An essential role for connexin43 gap junctions in mouse coronary artery development

Gap junctions are intercellular junctions containing membrane channels that mediate the movement of ions, metabolites and secondary messengers below 1.2 kDa. They are formed by a protein family of at least 16 members termed connexins (Bruzzone et al., 1996; Condorelli et al., 1998; Itahana et al., 1998; Manthey et al., 1999). Given these properties, gap junctions are thought to have a role in maintaining tissue homeostasis, and regulating cell proliferation and differentiation, and embryonic development (Goodenough et al., 1996; Kumar and Gilula, 1996; Lo, 1999; Newman, 1985). Mutations in connexin genes have been associated with a variety of human diseases, including heart anomalies, cataract, deafness, skin disorders and X-linked Charcot-Marie-Tooth disease (Britz-Cunningham et al., 1995; Kelsell et al., 1997; Richard et al., 1998; Shiels et al., 1998; Xia et al., 1998). The essential role of gap junctions also has been revealed by the analysis of connexin knockout mice (reviewed by White and Paul, 1999; Lo, 1999). Furthermore, connexin knock-in mouse models demonstrated that the phenotypes resulting from eliminating one connexin gene cannot be fully corrected by the replacement of another connexin gene (Plum et al., 2000). This suggests that each connexin isotype may have unique function that is cell- or tissue type-specific.

The connexin gene known as connexin43 or α₁ connexin (referred to as Cx43α1; Gja1 – Mouse Genome Informatics) plays a crucial role in cardiac development, as Cx43α1 knockout mice die shortly after birth due to pulmonary outflow obstruction (Reaume et al., 1995). Typically, the Cx43α1 knockout mouse heart exhibits two prominent pouches at the base of the pulmonary outflow tract, a region known as the infundibulum of the heart. Studies using transgenic mouse models to manipulate Cx43α1 function indicate that the outflow obstruction probably involves the perturbation of cardiac neural crest cell migration. Thus, when Cx43α1-mediated gap junction communication in cardiac neural crest cells was up- or downregulated in the CMV43 (overexpression) and FC (dominant negative) transgenic mice, respectively, right ventricular cardiac defects associated with outflow obstruction were observed (Ewart et al., 1997; Sullivan et al., 1998). In the CMV43 transgenic mice, an increase in gap junction communication was associated with an increase in the rate of neural crest cell migration and an elevation in the abundance of neural crest cells in the heart outflow tract (Huang et al., 1998a). By contrast, in the FC transgenic and Cx43α1 knockout mice, a reduction in coupling was associated with a reduction in cell migration rate and a lower abundance of neural crest cells in the heart (Huang et al., 1998a). Although the FC and CMV43 transgenic mice both exhibited defects involving the heart outflow tract, the same region affected in the Cx43α1 knockout mice, neither animal model exhibited the pouches typically seen in Cx43α1 knockout mouse heart (Ewart et al., 1997; Huang et al., 1998b; Sullivan et al., 1998). This was surprising, as the cardiac neural crest cells from the FC transgenic mice showed similar reductions in gap junction coupling and cell locomotion as that found in the Cx43α1 knockout mice (Huang et al., 1999a; Xu et al., 2002). Our present study suggests that this discrepancy may reflect the fact that formation of the pouches involves not only the cardiac neural crest cells, but also another extra-cardiac migratory cell population, the proepicardial cells.

The proepicardial cells originate from the proepicardial organ (PEO), a grape-like cluster of cells derived from the hepatic primordium and found at E9.0-E9.5 at the ventral region of the sinus venosus (Viragh and Challice, 1981; Viragh et al., 1993). These cells rapidly migrate over the primitive tubular heart to form the epicardium, and subsequently, cells from the epicardium undergo an epithelial-mesenchymal transformation and infiltrate the heart, giving rise to the vascular smooth muscle cells of the coronary arteries, and also fibroblasts in the heart (Dettman et al., 1998; Mikawa and Gourdie, 1996; Vrancken-Peeters et al., 1997; Vrancken-Peeters et al., 1999). Although proepicardially derived cells also have been suggested to contribute endothelial cells in the heart, this possibility remains unresolved (Poelman et al., 1993; Mikawa and Gourdie, 1996). When the epicardium is not maintained or does not form, such as in the α4 integrin and VCAM-1 knoockout mice, coronary arteries fail to develop (Kwee et al., 1995; Yang et al., 1996). The phenotype of these knockout mice strongly suggests that cell-cell adhesion and cell-matrix interactions play an important role in the invasion and subsequent deployment of the epicardial cells.

Our present study suggests that cell-cell interactions mediated by Cx43α1 gap junctions also play an important role in the deployment of the proepicardial cells. The cardiac defects in the Cx43α1 knockout mice are more subtle than that seen in the α4 integrin or VCAM-1 knockout mice, as formation of the epicardium does not appear to be affected by the loss of Cx43α1 function. In the Cx43α1 knockout mouse heart, coronary arteries form but are abnormally patterned. Histological analysis suggested that the coronary artery anomalies and the infundibular pouches involve the abnormal deployment of the proepicardially derived vascular smooth muscle cells. Analysis of the proepicardially derived cells revealed that they are functionally well coupled, and they also have an abundance of Cx43α1 gap junction contacts. In the Cx43α1-deficient proepicardial cells, coupling is greatly reduced, but in contrast to Cx43α1-deficient neural crest cells, this was associated with an elevation in the rate of cell locomotion and cell proliferation. These and other findings suggest that Cx43α1 gap junctions play an essential role in normal coronary artery patterning. We propose that this may involve a dual role for Cx43α1 gap junctions in modulating the deployment of the proepicardial and neural crest cells. We further discuss the possible involvement of Cx43α1 in human congenital cardiovascular anomalies and cardiovascular disease.

The Cx43α1 knockout mice are in a mixed B6/129 background, and offspring obtained from one generation outcross to CD1 were used for experiments. The FC transgenic mice, maintained in a B6/SJL background, were intercrossed with CD1 to obtain hemizygous transgenic and nontransgenic embryos for experiments. Genotyping was carried out by PCR analysis of yolk sac or tail DNA. For genotyping the Cx43α1 knockout mice, primers used were to the wild-type Cx43α1 allele and/or the neo insert in the Cx43α1 knockout allele (Huang et al., 1998a; Reaume et al., 1995). For the FC transgenic mice, the primers used were for the lacZ region of the Cx43α/lacZ fusion protein construct (Sullivan et al., 1998).

Hearts from mouse fetuses and neonates were fixed in 4% paraformaldehyde (PF) overnight, and after dehydration in a graded series of alcohol, they were paraffin embedded, cut into serial sections, de-paraffinized and subjected to immunostaining as previously described (Waldo et al., 1999; Epstein et al., 2000). Smooth muscle cells were identified using antibodies to a smooth muscle cytoskeletal protein, SM22α (Zhang et al., 2001) or smooth muscle myosin heavy chain (Biomedical Technologies, Stoughton, MA). Anti-proliferating cell nuclear antigen (PCNA) antibody (Zymed Laboratories, San Francisco, CA) were used to examine cell proliferation, and to examine apoptosis, a TUNEL assay was carried out using the APOPTAG kit from Intergen Company (Purchase, NY).

For immunohistochemical analysis of proepicardial organ (PEO) explant cultures, fixation and subsequent immunostaining were carried out as previously described (Li and Nagy, 2000). For whole-mount Cx43α1 immunostaining, E9.5 wild-type mouse embryos were incubated with primary and secondary antibodies for 48 hours and 20 hours, respectively. After washing, the embryos were dehydrated, plastic embedded, cut into 5 μm sections and mounted on slides using a Vectorshield mounting medium (Vector Laboratory, Inc.).

PEO explants were obtained from E9.0-E9.5 mouse embryos, with E0.5 designated as the morning a vaginal plug was found. Only the top part of the protruding PEO cluster was removed, in order to minimize contamination from hepatic primordium. The explants were plated on glass coverslips coated with Type I rat tail collagen and maintained at 37°C in Dulbecco’s modified essential medium (high glucose) supplemented with 10% fetal bovine serum, unless otherwise noted. To obtain epicardial explants, E11.5 mouse hearts were similarly plated onto collagen coated coverslips. A polyclonal anti-cytokeratin antibody (Dako) was used to examine the identity and purity of cultured PEO explants. For the study of proepicardial cell differentiation, a cocktail of two monoclonal antibodies against smooth muscle calponin was used (clone CP93 and hCP, Sigma).

To monitor dye coupling, coverslips containing 24 hours PEO explants were transferred into phosphate buffered L-15 medium (Sigma) supplemented with 10% fetal bovine serum, and maintained at 37°C on a heated stage of a Leitz Diavert microscope equipped for epifluorescent illumination. Single proepicardial cells were impaled with a glass microelectrode filled with 5% carboxyfluorescein and iontophoretic dye injection was carried out for 2 minutes using 1 nA hyperpolarizing current pulses of 0.5 second duration at 1 Hz. The number of dye-filled cells at the end of 2 minutes of iontophoresis was recorded. To analyze proepicardial cell motility, images of the PEO explant cultures were recorded at 5 minute intervals for 24 hours. Time lapse movies generated in this manner were used to trace the migration paths of individual proepicardial cells, and from which the speed and directionality of cell movement were determined using the Dynamic Imaging Analysis Software (Solltech, Oakdale, IA).

To examine the proliferation rate of proepicardial cells, 24 hours explant cultures were incubated with 10 μM BrdU for 1 hour, fixed in 4% PF for 30 minutes and immunostained with an anti-BrdU antibody (Calbiochem, San Diego, CA) and further counterstained using Hematoxylin. Images of the stained PEO explant cultures were captured digitally, and the labeled nuclei were quantitated using the Openlab software (Improvision, Coventry, UK).

Examination of pouches in the infundibulum of the fetal and newborn Cx43α1 knockout mouse heart by histology revealed an unusual subendocardial layer of α-cardiac myosin free tissue lining the pouches (Fig. 1A). In some regions, distal coronary artery branches were directly continuous with the pouches (Fig. 1F-H). Interestingly, the tunica media of the coronary artery, which comprises vascular smooth muscle cells, continued seamlessly into the tissue lining the pouch. Further examination of the coronary arteries revealed only a single major coronary artery (either right or left) in some knockout mouse hearts. In a few animals, we observed an anomalous origin of the coronary artery, i.e. where one of the two major coronary arteries arose from the pulmonary rather then aortic trunk (Fig. 1C-E). In some cases, the stem of the coronary artery was continuous with a thin-walled sinus that fed into a complex of intertrabecular sinusoids (Fig. 1B). As the infundibulum is also the region of the heart where the major coronary arteries arise from the aorta, these observations suggest the possibility that formation of the pouches may be related to defects in the deployment of the coronary arteries.

To examine coronary artery anomalies further in the Cx43α1 knockout mice, we carried out a more detailed analysis of coronary arteries and their pattern of deployment in the E16.5 and E17.5 mouse heart. For these studies, vascular smooth muscle myosin (VSMM) was used as a marker to label the coronary arteries. Unexpectedly, we found a reduction in VSMM expression in the coronary arteries of the Cx43α1 knockout mice (Fig. 2). This can be seen in the left and right branches of the coronary arteries (Fig. 2A-D), as well as in the septal and mural coronary arteries in the left and right ventricular wall (Fig. 2E-J). In addition, we noted discontinuities in the VSMM staining, suggesting that vascular smooth muscle cells may not completely line the wall of the coronary arteries in the Cx43α1 knockout mouse heart (for example, compare Fig. 2C with Fig. 2D and Fig. 2E with Fig. 2F). Also striking was the finding of a significant narrowing of the mouth and stem of the coronary arteries as they exited the aorta (compare double headed arrows in Fig. 2B with those in 2A), and a general reduction in the size of the coronary arteries (Fig. 2). It was not possible to analyze this quantitatively, given the combination of the very small size of the coronary arteries and the difficulty in achieving identical alignment of the sectioning planes of the wild-type and knockout mouse hearts. The small size of the coronary arteries also excluded the use of a casting method for three-dimensional reconstructions.

Nevertheless, a systematic analysis of serial sections of wild-type, and heterozygous and homozygous Cx43α1 knockout mouse hearts revealed a variety of coronary artery patterning defects. In wild-type mice, the right and left coronary arteries exited the aorta from the right and left aortic sinuses respectively (R, L in Fig. 3), while no coronary artery arises from the third aortic (non-coronary) sinus (N in Fig. 3). Each coronary artery stem divides into major branches that feed the ventricular septum (S in Fig. 3), the free walls of the heart (M in Fig. 3), the base of the heart (C in Fig. 3) and the atria. In the heterozygous and homozygous knockout hearts, there were numerous alterations in the pattern of coronary artery development. For example, in addition to a main coronary artery, there were often small accessory coronary arteries either exiting from the non-coronary aortic sinus (8065, 8464, 8462 and 8463 in Fig. 3), or the right or left coronary sinus (8065 in Fig. 3). The right coronary artery sometimes tunneled some distance caudally through the wall of the aorta before exiting (8048, 8462, 8463 and 8065 in Fig. 3). As described above, occasionally a stem would be missing and only one coronary artery arose from the aorta (7949 and 8489 in Fig. 3). The ostia or mouths of the coronary artery were often very small (8465 in Fig. 3), and in some cases, the main branches of a coronary artery exited the aorta separately rather than branch from a main stem (8065 in Fig. 3), or the stem was so short that it divided in the wall (8064 in Fig. 3). In several hearts, a branch of the coronary artery became sinusoidal with very thin walls (8489 and 8488 in Fig. 3). Surprisingly, we found the same set of coronary arterial anomalies in the heterozygous and homozygous knockout mouse hearts (Fig. 3). The data summarized in Fig. 3 represents 85% of the hearts analyzed and are representative of the defects typically observed (except for the heart with anomalous origin of the coronary artery, which is illustrated only in Fig. 1B). Given these observations, it is interesting to note that although the infundibular pouches are usually seen only in homozygous knockout mice, pouches were also observed in a few heterozygous knockout mouse hearts (Fig. 4H).

To further determine the nature of tissue lining the pouches and its possible relationship to the coronary artery anomalies, we examined the expression of myocyte and vascular smooth muscle differentiation markers in fetal (E14.5 to E17.5) and newborn Cx43α1 knockout mouse hearts. These studies showed that the walls of the pouches are not only cardiac myosin free (Fig. 4A), but also deficient in MLC-2V, another cardiomyocyte marker (Fig. 4B). Instead, we detected expression of SM22α, an early marker of vascular smooth muscle cells (Fig. 4D) (Li et al., 1996). This was observed in pouches of E16.5 and E17.5 knockout mouse heart. In the newborn knockout mouse heart, we also detected a low level expression of VSMM (Fig. 4F). These results are consistent with the notion that the lining of the pouch may contain vascular smooth muscle-like cells, although we note that SM22α and VSMM expression was observed only in a thin layer of tissue immediately subjacent to the endocardium. Further analysis of the knockout mouse heart by Masson Goldner trichrome stain for extracellular matrix deposition indicated the presence of heavy matrix deposition in the pouches, but unlike the SM22α and VSMM staining, this was distributed a few cell layers beneath the endocardium (Fig. 4G). This suggests the abnormal presence of fibroblasts in the pouches. Immunostaining also revealed strong PCNA expression in pouch tissue subjacent to the endocardium, suggesting an elevation of cell proliferation (Fig. 4C). In this same region, TUNEL labeling indicated a modest increase in apoptosis (Fig. 4E). Together with the coronary artery anomalies observed above, these observations suggest that formation of the pouches in the Cx43α1 knockout mouse probably involves the abnormal distribution of vascular smooth muscle and fibroblast cells, both of which are derived from the PEO.

To determine if Cx43α1 is expressed in the PEO, we carried out whole-mount immunostaining of E9.5 mouse embryos using a rabbit polyclonal Cx43α1 antibody. Semi-thin plastic sections of the immunostained embryos were examined by darkfield immunofluorescence microscopy. This analysis revealed that the PEO consists of mesenchymal cells with a dendritic morphology (Fig. 5A). They exhibit punctate cell surface immunolabeling typical of gap junctions. In some regions, particularly along cell processes, very long profiles of Cx43α1 immunostaining were observed (Fig. 5B,C). As the PEO is a transient structure that quickly delaminates to form the epicardium, we further examined the epicardium of E12.5 mouse embryos, and found the continued expression of Cx43α1 gap junctions in the epicardial cells (purple in Fig. 6A,B). A similar analysis of the Cx43α1 knockout mouse heart showed that the epicardium is present and exhibits normal tissue morphology, including expression of cytokeratin (blue in Fig. 6C,D), a marker specific for epicardial cells (Dettman et al., 1998).

We also examined Cx43α1 expression in proepicardial cells derived from PEO explants plated on a collagen substratum (Fig. 7). Within 24 hours, a monolayer of epithelial cells emerged from the explant expressing cytokeratin, confirming that they are indeed proepicardial cells (Fig. 7B). As in the PEO, Cx43α1 gap junctions were found in abundance in the explanted proepicardial cells, being localized along the cell borders (Fig. 7C). Cx43α1 expression was reduced in proepicardial cells derived from the heterozygous knockout embryos (not shown), and completely absent in homozygous knockout proepicardial cells (Fig. 7D). Note, in contrast to the mesenchymal morphology of cells in the PEO, the proepicardial cells in the explant cultures were epithelial, reflecting the mesenchymal to epithelial transition associated with formation of the epicardium. An examination of PEO explants derived from the Cx43α1 knockout mouse showed no change in this epithelial cell transition nor in the apparent packing of cells in the epithelial sheet (Fig. 9).

Gap junction communication in the proepicardial cells was examined by monitoring dye coupling. This entailed using microelectrode impalement and the iontophoretic injection of 6-carboxylfluorescein to quantitate the extent of dye spread. This analysis showed that wild-type proepicardial cells are functionally well coupled by gap junctions (Fig. 8). Significantly, the level of dye coupling is reduced in a stepwise manner in the heterozygous and homozygous knockout proepicardial cells (Table 1). These observations indicate that Cx43α1 plays a significant role in mediating gap junctional communication in the proepicardial cells. However, as a low level of dye coupling remained, it is likely that at least one other connexin isotype is expressed in the proepicardial cells.

Given our previous studies showing abnormal migration of cardiac neural crest cells in the Cx43α1-deficient embryos, we analyzed the migration paths of individual proepicardial cells (Fig. 9). This was carried out by capturing images of PEO explant cultures over a 20 hour interval and using the resulting time lapse movies to determine the speed and directionality of cell movement. Directionality reflects the frequency of direction change during migration and is defined as the net distance divided by the total distance traveled.

We observed that the migration of proepicardial cells usually began with the extension of lamellipodia, and occasionally filopodia were seen. After emergence from the explants, the proepicardial cells migrated as a coherent sheet, with cells at the periphery maintaining their leading position (Fig. 9). A significant elevation in the speed of cell locomotion was observed in both the heterozygous and homozygous Cx43α1 knockout proepicardial cells. It is important to note that the same increases in speed were observed in the heterozygous and homozygous Cx43α1 knockout proepicardial cells, whether cell motility was examined over the first 4 hours or the entire 20 hour interval of the experiment (Table 2).

Given the observed increased PCNA immunostaining in cells lining the pouches of the Cx43α1 knockout mouse heart, we also used BrdU incorporation to assess the rate of cell proliferation in the PEO explants. A significant increase in BrdU incorporation was observed in the heterozygous and homozygous Cx43α1 knockout proepicardial cells (Table 3). Again, the heterozygous and homozygous knockout proepicardial cells exhibited the same elevation in cell proliferation rate (Table 3). We also monitored the cell density in these explant cultures, but observed no consistent change (Table 3). These results indicate that two copies of the Cx43α1 gene are required for the normal regulation of proepicardial cell motility and cell proliferation.

To determine whether vascular smooth muscle cell differentiation might be altered by the loss of Cx43α1 function, we examined the PEO explant cultures for evidence of vascular smooth muscle cell differentiation. In quail and chick PEO explants, cells stained with antibodies against calponin, a smooth muscle marker, appeared after only 2 days in culture (Landerholm et al., 1999). Such cells exhibited a disruption of cell-cell contact resulting from their epithelial to mesenchymal cell transformation. However, in the mouse PEO explants, no calponin positive cells were found even after 8 days in culture (not shown). However, vascular smooth muscle cells were found in epicardial explants obtained from E11.5 fetal mouse hearts. Epicardial cells emerged soon after the establishment of stable heart-substratum adhesion. After 5 days in vitro, small, round or spindle shaped cells were observed (Fig. 10). These cells were immunostained by calponin (Fig. 10A,B) and SM22α antibodies (not shown), suggesting that they are smooth muscle cells derived from progenitors in the epicardium. A parallel analysis of epicardial outgrowths from Cx43α1-deficient fetal heart explants showed no discernible difference in the appearance or abundance of these calponin/SM22α-positive cells, suggesting that vascular smooth muscle cell differentiation is not altered by the loss of Cx43α1 function (Fig. 10C).

To distinguish between the contribution of neural crest versus proepicardial cells to the cardiac phenotype of the Cx43α1 knockout mouse, we examined proepicardial cells from the FC transgenic mouse model. The FC transgenic mice express a dominant negative Cx43α1/β-galactosidase fusion protein in subpopulations of neural crest cells, and although they exhibit outflow obstruction and right ventricular heart defects, infundibular pouches have never been observed (Sullivan et al., 1998). We further examined the proepicardial cells from the FC transgenic mice and found no significant alteration in dye coupling, proliferation rate or motility (Table 4). These results are consistent with the notion that proepicardial cell defects play an essential role in pouch formation in the Cx43α1 knockout mouse heart.

These studies showed for the first time that heart malformations in the Cx43α1 knockout mouse is associated with a variety of coronary artery defects. This was accompanied by an elevation in the rate of proepicardial cell migration and proliferation. Interestingly, heterozygous and homozygous knockout embryos exhibited the same coronary artery anomalies, and also similar increases in the rate of cell migration and proliferation. These findings are consistent with a previous study showing that proliferating vascular smooth muscle cells have diminished gap junction permeability (Kurjiaka et al., 1998). In the Cx43α1 knockout fetal hearts, VSMM expression in the coronary arteries was attenuated and patchy. However, the epicardium appeared normal, and analysis of epicardial explant cultures showed no change in vascular smooth muscle cell differentiation. Together these results suggest that the coronary artery defects in the Cx43α1 knockout mouse are not likely to be due to a deficiency in proepicardial or epicardial cells, nor to a defect in smooth muscle cell differentiation, but rather the abnormal proliferation and deployment of the proepicardially derived cells.

It is important to consider these findings in light of previous work showing that formation of the coronary arteries is also dependent on neural crest cells (Bartelings et al., 1993; Hood and Rosenquist, 1992; Hyer et al., 1999; Waldo et al., 1994). Unlike the proepicardial cells, neural crest cells do not contribute structurally to the coronary arteries and thus their role is presumably a regulatory one. In fact, our earlier studies had indicated the involvement of neural crest perturbations in the cardiac anomalies of the Cx43α1 knockout mouse (Ewart et al., 1997; Huang et al., 1998a; Huang et al., 1998b). We observed neural crest cells migrate into the outflow tract of the Cx43α1 knockout mouse heart, but in reduced abundance. In contrast to our findings with the proepicardial cells, significant alteration in cell locomotory behavior was observed only in the homozygous knockout neural crest cells (Xu et al., 2002). Together, these findings suggest that the coronary artery anomalies in the Cx43α1 knockout mouse cannot be solely dependent on neural crest perturbations, but rather may arise from the combined effects of perturbations involving the neural crest and proepicardial cells.

Given that neural crest and proepicardial cells are both found at the base of the heart where the coronary artery stems form by tunneling through the base of the aorta, perhaps signaling mediated through interactions between the neural crest and proepicardial cells may play a role in modulating the initial patterning of coronary arteries. The altered motility of the proepicardial and neural crest cells in the Cx43α1 knockout mouse potentially could lead to an excess of proepicardial cells and a reduction in neural crest cells. Such an imbalance between the two cell populations may alter cell signaling that is important in the patterning of coronary arteries. This signaling could involve gap junction-mediated communication between crest and proepicardial cells, or paracrine signaling not requiring direct cell-cell contact. It is also possible that the role of neural crest may be mediated via secondary interactions involving the myocardium or endocardium. Various hypotheses have been put forward regarding the role of neural crest cells in coronary artery development, including ensuring the survival of persisting coronary arteries by laying down the cardiac ganglia at the base of the heart (Waldo et al., 1994). Another possibility is that neural crest cells involved in peripheral innervation in the heart may provide signals that can modulate proepicardial cell deployment and thus help pattern coronary artery development. Regardless of the precise mechanism of proepicardial and neural crest interactions, our findings suggest that normal coronary artery development may be dependent on a precise balance of cardiac neural crest versus proepicardial cells, not merely whether neural crest or proepicardial cells are present in the heart.

We also investigated formation of the pouches in the Cx43α1 knockout mouse heart. Such pouches are not observed in neural crest-ablated embryos, and have never been seen in transgenic mouse models with Cx43α1 perturbations targeted to neural crest cells. We observed that these pouches are lined by a tissue deficient in the expression of cardiomyocyte markers. Although the precise identity of cells in this tissue is not known, we have previously shown that it expresses smooth muscle α-actin (Huang et al., 1998b; Lo and Wessels, 1999), which normally is turned off in the working myocardium at this stage of development. This would suggest that the pouch comprises cells that are in early stages of vascular smooth muscle cell differentiation. Consistent with this possibility, we observed a subendocardial layer of cells expressing vascular smooth muscle markers. However, subjacent to these smooth muscle-like cells, we also detected the presence of fibroblast cells as indicated by heavy trichrome staining. These observations suggest that formation of the pouches may involve the proepicardial cells, which are progenitors of all vascular smooth muscle and fibroblast cells in the heart. Thus, the enhanced rate of proepicardial cell migration and proliferation potentially could result in an excess of proepicardial cells at the base of the heart, thereby playing a role in pouch formation.

Consistent with this model is our observation that the FC transgenic mice never exhibit pouches, but have outflow obstruction like that of the Cx43α1 knockout mice. Previous studies of neural crest cells from FC transgenic embryos have shown reductions in dye coupling and migration rate to levels like that of the Cx43α1 knockout mouse. By contrast, FC transgenic proepicardial cells showed no change in dye coupling, nor in the rate of cell migration or proliferation. These observations suggest that Cx43α1 deficiency in the proepicardial cells is necessary for pouch formation.

Nevertheless, it is likely that other cells or tissues are involved in pouch formation. This is indicated by the fact that although predominantly only homozygous knockout animals have pouches, proepicardial cells from both the heterozygous and homozygous Cx43α1 knockout mice showed similar cell migration and proliferation alterations. Our finding that distal coronary artery branches were directly continuous with the pouches in some Cx43α1 knockout animals suggests that coronary artery anomalies have a role in pouch formation. We hypothesize that the infundibular pouches, like the coronary artery anomalies, may arise from perturbations involving both the cardiac neural crest and proepicardial cells. We note that pouch formation probably does not arise from the perturbation of endothelial cells, as cardiac malformations were not observed when tie2 promoter Cre was used to target the deletion of a floxed Cx43α1 allele in endothelial cells (Theis et al., 2001).

The present study together with our previous work suggests that Cx43α1 gap junctions play an important role in modulating the motility of two migratory cell populations essential for cardiovascular development, the neural crest and proepicardial cells. In contrast to neural crest cells, which showed a decrease in the apparent rate of cell migration in the absence of Cx43α1 function, the opposite was observed for the Cx43α1-deficient proepicardial cells. The proepicardial cells also showed increased cell proliferation, which was not observed in the Cx43α1-deficient cardiac neural crest cells. These differences perhaps reflect different signal transduction pathways functioning in neural crest versus the proepicardial cells, or the differential permeability of Cx43α1 gap junction channels to different signaling molecules active in these cell signaling pathways. In the proepicardial cells, although dye coupling was found to decrease in line with the loss of one versus both Cx43α1 alleles, cell migration and cell proliferation were identically altered in the heterozygous and homozygous knockout proepicardial cells. These observations raise the possibility that the role of Cx43α1 in cell motility may not be mediated simply by its channel-forming capacity, a possibility that is also suggested by our recent study of neural crest cell motility in other knockout mouse models (Xu et al., 2002).

The infundibular pouches observed in the Cx43α1 knockout mice share many features in common with congenital aneurysm of an aortic sinus of Valsalva in humans (Boutefeu et al., 1978; Barragry et al., 1988). Multiple other coronary anomalies observed in the Cx43α1 knockout mice also have been described in humans. For example, a high anterior origin of the right coronary artery (as seen in 8048, 8462, 8463 and 8065 in Fig. 3) is observed in 2% to 6% of individuals undergoing routine coronary angiography. Anomalous origin of the left coronary artery from the pulmonary artery (similar to Fig. 1E-H) is associated with increased morbidity and mortality (Levin et al., 1978). Anomalous origin of the right coronary artery from the pulmonary artery also has been observed in humans but is relatively rare. Anomalous origin of the left coronary artery from the right coronary artery or a single right coronary artery, similar to that observed in the Cx43α1 knockout mice (see Fig. 3, mouse 7949), is associated with a high incidence of sudden cardiac death in young individuals (Cheitlin et al., 1974; Liberthson et al., 1979; Roberts, 1986; Kragel and Roberts, 1988). In light of these findings, we propose that Cx43α1 should be evaluated as a candidate gene for mutations that may play a role in human coronary artery anomalies.

We thank John Burch for helpful suggestions regarding the proepicardial explant cultures. Supported by grants from NIH HD36457, HL36059, HD39946, NSF IBN 9905067 and Mid-Atlantic AHA Fellowship 0020280U.