Engineering Cardiac Cell Junctions In Vitro to Study the Intercalated Disc

Abstract

This review article discusses a recent work using engineered cardiac cells to study the function of the intercalated disc putting emphasis on mechanical and electrical coupling.

Keywords::

• cell engineering
• intercalated disc function
• mechanical cell–cell coupling
• electrical cell–cell coupling

INTRODUCTION: ENGINEERING CARDIAC TISSUE

The myocardium consists of a complex three-dimensional network of excitable cells, which are excited by a propagating electrical impulse, which in turn acts as a trigger for and a modulator of contraction of individual myocytes. The electrical impulse (action potential) is created by a complex interplay of ion currents flowing through multiple ion channels. Exact understanding of the process of impulse propagation requires knowledge not only about the mechanism of action potential generation but also about so-called passive electrical properties of the myocardium which include cell dimensions, cell membrane capacity and resistance, the resistance of the cytoplasmic space, and the gap junction plaques containing low-resistance gap junction channels. (Kleber & Rudy, 2004) There have been many experimental attempts to determine these properties, which mostly used experimental and theoretical models derived from nerve tissue. Inherent to all these models is the specific selection of a cardiac structure (Purkinje fiber segments and cylindrical papillary muscles) with the aim to use a defined arrangement of myocytes that would allow deduction of cellular electrical parameters from experiments involving the passage of electrical currents across a multicellular, functional syncytium (Kleber & Riegger, 1987).

In the 1970s Lieberman et al. (1972, 1975) produced “1-dimensional” artificial strands of cardiac myocytes growing along nylon filaments or in agar grooves to determine passive electrical properties and propagation. Subsequently, two-dimensional engineered cell cultures fabricated by photoresist microlithography were used by Rohr et al. (1991) and later in many studies of our laboratory to engineer defined cardiac tissue geometries in cell cultures for studying cardiac electrical impulse propagation and defining basic biophysical rules which determine the effect of extracellular defibrillation shocks (Fast & Kleber, 1993; Kucera et al., 1998, 2001; Rohr et al., 1997, 1998; Gillis et al., 1996; Fast et al., 1998). A quantum step in terms of spatial resolution in tissue and cell engineering was made by the introduction of soft-microlithography (Singhvi et al., 1994). This technique, which enables the engineering of single cells and multicellular preparations with micrometer resolution on defined extracellular substrates, was applied to myocardial tissue and cells by Parker et al. (Feinberg et al., 2007; Bray et al., 2008; Geisse et al., 2009; McCain & Parker, 2011) to determine shape and arrangement of contractile proteins and the dependence of contractile force generation on cell geometry (length/width ratio) and the organization of the contractile apparatus.

This brief review article focuses on engineering cardiac myocyte pairs and the cardiac cell junction using soft-microlithography to study mechanotransduction at the intercalated disc (ID) and the effects of genetic connexin ablation and heterogeneous connexin expression on electrical impulse transfer. In the classical view, it was perceived that the cardiac cell junction or ID would contain two important functional elements, the mechanical junctions (fascia adherens, desmosomal complexes, and area composita) (Li & Radice, 2010) and gap junctional plaques and ribbons (Hoyt et al., 1989) consisting of gap junction channels. These channels, formed by connexin proteins, provide a low resistance pathway to electrical currents and small regulatory molecules (Harris, 2001; Valiunas et al., 2005) and thus play a key role in electrical cell-to-cell impulse transfer. More recently it has become evident that the ID hosts ion channels as a third class of functional proteins, critically important to cell function (Maier et al., 2004; Petitprez et al., 2011; Milstein et al., 2012; Balse et al., 2012). Inherent to this triad of functional protein complexes is mutual regulatory interactions. Thus it has been shown that ablation of plakophyllin-2 (PKP-2), a desmosomal protein containing armadillo repeats, decreases the inward Na+ current carried by Nav1.5 (Sato et al., 2009). Likewise, genetic ablation of connexin also modulates Nav1.5, (Desplantez et al., 2012) as will be discussed in the third section of this article. A further aspect of this fascinating story of interaction between closely apposed protein complexes of the ID, exerting entirely different functions, relates to arrhythmogenic cardiomyopathy. This disease is primarily due to mutations in desmosomal proteins, (Campuzano et al., 2013) but leads to change in expression and/or malfunction of the partner protein complexes in the ID (Asimaki et al., 2009; Noorman et al., 2013). This is a further interesting example showing that the analysis of a disease mechanism can amend to new knowledge about basic cellular function. Furthermore, IDs are exposed to highly dynamic mechanical forces because they exist at the interface of actively contracting cardiac myocytes. Studies are increasingly revealing that mechanical forces are important regulators of cell–cell coupling, (Zhuang et al., 2000; Yamada et al., 2005; Salameh et al., 2010) motivating new hypotheses about the mechanosensitivity of processes related to ID assembly, structure, and function.

ENGINEERING CARDIAC CELL PAIRS AND THE INTERCELLULAR JUNCTION

Mechanical coupling between cardiac cells is assumed by adherens junctions and desmosomes, (Li & Radice, 2010; Borrmann et al., 2006) which couple extracellularly to corresponding complexes of neighboring cells and intracellularly to myofibrils and intermediate filaments. Thus, forces generated by the cytoskeleton are exerted against the ID (and vice versa), which must maintain the structural integrity of the tissue while exposed to mechanical stress. The specific location, diverse composition, and multiple inputs and outputs of the ID suggest that there are complex biophysical factors and molecular feedback mechanisms that regulate its structure and function. Although the structure of the ID has been thoroughly studied in explanted animal (Fishman et al., 1990; Hirschy et al., 2006; Oxford et al., 2007) and human (Asimaki et al., 2009; Peters et al., 1994; Sepp et al., 1996) hearts, ex vivo systems have limitations in identifying structure–function relationships and underlying mechanisms due to heterogeneities and concurrent remodeling processes that occur in the whole heart. Thus, in vitro models are essential for isolating the effects of individual stimuli on the structure and function of the ID. By leveraging techniques in tissue engineering, we and others have begun to develop more sophisticated, quantitative in vitro systems to reveal not only novel structure–function relationships at the ID, but also how intracellular and extracellular factors, such as cytoskeletal architecture or extracellular matrix elasticity, regulate these relationships.

Adherens junctions have been primarily studied in non-cardiac cell types, such as endothelial cells, but many of the lessons learned from other cell types can be applied to the cardiac ID. In cardiac myocytes and other adherent cells, mechanical forces generated by the cytoskeleton are exerted against either the extracellular matrix via cell-matrix adhesions (Schiller & Fassler, 2013) or neighboring cells via cell–cell adhesions (Twiss & de Rooij, 2013). The field of cellular biomechanics has developed tools to quantify the magnitude and distribution of forces generated by a single cell against its extracellular matrix by culturing cells on compliant hydrogels (Butler et al., 2002) or arrays of flexible microposts (Sniadecki & Chen, 2007). By deriving a simple force balance between cell-matrix and cell–cell adhesions, these same techniques have been adapted to quantify cellular forces transmitted to the cell–cell junction in cell pairs (Maruthamuthu et al., 2011; Liu et al., 2010). Results have shown that, similar to focal adhesions, cell–cell junctions are mechanosensitive and grow or shrink in response to increasing or decreasing cytoskeletal tension, respectively (Maruthamuthu et al., 2011; Liu et al., 2010, le Duc et al., 2010). Also similar to focal adhesions, cell–cell adhesions are active sites of signaling, (Weber et al., 2011) indicating that, in addition to mechanically coupling cells together to maintain tissue structure, cell–cell junctions are also sites of dynamic mechanotransduction. Because adherens junctions proteins are also expressed in cardiac myocytes, these results can be extrapolated to suggest that IDs also translate mechanical forces into biochemical signals to regulate processes inside the myocyte.

Because adherens junctions are mechanosensitive, myocyte contractility is likely an important regulator of ID development. The ID co-develops with myofibrils during prenatal and postnatal growth, not reaching full maturity in humans until six years of age (Peters et al., 1994). Early in development, cell–cell junction proteins are found along all cell borders with minimal spatial organization. Over the course of development, cell–cell junction proteins localize to a major extent to longitudinal borders while cell-matrix adhesions form costameres registered with the Z-lines (Hoyt et al., 1989; Hirschy et al., 2006; Peters et al., 1994; Wu et al., 2002; Angst et al., 1997). Because much of this organization occurs during post-natal growth, while myofibrils are also maturing and generating increasing magnitudes of force, cardiac myocytes must be remodeling adhesions while actively contracting. The temporal synchrony of the maturation of the contractile apparatus with the ID indicates that adhesion remodeling is a dynamic process where mechanics likely plays an important role. To better understand the spatiotemporal co-development of ID structures with the cytoskeleton, we cultured pairs of cardiac myocytes on micropatterned islands of fibronectin on compliant polyacrylamide gels with physiological elastic moduli (Figure 1A). With this technique, we could independently modulate tissue geometry and extracelluar matrix elasticity and monitor cytoskeletal maturation, force generation, and cell–cell junction formation over time in culture in a controlled setting (McCain et al., 2012a). At early time points in culture, myocytes were contracting but not well-coupled, as contractile forces were transmitted through focal adhesions to the extracellular matrix near the cell–cell interface. Paired myocytes were also not completely synchronous. Over time, the magnitude of contractile force increased while, concurrently, focal adhesions disassembled at the cell–cell interface and adherens junctions developed, causing forces to be transmitted to the newly formed cell–cell junction instead of the extracellular matrix. These data suggest that there is a hierarchical relationship between focal adhesions and cell–cell junctions, where focal adhesions attach the longitudinal ends of myocytes to the extracellular matrix early in development but eventually disassemble as cell–cell junctions mature and transmit contractile forces across the ID.

Figure 1. Engineering cardiac cell pairs and the intercellular junction. Panel A: Pairs of neonatal rat ventricular myocytes cultured on micropatterned fibronectin islands on polyacrylamide gels were used to study force transmission at the intercalated disc (ID). After one day of culture, myocytes expressed cell–cell junction proteins (i), but were not fully mechanically coupled, as evidenced by traction stresses generated against the extracellular matrix near the cell–cell interface (ii). After four days of culture (iii), myocytes generated insignificant amounts of traction stresses near the cell–cell interface (iv), indicating more complete mechanical coupling at the ID. i, iii: red: actin, blue: nuclei, green: β-catenin. Bars = 10 μm. Panel B: Pairs of neonatal rat ventricular myocytes cultured on micropatterned fibronectin islands were used to study electrical cell–cell coupling at the ID. Dual voltage clamp measurements paired with quantitative immunofluorescence revealed that pairs with lower length to width aspect ratios (i) had lower intercellular conductance values and less Cx43 immunosignal at the cell–cell interface compared to pairs with higher length to width aspect ratios (ii). Volumetry of Cx43 immunosignals was performed from stacks of high-resolution confocal slices after wide-field deconvolution. Panel C: Relationship between Cx43 immunosignal and intercellular conductance. The intersection with the y axis shows a lower limit for immunosignal detection at low levels of intercellular conductance (approximately 10 nS). Modified from references Weber GF, et al. (2011).and Matsushita T, et al. (1999).

In addition to intracellular remodeling, the mechanical properties of the extracellular matrix are also undergoing dynamic changes during cardiac development, which could affect cell–cell coupling. The elastic modulus of prenatal hearts has been measured at approximately 1 kPa and increases to about 10 kPa in the adult heart (Engler et al., 2008). After a myocardial infarction, the scar tissue that remains has an elastic modulus that can exceed 50 kPa (Berry et al., 2006). Diseases such as hypertrophic cardiomyopathy are associated with increased fibrosis, (Ho et al., 2010) which also stiffens the myocardium. In vitro studies have shown that myocyte morphology, sarcomere structure, force generation (Engler et al., 2008; Majkut et al., 2013), and calcium cycling (Jacot et al., 2008) are each affected by the elastic modulus of the extracellular matrix, indicating that cardiac myocytes are highly sensitive to the mechanical properties of their microenvironment. Because of the crosstalk between cell-matrix and cell–cell adhesions, we asked how changing the elastic modulus of the extracellular matrix affected cell–cell adhesion (McCain et al., 2012a). On polyacrylamide gels with elastic moduli mimicking developing and mature hearts, pairs of cardiac myocytes coupled together with minimal focal adhesion attachment at the cell–cell interface, indicating that contractile forces were essentially transmitted completely to the cell–cell junction. However, on stiff polyacrylamide gels with elastic moduli in the range of fibrosis, focal adhesions remained attached to the cell–cell interface, indicating that forces were transmitted both to the extracellular matrix and the cell–cell junction. Thus, on gels with high elastic moduli, myocytes were not as well-coupled, which could lower the efficiency of contractile force transmission across the cell–cell interface and potentially disrupt mechanoelectrical coupling. Our results combined with in vivo observations that cell–cell and cell-matrix adhesions undergo significant remodeling in the infarct border zone (Peters et al., 1997; Matsushita et al., 1999) suggest that altered mechanical properties of the microenvironment could contribute to dysfunctional cell–cell coupling in pathological settings.

Many cardiac pathologies are also associated with distinct changes in myocyte geometry. For example, concentric hypertrophy is characterized by decreased length to width aspect ratio, while eccentric hyper- trophy is characterized by increased length to width aspect (Gerdes, 2002). By combining traction force microscopy with microcontact printing, it has been shown that myocyte shape regulates myofibril alignment (Bray et al., 2008) and force generation in isolated myocytes (Kuo et al., 2012), implying that myocyte shape remodeling impacts contractility in vivo. Using the same cell pair assay described above, we quantified how myocyte shape regulates the amount of force generated against the cell–cell junction (McCain et al., 2012a). Similar to results from single cell studies, the amount of force generated against the cell–cell junction was regulated by myocyte shape, as myocyte pairs with physiological length to width aspect ratios generated more force against the cell–cell junction compared to those with lower aspect ratios. Because cell–cell junctions are mechanosensitive, our study suggests that remodeling of myocyte shape not only affects contractility, but likely also cell–cell communication.

To elucidate the effects of myocyte shape remodeling on electrical cell–cell coupling, we performed dual voltage clamp studies to correlate cell pair geometry to gap junction structure and function (McCain et al., 2012b). By seeding cardiac myocytes on micropatterned islands of fibronectin, we controlled cell dimensions (length, height, and width) and junction geometry, as myocyte pairs consistently formed sigmoidal cell–cell junctions with relatively little variability in junction length or height. By performing quantitative immunostaining after dual voltage clamp experiments (Figure 1B), we identified a positive correlation between intercellular conductance and connexin43 (Cx43) immunosignal (Figure 1C), complementing measurements in engineered multi-cellular tissues showing correlations between Cx43 immunosignal and propagation velocity (Beauchamp et al., 2004). We also found that myocyte pairs with intermediate length to width aspect ratios had the highest intercellular conductance and Cx43 density, likely because this aspect ratio provided optimal contact at the cell–cell interface for gap junction formation. Thus, gap junction density seemed to be mostly a function of cell–cell contact area. Extrapolating our results also suggested that low levels of electrical coupling are present in the absence of measureable Cx43 immunosignal. This suggests that there is a lower detection threshold for gap junction immunostaining, where low levels of Cx43 gap junction channels can be present and functioning (as evidenced by whole cell dual voltage clamp), but are too small or diffuse to be identified with immunostaining.

GENETIC DELETION OF Connexin IN ENGINEERED ATRIAL CELL PAIRS

Atrial cells are connected by Cx43 and connexin40 (Cx40). In addition a low amount of Cx45 can be detected by immunofluorescence (Vozzi et al., 1999; Severs et al., 2008). Both Cx40 and Cx43 make pores of relatively large electrical conductance, if expressed separately in heterologous expression systems (Harris, 2001; Cottrell & Burt, 2005; Cottrell et al., 2002; Desplantez et al., 2011). Electrical propagation velocity in atrial tissue is markedly heterogeneous with a relatively high velocity in the Crista terminalis (forming a “propagation highway” between the sino-atrial node and the atrio-ventricular node) and Bachmann's bundle (forming a main connection between the right and left atria) (Kleber et al., 2003). The remainder of the atria shows a relatively low velocity, lower than ventricular myocardium, despite the presence of Cx43 and Cx40 (Kleber et al., 2003). The role of the individual connexins in atrial propagation, especially the question of Cx40 and Cx43 co-expression, has not been fully elucidated and is a matter of ongoing debate.

A way to resolve the functions of Cx43 and Cx40 in the atria is to measure propagation velocity in presence or absence of Cx43 or Cx40 in mice, in which one of the connexins has been ablated. In whole mice, ablation of Cx40 has shown conflicting results. In some studies, a decrease of atrial conduction time was observed and/or lengthening of the P-wave. This was taken to indicate a simple rule, namely that ablation of a connexin with a high pore conductance would lead to a decrease in cell– to-cell coupling, and accordingly to a decrease in velocity. However, in other studies it was shown that P-wave length did not increase with Cx40 deletion, and propagation in the right atrium became inhomogeneous (see discussion in reference Desplantez et al. 2012). Moreover, atrial malformations were observed with Cx40 haplo- insufficiency, suggesting that macroscopic excitation spread in such hearts might be affected by factors other than the change of connexin expression (Gu et al., 2003).

In an earlier study, we engineered strands from murine atrial tissue using photomicrolithography to show that linear propagation velocity at a cellular level was markedly decreased with Cx43 ablation and significantly increased versus WT with Cx40 ablation (Beauchamp et al., 2006). These findings were in seeming conflict with some of the aforementioned studies in whole mice, but in accordance with studies in human hearts. In fact, studies in human hearts suggest a complex, reciprocal relationship between propagation velocity and connexin Cx43 and Cx40 expression (Kanagaratnam et al., 2002). This complexity might have several components: There could be interaction (1) between the atrial connexins at the level of expression at the ID, (2) at the level of gap junction channels with formation of heteromeric hemichannels (or connexons) and mixed gap junctions channels composed of all three atrial Cx subspecies, and (3) between connexins and expression of other functional complexes, such as ion channels. Thus, in our earlier work, we could show that the area occupied by Cx40 at the ID decreased with Cx43 ablation, whereas the Cx43 immunofluorescence signal increased with Cx40 ablation (Beauchamp et al., 2006). These changes occurred with a constant cellular pool of Cx40 and Cx43 proteins, as demonstrated by whole cell Western blots, suggesting the possibility of interaction at the level of trafficking.

To perform high-resolution volumetric analysis of connexin immunosignals in defined atrial cell–cell junctions and assessment of intercellular conductance by dual voltage clamp, we engineered atrial cell pairs from wild type (WT) and Cx43 knock out (Cx43 KO) atrial myocytes. In contrast to ventricular pairs, the atrial ID is mostly oriented perpendicular to the main cell pair axis (Figure 2A). As a major finding, and in line with our previous study, we could show that germline ablation of Cx43 leads to a decrease of the immunosignals of partner connexins, Cx40 and Cx45 at the IDs (Figure 2B). This result points to an interaction between Cx40, Cx45, and Cx43 at the level of trafficking. This is because whole cell levels of Cx43 and Cx40 proteins remained unchanged in the case of either Cx43 or Cx40 genetic ablation (Beauchamp et al., 2006). A defect of Cx40 and Cx43 trafficking has been shown in experiments involving the co-expression of mutated Cx40 protein (rare somatic mutation found in patients with lone atrial fibrillation) with wild-type Cx43 (Gollob et al., 2006). In this case, the mutated protein prevented the integration of the wild-type partner connexin into the ID, again suggesting mutual interaction of Cx40 and Cx43 at the level of trafficking. While very recently a number of basic papers on connexin trafficking have appeared (Zhang et al., 2014), the possible mechanism underlying Cx43/Cx40 interaction in atria still remains to be elucidated.

Figure 2. Engineering atrial cell pairs. Panel A: Engineered atrial cell pair. Left: Wide-field image of engineered murine atrial cell Cx43^+/+ pair overlaid with the immunofluorescence of N-Cadherin (green) and Cx40 (red). Note the sharply defined interface between the two myocytes (x/z plane). Right: 3D surface rendering of Cx40 (red) immunofluorescence in the y/z plane showing plaque-like arrangement of Cx40. Image deconvolution was applied to confocal signals before reconstruction. Panel B: Relative levels of Cx43, Cx40, and Cx45 immunofluorescence signals in Cx43^+/+, Cx43^+/−, Cx43^−/− fetal mice myocytes (levels in Cx43^+/+ pairs are taken as unity). Panel C: Intercellular electrical conductances in pairs of Cx43^+/+, Cx43^+/−, Cx43^−/− fetal (1 day before birth) mice myocytes. Modified from reference Desplantez T, et al. (2012).

At the level of electrical cell–cell coupling, whole cell dual voltage clamp showed a massive decrease in electrical intercellular conductance by 70% associated with Cx43 ablation (Figure 2C). Analysis of the biophysical properties of gap junction channels was performed by assessing the bell-shaped relationship between the voltage across the gap junction channel, V_j and the relative gap junctional conductance, g_rel = g_j,ss/g_j,inst. With Cx43 ablation, the dependence of g_rel on V_j showed a change in shape typical for Cx40 dominating the gap junction channel properties. Determination of unitary channel conductances, γ_j showed not only a general shift toward higher values, but also an appearance of a bimodal distribution, with a peak at 20–30 pS. Although the explanation for these changes, and especially the appearance of a peak at low conductance values, remains in part speculative, it may indicate emergence of a population of homomeric-homotypic Cx45, (Desplantez et al., 2011; Moreno et al., 1995; Elenes et al., 2001) corresponding to the Cx45 pool, which in wild-type cells forms heteromeric connexons with Cx43. While Cx45 forms heteromers with Cx43 and possibly Cx40, (Cottrell & Burt, 2005; Martinez et al., 2002; Desplantez et al., 2004, 2007) experiments on the question of Cx43/Cx40 interaction (formation of Cx40/Cx43 heteromeric connexons and heterologous gap junction channels) have produced controversial results ([Burt et al., 2001; Valiunas et al., 2001] and discussion in reference [Desplantez et al., 2012)]). While overlap of Cx43 and Cx40 immunosignals in atrium is undisputed, it remains to be shown, whether and how Cx43 and Cx40 combine in individual gap junction channels.

A third point, increasing the complexity of protein regulation at the atrial cell junction even more, is the observation of a significant decrease of INa current with ablation of Cx43 (Desplantez et al., 2012). The decrease of INa current by approximately 50% may indicate that almost the whole ID fraction of Nav1.5 is affected by Cx43 ablation. It has recently been shown that a significant fraction of cardiac Nav1.5 channels is located at the ID, (Petitprez et al., 2011) together with the K+ inwardly rectifying channel Kir2.1 ion channels (Milstein et al., 2012). Interestingly, silencing of plakophyllin-2 (a desmosomal protein) was also associated with a decrease of INa (Sato et al., 2009). In this latter case however, a shift in channel kinetics (shift of activation curve) was mentioned as the underlying cause.

In summary, the observations made in several recent studies on a multitude of interactions among the proteins at the ID forming either desmosomal complexes, ion channels or gap junctions underline the importance of complex regulatory system(s) for normal cardiac electrical (and probably mechanic) function and disease mechanisms. At the level of mechanistic explanations, these experiments show that simple hypotheses and models are not adequate to explain the physiology of the ID, and major questions remain unanswered.

ELECTRICAL CELLULAR INTERACTIONS AND PROPAGATION IN TISSUE WITH HETEROGENEOUS Cx EXPRESSION

Remodeling of gap junctions and intercellular electrical communication has been implicated in arrhythmogenesis in the setting of ventricular and atrial arrhythmias and in aging myocardium. A key and basic question concerning cardiac electrical cell-to-cell transfer relates to the theoretical possibility that, besides the transfer via gap junctions (representing an electrical resistance), a further mechanism could play a role, which would implicate so-called “field-effect” or “ephaptic” transmission, a mechanism which does not necessitate any low resistance connection (gap junction channels) between cardiac cells. This concept and the dispute over these two mechanisms is almost than five decades old (Weidmann, 1966), but it has recently received new attention by (i) unexplained experimental findings, (ii) the observation that ion channels are present in the ID, and (iii) the possibility to design theoretical models of the ID containing all major functional elements (Kucera et al., 2002; Mori et al., 2008). These models can show ephaptic transmission in almost totally uncoupled tissue indeed, however, only within a very narrow setting of critical model parameters. Especially intriguing is the experimental observation that cardio-specific ablation of Cx43, the major connexin forming gap junction channels in the ventricular myocardium may only moderately affect propagation velocity. Thus in artificial strands of neonatal rat ventricular myocytes, heterogeneous germline ablation of Cx43 had only a very minor effect on propagation (Thomas et al., 2003). A further observation was made in mouse hearts with cardio-specific conditional ablation of Cx43. These hearts show a variable and inhomogeneous expression of Cx43, probably as a consequence of a subtotal efficiency of the Cre/flox system. In such hearts, a massive decrease of Cx43 by approximately 90%, as judged from Cx43 immunofluorescence, produced only a moderate decrease of propagation velocity (by about 50%) (Gutstein et al., 2001). In a further mouse model of conditional Cx43 ablation it was shown that Cx43 immunofluorescence decreased with age and became markedly heterogeneous, whereby only a marked decrease was associated with the occurrence of ventricular arrhythmias (Danik et al., 2004).

These observations prompted these authors to recur to the ephaptic impulse transmission hypothesis, which was also tested in a theoretical model (Mori et al., 2008). In order to analyze the role of heterogeneous Cx43 expression at cellular resolution, we developed a model of engineered neonatal mouse ventricular tissue using two different cell populations (Beauchamp et al., 2012). In a first stage we engineered cell pairs from two population of cells, one of wild-type ventricular myocytes carrying a GFP reporter system (Okabe et al., 1997), the other consisting of Cx43^−/− myocytes (further denoted as Cx43KO). Figure 3A demonstrates that the ID between the two WT cells shows, as expected, immunofluorescence signals for Cx43. Importantly, the engineered interface between a WT cell and a Cx43KO cell is negative for either Cx43 immunofluorescence, indicating that gap junction channels are below immunodetection threshold (McCain et al., 2012b). Intercellular electrical conductance between these two types of cells is depicted in Figure 3C. It can be seen, that all three combinations of IDs conduct electrical current indicating the presence of functional gap junction channels in all three cases. However, the ID between the WT–Cx43KO and Cx43KO–Cx43KO cells show a marked decrease in conductance. The bell-shaped current–voltage relationships of the three types of junctions were typical for the respective cell types, which contribute either Cx43/Cx45 or only Cx45 connexon hemichannels to the respective gap junction channels. Thus the I/V relationship of WT–Cx43KO pairs was typically asymmetric indicating the presence of only Cx45 homomers in one half of the channel and Cx43/Cx45 heteromers in the other (Figure 3B).

Figure 3. Engineering cell pairs with different Cx43 expression. Top view of murine ventricular cell pairs depicted from an average of superimposed confocal slices. Green: GFP; blue: DAPI; orange: Cx43. Panel A: Top: pair of Cx43^+/+/Cx45^+/+/GFP^+/− cells. Note two large junctional plaques and smaller but abundant Cx43 fluorescence in the cytoplasmic space. Bottom: Pair composed of a Cx43^+/+/Cx45^+/+/GFP^+/− cell (right) and a Cx43^−/−/Cx45^+/+ cell (left). Only one small spot of Cx43 fluorescence is present at the junction (arrow). Panel B: I/V curve of the heterogeneous junction in pair B showing asymmetry with increased voltage-dependent gating if Cx43^−/− cell is hyperpolarized (typical for homomeric Cx45 connexons). Panel C: Comparison of intercellular conductance among pairs of Cx43^+/+, Cx43^−/+ and Cx43^−/− murine ventricular myocytes. Modified from reference Thomas SP, et al. (2003).

The presence of Cx43 and Cx45 in these channels was also verified by measurements of single-channel conductances. What are the implications of these findings for the mechanism of impulse transfer? First, all the cells including those with full Cx43 ablation were able to transfer the electrical impulse through gap junction channels. Second, electrical transfer is possible in absence of immunofluorescence for Cx43. Third, it can be inferred that the percentage decrease in Cx43 immunofluorescence in a cardiac cellular network with heterogeneous Cx43 expression does not reflect the percentage of WT cells. This is because only the interfaces between WT cells are expected to show Cx43 fluorescence, and there is still substantial electrical coupling at the threshold of Cx43 immunodetection. This predicts that in mixed tissue, depending on the size of the WT and Cx43^−/− cell clusters in the mixture, the percentage of Cx43 immunofluorescence may largely overestimate the degree of Cx43 ablation. The consequences of heterogeneous Cx43 for impulse propagation at a cellular level was verified by multisite high-resolution optical mapping of membrane potential in synthetic strands engineered from mixtures of WT and Cx43^−/− cells. In accordance with the measurements in isolated mouse hearts, propagation velocity decreased to about 50% of normal in mixtures of 50% Cx43^−/− cells. Further decrease will occur only with a higher degree of Cx43 ablation. The preservation of relatively high velocities is due to a “meandering type” of propagation along a fast path of connected WT cells, while Cx43^−/− cells show large local delays in excitation. Overall, these data demonstrate that (1) a very low level of Cx43 immunofluorescence is compatible with relative rapid propagation, and (2) cardiac impulse propagation in normal and pathological states does not necessitate the presence of ephaptic transmission, although co-existence of such a mechanism cannot entirely be ruled out (Beauchamp et al., 2012).

OUTLOOK

New technological developments in fields such as tissue engineering, imaging, biomaterials, and microelectronics continue to enhance and refine our experimental capabilities for understanding cell–cell coupling in cardiac myocytes. In addition to the studies described above, micropatterned cardiac cell pairs have been used as a model system for understanding gap junction trafficking (Zhang et al., 2014) and coupling between cardiac myocytes and other cell types, such as mesenchymal stem cells (Pedrotty et al., 2008). New microdevices have been developed to selectively apply electrical stimulation to only one myocyte in a pair, a tool useful for studying the efficiency of gap junction coupling (Zhang et al., 2013). Instead of culturing pairs of cardiac myocytes on extracellular matrix proteins, another approach has been to culture isolated myocytes directly on substrates coated with cadherin such that myocytes form pseudo-cell–cell junctions directly with the substrate (Ganz et al., 2006; Chopra et al., 2011). This method allows investigators to more directly compare the biophysical properties of cell–cell versus cell-matrix adhesions. Investigators have also moved from two to three dimensions by culturing cardiac myocytes in hydrogels as another model system for understanding the spatiotemporal development of the ID (Zhou et al., 2013). The impact of each of the studies described in this review are dependent on constant feedback between in vivo and in vitro systems. In vivo systems are important for identifying characteristics of physiological and pathological cell–cell junction remodeling in functioning hearts, while in vitro systems are necessary for isolating individual biophysical factors to test their effects in a controlled setting. This synergy we gain with in vivo and in vitro systems enriches our understanding of IDs in health and disease and has potential to contribute to the development of novel therapeutic strategies focused on the cell-cell junction.

Declaration of interest: The authors report no declarations of interest. The authors alone are responsible for the content and writing of the paper.