Organization of fibroblasts in the heart



Cardiac fibroblasts are organized into a three-dimensional network in the heart. This organization follows the endomysial weave network that surrounds groups of myocytes. Reverse transcriptase-polymerase chain reaction, Western blots, and immunohistochemistry were used to show that discoidin domain receptor 2 (DDR2) was specific for cardiac fibroblasts and not expressed on endothelial cells, smooth muscle cells, or cardiac myocytes. DDR2 is expressed early in development and in the adult heart. High voltage electron microscopy (HVEM), scanning electron microscopy, and laser scanning confocal microscopy document the three-dimensional organization of fibroblasts in the heart. Antibodies against connexin 43 and 45 showed different patterns but confirmed, along with HVEM, that fibroblasts are connected to each other as well as cardiac myocytes. The implications of this arrangement of fibroblasts can be important to cardiac function. The signaling of DDR2 and the expression of matrix metalloproteinase 2 in relation to collagen turnover and remodeling is discussed. Developmental Dynamics 230:787–794, 2004. © 2004 Wiley-Liss, Inc.


The transmission of mechanical force is one of the principal functions of the connective tissue network in the heart. The organization of this network has been well described in normal hearts and in a variety of cardiac disease models (Bishop and Laurent, 1995). Studies have shown that collagen within the extracellular matrix (ECM) forms a three-dimensional network around bundles or laminae of myocytes (Caulfield and Borg, 1979). The formation of the connective tissue network is associated with increases in pressure and/or volume (Borg and Caulfield, 1981; Borg, 1982) and inhibition of collagen network formation results in abnormally shaped hearts or rupture of the ventricular wall (Borg et al., 1985; Lohler et al., 1984). To function as a stress tolerant network, the connective tissue must be attached to cardiac myocytes. Previous studies have shown that the integrin family of cell surface receptors serves to anchor collagen and other ECM components to the surface of cardiac myocytes and fibroblasts (Ross and Borg, 2001).

Although much is known concerning the organization and function of the connective tissue network, comparatively little is known about the cells that produce this network. The cardiac fibroblast is the principal cell type responsible for producing components of the ECM (Manabe et al., 2002). They are the predominant cell type of the heart by number but obviously not by volume (Manabe et al., 2002). Studies, principally in hypertrophying hearts, indicate that fibroblasts are responsible for the deposition of collagen and that this increase in collagen is critical to cardiac function (Weber, 1989).

One of the limiting factors in studying the cardiac fibroblasts in vivo is the lack of a specific marker. Attempts to show fibroblast-specific markers in the heart, such as fibroblast-specific factor (FSP-1), have not shown favorable results (Strutz et al., 1995). The discoidin domain receptors DDR1 and DDR2 represent a relatively new family of collagen-specific receptor tyrosine kinases (Shrivastava et al., 1997; Vogel et al., 1997). Receptor tyrosine kinases are a family of proteins involved in the conversion of extracellular stimuli into cellular response (Schlessinger, 1997). These receptors mediate a variety of cell functions, including growth, migration, morphology, and differentiation. Although the tissue distribution of DDR1 and DDR2 varies and can be mutually exclusive (Alves et al., 1995; Johnson et al., 1993), DDR2 expression has been detected in both rat and mouse heart (Lai and Lemke, 1991, 1994).

Studies are just beginning to elucidate the functional role of DDR2. Temporal expression of DDR2 in dermal fibroblasts was found to remain constant during development, despite increases in collagen expression (Chin et al., 2001). Stimulation of DDR2 by collagen type I revealed that this receptor is capable of up-regulating the expression of matrix metalloproteinase 1 (MMP-1; Vogel et al., 1997), an enzyme involved in the degradation of collagen types I, II, III, VII, and X (Cleutjens, 1996). The purpose of this study was to determine the cell-specific expression of DDR2 in the heart. Expression of DDR2 was shown to be specific for cardiac fibroblasts and DDR2 was not detected on myocytes, endothelial cells, or smooth muscle cells. By using DDR2 as a marker, fibroblasts were observed to be interconnected by gap junctions and formed a three-dimensional network intimately associated with myocytes and the connective tissue network of the heart.


Critical to the examination of fibroblast organization in the heart was the characterization of a specific marker for fibroblast. To determine whether DDR2 represented such a marker, initial experiments were conducted to determine at what point during cardiac development DDR2 was expressed. RNA was isolated from developing rat hearts and reverse transcriptase-polymerase chain reaction (RT-PCR) carried out by using primers specific for rat DDR2 revealed that this receptor is expressed in the heart from embryonic day (ED) 11.5 through ED 20 and that expression of this receptor continues in the neonatal heart and in normal adult hearts (Fig. 1A).

Figure 1.

Expression of DDR2 in the heart. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) of RNA isolated from hearts at different time points during development revealed that discoidin domain receptor 2 (DDR2) is expressed throughout cardiac development and persists in adult animals. ED, embryonic day. B: Cell-specific expression of DDR2 was determined by both RT-PCR (left panel) and RNase protection assays (right panel). DDR2 expression was detected in either whole heart (5d Hrt) or fibroblast (NHF) RNA samples but not in myocyte (Myo) RNA. C: Western blot demonstrating that DDR2 protein expression is confined to fibroblasts and that the receptor is not expressed on myocytes (Myo), endothelial (Endo) or smooth muscle (SMC) cells. Blk lane is buffer blank. Arrow indicates position of DDR2.

To examine the cell-specific expression of DDR2, RNA was extracted from isolated neonatal day 4 cardiac myocytes and fibroblasts. Analysis by both RT-PCR and RNase protection assays (RPA) indicated that DDR2 is expressed only on cardiac fibroblasts (Fig. 1B). No expression of this receptor was detected on cardiac myocytes. Likewise, fibroblast, myocyte, endothelial, and smooth muscle cell protein lysates were also screened for DDR2 expression by Western blotting and again receptor was only found on fibroblasts (Fig. 1C). Previous studies have indicated that DDR2 is a marker of mesenchymal cells (Alves et al., 1995), supporting the finding of DDR2 expression only on cardiac fibroblasts.

Scanning electron microscopy (SEM) confirmed the lamellar structure of the heart in which groups of myocytes were organized two to five cells thick between the surrounding endomysial collagen weave network (Fig. 2A). In addition, associated with the endomysial collagen was the presence of fibroblastic appearing cells (Fig. 2B). The fibroblastic cells had long filopodia and, in some cases, appeared to be in contact with each other and with myocytes. Fibroblasts were not found between groups of myocytes. By normal transmission microscopy, fibroblasts usually appeared as isolated cells lying between or near myocytes (data not shown); however, by high voltage transmission electron microscopy (HVEM), the fibroblasts appeared to interact with other fibroblasts, myocytes, and endomysial collagen (Fig. 3). Because HVEM used thicker specimens (0.25 μm), cells that appeared as fibroblasts by SEM were confirmed this technique. Fibroblasts had contacts with the endomysial collagen, with adjacent fibroblasts, and with cardiac myocytes (Fig. 3).

Figure 2.

Localization of fibroblasts in the heart. A: Scanning electron photomicrograph showing the cardiac fibroblasts (F) in the endomysial collagen network surrounding cardiac myocytes (M). Long filopodia of the fibroblasts (arrows) appear to be in contact with collagen (C) and with the myocyte surface. B: Scanning electron photomicrograph showing fibroblasts (F) associated with the endomysial collagen weave (arrows).

Figure 3.

High-voltage electron microscopy showing collagen–myocyte–fibroblast interactions. High-voltage electron photomicrograph showing the association of collagen (arrows) with fibroblasts (F) and myocytes (M).

To correlate the images from SEM and the HVEM, confocal microscopy was used to localize fibroblasts and to characterize their interactions with endomysial collagen, other fibroblasts, and cardiac myocytes. By using immunofluorescent tags for collagen, DDR2, and cardiac myocytes, the organization of cardiac fibroblasts within the heart could be discerned (Fig. 4). Confocal analysis confirmed that cardiac fibroblasts were interspersed in the connective tissue of the endomysial layer surrounding myocytes and that they lie adjacent to the cardiac myocytes with long filopodia extending between myocytes and connecting fibroblasts.

Figure 4.

Confocal microscopy showing collagen-myocyte-fibroblast association. A: Confocal micrograph of cardiac muscle showing 488-phalloidin (green) staining of myocytes and no secondary antibody staining (negative control). B: Single channel confocal image showing discoidin domain receptor 2 (DDR2) localization (blue) on cardiac fibroblasts. C: Photomicrograph showing the collagen network pattern (red) in the endomysium around myocytes (green) and DDR2 positive fibroblasts (blue). Colocalization of collagen- and DDR2-positive fibroblasts is indicated by purple coloration. This finding supports the scanning electron micrograph images (Figs. 2, 3), which also show that the fibroblasts lie in the collagen network around myocytes. Scale bar = 25 μm in A (applies to A–C).

The confocal as well as HVEM analysis indicated that fibroblasts made contact with each other as well as with cardiac myocytes. To confirm this observation, triple labeling for the gap junction proteins connexin 43 and 45 (Cx-43 and Cx-45) was used. Localization of Cx-43 indicated that gap junctions were on both myocytes and fibroblasts (Fig. 5). In both frozen and polyacrylamide-embedded sections, staining showed that the Cx-43 appeared to connect adjacent fibroblasts to one another as well as to cardiac myocytes on the lateral margin and intercalated disks (Fig. 5). Whereas these photomicrographs are suggestive of specific contact, higher resolution electron photomicrographs with gold antibody labeling will be necessary to resolve the localization. However, localization of Cx-45 showed a different pattern (Fig. 6) in which the labeling was primarily associated with the cardiac fibroblasts. In some areas, the Cx-45 appeared to stain both fibroblasts and myocytes (Fig. 6C); however, the greatest localization was on the fibroblast. By using Western blots against isolated myocyte and fibroblast protein lysates, both antigens could be detected on each cell population (Fig. 7) with slightly higher expression of both connexins on myocytes.

Figure 5.

Confocal microscopy of connexin 43 (Cx-43) staining. A: Confocal photomicrograph of cardiac muscle showing 488-phalloidin (green) staining of myocytes and no secondary antibody staining (negative control). B,C: Longitudinal (B) and cross-section (C) of cardiac muscle showing the colocalization (arrows) of Cx-43 (red), fibroblasts stained with discoidin domain receptor 2 (blue), and 488-phalloidin staining of myocytes (green) on frozen heart sections. Arrows indicate Cx-43 labeling. Scale bar = 25 μm in A (applies to A–C).

Figure 6.

Confocal microscopy of connexin 45 (Cx-45) staining. A: Longitudinal frozen section showing negative control stained with 488-pahlloidin (green) to visualize f-actin in myocytes. B,C: Cross-section (B) and longitudinal section (C) showing the localization of Cx-45 (red), discoidin domain receptor 2 (blue), and f-actin (green), indicating that Cx-45 is localized primarily to the cardiac fibroblasts. Arrows indicate Cx-45 labeling. Scale bar = 25 μm in A (applies to A–C).

Figure 7.

Western blot to compare connexin (Cx) 43 and 45 distribution. A,B: Western blotting of myocyte (M) and fibroblast (F) protein lysates demonstrated that Cx-43 (A) and Cx-45 (B) are predominantly expressed on myocytes compared with fibroblasts. Membranes were also probed for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a loading control.

The data presented herein demonstrate that fibroblasts in the heart are organized into a three-dimensional interconnected network analogous to the arrangement of fibroblasts in collagen gels (Grinnell, 2000). In this in vitro system, fibroblast can migrate within and contract the collagen gel. In vivo fibroblasts are enmeshed in a collagenous endomysial network that surrounds groups of myocytes. This structure, described as laminae of two to five myocytes in thickness, is fundamental to the organization and function of the heart (Legrice et al., 1997). Cardiac fibroblasts have been shown to be important components of the ECM in the heart because of their ability to synthesize and express several types of collagens (Carver et al., 1993), MMPs (Spinale, 2002), growth factors (Diaz-Araya et al., 2003), and their role in cytokine signaling (Smith et al., 1997). Indeed, fibroblasts have been termed sentinel cells because of the many functions they mediate in the heart (Smith et al., 1997). However, the lack of a fibroblast marker has made studying fibroblast ontogeny and organization in vivo difficult.

DDR2, a member of the receptor tyrosine kinases family of proteins, represents a good marker for identifying cardiac fibroblasts in vivo. Receptor tyrosine kinases are involved in the conversion of extracellular stimuli into cellular response and mediate a variety of cell functions including growth, migration, morphology, and differentiation (Schlessinger, 1997). Although multiple types of collagen can activate the tyrosine kinase domain in DDR2, studies have shown that phosphorylation of DDR2 is significantly reduced after deglycosylation of collagen or if the triple helical structure of collagen is altered (Vogel et al., 1997), suggesting a potential role for this protein in detecting the physical integrity of collagen within the ECM. DDR2 has also been associated with MMP-2, which is an important component in the turnover and remodeling of the ECM of the heart (Spinale et al., 1998). Given that fibroblasts are responsible for production of the structural collagens (types I and III) and that DDR2 is expressed only on fibroblasts in the heart, this finding suggests a potential role for DDR2 in modulating collagen production/remodeling.

Gap junctions form a communication system for small molecules and transmit electrical activity in a variety of cell types (van Veen et al., 2001). Whereas there are numerous types of gap junctions, it appears that Cx-40, Cx-43, and Cx-45 are the predominant gap junction proteins in the heart (Gourdie et al., 1993; Coppen et al., 1999). The localization of Cx-43 and Cx-45 has been reported previously to be on both cardiac myocytes and fibroblasts (Gudesius et al., 2003). Immunolocalization confirms that, although Cx-43 and Cx-45 are not cell-specific, they do show different distribution patterns (Figs. 5, 6) and slightly different expression levels (Fig. 7). Whereas much is known about Cx-43 and Cx-45 in myocytes, little is known concerning the role of these gap junction proteins in fibroblasts. The presence of gap junctions, interconnecting fibroblasts and coupling these cells to cardiac myocytes, is likely to have significant functional implications. One of the roles of fibroblasts is to electrically connect myocytes (Gudesius et al., 2003). This role could have significant importance in fibrotic tissue associated with hypertrophy and/or infarction (Gudesius et al., 2003). In addition, studies have shown that inhibition of Cx-43 on fibroblasts blocks collagen gel contraction (Ehrlich et al., 2000). This finding leads to the speculation that fibroblasts within the endomysial network might be able to exert force on the myocytes within the laminae by contracting the surrounding collagen network. The described interconnected arrangement of fibroblasts, through Cx-43 and Cx-45, would allow these cells to be more resistant to mechanical deformation as well as being able to generate mechanical force during contraction.

The results presented herein demonstrate that, although DDR2 is expressed in the heart throughout development, in the adult heart this receptor can serve as a marker for fibroblasts. This marker should be useful in delineating the lineage of fibroblasts in the heart as well as potential roles in normal cardiac function as well as in disease. Further studies will be necessary to define the functional role for this receptor on cardiac fibroblasts. DDR2 may have important structural and functional implications for the heart during collagen deposition and remodeling in both heart development and disease. Future studies will examine the interaction of this receptor with other ECM receptors, such as integrins, to determine whether these receptors cooperate to regulate cardiac fibroblast function during development and cardiac disease.


Isolation and Culture of Cells

Neonatal cardiac myocytes and fibroblasts were obtained from collagenase digestion of 3- to 4-day neonatal rat hearts as previously described (Borg et al., 1984). Fibroblasts were separated from myocytes by selective cellular attachment, and myocytes were further purified by using a Percoll density gradient (Sigma Chemical Co., St. Louis, MO). Fibroblasts were cultured in Dulbecco's modified Eagle medium (DMEM; Sigma) supplemented with 10% newborn bovine serum (NBS; Gibco BRL, Rockville, MD), 5% fetal bovine serum (Atlanta Biological, Atlanta, GA), 100 U/ml penicillin G, 100 μg/ml streptomycin, 1 μg/ml amphotericin B, and 10 μg/ml gentamicin (Sigma). All fibroblasts used in the experiments described were at passage 3. Myocytes were maintained in DMEM containing 8% horse serum (Gibco), 5% NBS, 100 U/ml penicillin G, 100 μg/ml streptomycin, 1 μg/ml amphotericin B, and 2 μg/ml cytosine arabinoside (Sigma).

RNA Isolation, RT-PCR, and RNase Protection Assays

Total RNA was extracted from freshly isolated hearts from embryonic and adult rat and cultured cells (myocytes and fibroblasts) by using RNA STAT-60 (Tel-Test, Friendswood, TX) according to the manufacturer's recommendations. For developmental PCR, 2 μg of RNA were analyzed by RT-PCR (GeneAmp RNA PCR; Perkin Elmer, Norwalk, CT) by using the gene-specific primers rt12 (5′-TAAGGAGGTCCAATGTTACTTCCGCTC-3′) and rt13 (5′-ATCAGTCTGGATGGCTCCTGGTAGTC-3′) and the following cycling conditions: denaturation at 94°C, 5 min; gradient annealing temperatures beginning at 70°C and decreasing 2°C every two cycles to a final annealing temperature of 58°C; extension at 72°C for 7 min. For cell-specific studies, primers rt1 (5′-TGCCATCAAGTGCCAATACC-3′) and rt2 (5′-TGGAGATTCACTATGTCGGC-3′) were used with a single annealing temperature of 64°C. PCR products were analyzed on a 0.8% agarose gel (Sigma) stained with ethidium bromide (Bio-Rad; Hercules, CA).

A 413-bp fragment of rat DDR2 generated by RT-PCR using gene-specific primers rt6 (5′-AATGATCCCGATTCCCAGAATG-3′) and rt15 (5′-TCCCATGTCGGTTACGCCAG-3′) was cloned into the pCRII vector (Invitrogen; Carlsbad, CA) was used as a probe for RPA. Radiolabeled RNA probe was generated by using Sp6 polymerase (Maxiscript; Ambion, Austin, TX) and purified after electrophoresis on a 6% denaturing acrylamide gel. Purified probe was hybridized overnight at 42°C with 10 μg of target RNA and digested with RNase A/T1 (RPA III; Ambion). Protected fragments were resolved on a 6% denaturing acrylamide gel and exposed to film overnight at −80°C.

Confocal and Electron Microscopy

Immunofluorescence studies were carried out by using confocal fluorescence microscopy to determine the localization and cell-specific expression of DDR2 using 4-day neonatal rat hearts. Briefly, 120-μm Vibratome sections from 4-day neonate rat hearts fixed in 2% paraformaldehyde were stained with a primary antibody against DDR2, Connexin 43, and Connexin 45 (SC-7555; Santa Cruz Biotechnology, Inc., Santa Cruz, CA; Sigma C-6219; Chemicon MAB 3100, Temecula, CA, respectively). Sections were rinsed with 1% bovine serum albumin/phosphate-buffered saline (PBS; 3 mmol/L Na2HPO4, 3 mmol/L NaH2PO4, 136 mmol/L NaCl, pH 7.2) and then incubated with a secondary Cy 5-conjugated rabbit anti-goat IgG (diluted 1:100; Jackson ImmunoResearch Laboratories, Inc.) to visualize DDR2. Alexa 488 phalloidin (diluted 1:100; Molecular Probes, Eugene, OR) was used to visualize F-actin. Sections were rinsed with PBS followed by mounting in a 1:3 mixture of glycerin:PBS and 10 mg/ml DABCO (Sigma), and images were collected on a Bio-Rad MRC 1024 confocal scanning laser microscope.

Samples for high-voltage electron microscopy (HVEM) were prepared as previously described (Borg et al., 1983). Briefly, samples were fixed in 4% buffered glutaraldehyde, treated with 2% OsO4, dehydrated in acetone, and embedded in Epon–araldite resin. Sections of 0.25 μm were cut, stained with lead citrate, and examined on a JEOL million-volt HVEM located in Boulder, Colorado. Samples for SEM were prepared as previously described (Borg et al., 1983). Briefly, hearts were rapidly removed from the anesthetized rats, rinsed in PBS, fixed in 2% glutaraldehyde, followed by 2% buffered OsO4, cut into slices perpendicular to the septum, dehydrated, critical point dried, and coated with gold. Specimens were examined in a JEOL 35 SEM at 15 kV.

Isolation of Fibroblast Protein and Western Blotting

Neonatal cardiac fibroblasts were lysed (50 mmol/L Tris-HCl pH 7.2, 10% glycerol, 150 mmol/L NaCl, 0.5% Triton X-100, 2 mmol/L ethylenediaminetetraacetic acid containing CompleteMini protease inhibitor; Roche), and protein concentrations were determined by using the BCA assay (Pierce). Ten micrograms of protein was loaded onto 4–15% gradient gels (Bio-Rad) and subjected to SDS-PAGE followed by Western blot analysis. Membranes were probed with 5 μg/ml anti–EC-DDR2 antibody and bands visualized by using the ECL chemiluminescence system (Amersham).


The authors thank Cheryl Cook, Josh Hastings, and Jeff Davis for technical assistance.