Mutations of the transcription factor Nkx2-5 cause human congenital heart disease (Schott et al., 1998; Benson et al., 1999). Loss-of-function mutations, as defined by loss of DNA-binding activity, are associated with atrioventricular conduction defects (Benson et al., 1999; Kasahara et al., 2000; Goldmuntz et al., 2001). In children, important causes of conduction defects include damage to the conduction tissues, as caused by maternal lupus antibodies or by intracardiac repair of congenital heart defects. Drugs or mutations that perturb the function of proteins that generate or propagate the cardiac action potential are also common causes. Such proteins were hence investigated to define the basis of conduction defects in Nkx2-5 mutation. A quantitative survey of connexins, channels and pumps, and whole cell patch clamp experiments in wild-type and heterozygous Nkx2-5 knockout hearts yielded no obvious culprits, however (Tanaka et al., 2002). We thus considered an alternative hypothesis rooted in development.
The cardiac conduction system develops by the recruitment of multipotent embryonic myocytes into a node or fiber (Gourdie et al., 1995; Cheng et al., 1999). Once recruited, the cells exit the cell cycle. Genes that may have a role have been identified in a few cases by in vitro embryonic manipulation. In the chick embryo, the coronary arterioles secrete endothelin, which induces the conversion of myocytes into Purkinje cells. Purkinje fibers thus develop adjacent to the arterioles (Gourdie et al., 1998). In the mouse, the Purkinje fibers develop in a subendocardial distribution. Cultured mouse embryo experiments suggest that neuregulin-1 may be an inductive signal (Rentschler et al., 2002). The signals that pattern development of the central conduction system, i.e., AV node, His bundle, and proximal bundle branches, are unknown.
Interestingly, Nkx2-5 is upregulated during periods of recruitment into the central and peripheral conduction systems in the chick, mouse, and human heart (Thomas et al., 2001). The observation supports the hypothesis that the Nkx2-5 mutant conduction phenotype could result from a defect in embryonic development. Clinical observations provide additional evidence that highlight the importance of development in conduction system function.
INBORN ERRORS OF THE DEVELOPMENT OF THE CONDUCTION SYSTEM AS A CAUSE OF POSTNATAL PHYSIOLOGIC DEFECTS
Pediatric cardiologists have long recognized that certain cardiac malformations are associated with atrioventricular conduction defects. To illustrate the point, we searched the database of the Department of Cardiology at Children's Hospital, Boston, which records the diagnoses and procedures of the 59,832 patients seen in the department between 1 January 1988 and 1 January 2002. Of 360 patients who ever had a diagnosis of complete heart block and an anatomic defect, 13 were confirmed to have congenital heart block based on electrocardiogram, physician documentation, or both (Table 1). None of the 13 had an electrocardiogram or documentation that indicated residual atrioventricular conduction. Most of the remainder had transient or acquired heart block. A few may have had congenital heart block but confirmatory evidence was not found.
Table 1. Congenital heart defects associated with congenital heart block in cardiology patients seen at Children's Hospital, Boston, 1988–2001*
The most commonly associated defects were l-TGA and heterotaxy, specifically left atrial isomerism. The number of patients with an interrupted inferior vena cava, a sign of left atrial isomerism, is noted in parentheses. Tetralogy of Fallot, especially as a feature of Down syndrome, is not usually associated with congenital heart block but has been reported (Machado et al., 1988). The two patients with a muscular ventricular septal defect and secundum atrial septal defect probably developed congenital heart block because of maternal lupus antibodies, i.e., anti-SSA/Ro and SSB/La, which do not cause congenital heart defects.
Tetralogy of Fallot
Ventricular septal defect
Maternal lupus Ab+
Atrial septal defect
Maternal lupus Ab+
The search revealed two defects, l-transposition of the great arteries (l-TGA) and heterotaxy with left atrial isomerism, which others have previously shown to be associated with congenital heart block (Pinsky et al., 1982; Machado et al., 1988). In l-TGA, the right and left ventricles are reversed in their relative positions to become the systemic and pulmonary pumping chambers, respectively. In heterotaxy, the normal left- and right-sided asymmetric features of the atria, lungs, and abdominal viscera are lost. Heterotaxy patients have atria, structures, or organs on one side that resemble the opposite side, hence the terms right or left atrial isomerism. Interruption of the segment of the inferior vena cava between the liver and kidneys, as in the three patients discovered in the search, is a hallmark of left atrial isomerism. Histopathologic analyses of hearts with l-TGA (Lev et al., 1963; Anderson et al., 1974a, 1974b) or left atrial isomerism have demonstrated discontinuities in the atrioventricular conduction axis as a cause of congenital complete heart block (Ho et al., 1992).
The cases highlight the relevance of specific, albeit unknown, developmental pathways in the pathogenesis of some forms of congenital heart block. The absence of heart block in other major heart defects that are frequently seen at Children's Hospital suggests the specificity of the etiologies. For example, among almost 3,000 patients who had hypoplastic left heart syndrome (n = 402), coarctation of the aorta (n = 1,271), or endocardial cushion defects (n = 1,271), none had congenital heart block. Thus, the conduction defect in l-TGA is likely related to abnormal looping rather than associated defects such as ventricular hypoplasia, ventricular septal defect, or valvular abnormalities. Likewise, endocardial cushion defects such as partial or complete common atrioventricular canal defects are a feature of heterotaxy. The pathogenesis of the conduction defect in left atrial isomerism must therefore be specific to the development of bilateral left-sidedness and not to the abnormal development of the endocardial cushions or atrioventricular canal.
Although the database search focused on patients who had complete heart block, infants with l-TGA or left atrial isomerism commonly have less severe forms of heart block that can progress. If abnormal development of the conduction system can cause postnatal conduction defects, it is not necessarily an all-or-none phenotype. The initial severity of the conduction defect is variable and can become worse. What might be considered acquired heart block could have a congenital basis.
PHYSIOLOGIC CONDUCTION DEFECTS IN NKX2-5 HAPLOINSUFFICIENT HUMANS AND A MOUSE MODEL
DNA-binding mutations of Nkx2-5 cause conduction defects in man (Table 2). The mutations were first discovered in families with heritable atrial septal and atrioventricular conduction defects. The heart block varied in severity from mild prolongation of the PR interval (first-degree block) to complete (third-degree block) and was progressive in one family (Schott et al., 1998). Benson et al. (1999) subsequently identified a patient who had advanced second-degree block but no cardiac malformation. Based on electrophysiologic studies in one family, the conduction defect was localized to the AV node (Schott et al., 1998).
Table 2. Conduction defects and electrophysiologic abnormalities*
≥ 7 weeks
Reported in humans (Schott et al., 1998) and mice (Jay et al., 2004) with heterozygous Nkx2-5 loss of function. The presence of AH and absence of HV block in human cases are inferred based on the description of electrophysiologic abnormalities reported in one family (Schott et al., 1998). +, reported as present; −, reported as absent; ?, not reported.
AV block, degree
Diminished His potential
Two groups have reported the effects of homozygous loss of function in Nkx2-5 knockout mice (Lyons et al., 1995; Tanaka et al., 1999). The discovery of the human mutation motivated investigation of the Nkx2-5 heterozygous knockout (Nkx2-5+/−) mouse. The first report described mild first-degree block in female mutant mice only (Biben et al., 2000). Our group has observed it in both sexes at age 7 weeks and older (Tanaka et al., 2002; Jay et al., 2004). The cause of the discrepant results is unknown but could be related to the temporal resolution of electrocardiogram measurements. Neither group has observed second- or third-degree heart block (Table 2).
As in humans, a defect in the AV node caused the prolongation of the PR interval in the Nkx2-5+/− mouse, based on prolongation of the AH interval on intracardiac electrogram and normal P-wave duration on the surface electrocardiogram. Rapid atrial pacing at high rates provided further evidence for reduced AV node function in the mutant. Interestingly, first-degree block and other signs of decreased AV node function were not present in 3-week-old mutant mice but were at 7 weeks (Jay et al., 2004). The delayed development of first-degree block in the mutant mice is reminiscent of the progression of block in humans with Nkx2-5 mutation or congenital heart defects such as l-TGA and heterotaxy.
After a delay in the AV node, conduction progresses through the His bundle. Using an octapolar electrode catheter that passed from the internal jugular vein to the right atrium and ventricle, we could measure atrial, His, and ventricular depolarization simultaneously. In blinded studies, the His depolarization signal was noted to be markedly diminished or absent in a subset of mice (Table 2) (Jay et al., 2004). The small or absent His signal correlated exactly with the Nkx2-5+/− genotype in studies of about 100 animals ranging in age from 3 weeks to 1 year; all wild-type mice had clearly detectable His signals (Fig. 1).
The His bundle bifurcates into the left and right bundles, which in turn branch into the peripheral Purkinje network. Depolarization of the ventricles begins at the Purkinje-contractile myocyte junction. Conduction through the His-Purkinje system is represented by the HV interval on the intracardiac electrogram. The QRS interval on the surface electrocardiogram, which begins with the onset of the ventricular spike on the intracardiac electrogram, marks the time required to depolarize the entire contractile myocardium after the Purkinje system. Conduction velocity through the Nkx2-5+/− bundle branches and to the Purkinje-myocyte junction was normal, as indicated by the normal HV interval. The QRS was mildly prolonged in mutant mice at all ages examined. The HV and QRS data together indicate that the delay results from an abnormality at or distal to the Purkinje-contractile myocyte junction.
ANATOMIC DEFECTS OF THE CONDUCTION SYSTEM IN AN NKX2-5 HAPLOINSUFFICIENT MOUSE
Physiologic defects found at each level of the conduction system in the Nkx2-5+/− mice correlated with cellular hypoplasia of the relevant structures. The number of conduction cells was assessed qualitatively and quantitatively using two separate sets of molecular markers, minK-lacZ and connexin40 and -45 (Cx40, Cx45) (Gourdie et al., 1993a, 1993b; Kupershmidt et al., 1999). Both sets of markers clearly demonstrated the hypocellularity of the conduction system in Nkx2-5 heterozygotes (Jay et al., 2004). For example, one of the authors (B.S.H.) could determine the genotype of an unknown heart based on the number of Cx40-positive cells present in a section of the AV node, His bundle, or Purkinje system. Anatomic abnormalities of the AV node likely contribute to first-degree AV block in association with unknown factors related to maturation, and hypoplasia of the His bundle and Purkinje system can entirely explain the defects associated with these structures.
The AV node forms at the inner curvature of the looping heart tube (Viragh and Challice, 1977). MinK-lacZ-positive cells mark the region in wild-type E9.5 embryos, but no such cells are present in the atrioventricular region of Nkx2-5 null mutant embryos (Fig. 2a and b), suggesting that Nkx2-5 is essential to establish the AV node (Jay et al., 2004). In the Nkx2-5+/− adult, the AV node is smaller and contains fewer cells than the wild type (Fig. 2c and d). A direct correlation thus appears to exist between Nkx2-5 gene dosage from nil to normal and cellularity of the node. In addition, the Nkx2-5+/− AV node completely lacked the proximal portion of the node, which expresses Cx45 but not Cx40 (Jay et al., 2004).
Since Nkx2-5+/− mice did not develop first-degree AV block until after age 4 weeks, neither hypoplasia nor absence of the proximal AV node is sufficient to cause a conduction defect in young mice but could contribute to the pathogenesis as the mice age. We suspect that unknown factors related to maturation could cause Nkx2-5+/− mice to develop exaggerated PR prolongation given their abnormal AV nodes. For example, mice double in size between 3 and 7 weeks of age, whereas the AV node and His bundle have a fixed number of cells that could be stretched thin with growth of the heart. Remodeling of gap junctions between birth and young adulthood could also be a factor (Litchenberg et al., 2000). Interestingly, conditional knockout of Nkx2-5 using a Cre transgenic mouse driven by the MLC2V promoter supports the role of AV node hypoplasia in the pathogenesis of first-degree AV block. MLC2V-Cre is expressed in a subset of cells in the AV node, which presumably deleted Nkx2-5 in those cells and caused a hypoplastic AV node and first-degree block in adult mice (Pashmforoush et al., 2004).
The His bundle in Nkx2-5+/− mice was thin and wispy compared to the wild type (Fig. 2e and f). The hypoplastic His bundle can directly explain the low-amplitude His depolarization signal measured in Nkx2-5+/− hearts. The low-amplitude signal in 3-week-old Nkx2-5+/− mice is consistent with congenital hypoplasia of the His bundle (Jay et al., 2004).
The peripheral Purkinje network of fibers in Nkx2-5+/− mice appeared pruned compared to the wild type. Whole mount inspection of mink-lacZ-stained hearts demonstrated dense blue staining of the Purkinje network in the interventricular septum of the wild type (Fig. 2g), whereas individual fibers could be discerned in the heterozygotes because there were fewer cells and fibers (Fig. 2h). Qualitative and quantitative confirmation was obtained by examination of frozen sections where Purkinje cells were identified by Cx40 expression. Fibers were fewer and smaller, and half as many Purkinje cells were present in Nkx2-5+/− hearts as in the wild type (Jay et al., 2004). This correlated exactly with a reduction of mink-lacZ enzymatic activity in Nkx2-5+/− embryos at E14.5, suggesting that the hypoplastic Purkinje network arose from abnormal embryonic development rather than postnatal loss. A beautiful confirmation of the results was also obtained from hearts expressing GFP driven by the Cx40 promoter. Images of the left ventricular septal surface demonstrating the fluorescent radiations of the left bundle branch clearly showed that Nkx2-5+/− hearts fewer Purkinje fibers than the wild type (D. Gros, Conduction System Meeting, April 2004, Washington, DC).
The hypocellular Purkinje network in Nkx2-5+/− mice can explain the prolonged QRS duration. Since conduction through the contractile myocardium is slower than the Purkinje system, depolarization of the entire ventricular myocardium would be prolonged if fewer Purkinje-myocyte junctions were present to activate a proportionately larger volume. Comparison with the MLC2V-Cre conditional knockout of Nkx2-5 suggests that the number of peripheral Purkinje cells could be further reduced in the absence of Nkx2-5. Rhythm tracings from the adult conditional knockout mice seem to demonstrate Mobitz type II, second-degree AV block (Pashmforoush et al., 2004), which localizes the conduction defect at or distal to the His bundle (Damato et al., 1969; Rosen et al., 1971, 1972). It would therefore be interesting to know whether a paucity of Purkinje cells could explain the progressive AV block phenotype in the ventricular-restricted knockout experiment, as the heart outgrew the capacity of the His bundle or proximal bundle branches to depolarize the entire myocardial mass.
Haploinsufficiency of Nkx2-5 appears to disturb a relationship across large and small species between QRS duration and the distance from the Purkinje-myocyte junction to the farthest contractile myocyte in its region of excitation. Mice and humans, whose hearts are approximately 10 times thicker than the mouse, have a predominantly subendocardial distribution of Purkinje cells. Consistent with their relative ventricular thicknesses, the human QRS is about 10 times longer than the mouse. Similarly, the bovine heart is four times larger than the human, but the Purkinje network extends almost to the subepicardium. Hence, cattle have the same QRS duration as humans (Oosthoek et al., 1993).
ROLE OF NKX2-5 IN CONNEXIN40 EXPRESSION
Haploinsufficiency of Nkx2-5 clearly causes physiologic and anatomic conduction defects in the mouse, which can explain the pathogenesis of the human Nkx2-5 phenotype. The analysis of heterozygous Nkx2-5 knockout mice raised an additional issue regarding the regulation of Cx40 by Nkx2-5. Downregulation of Cx40 in Nkx2-5 mutation was considered a potential cause of the conduction defect based on two general observations. First, the Cx40 promoter contains Nkx2-5 binding sites, and in vitro transactivation assays show that Nkx2-5 can activate expression of a Cx40 promoter-reporter gene construct (Bruneau et al., 2001). Second, homozygous Cx40 knockout mice have first-degree atrioventricular block and prolonged AH and HV intervals (Tamaddon et al., 2000; VanderBrink et al., 2000).
Quantitative confocal immunohistochemisty revealed the same cellular expression levels of Cx40 in wild-type and Nkx2-5+/− Purkinje cells, however (Jay et al., 2004). No qualitative difference was discerned either in the AV node or His bundle of Cx40 and 45 expression. Thus, there are fewer cells in Nkx2-5+/− conduction system, and each cell has a normal amount of connexins. Half-normal levels of Nkx2-5 do not significantly affect the expression of Cx40. Subtle quantitative reductions of Cx40 might have escaped detection, but it could not have caused a conduction defect since heterozygous Cx40 knockout mice, which have half-normal protein levels, have normal conduction (Kirchhoff et al., 1998).
Two mouse models raise questions about the role of Nkx2-5 in the regulation of Cx40 expression outside the normal pathophysiologic range of transcription factor activity. Kasahara et al. (2001) reported the phenotype of a mouse that overexpressed a mutant Nkx2-5, I183P, which cannot bind DNA. The mice developed progressive atrioventricular block and showed concomitant downregulation of Cx40 and -43. Abnormal connexin expression likely contributed to the conduction defect, but one could not conclude whether it was a direct or indirect effect of the mutant transgene. In the MLC2V-Cre knockout of Nkx2-5, fewer Cx40 gap junction plaques were seen in images of the AV node and His bundle, but it was unclear whether the level of either Cx40 or Nkx2-5 was reduced in individual cells (Pashmforoush et al., 2004). These results highlight the difficulties of interpreting expression data in mutant animals even when in vitro transactivation assays suggest that a particular gene is a direct transcriptional target. Extrapolation of conclusions from overexpressing or complete knockout models to mechanistic models of heterozygous or haploinsufficient transcription factor mutant phenotypes can be similarly problematic.
DEFINING THE ROLE OF NKX2-5 IN DEVELOPMENT OF THE CONDUCTION SYSTEM
The anatomic and conduction defects associated with Nkx2-5 mutation are well described, but the precise mechanistic role of Nkx2-5 in the pathogenesis of any particular defect remains nebulous. The discovery that the number of cells in the conduction system is related to Nkx2-5 gene dosage yields an important clue. Based on current understanding of the development of the conduction system, four general mechanisms through which Nkx2-5 could act are theoretically possible. Since there is no significant proliferation once cells are recruited to the conduction system, the net reduction in conduction cells in Nkx2-5 mutant animals must result either from diminished recruitment or from increased loss.
Diminished recruitment in Nkx2-5 insufficiency could result from one of three possibilities. Nkx2-5 might regulate the specification of a small population of cells that form the nidus for recruiting other myocytes (Fig. 3a). No compelling evidence for such a population has been found, but one could speculate that such prespecified cells might exist to form the AV node, based on the expression pattern of a cGATA6-lacZ transgene in the precardiac mesoderm and atrioventricular canal myocardium (Davis et al., 2001). Alternatively, Nkx2-5 might regulate the expression of a putative inductive signal (Fig. 3b), which instructs pluripotent myocytes to join the developing node or fiber, or the response to the signal (Fig. 3c). Nkx2-5 gene dosage would therefore be important either in the developing conduction system or in the pluripotent myocyte.
Once recruited, some myocytes in or around the developing conduction system undergo apoptosis (Cheng et al., 2002). Nkx2-5 haploinsufficient hearts might lose myocytes once they have entered the conduction system (Fig. 3d). Apoptosis could explain the progression from mild to complete atrioventricular block reported in patients with Nkx2-5 mutation (Schott et al., 1998). No direct evidence has been reported, however, that demonstrate more apoptosis or loss of conduction myocytes in Nkx2-5 heterozygotes or in ventricular-restricted Nkx2-5 conditional knockout mice (Pashmforoush et al., 2004). Confirmation of the theoretical possibility would require examination by TUNEL staining, for example, or counts of the number of cells in the conduction system of embryos and mice at various ages. As the number of cells in the conduction system is likely fixed at birth, growth of the postnatal heart could cause what appeared to be loss of cells or shrinkage of the AV node, bundles, and fibers.
Beside developmental processes, one could also ask what specific genes Nkx2-5 regulates. Few genes are known that might be important in the development of the conduction system, and we know of none that is an Nkx2-5 target. Preliminary northern analysis indicates that Wnt11, which is upregulated in the developing chick conduction system (Bond et al., 2003), is not differentially expressed in the hearts of Nkx2-5+/− embryos (data not shown). Msx2, which is expressed in the developing chick conduction system and atrioventricular myocardium of the mouse, is abnormally upregulated in Nkx2-5 null mutant embryos (Chan-Thomas et al., 1993; Tanaka et al., 1999). Absence of Msx2, however, does not affect cardiac conduction or rescue the Nkx2-5+/− conduction defect (Jay et al., in press).
Significant progress has been made in the past decade in understanding the development of the cardiac conduction system. It is undisputed now that conduction cells arise from a multipotent cardiac myocyte lineage. New molecular markers and mouse strains like the minK-lacZ and cGATA6-lacZ mouse have and will greatly facilitate complicated analyses (Kupershmidt et al., 1999; Davis et al., 2001). Finally, multiple lines of evidence in various species and model systems have identified functionally important genes, providing footholds to dissect molecular mechanisms. These advances will make possible future experiments that address more detailed mechanistic questions regarding the function of Nkx2-5.
The authors thank Maria Rivera for her assistance and Dr. Tom Schultheiss for his thoughtful input throughout this work. The cardiology patient database search was approved by the Committee on Clinical Investigation, Children's Hospital. Supported by grants from the National Heart, Lung, and Blood Institute (NHLBI)/National Institutes of Health (to P.Y.J., R.G.G., C.I.B., and S.I.), the Charles H. Hood Foundation and the Marram and Carpenter Fund for Innovation (to P.Y.J.), and the Deutsche Akademie der Naturforscher Leopoldina (to A.B.).