Distinct Roles of Glycogen Synthase Kinase (GSK)-3α and GSK-3β in Mediating Cardiomyocyte Differentiation in Murine Bone Marrow-derived Mesenchymal Stem Cells*

The signaling mechanisms facilitating cardiomyocyte (CM) differentiation from bone marrow (BM)-derived mesenchymal stem cells (MSCs) are not well understood. 5-Azacytidine (5-Aza), a DNA demethylating agent, induces expression of cardiac-specific genes, such as Nkx2.5 and α-MHC, in mouse BM-derived MSCs. 5-Aza treatment caused significant up-regulation of glycogen synthase kinase (GSK)-3β and down-regulation of β-catenin, whereas it stimulated GSK-3α expression only modestly. The promoter region of GSK-3β was heavily methylated in control MSCs, but was demethylated by 5-Aza. Although overexpression of GSK-3β potently induced CM differentiation, that of GSK-3α induced markers of neuronal and chondrocyte differentiation. GSK-3 inhibitors, including LiCl, SB 216743, and BIO, abolished 5-Aza-induced up-regulation of CM-specific genes, suggesting that GSK-3 is necessary and sufficient for CM differentiation in MSCs. Although specific knockdown of endogenous GSK-3β abolished 5-Aza-induced expression of cardiac specific genes, surprisingly, that of GSK-3α facilitated CM differentiation in MSCs. Although GSK-3β is found in both the cytosol and nucleus in MSCs, GSK-3α is localized primarily in the nucleus. Nuclear-specific overexpression of GSK-3β failed to stimulate CM differentiation. Down-regulation of β-catenin mediates GSK-3β-induced CM differentiation in MSCs, whereas up-regulation of c-Jun plays an important role in mediating CM differentiation induced by GSK-3α knockdown. These results suggest that GSK-3α and GSK-3β have distinct roles in regulating CM differentiation in BM-derived MSCs. GSK-3β in the cytosol induces CM differentiation of MSCs through down-regulation of β-catenin. In contrast, GSK-3α in the nucleus inhibits CM differentiation through down-regulation of c-Jun.

Ischemic cardiomyopathy and myocardial infarction are accompanied by an irreversible loss of cardiomyocytes, endothelial cells, and smooth muscle cells, essential components of the heart (1). Cell-based cardiac repair offers the promise of rebuilding the injured heart from its component parts (reviewed in Refs. 2 and 3). Although remarkable progress in the field has clearly proven the concept of "cell-based cardiac repair," initial clinical studies using adult stem cells have shown that the salutary effects mediated by cell transplantations are generally modest (4 -7). A major challenge for cardiac regeneration therapy using adult stem cells may be to enhance stem cell differentiation into cardiomyocytes.
Among several important signaling mechanisms generally involved in cardiomyocyte differentiation, Wnt/␤-catenin (canonical Wnt) and non-canonical Wnt signaling have been suggested to have a critical role in cardiogenesis (8). Glycogen synthase kinase (GSK) 2 -3 is a key component of the canonical Wnt signaling pathway. GSK-3 phosphorylates ␤-catenin, and phosphorylated ␤-catenin is then subjected to ubiquitin proteasome degradation. However, upon Wnt binding to its receptors, Frizzled and low-density lipoprotein receptor-related protein, ␤-catenin phosphorylation by GSK-3␤ is inhibited and ␤-catenin is stabilized. Stabilized ␤-catenin translocates into the nucleus and induces target gene expression. Although both the canonical and non-canonical Wnt pathways are important in mediating cardiomyocyte differentiation in stem cells and cardiac progenitor cells (9 -15), the role of downstream components of the Wnt pathway, and, in particular, the role of GSK-3, in mediating cardiomyocyte differentiation is not yet fully understood.
GSK-3 is a serine/threonine kinase that has a wide variety of functions in cells. GSK-3 phosphorylates many known intracellular targets, including ␤-catenin, glycogen synthase, elF2B⑀, GATA4, myocardin, c-Jun, cyclin D1, and N-Myc, thereby regulating various cellular functions, including hypertrophy and apoptosis in cardiomyocytes (16). GSK-3 has two major isoforms, GSK-3␣ and GSK-3␤, which have 97% identical amino acids in the catalytic domain but differ substantially in the N and C termini. Increasing lines of evidence suggest that GSK-3␣ and GSK-3␤ both have common and non-overlapping functions (17). For example, both GSK-3␣ and GSK-3␤ phosphorylate/degrade ␤-catenin in embryonic stem cells, but GSK-3␣ and GSK-3␤ play distinct roles in cardiac development in mice (18,19). Importantly, the isoform-specific functions of GSK-3␣ and GSK-3␤ during cardiomyocyte differentiation are not well understood in mesenchymal stem cells (MSCs). 5-Azacytidine (5-Aza) is a chemical analogue of cytidine that removes methyl groups from DNA, thereby inducing gene expression. 5-Aza is a potent inducer of cardiomyocyte differentiation in bone marrow-derived MSCs (20). MSCs show potential for clinical application with evidence of tissue regeneration, including myocardial regeneration (21). We reasoned that by studying the function of signaling molecules modulated during cardiac differentiation, we should be able to elucidate the signaling mechanisms involved in stimulating cardiomyocyte differentiation in adult stem cells. Through initial screening of signaling molecules modulated by 5-Aza during cardiomyocyte differentiation in MSCs, we found that GSK-3 plays an important role in regulating this process.
Thus, the goals in this study were to elucidate whether GSK-3␣/␤ is affected during differentiation of MSCs into the cardiomyocyte lineage in response to 5-Aza treatment, and if so, to examine whether modulation of GSK-3 plays a causative role in mediating cardiomyocyte differentiation in MSCs. Furthermore, we evaluated whether GSK-3␣ and GSK-3␤ play distinct roles in mediating cardiomyocyte differentiation and whether regulation of ␤-catenin is involved in modulation of cardiomyocyte differentiation by GSK-3␣/␤.
5-Aza Treatment-MSCs were treated with 5-Aza (Sigma, 5 M) for 24 h without serum and then cultured in serum containing culture medium. Identically prepared MSCs without 5-Aza treatment were used as a control. Inhibition of GSK-3 kinase activity was achieved by maintaining MSCs in culture medium containing LiCl (10 mM), BIO (0.5 g/ml), or SB216743 (2 g/ml).
Reverse Transcriptase-PCR-Total RNA was extracted using TRIzol (Invitrogen) and 1 g of RNA was used for cDNA synthesis (Thermoscriptase, Ambion). The RT-PCR mixture (Promega) was incubated at 95°C for 5 min followed by 95°C for 30 s, 59°C for 1 min, and 72°C for 30 s for 34 cycles and then incubated at 72°C for 7 min. PCR primers used in this study are shown in supplemental Table S1.
Luciferase Assay-Plasmids harboring TOP flash (TCF-luciferase plasmid, Millipore) and FOP flash (mutant TCF-luciferase plasmid, Millipore) were transfected into MSCs using Lipofectamine reagent (Invitrogen), and the luciferase activity was measured with the Luciferase Assay System (Promega) after cell lysis with Passive Lysis Buffer (Promega).
Statistical Analyses-All values are expressed as mean Ϯ S.E. Statistical analyses were performed using analysis of variance and Newman-Keuls multiple comparison test with a p Ͻ 0.05 considered significant.

Murine Bone Marrow-derived MSCs Express Pluripotent Markers and 5-Aza Induces Expression of Cardiac Marker Genes in MSCs-
MSCs at the third passage expressed Oct4 and Rex1 mRNA, pluripotent stem cell markers (Fig. 1A). Ninetyfive percent of MSCs at the third passage were Oct3/4 positive (Fig.  1B). Although untreated bone marrow-derived mouse MSCs do not express cardiac marker genes, 5-Aza treatment (5 M for 24 h), an established method of inducing bone marrow MSC differentiation into cardiomyocytes (20), induced mRNA expression of Nkx2.5, one of the earliest cardiac markers, and ␣-myosin heavy chain (␣-MHC), a contractile protein (Fig. 1C). Although no cardiac troponin I (cTnI) positive cells were observed in control MSCs, 5-Aza induced a premature but clear striation pattern of cTnI in MSCs (Fig. 1D).

5-Aza Induces Cardiomyocyte Differentiation through an Increase in GSK-3␤ Protein and mRNA
Expression-GSK-3␤ and ␤-catenin are important components of the canonical Wnt signaling pathway. To examine the effect of 5-Aza on the canonical Wnt signaling pathway, protein expression of GSK-3 isoforms and ␤-catenin was evaluated in 5-Aza-treated (5 M for 24 h) and control MSCs. Expression of GSK-3␣ and GSK-3␤ was detectable but low. On the other hand, ␤-catenin was expressed abundantly in unstimulated MSCs. 5-Aza treatment increased expression of GSK-3␣ and GSK-3␤ in a time-dependent manner in MSCs, although induction of GSK-3␣ by 5-Aza was milder than that of GSK-3␤ (Fig. 2, A and B). Expression of GSK-3␤ at day 5 was significantly greater in 5-Aza-treated MSCs than in untreated MSCs (supplemental Fig. S1A). 5-Aza treatment did not induce up-regulation of GSK-3␤ in COS-7 cells, suggesting that the effect of 5-Aza is cell type-specific (supplemental Fig. S1B). Up-regulation of GSK-3␣ and GSK-3␤ by 5-Aza was also observed at the mRNA level (supplemental Fig. S1C). The promoter regions of GSK-3␣ and

Distinct Roles of GSK-3 Isoforms in Cardiac Differentiation
GSK-3␤ contain CpG islands (supplemental Fig. S2). The CpG islands are methylated in untreated MSCs, but are demethylated after 5-Aza treatment (Fig. 2C). Although protein expression of ␤-catenin gradually increased in control MSCs, a progressive decrease in ␤-catenin was observed in 5-Aza-treated MSCs (Fig. 2, A and D), accompanied by decreases in TCF/LEF transcriptional activity as determined by TOP flash/FOP flash reporter gene assays (Fig. 2E).
To achieve up-regulation of GSK-3␤ by an alternative method, MSCs derived from Tet-GSK-3␤ transgenic mice were transduced with Ad-tTA (the Tet-Off system) or Ad-rtTA (the Tet-On system), and then treated with or without Dox. GSK-3␤ expression was induced by withdrawing Dox in the Tet-Off system and by adding Dox in the Tet-On system (Fig. 4, A and  B). The effect of Dox upon transgene expression was reversible (Fig. 4C), suggesting that GSK-3␤ expression can be regulated by Dox treatment in these MSCs. In both the Tet-On and Tet-Off systems, up-regulation of GSK-3␤ induced expression of Nkx2.5 and ␣-MHC mRNA in MSCs (Fig. 4D). Alternatively, Tet inducible GSK-3␤ transgenic mice were crossed with transgenic mice harboring CMV-tTA, and then MSCs were prepared from the bone marrow of bigenic mice. In the absence of Dox, MSCs prepared from the bigenic mice expressed more GSK-3␤ and less ␤-catenin than MSCs from control mice (Fig.  4E). Culturing MSCs prepared from Tet-GSK-3␤ and CMV-tTA bigenic mice in Dox-free medium induced ␣-MHC expression, which was completely suppressed in the presence of Dox (Fig. 4F). These results support the notion that cardiomyocyte differentiation of MSCs can be stimulated by drug-regulatable up-regulation of GSK-3␤.
Because both canonical and non-canonical Wnt pathways induce differentiation of stem cells into the cardiomyocyte lineage, we compared the effect of GSK-3␤ upon MSC differentiation with that of Wnt agonists. Using adenovirus transduction, we overexpressed either Wnt11, an agonist for the non-canonical Wnt pathway, or Wnt3a, an agonist for the canonical Wnt pathway, in MSCs (Fig. 5, A and B). Wnt11 induced activation  DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52

JOURNAL OF BIOLOGICAL CHEMISTRY 36651
of protein kinase C and Wnt3a caused a significant accumulation of ␤-catenin, suggesting that the non-canonical and canonical Wnt pathways were stimulated in MSCs in our experimental conditions (Fig. 5C). Up-regulation of GSK-3␤ induced the frequent appearance of tubular structures in MSCs (Figs. 5D and supplemental S5). Tubular structures are induced when MSCs are differentiated into the cardiomyocyte lineage (20,27). Although up-regulation of GSK-3␣ and Wnt11 also induced some tubular structures, they were not as prominent as those induced by GSK-3␤. Tubular structures were not observed in MSCs treated with Ad-LacZ, Ad-Wnt3a, or Ad-DN-GSK-3␤ (Figs. 5D and supplemental S5). Wnt11 induced mRNA expression of ␣MHC, but little or no expression of Flk-1, Nkx2.5, or cTnC, in MSCs (Fig. 5E), whereas Wnt3a did not induce mRNA expression of any of these marker genes (Fig. 5F). These results suggest that GSK-3␤ induces cardiomyocyte differentiation of MSCs more potently than the Wnt agonists.

Expression of GSK-3 Is Required for Cardiomyocyte Differentiation in
MSCs-Because GSK-3 was up-regulated by 5-Aza and up-regulation of GSK-3␤ potently induced cardiomyocyte differentiation in MSCs, we next examined whether GSK-3 is required for induction of cardiomyocyte differentiation by 5-Aza in MSCs. MSCs were stimulated with 5-Aza in the presence or absence of known inhibitors of GSK-3. Treatment of MSCs with LiCl suppressed 5-Aza-induced down-regulation of ␤-catenin, suggesting that LiCl suppresses GSK-3 activity under 5-Aza treatment in MSCs (Fig. 6A). Although treatment of MSCs with LiCl induced expression of GATA4, LiCl alone did not induce expression of other cardiomyocyte marker genes. LiCl did, however, inhibit 5-Aza-induced up-regulation of cardiomyocyte marker mRNAs, including Nkx2.5, atrial natriuretic factor, cTnC, and cTnI (Fig. 6B), as well as cTnI protein expression (Fig. 6C). Treatment of MSCs with other GSK-3 inhibitors, including
Down-regulation of GSK-3a Induces Cardiomyocyte Differentiation of MSCs through Up-regulation of c-Jun-c-Jun plays an important role in mediating cardiomyocyte differentiation in bone marrow mononuclear cells (9). Down-regulation of GSK-3␣, but not GSK-3␤, up-regulated c-Jun expression in the nucleus in MSCs (Figs. 10, A and  B, and supplemental S6D). Transduction with adenovirus harboring shRNA-c-Jun reversed the up-regulation of c-Jun, and mesoderm and cardiomyocyte markers induced by down-regulation of GSK-3␣ in MSCs (Fig. 10, C and D), suggesting that c-Jun plays an important role in mediating cardiomyocyte differentiation of MSCs induced by down-regulation of GSK-3␣. Down-regulation of c-Jun up-regulated mRNA expression of nestin, suggesting that c-Jun negatively regulates neuronal differentiation in MSCs (Fig. 10E).

DISCUSSION
Expression of GSK-3 remains low when MSCs are uncommitted. However, GSK-3␤ is up-regulated when cardiomyocyte differentiation of MSCs is initiated by 5-Aza. Furthermore, upregulation of GSK-3␤ is both necessary and sufficient for cardiomyocyte differentiation initiated by 5-Aza in MSCs. Unexpectedly, down-regulation, rather than up-regulation, of endogenous GSK-3␣ stimulated cardiomyocyte differentiation in MSCs. These results suggest that GSK-3␤ is an endogenous regulator of MSC differentiation and that GSK-3␣ and GSK-3␤ have opposite effects upon cardiomyocyte differentiation in MSCs.
5-Aza is a cytosine analogue and a demethylating agent that induces changes in chromatin structure, gene expression, cellular morphology, and survival in mammalian cells. Because phosphorylation of GSK-3␤ is not significantly affected by 5-Aza, 5-Aza must increase the total activity of GSK-3␤ primarily through up-regulation of GSK-3␤ mRNA. The promoter region of GSK-3␤ contains a prominent CpG island that is methylated in unstimulated MSCs, suggesting that 5-Aza induces up-regulation of GSK-3␤ through epigenetic modification of the promoter. Because GSK-3␤ stabilizes DNA methyltransferases, 5-Aza may initiate a positive feedback loop of GSK-3␤ promoter demethylation (28). At present, whether or not demethylation of the GSK-3␤ promoter is an endogenous mechanism for differentiation of adult stem cells into the cardiomyocyte lineage remains to be elucidated. In any event, GSK-3␤ may substitute for 5-Aza for induction of cardiomyocyte differentiation in MSCs, because clinical use of 5-Aza would be limited due to its nonspecific effects and obvious teratogenic actions.
GSK-3␤ is a central component of the Wnt pathway and negatively regulates ␤-catenin through phosphorylationdependent proteolytic degradation (29). Although previous studies have shown that stimulation and inhibition of the Wnt signaling mechanism affect cardiomyocyte differentiation (8), up-regulation of GSK-3␤ induced cardiomyocyte markers more strongly than stimulation of either the canonical or noncanonical Wnt signaling pathways with Wnt3a and Wnt11, respectively. Because down-regulation of ␤-catenin alone also potently induces cardiomyocyte markers, modulating downstream components of the Wnt signaling pathway may induce cardiomyocyte differentiation more efficiently than stimulating the Wnt pathways at the receptor level.
It should be noted that GSK-3␤ not only regulates the Wnt pathway but also modulates a wide variety of signaling pathways, including other signaling cascades known to regulate stem cell differentiation, such as the Notch (30) and Hedgehog (31) pathways. Thus, up-regulation of GSK-3␤ may have a broader effect than selective stimulation of the Wnt pathway by the Wnt receptor ligand.
Increasing lines of evidence suggest that GSK-3␣ and GSK-3␤ have distinct cellular functions, despite the fact that they share 97% identity in their kinase domains and 36% identity overall. Our results suggest that GSK-3␣ and GSK-3␤ have distinct effects upon cardiomyocyte differentiation. Up-regulation of GSK-3␤ has a stronger effect upon cardiomyocyte differentiation in MSCs than up-regulation of GSK-3␣. Although overexpression of GSK-3␣ slightly induces expression of cardiomyocyte markers, this effect may be mediated through promiscuous phosphorylation of GSK-3␤ substrates due to overexpression. In fact, adenovirus-mediated overexpression of GSK-3␣ substantially altered subcellular localization of GSK-3␣ in MSCs (see below). Importantly, although down-regulation of GSK-3␤ inhibited 5-Aza-induced cardiomyocyte differentiation, down-regulation of GSK-3␣ stimulated cardiomyocyte differentiation in MSCs. Previous studies have suggested that GSK-3␣ and GSK-3␤ could differentially affect cardiac development. Although GSK-3␤ knock-out mice exhibit cardiac defects consisting of malformation of the cardiac outflow tract and markedly thickened ventricular walls, contributing to their early mortality, GSK-3␣ knock-out mice show no significant cardiac defects (18). However, to our knowledge, the fact that GSK-3␣ and GSK-3␤ have opposite effects upon differentiation of stem cells has not been shown previously.
One possible explanation for the potential difference in their functions is that GSK-3␣ and GSK-3␤ exist in distinct subcel-lular localizations. For example, in adult mouse hearts, GSK-3␣ is localized primarily in the nucleus, whereas GSK-3␤ exists primarily in the cytosol (17). GSK-3␣ localized in the nucleus phosphorylates and induces nuclear exit/proteolytic degradation of G 1 cyclins, whereas endogenous GSK-3␤ localized in the cytosol does not induce nuclear exit of G 1 cyclins in the mouse heart (17). Immunostaining of the GSK-3 isoforms suggest that GSK-3␣ is localized primarily in the nucleus, and GSK-3␤ pri-marily in the cytosol but also in the nucleus in MSCs. Forced expression of GSK-3␤ in the nucleus did not induce cardiomyocyte differentiation, whereas overexpression of GSK-3␣ induces cytosolic expression of GSK-3␣ and partially stimulates cardiomyocyte differentiation. We therefore speculate that endogenous GSK-3␤ more efficiently phosphorylates ␤-catenin in the cytosol, thereby inducing efficient proteolytic degradation, whereas endogenous GSK-3␣ primarily localized in the nucleus may regulate other targets, presumably transcription factors. Because GSK-3␣ and GSK-3␤ equally affect expression of ␤-catenin in some cell types, such as ES cells (19,32), we speculate that subcellular localization of GSK-3␣/␤ may be cell type-or developmental stage-dependent.
Importantly, up-regulation of GSK-3␤ and down-regulation of GSK-3␣ have additive effects upon cardiomyocyte differentiation in MSCs, consistent with the notion that they mediate cardiomyocyte differentiation through distinct cellular mechanisms. Our results suggest that down-regulation of ␤-catenin plays an important role in mediating the effect of GSK-3␤ upon MSC differentiation into the cardiomyocyte lineage. On the other hand, down-regulation of GSK-3␣ induces cardiomyocyte differentiation through up-regulation of c-Jun in MSCs. Because GSK-3 phosphorylates ␤-catenin and c-Jun (19,33), thereby stimulating their degradation, it is likely that GSK-3␤ in the cytosol may phosphorylate ␤-catenin, whereas GSK-3␣ in the nucleus may phosphorylate c-Jun, thereby regulating cardiomyocyte differentiation in MSCs (supplemental Fig. S7). Interventions to selectively stimulate GSK-3␤ or inhibit GSK-3␣ may be considered independently or in combination with other methods to facilitate cardiomyocyte differentiation of MSCs for cell-based therapy in vivo.
We have successfully engineered MSCs that conditionally express GSK-3␤ through either withdrawal or application of Dox. Phasic modulation of the Wnt/␤-catenin signaling mechanism effectively stimulates differentiation of progenitor cells into the cardiomyocyte lineage (10, 34). Activation of ␤-catenin is required to maintain and expand cardiac progenitor cells (11,35) but must be repressed to induce cardiomyocyte differentiation from cardiac progenitor cells (8). Thus, it would be interesting to test whether ex vivo engineered MSCs, in which the timing of expression of GSK-3␤ can be regulated by Dox treatment, enhance the efficacy of cell therapy in vivo.