Herein we report that ablation of GSK-3β led to impaired cardiomyocyte differentiation in ES cells and to pronounced hyperplasia of cardiomyocytes during embryonic development, culminating in a ventricle filled with myocytes. In striking contrast, defects of cardiomyocyte differentiation of Gsk3a–/– ES cells were less severe, and cardiac development in the Gsk3a–/– mouse was normal. These data clearly demonstrate distinct developmental roles of the two GSK-3 isoforms. Furthermore, our studies identify disordered regulation of GATA4, cyclin D1, and c-Myc, three factors known to play a role in cardiomyocyte proliferation during development, in the Gsk3b–/– heart and implicate these factors in the hyperproliferative phenotype.
We detected similar expression of GATA4 and Nkx2.5 in GSK-3–deficient EBs, which suggests that regulation of early cardiomyocyte differentiation is not specific to either GSK-3 isoform. This finding may be explained by the fact that β-catenin signaling, which promotes induction of the cardiomyocyte fate during the early stage of ES cell differentiation (42
), is preserved in GSK-3β– and GSK-3α–deficient EBs. Indeed, the lack of an increase in cytosolic or nuclear β-catenin in cardiomyocytes of Gsk3b–/–
animals is consistent with the finding that Gsk3a
can fully compensate for loss of Gsk3b
in the regulation of β-catenin (22
). By comparison, EB contraction and expression of late markers of cardiomyocyte differentiation were prominently reduced in GSK-3β–deficient compared with GSK-3α–deficient and WT EBs. The pronounced inhibition of cardiomyocyte differentiation in the setting of GSK-3β deficiency suggests that GSK-3β–selective regulation of transcription factors or cell cycle regulators is critical for the late stage of cardiomyocyte differentiation.
Differences in GSK-3α and -3β signaling in cardiomyocyte differentiation are even more apparent in heart development. Although atrioventricular and semilunar valve development was normal in the Gsk3b–/– mouse, outflow tract and septation defects were evident (DORV and VSD). The finding of DORV in GSK-3β–deficient embryos may be interpreted as a nonspecific finding, as septation of the great vessels, creation of AV valves and outflow tracts, and closure of the VSD are regulated by cells of the endocardial cushions and neural crest. While abnormalities of neural crest could explain the DORV in GSK-3β–deficient embryos (and only a tissue-specific knockout in neural crest cells could definitively address this), we saw no evidence of improper septation of the great vessels, which is a common feature of neural crest defects. Furthermore, spatial migration of neural crest cells appeared normal. Enlargement of the endocardial cushions that impair normal rotation of the developing outflow tract can produce DORV. However, since there was no abnormality in valve development (which requires intact β-catenin signaling for the EMT), we consider a selective defect in the cushion cells to be unlikely. We considered that the marked distortion of LV geometry secondary to the cardiomyocyte hyperproliferation may have given rise to the DORV by impairing normal rotation of the outflow tract; however, the DORV was present before noticeable differences in ventricular wall thickness (data not shown).
Our finding that GSK-3α is not required for mouse development is in contrast to observations in zebrafish, in which morpholino oligonucleotide–mediated knockdown of GSK-3α produced severe axial developmental defects and death (43
). Furthermore, while knockdown of either zebrafish GSK-3 isoform produced thin hearts and pericardial edema, we only found heart defects in Gsk3b–/–
embryos. Finally, while zebrafish GSK-3α is required for cardiomyocyte survival, in striking contrast, we found that Gsk3b–/–
mouse embryos showed increased cardiomyocyte proliferation. We cannot reconcile the differences between the zebrafish and mouse loss-of-function studies; however, we speculate that differences in amino acid sequences flanking the serine/threonine kinase domain (44
), altered substrate availability, or differences in the developmental expression pattern may account for unique roles for the GSK-3 isoforms in zebrafish compared with mammals. Despite these differences, morpholino oligonucleotide targeting of zebrafish GSK-3β produced abnormalities of heart tube jogging and outflow tract positioning (43
) that may provide insight into the DORV observed in Gsk3b–/–
mice. The zebrafish outflow abnormalities correlated with abnormal lefty and BMP4 expression, suggesting an important role for GSK-3β in left-right patterning. Our data demonstrating DORV likewise suggest a requirement for GSK-3β in outflow tract development of the mammalian heart.
The striking enlargement of ventricular walls in late gestational periods, virtually filling the ventricular cavities, must be viewed in comparison to findings of studies by our group (34
) and others (35
) that have employed overexpression approaches to suggest that GSK-3β is a potent negative regulator of cardiomyocyte growth. The growth that has been studied to date is postnatal growth, the vast majority of which is via hypertrophy rather than hyperplasia, since postnatal cardiomyocytes are generally unable to reenter the cell cycle except under extreme circumstances. Indeed, because cardiomyocyte cross-sectional areas in Gsk3b–/–
and control E17.5 embryos were similar, the increased ventricular wall thickness is not the result of cardiomyocyte hypertrophy. Herein we identify a key role for GSK-3β in regulating hyperplastic growth of the developing heart. Because both the left and right ventricles were excessively cellular, relative to the chamber dimension, this finding is much more likely to be due to effects of deletion of GSK-3β than a response to an epigenetic influence (e.g., constrained outflow). Similarly, pharmacologic inhibition of GSK-3 also increased cardiomyocyte proliferation. Withdrawal from the cell cycle is required for the late and terminal stages of cardiomyocyte differentiation (47
). Thus, active cycling of the Gsk3b–/–
embryo cardiomyocytes is likely one important factor contributing to inhibition of their terminal differentiation.
The most striking consequences of GSK-3β deficiency that likely drive cardiomyocyte proliferation are increases in nuclear GATA4, cyclin D1, and c-Myc, all of which, based on our findings, appear to be targets of GSK-3β in vivo. GSK-3–mediated phosphorylation of cyclin D1 at Thr286 has been proposed to trigger nuclear export of cyclin D1 and priming of cyclin D1 for ubiquitination and degradation (23
), though this is a subject of debate (25
). Our data, from studies employing genetic loss of function of GSK-3β in vivo, support a role for GSK-3β as a regulator of cyclin D1 nuclear/cytoplasmic partitioning and degradation. Of note, the phenotype of the GSK-3β knockout is virtually the opposite of that of mice deleted for cyclins D1/D2/D3 or deleted for cyclin D partners Cdk2/Cdk4, which have marked thinning of the walls (48
). Interestingly, most tissues in the cyclin D1/D2/D3–deficient mouse had normal proliferation rates, while cardiomyocytes and hematopoietic stem cells appeared to be relatively unique in their dependence on D-type cyclins for driving proliferation (48
Although GSK-3β deficiency affects several cell cycle regulators, its regulation of GATA-dependent signaling may be a critical contributor to cardiomyocyte proliferation. GATA4/GATA6 compound heterozygosity (50
) and cardiomyocyte-specific knockout of GATA4 (11
) produce heart developmental abnormalities characterized by reduced cardiomyocyte proliferation. Because the increased wall thickness in Gsk3b–/–
embryos is related to hyperplasia, rather than hypertrophy, we propose that increased GATA4 nuclear accumulation in Gsk3b–/–
embryos can be expected to drive proliferation. Indeed, deletion of GSK-3β likely produces a GATA4 gain of function, in vivo, which drives proliferation. Our conclusion is supported by the finding that forced GATA4 expression (and GSK-3 inhibition) increases neonatal cardiomyocyte proliferation. Our results suggest that GSK-3β–mediated phosphorylation and inhibition of GATA4 and other regulators of cellular proliferation are necessary to restrict the number of cardiomyocytes during late gestational periods of heart development, which allows for completion of cardiomyocyte differentiation.
GSK-3 has been proposed as a viable target in numerous disease states, including ischemic injury of the heart and brain (51
), inflammation (52
), diabetes (53
), Alzheimer disease (51
), and bipolar disorders (29
). While lithium is currently the only available approved drug that can inhibit GSK-3, numerous inhibitors with greater potency and specificity are in development (54
). The issue of whether lithium, and by extension GSK-3 inhibition, causes congenital heart defects has been debated in the literature for more than 20 years because early retrospective analyses that reported high rates of congenital heart defects in the setting of lithium treatment were flawed (30
). A more recent report confirmed a marginally increased risk for a cardiac defect after maternal use of lithium (55
), and administration of lithium to animal models at very early gestational stages produces heart defects (56
). However, it has been difficult to determine whether the heart defects observed in these lithium studies can be extrapolated to newer, more potent GSK-3 inhibitors for several reasons. First, lithium levels achieved in patients are only approximately 1 mM, and this causes (at physiological levels of Mg2+
) only an approximately 25% inhibition of total GSK-3 activity (i.e., approximately equal to deletion of 1 allele of either GSK-3α or GSK-3β) (57
). Second, lithium has several targets in addition to GSK-3, including Akt (which is activated by lithium and obviously has numerous targets in addition to GSK-3) and inositol monophosphatases (58
). Indeed, depletion of intracellular inositol has been suggested to mediate the anti-manic effects of lithium (59
). More recently, lithium has been shown to disrupt a complex consisting of Akt, GSK-3, protein phosphatase 2a, and β-arrestin (60
). Thus, lithium has effects well beyond simply inhibiting GSK-3, making it nearly impossible to ascribe specific effects of lithium to one specific target and completely impossible to ascribe effects to one specific isoform. We believe that our findings using genetically targeted mice bring clarity to the central issue: that inhibition of GSK-3β in the developing heart might be expected to lead to cardiac developmental abnormalities.
In summary, we have identified GSK-3β as a central regulator of cardiomyocyte proliferation and differentiation during embryonic heart development, mediated, at least in part, through its effects on 2 transcription factors and 1 cell cycle regulator. Deficits of ventricular remodeling of Gsk3b–/– animals are likely secondary to the requirement of GSK-3β to restrict myocardial cell number, but GSK-3β may also have a more direct role in maturation of the ventricle. The requirement for GSK-3β in heart development raises the consideration that mutations of GSK-3β may contribute to cases of DORV and VSD as well as LV hypertrophy in humans.