus was added at a multiplicity of infection of 10:1 and cells were cultured for 2 days (for Adhα4V5 and Adhα4ΔCV5) and 3 days (for AdGFPlacZ and AdGFPlacZα4AS). Twenty-four well dishes were coated with human vitronectin (2.5 μg/ml in serum-free M199 medium) for 2 hr at 37°C, blocked in heat-inactivated bovine serum albumin (10 μg/ml) for 30 min, and then 300 μl of M199 supplemented with FBS (10%v/v) was added to each well. Infected cells were harvested with 200 μl of trypsin for 1 min followed by 400 μl of M199 supplemented with FBS (10% v/v). Cells were mechanically disrupted by pipetting and complete disruption of cells was monitored on a phase contrast microscope. Cells were pelleted by centrifugation (5,000 rpm), resuspended in 320 μl of M199 supplemented with FBS (10%v/v), and placed on ice. Cells were counted on a hemacytometer and diluted to a concentration of 300 cells/μl. An equal number of cells (30,000) were added to individual wells, the plate was briefly swirled and placed in the 37°C incubator for 30 min. Wells were washed three times with PBS followed by aspiration and then fixed in formaldehyde (4% v/v) for 10 min at room temperature. Adhered cells were stained with anti-V5, DAPI, or Texas Red-X phalloidin (Invitrogen). Adhered cells were counted in four fields per well using the 10× objective. This value was then averaged and divided by the area of a 10× field (3.46 mm2). To correct for variance caused by rate of infection by different adenoviruses, percent infection was determined by counting four fields per well using the 20× objective. The total number of cells per 20× field was determined by counting DAPI- or phalloidin-stained cells. Then the number of infected cells in the same field was determined by counting the number of GFP- or V5-expressing cells. Percent infected was calculated as the ratio of infected to total cells multiplied by 100. The number of adhered, infected cells per mm2 was calculated by multiplying the total number of cells per mm2 by the percent infected. The mean number of infected cells per mm2 was calculated from all the samples and data was entered into bar graph format. Error bars are standard error of the mean and P values were calculated using a Student's t-test. In 293FT cell adhesion assays, cells were infected with recombinant adenoviruses for 24 hr. Cells were removed with trypsin and counted as above. Cells (5×104/100 μl) were added to individual wells of a 96-well plate coated overnight as above. Cells were allowed to adhere for 30 min before wells were washed three times with PBS and fixed in formaldehyde (4% v/v) for 10 min. Bound cells were stained with crystal violet (0.1% w/v in absolute ethanol) for 10 min. Wells were washed with water until no residual dye remained and plates were allowed to dry. Dye was solubilized in 100 μl SDS (2% w/v in water) and absorbance was read at 540 nm on a Lab Systems MCC/340 multiscan plate reader.


We thank Michael E. Farrell for excellent technical assistance in the production of the viruses used in this study. We thank Dr. Ruth Chiquet-Erhismann for her generous gift of the U1α monoclonal antibody and Dr. Louis Reichardt for his gift of the α8-cyto antiserum. Other monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the Department of Biological Sciences, The University of Iowa, Iowa City, IA 52242. We thank Dr. Teng-Leong Chew for the GFP-paxillin expression plasmid and for valuable discussions during the course of this work. We thank Dr. Hans-Georg Simon, Dr. Sona Gasparian, and Dr. Paul Schumacker for critical reading of the manuscript. Funding for this work was by an American Heart Association Scientist Development Award to R.W.D. (AHA 0030412Z). We also thank the Women's Board of Children's Memorial Hospital, Chicago, for their generous support of the Neonatology Research Laboratory.