Reciprocal myocardial-endocardial interactions pattern the delay in atrioventricular junction conduction

Coordinated contraction of the heart requires precise regional patterning of cardiomyocyte conduction velocity. During development, the first cardiac contractions are stimulated by electrical impulses that initiate in the inflow of the primitive heart tube and uniformly propagate towards the outflow (Bressan et al., 2013; Kamino, 1991). This results in a simple peristaltic/suction mechanism for pumping fluids (Forouhar, 2006; Männer et al., 2010). Formation of electrical impulse delay in the atrioventricular junction (AVJ) of the heart represents a key transition in cardiac function. This ensures effective unidirectional blood pumping, as AVJ impulse delay allows for atrial systole to complete prior to the initiation of ventricular contractions. Accordingly, dysfunction in atrial-to-ventricular conduction represents a life-threatening component of several human congenital cardiac syndromes (Clark et al., 2006; Garg et al., 2003; Schott et al., 1998). However, despite more than a century of investigation into the atrioventricular conduction system (Tawara, 1906; Efimov et al., 2004), the mechanisms that restrict slower conduction velocity to this region remain poorly understood.

In parallel with the emergence of conduction delay, paracrine factors from the AVJ myocardium support the formation of the endocardial cushions, which will give rise to the atrioventricular cuspid valves (Butcher and Markwald, 2007; Combs and Yutzey, 2009; Ma et al., 2005; Person et al., 2005; Singh et al., 2012). This process involves swelling of the extracellular matrix (ECM)-rich cardiac jelly within the AVJ. Notably, swelling of the ECM also physically separates the endocardial layer of the heart from the slow-conducting AVJ myocardium, creating a primitive valvular system which prevents retrograde blood flow during the cardiac cycle.

Several endocardial-derived factors have previously been reported to be inducers of fast conduction fate in myocardial cells at later stages of heart development (Gourdie et al., 1998; Hall et al., 2004; Patel and Kos, 2005; Rentschler et al., 2002; Sedmera et al., 2008; Takebayashi-Suzuki et al., 2000). To date, however, it has yet to be examined whether the physical separation of endocardium from myocardium that occurs during cardiac jelly deposition plays a role in AVJ conduction patterning. Current models for how the AVJ becomes slow conducting are focused on the cell-autonomous role of myocardial expressed factors, including Bmp2, Tbx2 and Tbx3 (Christoffels et al., 2004b; Kokubo et al., 2007; Singh et al., 2005,, 2012). It is believed that these factors prevent myocardial differentiation in the AVJ, forcing these cells to retain the slow-conduction characteristics of juvenile myocytes (Christoffels et al., 2004a).

In the present study, the electrophysiological, molecular and morphogenic events that occur concomitant with the heart's transition from a peristaltic pump to a multi-chambered system were examined. Our data demonstrate that the physiological and molecular characteristics of the AVJ myocardium are distinct from juvenile myocytes. Furthermore, both gene expression and impulse propagation can be converted to a fast-conducting phenotype either by implanting AVJ myocytes adjacent to the endocardium of the cushions or through exposure to endocardial-derived paracrine factors. These data indicate that AVJ myocardium must be separated from endocardial-derived factors to become slow conducting, supporting a novel model for how electrical impulse delay is patterned in the AVJ. Accordingly, we propose that AVJ cardiac jelly deposition, which occurs at the initiation of valve formation, alters the local balance of signaling cues supporting slow conduction fate in the AVJ.

To examine how slow conduction is specifically assigned to the AVJ, we mapped changes of physiological and molecular parameters along the entire heart during early chick cardiogenesis from HH stages (st) 10 through 18 (Hamburger and Hamilton, 1951) (Fig. 1). Voltage dye-based optical mapping analysis (Bressan et al., 2013; Efimov, 2004; Kamino, 1991) revealed distinct conduction characteristics between the heart tube primordium at HH st 10, when it began contracting (Patten and Kramer, 1933), and myocardium at heart-looping stages (HH st 18), when the atria and AVJ have become morphologically distinguishable. At linear-tube stages, action potentials initiated in the left posterior inflow segment of the heart and propagated anteriorly towards the outflow with a uniform velocity of 0.23±0.08 cm/s (Fig. 1A-C; supplementary material Movie 1) (Bressan et al., 2013; Kamino, 1991). Conversely, conduction velocity became regionally heterogeneous during heart looping (Fig. 1A-C; supplementary material Movie 1) (de Jong et al., 1992; Bressan et al., 2013). At these stages, action potentials propagated across the presumptive atria at an average rate of 8.2 cm/s, rapidly decelerated to 0.85 cm/s in the AVJ and progressively accelerated to 16.8 cm/s in ventricular myocardium (Fig. 1D).

Both the early heart tube and the looping-stage AVJ displayed conduction velocities slower than the looping-stage atrial and ventricular myocardium. We therefore examined whether the mechanism for AVJ conduction delay was due to this region uniquely maintaining physiological features of primitive heart tube myocardium, as has been previously suggested (Christoffels et al., 2004a). However, detailed analysis of optical mapping data revealed that action potential characteristics of heart tube myocytes and AVJ myocardium differed greatly. AVJ conduction velocity averaged 3.7-fold higher, action potential upstroke velocity was 5.5-fold faster and action potential duration was 2.4-fold shorter in the AVJ compared with the heart tube (Fig. 1D-F). These data indicate that, as heart morphogenesis proceeds, each cardiac region differentiates with specific action potential characteristics and conduction velocities.

To examine the molecular components associated with regionalization of propagation velocity, we profiled the expression of gap junctions and ion channels previously associated with cardiomyocyte conduction velocity. At HH st 10, mRNAs for GJa1, GJa5 and GJa7, which encode for cardiac connexin 43 (Cx43), 40 (Cx40) and 45 (Cx45), respectively (Jansen et al., 2010; Teunissen, 2004), were detected at low levels just above background (Fig. 1G). By HH st 18, however, Cx40, the lowest-resistance cardiac gap junction (Kreuzberg et al., 2006), was expressed robustly in atrial and ventricular myocardium, ∼6.7- and 7.8-fold higher, respectively, than in AVJ myocytes (Fig. 1G,H). Cx43 and Cx45 expression, however, did not show similar increases in expression between HH st 10 and HH st 18. (Fig. 1G). Similar to Cx40, the ion channel Nav1.5 (encoded by Scn5a), which is responsible for the rapid upstroke component of action potentials in fast-conducting myocardium (Remme et al., 2009; Ruan et al., 2009), was differentially expressed during the diversification conduction velocity. Whereas Nav1.5 mRNA was barely detectable in the heart tube (Fig. 1G,I), at looping-heart stages its expression was robust in the fast-conducting atria and ventricles but 13.8- and 8.0-fold lower, respectively, in the slow-conducting AVJ. These data suggest that, as regionalization of conduction velocity is patterned during early heart morphogenesis, molecular markers of faster conduction fate, including Cx40 and Nav1.5, become enriched in the atria and ventricles but remain low in the AVJ.

Several paracrine factors have been identified as regulators of developmental conduction fate. These include endothelin-1 (Et1) and neuregulin 1 (Nrg1), both of which have been implicated as potential inducers of fast conduction (Gourdie et al., 1998; Hall et al., 2004; Milan et al., 2006; Patel and Kos, 2005; Rentschler et al., 2002; Sedmera et al., 2008; Takebayashi-Suzuki et al., 2000). Conversely, Bmp2 is highly expressed by the AVJ and is thought to be a negative regulator of Cx40 and Nav1.5 expression through its downstream targets Tbx2 and Tbx3 (Christoffels et al., 2004b; Hoogaars et al., 2004; Singh et al., 2012; Yamada et al., 2000; Boukens and Christoffels, 2012; van den Boogaard et al., 2012).

To test whether these factors could influence conduction patterning in the AVJ, looping-stage hearts were isolated and treated with the signaling factors mentioned above. Initially, hearts were treated for 30 min to confirm the activation of downstream signaling effectors in the myocardium (supplementary material Fig. S1). Following this, hearts were treated for 16 h and conduction velocity was analyzed (Fig. 2). Vegf treatment was used as a control for general receptor tyrosine kinase activation. Et1-treated hearts displayed the most prominent increase in AVJ conduction velocity following treatment, with five out of eight hearts displaying a decreased delay between atrial and ventricular activation (Fig. 2A-C; supplementary material Movie 2). Six out of 11 Nrg1-treated hearts exhibited abnormal activation patterns including intermittent AV block (2/11) and ectopic initiation sites (4/11), but AVJ conduction velocity did not differ from control hearts (Fig. 2C; supplementary material Movie 3). Surprisingly, Bmp2 treatment did not result in significant changes in conduction (Fig. 2A-C; supplementary material Movie 4). Similarly, the Bmp signaling antagonist LDN-193189 (Boergermann et al., 2010) did not influence propagation pattern or eliminate AVJ conduction delay (Fig. 2A-C; supplementary material Movie 4). Consistent with the above findings, upregulation of Cx40 mRNA in the AVJ was detected following Et1 treatment, but was not evident in other experimental groups (Fig. 2D,E). Interestingly, increased Cx40 expression was not coupled to increased Nav1.5 expression in the AVJ following Et1 treatment (Fig. 2D,E). General myocardial patterning was also preserved following Et1 treatment, as noted by examination of mRNA for Nkx2.5, ventricular myosin heavy chain 1 (Vmhc1) and atrial myosin heavy chain 1 (Amhc1) (supplementary material Fig. S2) (Yutzey et al., 1994). These data demonstrate that the AVJ myocardium is capable of adopting a faster-conducting phenotype and can upregulate fast conduction markers such as Cx40 when provided with appropriate paracrine cues.

The data shown above indicate that slow conduction is not a developmentally fixed feature of the AVJ and that this region remains capable of responding to inductive cues leading to increased conduction velocity. The ability of Et1 to decrease AVJ conduction delay suggested that lack of availability of endogenous inductive signals might be a contributing factor to how the AVJ escapes fast-conducting fate. To test this model, we examined the expression of endothelin signaling components in looping-stage hearts (HH st 17) (Fig. 3). Whereas endothelin receptor B (ETbR) was expressed predominately in the atria and ventricles, endothelin receptor A (ETaR) was expressed ubiquitously throughout the myocardium during heart looping (Fig. 3A) (Yanagisawa et al., 1998). Conversely, mRNA for preproEt1, the inactive precursor of Et1, and endothelin-converting enzyme-1 (Ece-1) (Masaki, 2004; Yanagisawa et al., 1998), which cleaves the inactive peptide of Et1 to produce the mature Et1 ligand, were absent from the myocardium and restricted to the endocardium in the heart (Fig. 3A,D). Importantly, the swelling of the cardiac jelly within the atrioventricular canal at this stage (Butcher and Markwald, 2007; Combs and Yutzey, 2009; Person et al., 2005) physically separated the endocardium from the myocardium along the AVJ (Fig. 3A,D). This led to the question of whether decreasing the distance between the myocardium and endocardium would be sufficient to increase AVJ conduction velocity.

To test this, the cardiac jelly, a hyaluronic acid-rich ECM between the endocardium and myocardium of the heart, was enzymatically digested in vivo. Hyaluronidase (Baldwin and Solursh, 1989; Baldwin et al., 1994; Solursh et al., 1979) was injected directly into the pericardial space of HH st 12 embryos, and these embryos were allowed to develop until HH st 17-18. This broad degradation of hyaluronic acid resulted in morphological defects consistent with those reported in the hearts of hyaluronic synthase 2-knockout embryos (Camenisch et al., 2000). Specifically, treated hearts were enlarged when compared with control embryos (injected with 0.9% saline) (Fig. 4) and lacked uni-directional fluid movement. Alcian Blue staining confirmed that a hyaluronic acid-rich cardiac jelly was absent between the endocardium and myocardium in treated embryos (Fig. 4B,E). Importantly, despite the severe morphological and functional defects in treated hearts, conduction through the AVJ was significantly increased (Fig. 4C,F,G). In control embryos, ventricular activation occurred 82.9±8.3 ms (n=10) after the atria. Conversely, treated embryos displayed a delay of only 51.1±17.7 ms (n=11) (Fig. 4G). The pre-excitation of the ventricle was caused by rapid transmission of action potentials through the dorsal/inferior component of the AVJ. In addition, hyaluronidase-treated embryos displayed ectopic expression of Cx40 (Fig. 4) and downregulation of Bmp2 along the dorsal/inferior AVJ, extending from the ventricle to the level at which the AVJ was in contact with the myocardium of the cardiac inflow. These data demonstrate that functional and molecular characteristics of fast-conducting myocardium were induced in the AVJ following ablation of the cardiac jelly. However, as general cardiac morphology in these hearts was perturbed, the specific ability of the endocardium to influence AVJ conduction fate could not be evaluated.

Therefore, to test the ability of the endocardium to influence AVJ fate, we implanted AVJ myocardial cells from HH st 15-16 donor hearts into various regions within the superior endocardial cushion of stage-matched host embryos (Fig. 5). Following an additional 14 h of incubation the locations and the fates of the implanted cells were examined. Importantly, the overall function and morphology of these hearts was preserved following manipulation. Implanted cells were contracting in the cardiac jelly of host embryos (supplementary material Movie 5), suggesting that these cells were viable, electrically coupled to one another, but electrically isolated from the host heart. The implanted cells expressed Bmp2, Tbx2 and Tbx3, at levels comparable to the host AVJ myocytes, thereby also suggesting that the bulk of the implanted tissue did not acquire atrial or ventricular fates following transplantation into the cardiac jelly (Fig. 5G-I). By contrast, the expression of Cx40 in sister sections showed dramatic changes dependent on the location of implantation (Fig. 5J,K).

Implants that were positioned far from the endocardium expressed Cx40 at levels similar to the host AVJ myocardium (Fig. 6A,D). Conversely, Cx40 was upregulated in implants that were near to the host endocardium (Fig. 6B,C,E,F). This trend was further evaluated by quantifying staining intensity in implanted AVJ myocytes as a function of their distance from the endocardium. Cx40 staining intensity was highest adjacent to the endocardium and tapered as the distance from the endocardium increased (Fig. 6G, n=12). Importantly, this trend was not seen for Bmp2 staining, which was relatively uniform regardless of distance form the endocardium (Fig. 6H, n=7). These data demonstrate that AVJ myocytes can be induced to express markers of fast conduction in vivo and that proximity to the endocardium is a crucial regulator of conduction state.

The above data alone, however, cannot determine whether Cx40 induction in this system is due to a paracrine interaction with the endocardium, as mechanical deformation of the cardiac jelly during the contraction of the heart could result in stretch-induced activation of Cx40 (Buschmann et al., 2010) in implanted cells. Therefore, we isolated AVJ myocardial cells as described above and implanted them into mesenchymal populations outside the heart that would not be subjected to the same level of biophysical stretch as in the heart (Fig. 7). Cells were implanted either into the cranial mesenchyme between the first pharyngeal arch and the midbrain, a region of high endothelin signaling (Clouthier et al., 1998), or the mesenchyme of the limb buds, a region in which high endothelin signaling has not been described. In the limb bud mesenchyme, due to high native expression of Bmp2, the cardiac transcription factor Nkx2.5 was used to identify the site of injected cells in sister sections (Fig. 7K,L). In none of our limb mesenchyme implantations (n=0/4) did we observe induction of Cx40 mRNA (Fig. 7M). Conversely, within the cranial mesenchyme, robust Cx40 expression was activated, whereas Bmp2, Tbx2 and Tbx3 expression were again maintained (Fig. 7F-J, n=5/5). These data confirm that, given the correct inductive environment, Cx40 expression can be stimulated via stretch-independent paracrine signals. As our above data suggest that proximity to the endocardium is a crucial determinant of whether Cx40 is expressed, these results demonstrate that proper patterning of the AVJ myocardium requires that these cells be separated from inductive stimuli.

During heart development, cardiac muscle adopts regional differences in physiological features, including action potential conduction velocity. The upstream events that pattern slow conduction in the AVJ have remained poorly understood. Our data are consistent with a model in which AVJ myocytes are competent to acquire physiological characteristics and a molecular profile indicative of fast-conducting regions of the heart. Furthermore, our data indicate that increased physical distance between AVJ myocytes and endocardial cells is a crucial regulator of conduction velocity, preventing the activation of molecular determinants of faster conduction in AVJ myocytes. As this increased distance is the result of localized swelling of the cardiac jelly at the initiation of valve formation, these findings demonstrate that slowing of conduction velocity between the contractile chambers of the heart and prevention of retrograde blood movement, two crucial processes required for optimization of unidirectional flow, are directly linked during heart development.

When added exogenously to cultured hearts, at least one endocardial-derived factor, Et1, is sufficient to induce faster conduction characteristics in the AVJ. Et1 is thought to be a short-range signaling cue that requires local processing to activate its cognate receptors (Takebayashi-Suzuki et al., 2000; Yanagisawa et al., 1988; Yanagisawa et al., 1998; Yanagisawa, 1994). Consistent with previous studies, our data demonstrate that the transcript for the inactive precursor of Et1, ppEt1, is expressed by endocardial cells throughout the heart, as is the membrane-bound metalloproteinase which cleaves and activates Et1, Ece1 (Masaki, 2004). Additionally, the endothelin receptor, ETaR, is expressed throughout the myocardium during heart looping (Yanagisawa et al., 1998). These qualities make Et1 an attractive candidate factor for the induction of fast conduction phenotypes seen following enzymatic ablation of the cardiac jelly or implantation of AVJ myocytes near the endocardium. However, the current study cannot conclude which factor, or collection of factors, caused AVJ phenotypic conversion in our in vivo studies. The identity of the factors responsible and the exact molecular mechanism required for inducing faster conduction characteristics in the AVJ will require further study.

The process of cushion morphogenesis initially depends on the production and secretion of a hyaluronic acid-rich ECM, termed the cardiac jelly, which separates the endocardium from the myocardium along the AVJ. Following this, endocardial cells undergo EMT, invade the jelly and begin the remodeling process that give rise to the valve leaflets. Numerous signaling cascades are involved in these processes (Butcher and Markwald, 2007; Combs and Yutzey, 2009; Craig et al., 2010; Frieden et al., 2010; Garg et al., 2005; Hurlstone et al., 2003; Mercado-Pimentel et al., 2007; Person et al., 2005), with myocardial-expressed factors required for the expression of hyaluronic acid synthase 2, which directly modulates cardiac jelly expansion (Camenisch et al., 2000; Ma et al., 2005; Lagendijk et al., 2011; Shirai et al., 2009; Singh et al., 2012). Our work, therefore, suggests that conduction patterning in the AVJ is the result of a reciprocal myocardial-endocardial interaction, whereby myocardial-derived factors increase the space between the endocardium and myocardium, thus preventing signal reception of endocardial-derived factors in the AVJ myocardium.

This model would predict that modulation of the composition and size of the cardiac jelly would be a crucial component of how AVJ conduction fate is patterned. Indeed, enzymatic ablation of the cardiac jelly results in significant changes in AVJ conduction phenotype, leading to faster propagation through the AVJ. It should be noted, however, that general morphology in the heart was altered dramatically following hyaluronidase treatment, making it difficult to determine the precise interactions responsible for the changes in AVJ conduction.

Therefore, AVJ myocytes were implanted directly into the cardiac jelly of host hearts under conditions in which general cardiac morphology and function were preserved. In these studies, upregulation of the fast conduction marker Cx40 was observed only in implants that were in close proximity to the endocardium, thus demonstrating that local interaction with the endocardium was sufficient to alter AVJ fate. It should be noted, however, that these findings do not indicate that distance from the endocardium alone is sufficient to prevent fast conduction, as additional positive cues, including those related to biophysical force and hemodynamics, can induce increased conduction velocity and expression of conduction markers in both myocardial and endothelial cells (Buschmann et al., 2010; Zhuang et al., 2000).

Previous studies have demonstrated that intrinsic AVJ characteristics establish slow conduction fate in this region. For instance, it is well-documented that knockout of Bmp2, Tbx2 and Tbx3 result in ectopic expression of Cx40 in the AVJ myocardium (Ma et al., 2005; Singh et al., 2012). This is also true when Bmp2 and Tbx2 are lost as a result of either Tbx20 deletion or overexpression of Hesr1 and Hesr2 in the myocardium (Kokubo et al., 2007; Singh et al., 2005). Based on in vitro studies, ectopic Cx40 in these knockout models has been attributed to the ability of Tbx2 and Tbx3 to bind and suppress the activity of the Cx40 promoter. Our data, however, demonstrate that when challenged with positive cues, Cx40 induction can occur in the AVJ myocardium. It should be noted that neither overexpression nor knockdown techniques were utilized in these studies; therefore, observed responses in our in vivo implantation studies were the result of exposure to factors at native physiological levels. As such, this suggests that AVJ myocardium must be separated from positive inductive cues in order to establish a slow-conduction fate. Importantly, this does not preclude the involvement of the Bmp2-Tbx2/3 axis in the patterning of the AVJ cardiac conduction system, but does suggest that additional morphogenetic events are required. Proper patterning of the mature atrioventricular conduction apparatus, including the atrioventricular node and His bundle, is believed to be dependent on a host of transcriptional regulators, including Nkx2.5, Tbx5, Id2 and Gata4 as well as Tbx2 and Tbx3 (Kasahara et al., 2001; Moskowitz et al., 2007; Munshi et al., 2009; Schott et al., 1998). It remains to be determined whether cushion/valve morphogenesis contributes to the later stages of conduction velocity patterning in these structures.

In conclusion, we demonstrate that, given the correct inductive environment, the AVJ myocardium can adopt faster conduction characteristics, and that separation from those inductive cues is required for the AVJ to become slow conducting. This separation is the result of myocardial-derived factors that initiate valve formation. Therefore, the regional patterning of slow-conduction velocity, which is required for the heart to transition from a peristaltic or impedance pump to a multi-chambered system, is not due to a lack of capacity of the AVJ to become fast conducting. Instead, developing myocytes remain plastic, and morphogenetic events separate fast-conducting inductive cues from the myocardium, thereby restricting conduction delay to the AVJ.

Fertilized White Leghorn chicken eggs (Gallusgallus) were obtained from Petaluma Farms (Petaluma, CA, USA) and incubated at 38°C until the desired stage. All experiments performed conformed to the University of California animal welfare standards.

Procedures for optical mapping have been described previously (Bressan et al., 2013). For stripe analysis, regions of interest corresponding to 1 pixel (20 μm) by 60 pixels (1200 μm) were drawn along the inner curvature, middle and outer curvature of hearts. Conduction velocities were calculated as change in distance over change in time along the middle (red) line in each heart. Action potential upstroke velocity was calculated as change in fluorescence (ΔF) over change in time (ΔT) for the period from action potential take off to peak intensity (V_max). Action potential duration was calculated as time in milliseconds between 30% V_max and 70% repolarization.

For embryonic heart culture, HH st 17 hearts, including inflow, atria, AVJ, ventricle and three-quarters of the outflow tract, were isolated and transferred to DMEM/F12 media supplemented with 1% penicillin/streptomycin. Hearts were cultured for 30 min for immunohistochemistry or 16 h for physiological and expression analysis at 37°C and 5% CO₂. Et1 (15 μM; Sigma-Aldrich), 50 nM human-recombinant Nrg1 (R&D Systems), 25 nM human-recombinant Bmp2 (R&D Systems), 1 μM LDN-193189 (Stemgent) and 2.5 μM human-recombinant Vegf (Sigma-Aldrich) were added to culture media. These concentrations were chosen based on their ability to elicit activation of downstream signaling in whole-heart cultures (supplementary material Fig. S1). pERk1/2 was used to confirm the activity of Et1 and Nrg1 (Baliga et al., 1999; Clerk et al., 1994), and pSmad1/5/8 was used to confirm the activity of Bmp2 and LDN-193189 (Massague et al., 2005; Boergermann et al., 2010) Endothelin receptor antagonists BQ123 (Sigma-Aldrich) and BQ788 (Sigma-Aldrich) were used at a final concentration of 1 μM.

Hyaluronidase from Streptomyces hyalurolyticus (Sigma-Aldrich) was dissolved in a 0.9% saline solution. Between 1 and 2 μl of hyaluronidase at a concentration of 200 U/ml was microinjected into the pericardial space of HH st 12 embryos. Eggs were then sealed and re-incubated for 20 h. During optical mapping of hyaluronidase-treated embryos it was noted that prolonged exposure (longer than 10 min) to the staining solution (Bressan et al., 2013) led to increases in action potential duration and decreased action potential upstroke velocity. Therefore, all recordings analyzed were performed within 5 min of staining.

Twenty donor AVJ myocardium explants (HH st 17) were microdissected and pooled, dissociated in 0.05% trypsin for 10 min at 37°C, washed in trypsin-neutralizing solution and then spun down for 8 min at 3000 rpm. Cells were then resuspended in cell tracker 5 μM (Invitrogen) in sterile Tyrode's solution and labeled for 20 min at 37°C. Cells were then spun down, washed in Tyrode's solution and resuspended in 5 μl of Tyrode's solution. Cells were then backloaded into a silicon-coated glass micropipette (Sigmacote). For injections into the endocardial cushion and cranial mesenchyme host embryos were incubated to HH st 15-16, windowed and cells were pressure-injected into the targeted regions (see Figs 5 and 7). For injections into the forelimbs, host embryos were incubated until HH st 18-19 and cells were injected into the right limb bud. Eggs were then sealed and re-incubated for an additional 14 h. The implanted cells were photo-documented; embryos were then fixed in 4% paraformaldehyde for 45 min and processed for cryosectioning or paraffin embedding.

Whole-mount in situ hybridization (WISH) has been described previously (Bressan et al., 2013, Ishii et al., 2010). Probes for Cx40, Nkx2.5, AMHC1, VMHC1, Bmp2, Bmp10, Tbx2, ppEt1, Ece1 and ETbR have previously been described (Bressan et al., 2013; Hall et al., 2004; Ishii et al., 2010; Takebayashi-Suzuki et al., 2000). Chick Tbx3, ETaR and Scn5a (Nav1.5) were sub-cloned into the pCR II vector (Invitrogen), and Dig-labeled RNA probes were generated.

Where A₁ and A₂ are the areas of two regions within the implant, and X₁ and X₂ are the distance from the center of these regions to the edge of a line bisecting areas 1 and 2.

Tissue was excised from the heart tube or from looping-stage atria, AVJ or ventricles (see supplementary material Fig. S2B). RNA was isolated using RNAeasy kits (Qiagen), and cDNA was synthesized using iScript Select (Bio-Rad) with oligo(dT)s, according to the manufacturer's instructions. Real-time quantitative PCR (qPCR) was performed on a Step-One Plus thermocycler (Applied Biosystems). For qPCR analysis presented in Fig. 1G, in which expression was compared across different stages of development, a combination of Gapdh, α-Tubulin, γ-Tubulin and NKX2.5 were used for normalization with Step One Software (Applied Biosystems v2.1). All other qPCR data comparing expression within the same stage used Gapdh for normalization.

We would like to thank Dr Lisa Hua for her help and technical expertise and Dr Sara Venters for her careful reading of the manuscript. A part of this work was included in P.B.Y.'s master's thesis.