Glycogen synthase kinase-3beta regulates growth, calcium homeostasis, and diastolic function in the heart.

Glycogen synthase kinase (GSK) 3beta is a negative regulator of stress-induced cardiomyocyte hypertrophy. It is not clear, however, if GSK-3beta plays any role in regulating normal cardiac growth and cardiac function. Herein we report that a transgenic mouse expressing wild type GSK-3beta in the heart has a dramatic impairment of normal post-natal cardiomyocyte growth as well as markedly abnormal cardiac contractile function. The most striking phenotype, however, is grossly impaired diastolic relaxation, which leads to increased filling pressures of the left ventricle and massive atrial enlargement. This is due to profoundly abnormal calcium handling, leading to an inability to normalize cytosolic [Ca2+] in diastole. The alterations in calcium handling are due at least in part to direct down-regulation of the sarcoplasmic reticulum calcium ATPase (SERCA2a) by GSK-3beta, acting at the level of the SERCA2 promoter. These studies identify GSK-3beta as a regulator of normal growth of the heart and are the first of which we are aware, to demonstrate regulation of expression of SERCA2a, a critical determinant of diastolic function, by a cytosolic signaling pathway, the activity of which is dynamically modulated. De-regulation of GSK-3beta leads to severe systolic and diastolic dysfunction and progressive heart failure. Because down-regulation of SERCA2a plays a central role in the diastolic and systolic dysfunction of patients with heart failure, these findings have potential implications for the therapy of this disorder.

Glycogen synthase kinase-3 (GSK-3) 1 plays dual roles, both as a mediator of downstream effects of growth factor signaling and as a component of the Wnt/Wingless signaling pathway, which regulates dorsoventral patterning (1,2). Normally GSK-3 is active in un-stimulated cells and is inhibited when cells are exposed to either growth factors or to certain members of the Wnt family of secreted glycoproteins. GSK-3 can be inhibited by at least two mechanisms. In response to growth factors, generation of phosphoinositide 3-phosphates leads to the activation of Akt, also known as protein kinase B, which phosphorylates GSK-3 ␣ or ␤ on amino-terminal serine residues, Ser-21 for ␣ and Ser-9 for ␤ (3,4). This inhibits kinase activity directed toward so-called primed substrates, which have been previously phosphorylated at a residue several amino acids carboxyl-terminal to the GSK-3 site (5,6). Alternatively, GSK-3 can be inhibited by binding to Frat-1 (frequently re-arranged in T cell lymphomas-1; also known as GSK-3-binding protein or GBP) (7,8). This may sequester GSK-3 from its substrates. In addition to growth factors and Wnts, GSK-3 was recently found to be inhibited by several pathologic stimuli, including endothelin-1, Fas ligand, and ␣and ␤-adrenergic agents, all of which induce hypertrophic responses in cardiomyocytes in culture (9 -12). Furthermore, inhibition of GSK-3 was reported to be necessary for the hypertrophic response of cardiomyocytes to these pathologic stimuli since expression of GSK-3␤ with a Ser-9 to Ala mutation, preventing its inactivation by Ser-9 phosphorylation, abrogated the hypertrophic response (9 -11).
Postnatally, normal growth of the heart occurs almost exclusively via hypertrophic growth of cardiomyocytes, terminally differentiated cells that are unable to proliferate. Signals for normal growth of the heart require a signaling pathway involving phosphoinositide 3-kinase and phosphoinositide-dependent kinase-1 (13,14). Directly downstream of phosphoinositide-dependent kinase-1 a wealth of evidence points to Akt as a critical regulator of growth (15)(16)(17). Below phosphoinositide-dependent kinase-1 and Akt, the pathway bifurcates into one leading to the mammalian target of rapamycin (mTOR) and p70 ribosomal S6 protein kinase (p70S6K) and another leading to GSK-3. mTOR and p70S6K play major roles in regulating protein translation, and evidence from studies in a number of species implicate mTOR and p70S6K as central regulators of cell growth, including growth of cardiomyocytes (18,19). In contrast, it is not clear that the GSK-3 arm plays any role in normal growth or function of the heart in vivo (20). To address this question, we studied the effect of expressing GSK-3␤ specifically in the hearts of transgenic mice. Our findings indicate that GSK-3␤ negatively regulates physiologic concentric hypertrophy (normal growth) of ventricular cardiomyocytes, leading to a small heart with depressed contractility. Surprisingly, we also find that GSK-3␤ is a critical regulator of calcium handling in the heart. This impaired cardiomyocyte growth and altered calcium handling leads to severe systolic and diastolic dysfunction and heart failure.

MATERIALS AND METHODS
Creation of the GSK-3␤ Transgenic Mouse-The cDNA encoding wild type GSK-3␤ with a carboxyl-terminal hemagglutinin epitope tag was sub-cloned into the 5.5-kilobase murine ␣-myosin heavy chain promoter expression vector (a gift from Jeffrey Robbins, Children's Hospital, Cincinnati, OH). The injection fragment was gel-purified, reconstituted in 2 mM Tris-HCl, pH 7.5, 20 M EDTA, and injected into fertilized eggs from FVB/N mice. The eggs were then transferred to the oviducts of pseudo-pregnant females. Founders were then crossed into C57Bl6/J for at least five generations before characterization.
The immunoblotting protocol was modified slightly for the detection of SERCA2a. In brief, the lysate was heated in 1ϫ SDS sample buffer to only 55°C before running on SDS-PAGE, the primary antibody incubation was at 4°C for 14 h, and the secondary antibody incubation was for 3 h at room temperature. For all other immunoblots, lysates were heated in 1ϫ SDS sample buffer to 100°C, run on SDS-PAGE, and later incubated with primary antibodies overnight and with the appropriate secondary antibody for 1 h at room temperature. Antibody binding was detected with a peroxidase-conjugated goat anti-rabbit or anti-mouse IgG (DAKO) and chemiluminescence (Amersham Biosciences). Quantification of expression levels of various proteins was done by determining band density with a gel imaging system from Alpha Innotech. Expression of either calsequestrin or glyceraldehyde-3-phosphate dehydrogenase were used to normalize for protein loading.
Northern Blotting-Total RNA was purified from mouse tissues and NRVMs in-culture using the RNA isolation kit (Bio-Rad)). A 5.0-g aliquot of total RNA was electrophoresed in 1.2% denaturing formaldehyde agarose gels and blotted onto Hybond N (Amersham Biosciences). The radiolabeled probes were hybridized at 42°C in a hybridization solution (50% deionized formamide, 6ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate), 5ϫ Denhardt's solution, 0.5% SDS, 200 g/ml denatured salmon sperm DNA). Blots were washed twice at room temperature in 2ϫ SSC, 1% SDS for 5 min and twice at 50°C in 2ϫ SSC for 30 min. The probe for mouse SERCA2a was a kind gift from Dr. Roger Hajjar (Massachusetts General Hospital). For quantification, SERCA2a counts, obtained by phosphorimaging, were normalized to counts from 18 S ribosomal RNA.
Cytosolic Fractionation-Cytosolic fractions were prepared by hypotonic lysis as described (12). Immunoblotting with an anti-caveolin antibody (Cell Signaling) has previously confirmed that there is no contamination of cytosolic fractions with membrane components (12).
Pharmacological Studies-NRVMs were serum-starved for 24 h and then treated with the GSK-3 inhibitor I (Calbiochem) at the concentrations indicated for a further 24-h period.
Assay of GSK-3␤ Kinase Activity-Kinase assays were performed on lysates made from hearts that had been snap-frozen in liquid N 2 and then pulverized under liquid N 2 as described (10,21). To immunoprecipitate GSK-3␤, 3 g of antibody were added to the lysates for 2 h. Immune complexes were collected with protein G-agarose beads. After extensive washing, [␥-32 P]ATP (final concentration 100 M; ϳ3000 cpm/ pmol) and phospho-glycogen synthase peptide (25 M; Upstate Biotechnology Inc.) were added to the beads. After 20 min at 30°C, reactants were spotted onto P81 phosphocellulose squares, washed extensively, dried, and subjected to Cerenkov counting.
Echocardiographic Examination-Animals were lightly anesthetized with isoflurane (1.5%). Echocardiography was performed, and parameters were determined using standard techniques as described (23). For determination of wall stress using LaPlace's law [(systolic blood pressure ϫ radius)/(2ϫ posterior wall thickness)] animals were imaged in the awake, non-sedated state. Blood pressures were determined on three serial determinations of n ϭ 10 measurements each, 24 h after acclimatization to the Visitech Monitoring System (Durham, NC). Values were averaged over each set of measurements, and then a mean value for each animal was determined by averaging the mean values for the three sets of measurements.
Morphometric Measures-For determination of myocardial mass, animals were anesthetized and then were sacrificed with an injection of cardioplegia solution (at 4°C). The hearts were excised, and heart weight was determined. The LV and atria were then trimmed away from the right ventricle and weighed. Length of the right tibia was also determined. Heart weight, LV weight, and atrial weight were expressed relative to body weight and to tibial length.
Hemodynamic Assessment of the GSK-3␤ Transgenic-Animals were anesthetized with isoflurane (2.0%). The right carotid artery was exposed and a 1.4-French Millar catheter was advanced into the left ventricle. Left ventricular pressures were recorded and LV systolic pressure, LV end-diastolic pressure, ϩdP/dt, and ϪdP/dt were quantified using MacLab software.

Determination of Systolic Shortening and [Ca 2ϩ ] in Isolated
Cardiomyocytes-Adult mouse myocytes were loaded with fura-2/AM (Molecular Probes) as previously described (24). Fura-2-loaded myocytes were superfused with 1.8 mM Ca 2ϩ Tyrode solution at 37°C. Myocytes were subsequently stimulated at 5 Hz, and after stabilization, cell contractility and intracellular calcium transients were simultaneously measured using edge detection and fura-2 fluorescence respectively (IonOptix, Inc.) (24). A subset of myocytes from both groups was utilized to calibrate the fura-2 fluorescence ratio to [Ca 2ϩ ] i values in situ as described (25).
Determination of SERCA2 Promoter Activity-The Ϫ1 to Ϫ1824 SERCA/TK-CAT (MS-I CAT) construct was produced through ligation of a 1824-bp fragment of the SERCA2 promoter to the heterologous promoter (TK) linked to the CAT reporter gene in pBLCAT5 vector (26). pBL0-CAT (containing the minimal TK promoter linked to the CAT reporter gene) and empty plasmid constructs were used as negative controls. Human embryonic kidney 293 cells were transfected using FuGENE 6 (Roche Applied Science) with 5.0 g of plasmid per 1 ϫ 10 5 cells for 48 h in Dulbecco's modified Eagle's medium containing 1% fetal calf serum. Cells were co-transfected with 5.0 g of pcDNA3 or 5.0 g of pcDNA3 encoding either GSK-3␤ or GSK-3␤(S9A). Cells were also transfected with pcDNA3 encoding ␤-galactosidase to normalize CAT activity for transfection efficiency. CAT activity was determined using a commercially available assay system (Promega).

RESULTS
Expression and Activity of GSK-3␤-Four transgenic founders were produced. Two founders, who expressed the transgene at the highest levels, died within 6 weeks of birth from apparent heart failure. Two survived to breed successfully, and these lines, termed #57 and #44, are the basis of this report. We found that expression of endogenous GSK-3␤ was reduced in the transgenic and that, overall, the transgenic animal overexpressed GSK-3␤ (endogenous plus transgene product) by ϳ9-fold in the left ventricle of line 57 (Fig. 1A) and 12-fold in line 44 (not shown) compared with wild type littermates. The data to be presented will be from line 57, the lower expresser, unless otherwise noted.
Because the activity of the transgene product could be inhibited by phosphorylation at Ser-9, we examined the level of phosphorylation of GSK-3␤ on Ser-9. We found that phosphorylation of GSK-3␤ (endogenous and transgene product) on Ser-9 was very pronounced in the transgenic (Fig. 1A). Consistent with this, we found that although expression was ϳ9-fold greater in the transgenic compared with wild type, kinase activity in heart lysates from the transgenic animals, utilizing phospho-glycogen synthase as the substrate, was only 3.5-fold greater than wild type (Fig. 1B), compatible with extensive inhibition of the kinase via Ser-9 phosphorylation. These studies taken together demonstrate moderate overexpression but only modest activation of GSK-3␤ in the heart of the transgenic mouse.
Phenotype of the GSK-3␤ Transgenic Mouse-The results of the studies outlined above suggested that the GSK-3␤ transgenics should, if this pathway negatively regulates normal growth of the heart, have impaired growth. Indeed, we found a dramatic reduction in normal growth of the heart. The distinctive phenotype of both lines 44 and 57 was that of small, thin-walled left and right ventricles with foreshortened long axes ( Fig. 2A). Quantitation of cardiac mass by morphometry confirmed reduced growth in both male and female transgenic animals at 2 and 5 months (Tables I and II).
Echocardiographic data obtained in both transgenic lines were entirely consistent with the post-mortem measures above (Table III). At all time points, the end-diastolic dimension of the LV was similar between transgenic and wild type, but the anterior wall thickness and long axis length of the LV were significantly decreased in the transgenics. The disparity in wall thickness between transgenic and wild type continued to worsen with time (Table III) The gradual worsening of the disparity in wall thickness with advancing age confirmed that a significant part of the difference in LV mass was due to impaired post-natal growth. However, although there were major reductions in post-natal growth of the transgenic heart, we also wanted to determine whether cardiac development was significantly reduced in the transgenic since although the ␣-myosin heavy chain promoter activity increases markedly after birth, there is some expression in utero (27). To address this we examined heart size in transgenic and wild type neonatal mice. We found much more subtle and less consistent reductions in heart size in the transgenic neonates compared with wild type even when comparing neonates from line 44 to the wild type (Fig. 2B). Thus, although there may have been some reductions in cardiac growth in utero, these data, taken together with the post-mortem and echocardiographic measures that showed a continued and worsening disparity in growth throughout the first 5 months of life, suggest that the majority of the growth impairment was accounted for by differences after birth.
The similarity in heart size at birth between transgenic and wild type strongly suggested that the failure of normal growth of the GSK-3␤ transgenic heart was due to a defect in postnatal growth, which is predominantly hypertrophic growth, rather than to a reduction in cardiomyocyte number (i.e. a failure of cardiomyocyte proliferation in utero). To confirm this we determined cell diameter and cell length in cardiomyocytes isolated from the hearts of 2-month-old transgenic and wild type littermates. Cardiomyocyte diameter was markedly reduced in the transgenic, but cardiomyocyte length was normal (Fig. 2C). These data confirm that the growth defect is due at least in part (and probably in large part) to a failure of the cardiomyocytes to undergo normal concentric hypertrophy.
We next asked how GSK-3␤ might regulate cardiomyocyte growth. We first examined the activation state of putative positive growth regulatory pathways in the heart of the GSK-3␤ transgenic to determine whether down-regulation of these played any role in the impaired growth phenotype of the transgenic. One such pathway is that resulting in activation of Akt. At 2 months of age there was no difference between transgenic and wild type in the amount of phosphorylation of Akt on either of the activating phosphorylation sites, Ser-473 (Fig. 2D) or Thr-308 (not shown). At 5 months of age, Akt was more activated in the transgenic hearts compared with wild type, as evidenced by increased phosphorylation on Ser-473 (Fig. 2D). Activation of the ERK pathway, another positive regulator of growth, was also increased in the 5-month-old transgenics, although again, ERK1/2 phosphorylation was similar between The transgene product (HA-GSK-3␤) runs at a slightly higher molecular weight than endogenous GSK-3␤ due to the hemagglutinin (HA) epitope tag. B, GSK-3␤ kinase activity in the transgenic mouse heart. Kinase activity is expressed as fold increase over kinase activity in wild type hearts (n ϭ 3-4 hearts per condition). younger transgenic and wild type mice (Fig. 2D). In addition, the activity of c-Jun NH 2 -terminal kinases and p38 mitogenactivated protein kinases, which have been reported to be both positive and negative regulators of cardiac growth (21, 28 -32), was not different between the transgenic and wild type mice at either time point (data not shown). These data suggest that the reduced cardiac growth of the GSK-3␤ transgenic was not due to reduced activation of either of two pathways that are known to positively regulate cardiomyocyte growth (Akt and ERK1/ 2)(15-17,33) or to altered activity of the c-Jun NH 2 -terminal kinase or p38 mitogen-activated protein kinase pathways.
How then might GSK-3␤ negatively regulate cardiomyocyte growth? One target of GSK-3␤ is the nuclear factor of activated T cells (NF-AT) family of transcription factors. Although NF-ATs clearly play an important role in pathologic stress-induced growth, they are not believed to be major regulators of physiological hypertrophy (34). GSK-3␤ also negatively regulates the translation initiation factor, eIF2B, and therefore, we asked whether part of the growth reduction might be secondary to this. We found that phosphorylation of eIF2B⑀ was significantly enhanced in the GSK-3␤ transgenic (Fig. 2E). Because this phosphorylation leads to inhibition of protein translation Bottom, quantification of cardiomyocyte width (left) and length (right) in wild type versus transgenic littermates. *, p Ͻ 0.01 versus wild type. †, p Ͻ 0.05 versus wild type. D, activation state of Akt and ERKs in the GSK-3␤ transgenic mouse heart. Lysates from hearts of transgenic and wild type littermates at 2 and 5 months of age were immunoblotted for phospho (p)-Akt (Ser-473) and p-ERK1/2 or total Akt and ERK1/2. E, phosphorylation of eIF2B⑀ is increased in the GSK-3␤ transgenic heart. Upper panel, whole heart lysates from 5-week-old littermates were immunoprecipitated (IP) with anti-eIF2B⑀ and immunoblotted with anti-phospho-specific-eIF2B⑀. Lower panel, whole heart tissue lysates were also immunoblotted with anti-eIF2B⑀ to ensure equivalent protein expression and loading. F, cytosolic levels of the GSK-3␤ target, ␤-catenin, are reduced in the transgenic mouse heart. Cytosolic lysates from hearts of wild type, and transgenic littermates were immunoblotted with an anti-␤-catenin antibody (immunoblotting with anti-14-3-3 antibody showed equivalent protein loading, data not shown). RA, right atrium; LA, left atrium. (22), this can be expected to be one mechanism by which cardiac growth is reduced in the GSK-3␤ transgenic.
GSK-3␤ also negatively regulates the transcriptional co-activator, ␤-catenin, in Wnt signaling. More recently, inhibition of GSK-3␤ by hypertrophic stimuli has been reported to lead to an increase in ␤-catenin levels and transcriptional activating activity, and this has been implicated in the positive regulation of cardiomyocyte growth both in cultured cells and in vivo (12).
␤-Catenin is targeted for ubiquitination and subsequent degradation by the proteasome when phosphorylated by GSK-3␤. This blocks induction of various ␤-catenin-dependent genes, one of which, c-Myc, is also a positive regulator of cardiac growth (35). We found that ␤-catenin was de-stabilized in the transgenic heart since cytosolic levels of ␤-catenin were significantly reduced in the transgenic compared with wild type (Fig.  2F). Because non-myocytes make up a substantial percent of

TABLE III Echocardiographic data in GSK-3␤ transgenic (TG) vs. wild type (WT)
A mixed population of littermates (males and females of equivalent number) were used in each of the groups. EDD, end-diastolic dimension of the LV; ESD, end-systolic dimension of the LV; LAX, long axis length of the LV; AWT, anterior wall thickness of the LV; FS, fractional shortening; HR, heart rate; bpm, beats per minute.

and months
A mixed population of littermates (males and females of equivalent number) were used in each of the groups. The data are derived from n ϭ 8 animals/group. There were no significant differences in hemodynamics between males and females. Data are the means Ϯ S.E. SBP, systolic blood pressure; HR, heart rate; DBP, diastolic blood pressure; LVEDP, left ventricular end-diastolic pressure; ϩdr/dt, rate of rise of pressure in the LV; ϪdP/dt, rate of fall of pressure in the LV; bpm, beats per minute.

GSK-3␤ and Cardiac Growth
the left ventricular mass and the transgene is not expressed in these cell types, the magnitude of de-stabilization of ␤-catenin is likely underestimated by these studies using whole heart cytosolic fractions. In summary, GSK-3␤ is a profound negative regulator of post-natal cardiomyocyte growth. Furthermore, our findings suggest that this effect is likely mediated by GSK-3␤-induced negative regulation of protein translation factors and of transcription factors, including ␤-catenin and its targets, which are positive regulators of growth.
Abnormalities of Left Ventricular Systolic Function in the GSK-3␤ Transgenic-The second phenotype of the GSK-3␤ transgenic was impaired systolic function of the left ventricle. Echocardiographic fractional shortening, a measure of global LV function, was significantly impaired at 1 month and continued to deteriorate over time (Table III). These findings were supported by hemodynamic studies. Although heart rates were similar between transgenic and wild type, systemic blood pressures were significantly reduced in the transgenic (Table IV). ϩdP/dt, another measure of contractility, was also markedly reduced in the transgenic (Table IV). To determine whether the impairment in global LV contractile function was due to a primary defect in myocyte contractile function, we measured contractile function of myocytes isolated from the LV of transgenic and wild type littermates. Transgenic myocytes had normal contractile function when studied in an unloaded state, as demonstrated by normal shortening (Fig. 3).
LV systolic wall stress, the load placed on the LV as it contracts, is directly proportional to the radius of the LV cavity and inversely proportional to the wall thickness of the LV. Thus, we asked whether the impaired global contractile function could be due to a significant increase in wall stress secondary to the inability of the LV to normally hypertrophy and the consequent markedly reduced wall thickness. Because anesthesia can itself alter blood pressure and contractility, we determined wall stress in conscious, non-sedated transgenic and wild type mice. We found that wall stress was significantly increased in the transgenics, especially when one used endsystolic echocardiographic parameters in LaPlace's equation to estimate wall stress (Table V). Thus, increased wall stress accounts for at least part of the depressed contractile function of the left ventricle of the transgenic.
Abnormalities of Diastolic Function and Ca 2ϩ Handling in the GSK-3␤ Transgenic-The most striking cardiac phenotype of the GSK-3␤ transgenic mouse was massively enlarged atria (2-3 times greater atrial weight/tibial weight) in both male and female transgenics (Tables I and II). LVEDP was 2.2-3.9-fold higher in the transgenics, reaching values that are very elevated for the mouse heart (Table IV). This was not due to fibrotic changes in the left ventricle since staining with Masson's trichrome revealed no increase in fibrosis in the transgenic (not shown).
LVEDP, when chronically elevated, can be expected to lead to atrial dilatation. These data suggested that the ability of the transgenic heart to relax normally in diastole might be impaired. Consistent with this, ϪdP/dt, a hemodynamic measure of the ability of the ventricle to relax, was dramatically impaired (Table IV). Furthermore, in myocytes isolated from the hearts of 2-month-old transgenics, the time constant of relaxation, , and the time to 90% relaxation were markedly impaired (Fig. 4A).
Efficient re-uptake of calcium in diastole is essential for normal relaxation of cardiomyocytes. Therefore, we examined calcium handling in cardiomyocytes isolated from 2-month-old transgenic and wild type animals. These studies demonstrated that cytosolic [Ca 2ϩ ] in diastole was significantly elevated in the transgenic myocytes, and the rate of decline of [Ca 2ϩ ] in diastole ([Ca 2ϩ ]) was significantly slower than wild type (Fig.  4B). This suggests that the impaired diastolic relaxation of the transgenic was likely due to impaired re-uptake of Ca 2ϩ during diastole.
SERCA2a, which transports Ca 2ϩ into the sarcoplasmic reticulum, is primarily responsible for restoring cytosolic [Ca 2ϩ ] to normal in diastole (36). Therefore, we examined SERCA2a expression in hearts of transgenic and wild type littermates. These studies demonstrated that SERCA2a protein expression was significantly down-regulated in the transgenic (Fig. 5A). Although the difference in expression level between wild type and transgenic increased over time, a highly significant difference was present as early as 5 weeks of age (the earliest time point examined; Fig. 5A). In contrast, expression of the regulator of SERCA2a, phospholamban, was not altered, and phos-

FIG. 3. Quantification of contractility of adult mouse cardiomyocytes isolated from wild type versus transgenic littermates.
Cardiomyocytes were isolated from the hearts of 2-month-old wild type (WT) and transgenic (TG) mice, and then the percent cell shortening (%CS) was determined as described under "Materials and Methods." phorylation of phospholamban at Ser-16, which inactivates phospholamban, was only modestly reduced (data not shown). Thus, the marked down-regulation of SERCA2a likely accounts in large part for the grossly abnormal Ca 2ϩ handling in the transgenic myocytes.
When SERCA2a expression/activity is depressed, SR calcium stores are generally depleted, and calcium release in systole is impaired. However, the GSK-3␤ transgenic had normal peak systolic [Ca 2ϩ ] (Fig. 4B). This paradox may be explained by significant up-regulation of the sarcolemmal sodium calcium exchanger, acting in reverse mode, in the transgenic (Fig. 5B) (37, 38). Of note, expression of Na ϩ -Ca 2ϩ exchanger is inversely correlated with SERCA2a expression in failing human hearts and in various transgenic models (39,40).
The above abnormalities of calcium-handling proteins could simply be a compensation for the heart failure that the transgenic mice eventually develop. However, the abnormalities in Ca 2ϩ handling occurred very early in the lives of the transgenics and were generally more marked and led to much more profound alterations in diastolic function than that seen in various models of heart failure. Therefore, we asked whether GSK-3␤ might be directly responsible for any of the abnormalities in calcium handling proteins. We first found that the reduced SERCA2a expression in the transgenic heart was also apparent at the mRNA level (Fig. 5C). We then explored the effect of adenovirus-mediated gene transfer of GSK-3␤ on SERCA2a protein and mRNA levels in rat neonatal cardiomyocytes. We found that gene transfer of GSK-3␤(SA9), contain- ing a Ser-9 to Ala mutation that prevents inactivation of GSK-3␤ by serine 9 phosphorylation, significantly reduced SERCA2a protein expression (Fig. 5D). Similarly, at the mRNA level, gene transfer of GSK-3␤(SA9) reduced SERCA2a expression (Fig. 5D). We quantified the change in SERCA2a mRNA expression in NRVMs transduced with either GFP or GSK-3␤(SA9) and observed significant down-regulation of SERCA2a mRNA expression in the GSK-3␤(SA9)-transduced group. We next employed the GSK-3␤ inhibitor, LiCl, to determine whether we could reverse the inhibition of SERCA2a expression induced by gene transfer of GSK-3␤(SA9), and thus, exclude a nonspecific toxic effect of GSK-3␤(SA9) as the mechanism of the reduction in SERCA2a expression. However, LiCl itself had an inhibitory effect on SERCA2a expression as evidenced by its inhibition of SERCA2a in AdGFP-transduced cells. This effect of LiCl was not specific to SERCA2a since activity of various reporter constructs that did not contain the SERCA2 promoter (Rous sarcoma virus promoter-driven and minimal TK promoter-driven CAT constructs; data not shown) was also reduced. That said, LiCl did completely reverse the inhibition of SERCA2a mRNA expression by GSK-3␤(SA9), returning SERCA2a expression to levels equivalent to that of AdGFP-transduced cells treated with LiCl (Fig. 5D).
To define whether GSK-3␤ might directly regulate the SERCA2 promoter, we determined the effects of expressing GSK-3␤(SA9) on the activity of a SERCA2 promoter construct (pBLMS-CAT) in human embryonic kidney 293 cells (Fig. 5E). We observed an 82% inhibition of CAT activity in cells transduced with the GSK-3␤(SA9) construct relative to vector/ pBLMS-CAT controls. Next, we confirmed that the GSK-3␤ wild type construct produced an equivalent degree of inhibition (Fig. 5E). These studies demonstrated significant inhibition of SERCA2 promoter activity by GSK-3␤ and GSK-3␤(SA9). In contrast, expression of either of the kinases had no effect on activity of the Rous sarcoma virus promoter-driven CAT construct or pBL0-CAT controls, suggesting the inhibition was not due to toxicity, to an effect on the minimal TK promoter, or to a general inhibitory effect on transcription or translation (i.e. GSK-3␤-mediated inhibition of eIF2B) (22,41).
The above studies relied upon adenovirus-mediated gene transfer or transfection of GSK-3␤, and therefore, raise the possibility that our findings were an artifact of overexpression. To address this, we employed a small molecule inhibitor of GSK-3 (see "Materials and Methods") to determine whether inhibition of GSK-3 alone was sufficient to increase SERCA2a expression. First we confirmed that the inhibitor functioned as expected to inhibit GSK-3␤ as evidenced by its ability to stabilize ␤-catenin (Fig. 5F). The inhibitor, at concentrations that stabilized ␤-catenin, produced a dose-dependent increase in SERCA2a expression (Fig. 5F). These findings taken together with the studies employing gene transfer and the promoter studies discussed above suggest that GSK-3␤ can tonically inhibit SERCA2a expression and that inhibition of endogenous GSK-3␤ is sufficient to increase levels of SERCA2a. DISCUSSION The central findings of this paper are that GSK-3␤ regulates 1) normal growth of ventricular cardiomyocytes, 2) contractile function of the heart, and 3) calcium handling and diastolic function of the heart.
Role of GSK-3␤ in Regulating Normal Cardiac Growth-At birth differences in heart size between transgenic and wild type were not striking, but by as early as 1 month of age significant impairments in cardiac growth were apparent by echocardiography (Table III). The most striking abnormality was impairment in the ability of the heart to undergo concentric hypertrophy, resulting in marked reductions in cardiomyocyte width and overall ventricular wall thickness.
How does the growth impairment of the GSK-3␤ transgenic mouse compare with the growth impairment in other transgenic models expressing negative regulators of growth? The left ventricular weight/tibial length of the transgenic males was ϳ30% less than wild type. In contrast, whereas overexpression of the negative regulator of G-protein signaling, RGS4, or the dominant inhibitory mutant of G q signaling (the G q inhibitory peptide) reduced pressure overload-induced hypertrophy, they had no effect on normal growth (42,43). Similarly, a mouse overexpressing a dominant inhibitory mutant of phosphoinositide 3-kinase, which plays a critical role in organ growth, had only a 15% reduction in normal growth despite substantial (77%) reductions in basal phosphoinositide 3-kinase activity and even more marked inhibition of Akt/protein kinase B activity (14). Finally, overexpression of the endogenous calcineurin inhibitor, MCIP1, led to a 5-10% reduction in heart size (44).
One can also compare our findings with studies of mice deleted for positive regulators of growth. For example, deletion of the calcineurin A␤ gene led to an 80% reduction in total myocardial calcineurin activity, but this was associated with only a 12% reduction in cardiac growth (45). When one compares the growth defect to that of mice deleted for dominant regulators of normal heart growth including insulin receptor substrate-1 (ϳ37% smaller than wild type (46,47)), growth hormone receptor (40 -50% smaller than wild type (48)), or insulin-like growth factor-1 (30 -44% smaller than wild type (48,49)), the cardiac phenotype of the GSK-3␤ transgenic is almost as extreme. More recently, insulin has been reported to play a role in regulation of normal growth of the heart, and an insulin receptor-deficient mouse has a heart that is 22-28% smaller than wild type (17,50). Because GSK-3 is downstream of the receptors for both insulin and insulin-like growth factor-1 and is downstream of insulin receptor substrate-1, it is conceivable that at least part of the impaired growth in these models is due to greater activity of GSK-3␤. In addition, the similarity in cardiac growth phenotype between our mouse and control since levels were comparable between wild type and transgenic at these ages ( Fig. 2D and data not shown). Quantification of SERCA2a band density is shown below the immunoblot. a recently reported mouse deleted for phosphoinositide-dependent kinase-1 suggests that the altered activation of GSK-3␤ seen in the phosphoinositide-dependent kinase-1 knockout may be a key to explaining that phenotype (13). In summary, the GSK-3␤ transgenic has to our knowledge the most profound reduction in normal growth reported to date occurring in response to manipulation of a cytosolic signaling pathway.
It is not clear why we saw a major effect on normal left ventricular growth and function whereas a transgenic mouse expressing GSK-3␤ with the Ser-9 to Ala mutation had no significant reduction in left ventricular growth (see Fig. 3C in Antos et al. (20)) and was reported to be physiologically normal under non-stressed conditions (i.e. no systolic or diastolic dysfunction) (20). Strain differences cannot account for the differences since both studies employed C57Bl6 mice. Greater levels of expression of a transgene or enhanced activity of a transgene product often account for more extreme phenotypes in one transgenic line versus another. Although kinase activity was not measured in the mouse described by Antos et al. (20), GSK-3␤(S9A) is resistant to inhibition by phosphorylation. Furthermore, the intrinsic kinase activity of this mutant is believed to be at least as great if not greater than that of the wild type kinase. Thus, given the similar degree of overexpression in the two studies and the marked inhibition of kinase activity in our transgenic secondary to Ser-9 phosphorylation, greater kinase activity is not likely to account for the much more severe phenotype of our transgenic. It is conceivable that the mutation of Ser-9 to Ala might alter interactions of the mutant protein with regulators (e.g. Frat1/GSK-3-binding protein) or with critical growth regulatory substrates (e.g. eIF2B or ␤-catenin) by altering inhibitor/kinase or kinase/substrate interactions directly or by altering subcellular localization and access of GSK-3␤ to substrates. In any case the findings highlight the importance of using wild type transgenes where feasible.
It is not clear which of the many downstream targets of GSK-3␤ are critical for cardiomyocyte growth. As noted, NF-ATs do not appear to be central regulators of normal (as opposed to pathologic) hypertrophy. Our earlier studies suggested that ␤-catenin might play a role in promoting the growth of cardiomyocytes (12) and also that ␤-catenin is down-regulated in the transgenic heart. That this degree of down-regulation of ␤-catenin could affect cardiac growth is suggested by our findings in mice heterozygous for a mutation in the adenomatous polyposis coli (APC) gene. The APC gene product is necessary for GSK-3␤-mediated de-stabilization of ␤-catenin. These mice have very modest up-regulation of ␤-catenin (much less than the degree of down-regulation seen in the transgenic), but this is associated with a statistically significant increase of 5% in heart weight/body weight at 6 weeks of age. 2 Thus the downregulation of ␤-catenin likely plays a role in the reduced growth of the transgenic, but given the magnitude of the growth defect in the transgenic, there are likely to be other GSK-3␤ targets, such as eIF2B, c-Myc, and/or c-Jun, inhibition of which contributes to the impaired growth.
Mechanisms of Depressed LV Systolic Function in the GSK-3␤ Transgenic-The second obvious phenotype of our GSK-3␤ transgenic was impaired global LV function. Our data suggest that this was due at least in part to the marked increase in systolic wall stress, secondary to the failure of the transgenic heart to hypertrophy normally. Our studies in cardiomyocytes isolated from the GSK-3␤ transgenic heart suggest that the impaired global systolic function is not primarily due to a defect in contractility of the cardiomyocytes per se since total cell shortening of isolated myocytes, at least in the unloaded state when paced at 300 cycles/min, was normal. One additional contributor to the impaired systolic function could be that since diastolic relaxation was so impaired in the transgenics (see below), full relaxation could not be achieved in the transgenics at their intrinsic heart rates (ϳ600/min). In this case, pre-load (as determined by cardiomyocyte length at enddiastole) may have been reduced, leading to impaired systolic function.
Role of GSK-3␤ in Regulating Ca 2ϩ Handling, Diastolic Relaxation, and Atrial Hypertrophy-As profound as the impaired LV growth was, the most striking phenotype of the transgenic was massive atrial hypertrophy, reaching atrial weights 2-3fold greater than wild type in line #57 (Tables I and II) and up to 10-fold greater in line #44, in which atrial weights were equivalent to the left ventricular weight (data not shown). Histopathologically, the atria were characterized by extensive fibrosis and occasional organized thrombi. There are several possible reasons for the alterations in atrial growth. The NF-AT family of transcription factors appears to play a role in atrial development since overexpression of a dominant inhibitory mutant of the NF-ATs led to thinning of the atrial myocardium and sarcomeric disorganization (51). Because the GSK-3␤ transgene is expressed in the atrium, inhibition of NF-AT activity could have led to decreased atrial myocyte growth, and the thinner atrial walls could have contributed to the increased atrial volume (due to increased distensibility of the atrium). However, our data strongly suggest that the major cause of the atrial hypertrophy was the hemodynamic load (increased LVEDP) placed on the atrium. The marked increase in LVEDP was likely due to the dramatically altered Ca 2ϩ handling in diastole such that [Ca 2ϩ ] at end-diastole was ϳ100 nM greater than wild type, and [Ca 2ϩ ] relaxation () was substantially prolonged. This can be expected to lead to the markedly delayed relaxation of the LV that we observed.
SERCA2a is the major regulator of diastolic [Ca 2ϩ ] in the mouse heart (36). We found that SERCA2a protein and mRNA levels were significantly reduced in the transgenic. SERCA2a expression can decrease with advancing heart failure due to any cause, although the magnitude of the decrease is rarely if ever to the levels we observed here, which is greater than the 35-45% reduction seen in mice with a disruption of one copy of the SERCA2 gene (Refs. 52 and 53 and reviewed in Ref. 54). Furthermore, the reduction in SERCA2a occurred long before echocardiographic or clinical signs of heart failure developed. This suggested that down-regulation of SERCA2a might be directly mediated by GSK-3␤. Indeed, our data suggest that GSK-3␤ acts at the promoter to reduce expression of the SERCA2 gene (Fig. 5). Regulation of the SERCA2 gene is enormously complex with a number of transcription factors/ promoter elements having been implicated (26,(55)(56)(57). Our preliminary studies, based on gene transfer of activated mutants of NF-ATc4 and ␤-catenin, suggest that these known GSK-3␤ targets are not sufficient by themselves to induce SERCA2a gene expression (data not shown), and thus, the GSK-3␤ target regulating SERCA2a expression remains to be identified. That said, we believe these are the first data to implicate a cytosolic signaling pathway previously implicated in the hypertrophic response in the direct regulation of SERCA2a gene expression.
In conclusion, GSK-3␤ is an important negative regulator of normal concentric hypertrophic growth of the ventricle. The failure of the ventricle to grow normally results in impaired contractile function. In addition, activation of GSK-3␤ results in gross alterations in calcium handling, marked abnormalities of diastolic function, and massive atrial enlargement. This complex of abnormalities in the GSK-3␤ transgenic ultimately leads to advanced heart failure. These studies, especially when viewed in light of the direct down-regulation of SERCA2a expression by GSK-3, and the role of SERCA2a down-regulation in the systolic and diastolic dysfunction seen in patients with advanced heart failure, suggest GSK-3 inhibition could be an approach to the treatment of these patients.