Heart and Circulatory Physiology

Structural coupling of cardiomyocytes and noncardiomyocytes: quantitative comparisons using a novel micropatterned cell pair assay

Dawn M. Pedrotty, Rebecca Y. Klinger, Nima Badie, Sara Hinds, Ara Kardashian, Nenad Bursac


Well-controlled studies of the structural and functional interactions between cardiomyocytes and other cells are essential for understanding heart pathophysiology and for the further development of safe and efficient cell therapies. We established a novel in vitro assay composed of a large number of individual micropatterned cell pairs with reproducible shape, size, and region of cell-cell contact. This assay was applied to quantify and compare the frequency of expression and distribution of electrical (connexin43) and mechanical (N-cadherin) coupling proteins in 5,000 cell pairs made of cardiomyocytes (CMs), cardiac fibroblasts (CFs), skeletal myoblasts (SKMs), and mesenchymal stem cells (MSCs). We found that for all cell pair types, side-side contacts between two cells formed 4.5–14.3 times more often than end-end contacts. Both connexin43 and N-cadherin were expressed in all homotypic CM pairs but in only 13.4–91.6% of pairs containing noncardiomyocytes, where expression was either junctional (at the site of cell-cell contact) or diffuse (inside the cytoplasm). CM expression was exclusively junctional in homotypic pairs but predominantly diffuse in heterotypic pairs. Noncardiomyocyte homotypic pairs exhibited diffuse expression 1.7–8.7 times more often than junctional expression, which was increased 2.6–4.4 times in heterotypic pairs. Junctional connexin43 and N-cadherin expression, respectively, were found in 38.6 ± 7.3 and 39.6 ± 6.2% of CM-MSC pairs, 21.9 ± 5.0 and 13.6 ± 1.9% of CM-SKM pairs, and in only 3.8–9.6% of CM-CF pairs. Measured frequencies of protein expression and distribution were stable for at least 4 days. Described studies in micropatterned cell pairs shed new light on cellular interactions relevant for cardiac function and cell therapies.

  • micropatterning
  • coculture
  • stem cell
  • gap junction

myocardial infarction results in an irreversible cardiomyocyte loss followed by replacement fibrosis and, often, progression to congestive heart failure. Cellular cardiomyoplasty, the transplantation of exogenous cells into the damaged heart, was recently proposed as an alternative approach for the treatment of postinfarction disease (35). Despite the initial promise of the transplantation of skeletal myoblasts and bone marrow-derived stem cells (1, 32), the latest results of double-blind placebo-controlled clinical trials have been more ambiguous (45). One main recognized obstacle to understanding the mechanisms of repair and designing more efficient therapies has been the difficulty in tracking and systematically studying the structural and functional interactions between the implanted cells and host cardiomyocytes in situ. Importantly, although a number of novel cardiogenic cell types are being considered for future clinical trials (2, 3, 18, 24, 50, 53), standardized, well-controlled in vitro or in vivo assays for comparing the ability of different cells to directly integrate with cardiomyocytes and yield safe and efficient functional repair have yet to be developed.

Attempts to characterize interactions between cardiomyocytes and other cell types have been undertaken in a number of studies in vitro. Although the ability of bone marrow-derived mesenchymal stem cells to electrically couple with cardiomyocytes has been well established (6, 52), it remains controversial whether skeletal myoblasts (6, 20, 42, 48) or cardiac fibroblasts (6, 15, 25, 39, 44) can structurally and functionally couple with cardiomyocytes. One reason for the existing ambiguities may be the use of traditional cell coculture systems, which involve random cell shape and spatial distribution as well as the presence of multiple homo- and heterotypic contacts between interacting cells, all of which render the interpretation and quantification of the obtained results difficult. Furthermore, although previous studies have examined the occurrence of electromechanical coupling between different cell types and cardiomyocytes, no attempts were made to 1) quantify the frequency of this occurrence and its dependence on the length of the cell-cell contacts or 2) compare the ability of different cell types to couple with cardiomyocytes under identical and reproducible geometric conditions.

One approach to address these questions is the use of cell micropatterning techniques, which allow for control over cell morphology and position on the substrate, thus enabling systematic studies of cellular interactions at different spatial scales (7, 21, 3638). Using these techniques, we have developed a well-controlled in vitro assay for relatively high-throughput studies of cellular coupling in individual cell pairs with reproducible size, shape, and region of cell-cell contact. This assay was utilized to quantify and compare the probabilities of expression and the spatial distributions of the main electrical (connexin43) and mechanical (N-cadherin) coupling proteins in more than 5,000 homo- and heterotypic cell pairs made of cardiomyocytes (CMs), cardiac fibroblasts (CFs), skeletal myoblasts (SKMs), and bone marrow-derived mesenchymal stem cells (MSCs). Together, our results reveal that connexin43 and N-cadherin are expressed in different cell pairs with a frequency and spatial distribution that significantly depend on the coupling protein being studied, the length of the formed cell contact, and the cell types involved in coupling. The use of standardized in vitro assays for comparing the potential of different cell types to structurally and functionally integrate with cardiomyocytes is expected to enhance our understanding of heterocellular interactions in pathological cardiac states, such as fibrosis, and to facilitate the rational design of safer and more efficient cardiac cell therapies.


All animals were treated according to protocols approved by the Duke University Institutional Animal Care and Use Committee.

Cell isolations.

The utility of the micropatterned cell pair assay was demonstrated using the cell types shown to be important in cardiac function and fibrotic disease or currently used in cardiac cell therapies. For consistency, all studied cells were isolated from Sprague-Dawley rats as follows.

CMs were dissociated from the ventricles of 2-day-old neonatal Sprague-Dawley rats (Harlan) using trypsin (US Biologicals) and collagenase (Worthington) and then resuspended in culture medium supplemented with 10% calf serum and 10% horse serum, as previously described (8). A high purity of cardiomyocytes (>93%) was confirmed by positive staining for sarcomeric α-actinin (Sigma) (see Supplemental Fig. 1A). Supplemental data for this article is available online at the American Journal of Physiology-Heart and Circulatory Physiology website.

CFs were recovered from the CM isolation during the preplating steps and, after reaching confluence, were split and passaged at a 1:2 ratio and used at passage 1. A high purity of cells (>97%) was confirmed by positive staining for vimentin (Supplemental Fig. 1B) and negative staining for sarcomeric α-actinin, smooth muscle actin, and von Willebrand factor (not shown).

Adult rat SKMs were isolated from the hindlimb soleus muscle of female Sprague-Dawley rats using previously published methods (49). Briefly, a biopsy of the soleus muscle was minced in myoblast growth medium (DMEM, 20% FBS, and 50 μg/ml gentamicin) and transferred to a tissue culture-treated dish. Myoblasts that migrated from the minced tissue pieces and adhered to the dish were collected after 10 days. A high purity of cells (>90%) was confirmed by positive staining for desmin and MyoD (Supplemental Fig. 1C). Cells were cultured at below 70% confluence to prevent myotube formation, passaged without freezing, and used at passages 3–6.

Adult rat MSCs were isolated from the bone marrow of 3-mo-old male Sprague-Dawley rats as previously described (29). Briefly, whole bone marrow was collected from the femur and tibia using MSC growth medium (α-MEM, 20% FBS, and 50,000 units of penicillin/50,000 μg of streptomycin). Mononuclear cells were isolated using a Ficoll gradient and plated in tissue culture-treated dishes. Nonadherent cells were discarded at day 2, whereas the adhered cells were validated to be MSCs by positive staining for the cell surface marker CD90 and negative staining for the hematopoietic stem cell marker CD45 (Supplemental Fig. 1D). Cells were passaged without freezing and used at passages 3–6.

Protein micropatterning.

Protein micropatterns in the shape of single- (Fig. 1A) or two-trapezoid islands (Fig. 1B) were prepared as previously described (8). Briefly, silicon wafers were coated with a 10-μm layer of photoresist (SU-8 10; Microchem), exposed to UV light through a photomask (Fig. 1, A1 and B1), and developed to produce a negative of the desired pattern on the silicon wafer (Fig. 1, A2 and B2). Poly(dimethylsiloxane) (PDMS) stamps were cast against the wafer, coated with a protein solution (50 μg/ml fibronectin or 200 μg/ml collagen type IV), and used to transfer the protein pattern (Fig. 1, A3 and B3) onto a 22-mm-diameter PDMS-coated glass coverslip. Unstamped areas of the coverslip were coated with 0.2% (wt/vol) Pluronic F-127 (Molecular Probes), a polymer that prevents the cell and protein adhesion (Fig. 1, A4 and B4) (26). Pluronic surfactants have been used previously in a variety of cell micropatterning experiments, including those that examined roles of cellular interactions in biological and biophysical processes (19, 27, 36). For single-trapezoid islands, the length of the trapezoid top was set at 10 μm, whereas the base (B) and height (H) of the trapezoid were systematically varied in adjacent islands (Fig. 1A1). The total number of protein islands per coverslip was ∼8,000.

Fig. 1.

Micropatterning of single cells (A1–A4) and cell pairs (B1–B4). In single trapezoid islands (A1), the length of the trapezoid top was 10 μm, whereas the base (B) and height (H) were systematically varied in 5-μm increments. In symmetric 2-trapezoid islands (B1), trapezoid base, height, and top (border) were 35, 45, and 10 μm, respectively. A2 and B2: master silicon wafers (dark, photoresist; light, silicon wafer) used to cast poly(dimethylsiloxane) (PDMS) stamps. A3 and B3: stamped fibronectin islands (red) surrounded by cell-repellent area (black). A4 and B4: individual single cells and cell pairs formed on fibronectin islands. CF, cardiac fibroblasts; SKM, skeletal myoblasts; MSC, mesenchymal stem cells. (See text for further details).

Optimization of single-cell attachment and spreading.

Single-trapezoid fibronectin islands (Fig. 1A3) were seeded overnight with 7.5 × 104 cells in 2 ml of cell-specific medium. After 48 h, the cultures were immunostained to assess cell adhesion and spreading. For each cell type (CM, CF, SKM, or MSC), the total number of trapezoid islands that contained single spread cells (to >70% of the trapezoid area) was determined for the different trapezoid dimensions (i.e., different combinations of H and B). The dimensions that most frequently yielded a single spread cell (i.e., “optimal dimensions”), as well as the frequency of single-cell formation as a function of the trapezoid area and elongation ratio (H/B), were determined for each cell type.

Optimization of cell pair formation.

For cell pair studies, the following two-trapezoid protein islands were micropatterned: 1) asymmetric, where one trapezoid had the dimensions optimized for CMs (Fig. 2 A1) and the other had dimensions optimized for CFs, SKMs, or MSCs (Fig. 2, A2–A4), and 2) symmetric, where both trapezoids had the same dimensions optimized for all cell types pooled together (Fig. 2A5). In initial studies, the symmetric islands yielded higher numbers of heterotypic (consisting of a CM and a noncardiomyocyte) cell pairs for all cell types compared with asymmetric islands and were therefore used for all subsequent studies.

Fig. 2.

Efficiency of single-cell patterning as a function of trapezoid dimensions. The percentage of single cardiomyocytes (CMs; A1), CFs (A2), SKMs (A3), MSCs (A4), and all cell types pooled together (A5) is shown relative to the percentage of cells patterned on a trapezoid with optimal dimensions (circled in yellow). Darker color indicates a higher relative percent value. Insets: examples of immunostained micropatterned cells used for cell counting. DAPI, 4–6-diamidino-2-phenylindole. B1 and B2: percentage of successfully patterned single cells as a function of trapezoid area and elongation (H/B). For each of the 4 cell types, a total of 15,000 fibronectin islands were analyzed over 3 independent experiments (cell isolations). The total number of analyzed islands for all groups was 60,000.

For sequential cell seeding studies, cells of one type were seeded overnight (at 3 × 104, 5 × 104, 7.5 × 104, or 10 × 104 cells per coverslip) on two-trapezoid islands (Fig. 1B3), and after 48 h, the number of islands with 1) no cells, 2) a single cell attached to one of the two trapezoids, 3) a single cell spanning both trapezoids, 4) two cells, and 5) multiple (>2) cells were counted in immunostained cultures. Defined serum-free medium [DMEM/F-12 (GIBCO) with 50 U/ml penicillin, 50 μg/ml streptomycin, 2 μg/ml l-thyroxine, 0.1 μg/ml insulin, 0.5 μg/ml transferrin, 2.5 μg/ml ascorbic acid, 1 nM lithium chloride, and 1 nM sodium selenite (Sigma)], modified from Tuvia et al. (51), was used in all cultures to limit both the cell proliferation and spreading across the two trapezoids.

In addition, CMs (3 × 104/coverslip) were seeded overnight, and 48 h later, noncardiomyocytes (CFs, SKMs, or MSCs) were seeded overnight at different densities (3 × 104, 5 × 104, 7.5 × 104, or 10 × 104 cells per coverslip) in different media (serum free or cell specific). After an additional 48 h, the immunostained cultures were assessed for the formation of homotypic and heterotypic cell pairs. Finally, the efficiency of cell pair formation on fibronectin compared with collagen IV islands was assessed at a seeding density of 3 × 104 cell pairs per coverslip for each cell type. For all of the described studies, image analyses (recognition of different cellular patterns and cell pair types, cellular area coverage of the protein island, and counting of spread single cells and cell pairs) were performed in an automatic fashion from multicolor fluorescence images using custom-designed Matlab software.

Expression and spatial distribution of coupling proteins in cell pairs.

Cell contact in each cell pair was defined as either “side-side” (i.e., with 75–95 μm of contact length alongside 2 cells, each spanning both trapezoids longitudinally) or “end-end” (i.e., with 10–30 μm of contact length between 2 cells, each approximately occupying 1 trapezoid). For each cell type (CM, CF, SKM, or MSC) and each cell pair type (homotypic or heterotypic), the probability of expression of the coupling proteins (connexin43 or N-cadherin) was assessed by counting the number of positively immunostained cell pairs relative to the total number of formed cell pairs per coverslip. In addition, the distribution of positive staining within a cell pair was classified as either “junctional” (present at the site of cell contact) or “diffuse” (present within 1 or both cells but not at the site of cell contact). Positive staining both at the site of cell contact and inside the cytoplasm was also classified as junctional. Finally, to examine the effect of spontaneous CMs beating on the formation of junctions in heterotypic cell pairs, we also performed a set of experiments using the serum-free culture medium supplemented with 2 μM epinephrine.


Immunostaining was preformed as previously described (8). Briefly, cell cultures were fixed, permeabilized, blocked, and incubated with primary antibody overnight [anti-sarcomeric α-actinin (1:100; Sigma), anti-vimentin (1:1,000; Sigma), anti-myoD (1:50; Santa Cruz Biotechnology), anti-CD45 (1:100; Chemicon), anti-CD90 (1:100; Chemicon), anti-connexin-43 (1:100; Zymed), and anti-pan cadherin (1:100; Sigma)] and secondary antibody [anti-rabbit and anti-mouse Alexa Fluor IgGs (1:200; Molecular Probes)] for 2 h. 4–6-Diamidino-2-phenylindole (Sigma) or YOYO-1 (Molecular Probes) were used for visualization of the nuclei. Fluorescent wheat germ agglutinin (Molecular Probes), a lectin that binds to membrane-bound sugar residues, was used to visualize the cell membrane (12). FITC-conjugated phalloidin (Sigma) was used to assess cell shape and spreading. Immunostained cells were imaged using a fluorescent microscope (Nikon TE2000U), a charge-coupled device camera (Sensicam QE; Cooke), and IPLab image acquisition software or a confocal microscope (Zeiss LSM 510).

Statistical analysis.

Data for statistical analysis were collected from three to six independent experiments (cell isolations) and are expressed as means ± SD. A paired t-test was used when comparing two outcomes in the same cell type (e.g., homo- vs. heterocellular pair formation, adhesion to fibronectin vs. collagen IV islands, diffuse vs. junctional expression). A rank sum analysis was used when comparing the two outcomes collectively for all cell types (e.g., fibronectin vs. collagen IV adhesion, adhesion in serum vs. serum-free medium). A one-way ANOVA followed by post hoc Student's t-test was used when comparing different cell types with respect to a single outcome (e.g., frequency of junctional expression, adhesion to fibronectin islands, heterocellular pair formation) or when comparing multiple conditions or outcomes (e.g., seeding densities, spatial distributions of patterned cells) in the same cell type. For all statistical tests, differences were considered significant when P < 0.05.


Optimization of single-cell attachment and spreading.

Protein islands consisting of two abutting trapezoids (Fig. 1, B1–B4) were used to form individual micropatterned cell pairs. To facilitate pair formation, we first systematically varied the dimensions of the single trapezoids (Fig. 1, A1–A4) and assessed the efficiency of single-cell patterning for each of the four studied cell types (Fig. 2, A1–A5). The (H, B) (in μm) combinations that yielded the highest single-cell attachment and spreading were (30, 30) for CMs, (45, 35) for CFs, (35, 35) for SKMs, (45, 35) and (55, 45) for MSCs, and (35, 45) for all pairs pooled together. In general, single trapezoid-shaped cells were most efficiently formed for trapezoid areas of 650–980 μm2 for CMs and SKMs, 980–1640 μm2 for CFs, and 1,640–1970 μm2 for MSCs (Fig. 2B1) and for a trapezoid elongation ratio of 1.2–1.6 for all cell types (Fig. 2B2).

Optimization of heterotypic cell pair formation.

Cell culture conditions were further optimized with the goal to maximize the number of obtained heterotypic (consisting of a CM and a noncardiomyocyte) cell pairs per coverslip. In this optimization process, we first attempted simultaneous seeding of CMs and noncardiomyocytes (CFs, SKMs, or MSCs) in different ratios, since that would have been the simplest way to form the heterotypic cell pairs. However, freshly isolated CMs required ∼6 h to attach and start spreading on the micropatterned protein islands, whereas the attachment of all studied nonmyocytes required only ∼20–30 min. Therefore, simultaneous seeding of CMs and noncardiomyocytes yielded preferential attachment and spreading of noncardiomyocytes and the dominance of homotypic noncardiomyocyte pairs, whereas the number of obtained heterotypic cell pairs was negligible. Therefore, to promote the formation of heterotypic pairs, we developed the sequential seeding procedure, where one cell type was seeded and allowed to spread first, followed by the seeding of the other cell type. Initially, each cell type was seeded overnight at different densities on symmetric two-trapezoid fibronectin islands [(H, B) = (45, 35) μm, Fig. 1B3] and assessed after 48 h (Fig. 3). Seeding of CMs yielded a higher number of “singles” (a single cell spread on one trapezoid with the other trapezoid available for the attachment of a second cell type) than the seeding of any of the noncardiomyocyte types, which preferably spread on both trapezoids, leaving no room for the attachment of another cell (compare “singles” and “spanning” in Fig. 3, A and B–D). Consequently, cell pairs in all subsequent studies were formed by overnight seeding of 3 × 104 CMs per coverslip (i.e., 8 × 103 cells/cm2, yielding 17% of islands with singles, Fig. 3A), followed by noncardiomyocyte seeding after 48 h.

Fig. 3.

Spatial distribution of patterned cells on 2-trapezoid islands as a function of seeding density. Labels on x-axes sketch specific cellular patterns found on the 2-trapezoid islands 48 h after seeding of a particular cell type (AD). Note the significantly higher percentage of formed singles in CMs relative to other cell types. Seeding density is denoted in thousands (K). For each of the 4 cell types, a total of 7,500 fibronectin islands were analyzed for each of the 4 cell densities over 3 independent experiments (cell isolations). The total number of analyzed islands for all groups was 120,000.

Furthermore, the efficiency of heterotypic cell pair formation was not increased by increasing the seeding density of noncardiomyocytes from 3 × 104 to 10 × 104 cells per coverslip or by using cell-specific medium containing serum instead of the defined serum-free medium (Supplemental Fig. 2). Finally, the formation of homotypic CM pairs was favored on fibronectin vs. collagen IV islands (Fig. 4A), whereas homotypic SKM pairs formed more frequently on collagen IV (Fig. 4B), which is in agreement with recent studies of CM and SKM adhesion on different extracellular matrix proteins (28). In contrast, the frequency of heterotypic cell pair formation was independent of the type of adhesion protein (Fig. 4C). Since a trend toward a higher number of total cell pairs was found using fibronectin vs. collagen IV (Fig. 4D, P = 0.1), seeding of ∼8 × 103 CMs/cm2 on a substrate patterned with symmetric two-trapezoid fibronectin islands (with bases of 35 and 10 μm and a height of 45 μm), followed after 48 h by seeding of noncardiomyocytes at the same density in defined, serum-free medium, was adopted as the optimal patterning methodology for all subsequent cell pair studies.

Fig. 4.

Efficiency of cell pair formation on fibronectin and collagen IV islands. A: homotypic CM pairs in cocultures of CMs and CFs, SKMs, or MSCs. B: homotypic noncardiomyocyte pairs in the same cocultures as in A. C: heterotypic cell pairs in the same cocultures as in A. D: all formed cell pairs in the same cocultures as in A. For each of the 3 noncardiomyocyte cell types, a total of 30,000 fibronectin and 15,000 collagen IV islands were analyzed over 6 independent experiments (cell isolations). The total number of analyzed islands for all groups was 135,000. *P < 0.05, significantly different from fibronectin.

Coupling proteins in homotypic cell pairs.

The frequency of expression and the spatial distribution of connexin43 (Fig. 5A) and N-cadherin (Fig. 5B) were assessed in homo- and heterotypic cell pairs to determine the possibility of structural coupling between different cell types. As expected, connexin43 (Fig. 5C1) and N-cadherin (Fig. 5D1) were expressed in virtually all homotypic cardiomyocyte pairs. In contrast, connexin43 was expressed in 16.2 ± 3.1, 82.3 ± 4.0, or 53.5 ± 5.0% (Fig. 5C1) and N-Cadherin in 21.3 ± 1.9, 13.4 ± 2.8, or 39.7 ± 7.6% (Fig. 5D1) of the homotypic CF, SKM, or MSC pairs, respectively. Whereas homotypic cardiomyocyte pairs exclusively exhibited junctional protein distribution (at the site of cell contact), the homotypic noncardiomyocyte pairs exhibited diffuse distribution (i.e., multiple spots throughout the cell pair) 1.7–8.7 times more often than junctional, with no junctional N-cadherin distribution found in homotypic pairs of SKMs and CFs (Fig. 5, C1 and D1).

Fig. 5.

Frequency of expression of coupling proteins in cell pairs. A: examples of junctional, diffuse, or no (absent) connexin43 expression within homo- and heterotypic cell pairs with end-end or side-side intercellular contacts. Red, CM-specific sarcomeric α-actinin; blue, wheat germ agglutinin (WGA; membrane stain); green, nuclei and connexin43. B: examples of N-cadherin expression within different cell pairs. Red, sarcomeric α-actinin; blue, nuclei; green, N-cadherin. Enhanced contrast in insets demonstrates presence of cell-to-cell contacts, verifying heterotypic pair formation. Arrowheads in A and B indicate junctional protein expression. C and D: frequency of junctional and diffuse connexin43 (C1 and C2) and N-cadherin (D1 and D2) expression in homotypic (C1 and D1) and heterotypic cell pairs (C2 and D2). The total number of analyzed homotypic cardiomyocyte pairs was 450 for each of the 2 coupling proteins over 6 independent experiments (cell isolations). For each of the 3 noncardiomyocyte cell types, a total of ∼150 heterotypic and ∼300 homotypic cell pairs were analyzed for each of the 2 coupling proteins over 3 independent experiments (cell isolations). The total number of analyzed cell pairs for all groups was ∼3,600. *P < 0.05, significantly different from expression in corresponding heterotypic pairs. #P < 0.05, significantly different from the same type of expression in CF pairs. ^P < 0.05, significantly different from the same type of expression in SKM pairs.

Expression of coupling proteins in heterotypic cell pairs.

Interestingly, the two coupling proteins were expressed more frequently in heterotypic than in corresponding homotypic cell pairs (compare Fig. 5, C1 vs. C2 and D1 vs. D2). Specifically, connexin43 was found in 75.9 ± 2.9, 91.6 ± 2.3, or 90.2 ± 4.8% and N-cadherin in 89.5 ± 10.5, 85.6 ± 6.2, or 82.8 ± 3.9% of heterotypic CF, SKM, or MSC pairs, respectively. Junctional connexin43 distribution was found 4.4, 2.6, and 2.8 times more often in heterotypic vs. homotypic CF, SKM, and MSC pairs, respectively (Fig. 5, C1 and D1), whereas junctional N-cadherin distribution was found in 3.6 and 13.6% of heterotypic vs. 0% of homotypic CF and SKM pairs and 3.3 times more often in heterotypic vs. homotypic MSC pairs (Fig. 5, C2 and D2). Similarly, diffuse connexin43 distribution was found 4.7 and 1.3 times more often in heterotypic vs. homotypic CF and MSC pairs, respectively (Fig. 5, C1 and D1), whereas diffusely expressed N-cadherin was found 4.0, 5.4, and 1.6 times more often in heterotypic vs. homotypic CF, SKM, and MSC cell pairs, respectively (Fig. 5, C2 and D2). Overall, unlike MSCs and SKMs, CFs were found to express either connexin43 or N-cadherin in only a small portion of homo- and heterotypic pairs (up to 9.6% junctionally and 21.3% diffusely). Furthermore, the frequencies of junctional expression were unaffected by the epinephrine-induced increase in the fraction of spontaneously beating CMs (from <5 to ∼50%). Importantly, for all cell pair types, the junctional and diffuse expression frequencies remained the same over longer culture times (i.e., no differences were found at 2 vs. 4 days after noncardiomyocyte seeding).

Diffuse distribution of coupling proteins in heterotypic cell pairs.

Upon further examination of the diffuse protein distribution in cell pairs (Fig. 6), high-magnification confocal images revealed that the diffusely expressed proteins were located inside the cell cytoplasm. In contrast, proteins expressed at cell junctions appeared to span the entire thickness of the two contacting cell membranes (Fig. 6B). Furthermore, in the diffusely expressing heterotypic pairs (Fig. 6, A1 and A2), CMs always expressed the coupling proteins, either alone or together with the noncardiomyocytes (Fig. 6, C1 and C2). Therefore, although the diffuse protein distribution was found more frequently in heterotypic than in corresponding homotypic pairs (Fig. 5, C and D), this was a result of the occurrence of diffuse expression in the CMs rather than the increased frequency of diffuse expression in the noncardiomyocytes (Fig. 5, C and D2, and Fig. 6C2). Only SKMs expressed diffuse N-cadherin more frequently when in heterotypic than in homotypic cell pairs.

Fig. 6.

Diffuse distribution of coupling proteins in heterotypic cell pairs. A: examples of diffuse connexin43 (A1) and N-cadherin (A2) distribution within different cell pairs. Red, sarcomeric α-actinin; green, connexin43 or N-cadherin; blue, nuclei. Enhanced contrast in insets demonstrates presence of cell-to-cell contacts, verifying heterotypic pair formation. B: confocal z-stack image (made of 0.25-μm-thick optical slices) showing both junctional and diffuse connexin43 pattern in a single cell pair. Red, WGA; green, connexin43. Images at top and right are front and side views of the stack, respectively. Note that junctional expression occurs at the membrane, whereas diffuse expression is intracellular. C: percentage of heterotypic cell pairs with connexin43 (C1) and N-cadherin (C2) diffusely expressed in only CMs or in both CMs and noncardiomyocytes. For each of the 3 noncardiomyocyte cell types, a total of ∼100 heterotypic diffusely expressing cell pairs were analyzed for each of the 2 coupling proteins over 3 independent experiments (cell isolations).The total number of analyzed cell pairs for all groups was ∼600. The 3 groups for each cell pair type differ significantly.

Coupling protein expression as a function of the type of cell contact.

Finally, the side-side contacts in cell pairs were found to form ∼4.5–14.3 times more frequently than end-end contacts (Fig. 7A). Consequently, the majority of junctional connexin43 and N-cadherin were found in the cell pairs with side-side cell contacts. Furthermore, junctional connexin43 was more likely to be expressed in heterotypic pairs that contained side-side rather than end-end cell contacts for SKMs (3.3. times) and MSCs (3.5 times) but not CFs (Fig. 7A1). In contrast, junctional N-cadherin was expressed in a higher (albeit small) percentage of side-side than end-end contacts in CF heterotypic pairs, whereas the difference was insignificant for MSC and SKM pairs (Fig. 7A2).

Fig. 7.

Distribution of coupling proteins as a function of the type of cell-cell contact. A1: connexin43. A2: N-cadherin. Each bar represents the percentage of total cell pairs of a given type (CM-CM, CM-MSC, CM-SKM, and CM-CF) that expressed coupling protein and exhibited end-end (end) or side-side (side) cell contacts. Two subdivisions in each bar represent fractions with junctional (shaded) and diffuse (open) protein distributions, with the number next to the top of the shaded bar showing the percent fraction for junctional distribution. The total number of analyzed homotypic CM pairs was 180 for each of the 2 coupling proteins over 3 independent experiments (cell isolations). For each of the 3 noncardiomyocyte cell types, a total of ∼120 heterotypic cell pairs were analyzed for each of the 2 coupling proteins over 3 independent experiments (cell isolations).The total number of analyzed cell pairs for all groups was ∼1,080. *P < 0.05, significant difference between junctional and diffuse fractions. #P < 0.05, significant difference between junctional fractions in the pairs with end-end vs. side-side contacts. B1–B5: schematic of a 2-dimensional cardiac tissue (red) with integrated donor cells (yellow). In B1–B3, an integrated noncardiomyocyte donor cell is shown in the role of a passive current sink (B1), supporting transverse (B2) or longitudinal conduction (B3). In B4 and B5, an integrated homocellular group (cluster) of noncardiomyocytes is shown, supporting transverse (B4) or longitudinal (B5) conduction within cardiac tissue. Formed gap junctions are shown in green. Arrows indicate the conduction pathway through the donor cells.


One of the major hurdles in developing cell therapies that would provide safe, efficient, and sustained cardiac repair is the difficulty in systematically studying the fate of implanted cells inside the complex three-dimensional (3-D) heart tissue environment. In particular, the probability that a donor cell will structurally and functionally integrate with host cardiomyocytes depends on a number of factors, including the route of cell delivery, the spatial distribution and number of resident noncardiomyocytes at the delivery site, and local and systemic neurohumoral effects, among others. These mainly uncontrollable factors preclude 1) meaningful in situ comparison of the intrinsic potentials of different donor cell types (or their subsets) to electromechanically integrate with host cardiomyocytes and 2) identification of soluble and other factors that can modulate or improve this integration. We therefore developed a novel micropatterned cell pair assay for quantitative studies of structural coupling between cardiomyocytes and other cells, to enhance our understanding of cardiac pathophysiology and aid the design of future cardiac cell therapies.

Optimization of micropatterned cell pair assay.

Micropatterned homotypic cell pairs have been previously used for studies of VE-cadherin-mediated proliferation of endothelial cells (36) and ice propagation in HepG2 hepatoma cells (19). We, on the other hand, undertook a set of comprehensive studies (Figs. 14 and Supplemental Fig. 2) to maximize the formation of heterotypic cell pairs between cardiomyocytes and different types of noncardiomyocytes. Although our optimized methods yielded relatively efficient patterning of homotypic cell pairs (∼20%, Fig. 3), heterotypic pairs formed on only ∼1% of all fibronectin islands (Fig. 4C). Considering that with the current patterning density, ∼8,000 identical two-trapezoid islands were fabricated per 22-mm-diameter coverslip, the obtained ∼80 heterotypic cell pairs per coverslip still represent a statistically relevant sample size. Furthermore, comparative studies in this system are facilitated by the abundance of homotypic cardiomyocyte pairs (4–7% of all islands, Fig. 4A), which provide a common positive control for the quality and uniformity of the immunostaining within and among different coverslips. The relatively low seeding densities (8,000 cells/cm2) should enable similar studies with lower yielding cells, including different types of cardiac progenitors (2, 3, 24, 50, 53), ventricular cells from neonatal knockout mice (4), and atrial cells (5).

Interactions between cardiomyocytes and noncardiomyocytes.

The reproducible biochemical (defined serum-free medium) and geometric (cell shape, size, and area of contact) conditions in this assay enabled us to systematically quantify and compare the occurrence of distinct spatial patterns of connexin43 and N-cadherin in different cell pairs. Unlike cardiomyocytes, noncardiomyocytes in homotypic pairs predominantly exhibited diffuse (inside the cytoplasm) rather than junctional (at the site of the cell-cell contact) protein expression, which, in turn, was upregulated in heterotypic pairs in the presence of an abutting cardiomyocyte (Fig. 5, C and D). These results are consistent with studies that found upregulated expression of coupling proteins in SKMs (14) and MSCs (41) in cocultures with cardiomyocytes. Simultaneously, in the presence of a noncardiomyocyte in heterotypic pairs, cardiomyocytes expressed connexin43 and N-cadherin predominantly in a diffuse pattern (Fig. 5, C and D, and Fig. 6C), which otherwise was not present in either single cardiomyocytes (not shown) or in homotypic cardiac pairs (Fig. 5, C1 and D1). It remains to be elucidated whether the described bidirectional effects on connexin43 and N-cadherin expression in heterotypic cell pairs are the result of juxtacrine (cell contact mediated) and/or short-range paracrine signaling from the abutting cells. Furthermore, although a diffuse (dotted) pattern of connexin43 and N-cadherin has been previously reported in infarct border zones (30, 40) and regions with implanted cells (10, 46), further studies are warranted to investigate the precise distribution of coupling proteins at heterocellular interfaces within the heart.

It is also important to note that the diffusely expressed coupling proteins inside the cytoplasm (Fig. 6B) are not likely to support intercellular coupling, but rather represent newly produced, trafficking, or internalized connexin and cadherin molecules. The fact that the fractions of cell pairs expressing proteins in either a diffuse or junctional pattern remained constant over time in culture suggests that the two pattern types represent steady-state rather than transient stages in protein expression. Considering that purely diffuse (and likely nonfunctional) protein distribution was abundant in both heterotypic and homotypic noncardiomyocyte pairs (Fig. 5, C and D), relating positive immunostaining results to the existence of functional coupling in more complex settings, such as mono- or multilayer cell cocultures or 3-D tissues should be exercised with caution. Similarly, Western blot analyses of coupling proteins in heterocellular cardiac settings, when related to electromechanical function, should be limited to membrane, rather than the whole cell, fractions.

In this study, only a portion of noncardiomyocytes (MSCs, SKMs, or CFs) was found to form detectable electrical and mechanical junctions when in contact with cardiomyocytes (Fig. 5, C2 and D2). This result may reconcile previous conflicting reports on the ability of SKMs and CFs to couple with cardiomyocytes in vitro or in vivo. Although the previous studies may have examined a fraction of the formed cell contacts, our study was systematically performed in a large number of well-defined cell pairs in an attempt to better represent and compare the behaviors of the entire cell populations. It is important to note that the freshly isolated or passaged primary noncardiomyocytes used in this study inherently represent a mixture of heterogeneous cells. Whether the observed coupling in only a portion of these cells represents a probabilistic event or is characteristic of cells with a specific phenotype (i.e., at a specific developmental or differentiation stage, coming from a specific tissue region, or created by specific culture conditions) remains to be studied.

Interestingly, despite the limited space for longitudinal cell spreading (i.e., the 10-μm narrow connection between the 2 abutting trapezoids), cells in all pairs predominantly formed side-side (long) rather than end-end (short) contacts (Figs. 5A, 6A, and 7). Importantly, junctional connexin43 was expressed more often within side-side than end-end contacts (Fig. 7A), possibly due to the higher chance for junction formation in longer membrane appositions. In particular, the chance that connexin43 junctions were formed between cardiomyocytes and SKMs or MSCs increased respectively from 7.5 or 13% for 10- to 30-μm long-end-end contacts to 25 or 46% for 75- to 95-μm-long side-side contacts (Fig. 7A). Assuming linear scaling, a single SKM or MSC (with an optimal perimeter of 120 μm for SKMs and 156 μm for MSCs, Fig. 2, A3 and A4) would respectively have a 34 or 82% chance to electrically couple when fully surrounded with cardiomyocytes (Fig. 7B1). Furthermore, this single donor cell would have more chance to act as an ion current sink (by coupling to only 1 cardiomyocyte, Fig. 7B1) than as a conducting bridge (by having to couple to 2 cardiomyocytes on its opposite sides, Fig. 7, B2 and B3) (23). Multicellular regions (clusters) of SKMs or MSCs (Fig. 7, B4 and B5), on the other hand, would be less likely to electrically conduct within the cardiomyocyte network than single cells, because homotypic junctions within the cluster (necessary for conduction through the cluster) would form ∼2.7 times less often than heterotypic junctions between the cluster and surrounding cardiomyocytes (Fig. 5, C1 and C2).

Implications for cellular therapy.

Although extrapolating these results to a 3-D setting is difficult, it could be speculated that the obtained relative relationships would be maintained, whereas the results would simply scale up due to the larger cell area available for coupling in 3-D than in 2-D. In support, recent studies by Mills et al. (33) have shown that discrete cell clustering in conjunction with lower connexin expression of injected SKMs is more arrhythmogenic than the diffuse engraftment and higher connexin expression of infused MSCs. Furthermore, if implanted single cells or cell clusters were to elongate within anisotropic heart tissue, higher chances for electrical coupling to occur at longer cell sides (Fig. 7, B2 and B4) than shorter cell ends (Fig. 7, B3 and B5) might alter local electrical anisotropy relative to surrounding cardiac tissue with unknown consequences on arrhythmia induction. Therefore, although from our studies, MSCs appear to be a superior choice for cell therapy over SKMs with regard to their coupling potential (12–14% homotypic and 39–40% heterotypic MSC vs. a respective 0–8 and 14–22% of SKM pairs expressing junctions, Fig. 3, C and D), it remains unknown to what degree and over what spatial scales coupling between host cardiomyocytes and electrically passive (17, 34), immature (13, 47), or heterogeneous (16, 17) donor cells would be safe and advantageous for functional improvements in heart disease. Systematic studies in micropatterned cocultures of cardiomyocytes and other cell types, similar to those by Bursac et al. (8, 9, 22), could be used to further address these questions.

Study limitations.

One study limitation is the use of immunostaining, which may neglect the existence of small junctions made of a few protein molecules still capable of supporting electromechanical coupling between the cells (11, 48). At least in the case of homotypic cardiomyocyte pairs, the portion of coupled cells detected by immunostaining was ∼100% (Fig. 5, C1 and D1), a result likely to be obtained if cardiomyocyte pairs were studied functionally. Furthermore, our immunostaining results for heterotypic SKM pairs are in excellent agreement with recent dual-voltage clamp recordings used to evaluate the coupling frequency between C2C12 SKMs and adult rat cardiomyocytes (48). Nevertheless, it is possible that the coupling potential of each cell type was underestimated by a small portion of functionally coupled, but connexin43- and N-cadherin-negative, cell pairs. However, we believe that this underestimation would not significantly alter the comparative results among different cell types and cell pair settings. In addition, spontaneous cell fusion is not likely to affect the obtained results due to its reportedly low incidence [<0.01% of cells (31, 43)]. Importantly, although dye transfer and dual-voltage clamp techniques form conventional approaches for studies of cellular interactions, our assay complements these techniques by allowing relatively high-throughput quantitative experiments in a geometrically reproducible setting.

Another limitation of this study is the unavoidable use of neonatal cardiomyocytes, which exhibit uniform gap junctional distribution rather than the polar distribution characteristic of adult cardiomyocytes. Although the proportion of end-end vs. side-side heterotypic junctions may therefore differ in micropatterned cell pairs and in adult heart tissue, it is important to note that cardiomyocytes in the infarct border zone also exhibit a uniform, neonatal-like distribution of gap junctions (30, 40). Conceivably, cellular interactions between cardiomyocytes in the border zone and other cells (that may resemble the interactions described in the micropatterned cell pairs in this study) may prove to be highly relevant for the function of the infarcted heart and for the success and safety of cell implantation therapies.


In summary, we have developed and utilized a novel, reproducible micropatterned cell pair assay to quantitatively compare the intrinsic potential of different cell types to structurally couple with cardiomyocytes. This in vitro system will be further used to screen and systematically study the effects of different soluble factors and genetic manipulations on the formation of electromechanical junctions between cardiomyocytes and potential donor cells. We believe that the control of host-donor cell coupling may eventually allow us to guide the integration and differentiation fate of donor cells inside the heart and promote the design of safer and more efficient cardiac cell therapies.


This work was supported by American Heart Association Predoctoral Fellowship 0515377U, American Heart Association Scientist Development Grant 0530256N, and National Heart, Lung, and Blood Institute Grant HL083342.


We acknowledge Margaret White for help with image acquisition and cell counting and Dr. Vann Bennett for kindly supplying the serum-free media formulation.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract