Lithium inhibits cell cycle progression and induces stabilization of p53 in bovine aortic endothelial cells.

Lithium affects development of various organisms and cell fate through the inhibition of glycogen synthase kinase-3 beta and induction of the Wnt/beta-catenin signaling pathway. In this study, we investigated the effects of lithium on primary bovine aortic endothelial cells (BAEC). Lithium treatment of BAEC induced beta-catenin stabilization but failed to activate the transcriptional activity of the beta-catenin/T-cell factor complex. Lithium caused a sustained G(2)/M cell cycle arrest without affecting cell viability. Reversibility of this cell cycle arrest occurred up to 3 days after treatment but was reduced thereafter. Lithium-treated BAEC exhibited a senescent-like morphology with an increase in cells positive for the senescence-associated-beta-galactosidase activity. Lithium also increased the expression of p21(Cip), a cyclin-dependent kinase inhibitor, both at the protein and RNA levels. No change in p21(Cip) mRNA stability was observed, whereas the transcriptional activity of a p21(Cip) promoter-luciferase construct containing p53 binding sites was increased after lithium treatment. Furthermore, lithium caused increased transcription of a reporter gene under the control of a promoter containing the p53 consensus binding sites both in transiently transfected BAEC and in a stably transfected fibroblast cell line. Lithium caused accumulation of p53 protein in BAEC without affecting p53 mRNA levels. Finally, up-regulation of p21(Cip) in response to lithium did not occur in mouse embryonic fibroblasts that were null for p53 alleles, confirming the dependence on a p53 pathway for this lithium effect. These findings demonstrate for the first time that lithium induces also stabilization of the tumor suppressor p53 and reveal a new mechanism that may contribute to the neuroprotective effects of lithium.

Lithium affects development of various organisms and cell fate through the inhibition of glycogen synthase kinase-3␤ and induction of the Wnt/␤-catenin signaling pathway. In this study, we investigated the effects of lithium on primary bovine aortic endothelial cells (BAEC). Lithium treatment of BAEC induced ␤-catenin stabilization but failed to activate the transcriptional activity of the ␤-catenin/T-cell factor complex. Lithium caused a sustained G 2 /M cell cycle arrest without affecting cell viability. Reversibility of this cell cycle arrest occurred up to 3 days after treatment but was reduced thereafter. Lithium-treated BAEC exhibited a senescent-like morphology with an increase in cells positive for the senescence-associated-␤-galactosidase activity. Lithium also increased the expression of p21 Cip , a cyclindependent kinase inhibitor, both at the protein and RNA levels. No change in p21 Cip mRNA stability was observed, whereas the transcriptional activity of a p21 Cip promoter-luciferase construct containing p53 binding sites was increased after lithium treatment. Furthermore, lithium caused increased transcription of a reporter gene under the control of a promoter containing the p53 consensus binding sites both in transiently transfected BAEC and in a stably transfected fibroblast cell line. Lithium caused accumulation of p53 protein in BAEC without affecting p53 mRNA levels. Finally, upregulation of p21 Cip in response to lithium did not occur in mouse embryonic fibroblasts that were null for p53 alleles, confirming the dependence on a p53 pathway for this lithium effect. These findings demonstrate for the first time that lithium induces also stabilization of the tumor suppressor p53 and reveal a new mechanism that may contribute to the neuroprotective effects of lithium.
Lithium has profound effects on the embryonic development and patterning in various organisms including Dictyostelium and Xenopus (1). Inhibition of the development of the yolk sac vasculature by lithium has been observed in the chick embryo (2). Lithium is also the main therapeutic agent for the treatment of patient suffering from bipolar disorder (3). To date, the underlying mechanisms and the molecular targets of lithium action have not been established neither during development nor during therapy.
The developmental phenotype that is induced by lithium in Xenopus is similar to that caused by alteration in the expression of Wnt proteins (1,4,5) and of Wnt signaling components such as ␤-catenin (6) and the transcription factor T-cell factor (TCF) 1 (1,7). Lithium has been shown to be a specific and noncompetitive inhibitor of glycogen synthase kinase-3␤ (GSK-3␤) activity in vitro (8,9) and in vivo (10). GSK-3␤ is a serine/ threonine kinase that controls cell survival and cell fate through its involvement in multiple signaling pathways (11). Based on these findings, lithium is commonly used both as a potent mimetic of Wnt signaling and as a specific inhibitor of GSK-3␤ (1,(5)(6)(7)(8)(9)(10)(11). ␤-Catenin, both a scaffolding protein in cadherin-mediated cell adhesion and a signaling molecule in the Wnt signaling pathway, is a target of GSK-3␤ (12,13). In the absence of a stimulatory signal, low levels of ␤-catenin are tightly regulated via phosphorylation by GSK-3␤. This represents a signal for ␤-catenin degradation via the ubiquitin-dependent pathway (12,13). Inactivation of GSK-3␤ by Wnt signaling or by lithium leads to stabilization and nuclear translocation of ␤-catenin where it associates with TCF transcription factor members (14) and activates transcription of genes involved in cell adhesion (15) and cell proliferation (16,17).
Abnormalities in the regulation of ␤-catenin have been implicated in colorectal and melanoma tumorigenesis (12,13). However, in vivo overexpression of an N-terminally truncated form of ␤-catenin in intestinal epithelium did not increase tumorigenesis but increased both proliferation and apoptosis of undifferentiated cells without inducing cell differentiation (18). The proliferation of terminally differentiated cells within the same intestinal crypt was not affected (18). Similarly, only the proliferation of progenitor cells within the intestinal crypt was impaired in mice null for the transcription factor TCF4 alleles (19). Overexpression of ␤-catenin, either wild-type or an S37A mutant, in different cell backgrounds yielded various results. Proliferation of the epithelial Madin-Darby canine kidney cells has been observed (20), whereas apoptosis appeared to be the major effect in Drosophila retinal neurons (21), in mouse neurons (22), and in fibroblasts NIH3T3 (23). In the last, the apoptotic effects of ␤-catenin appeared independent of its transactivation function and interaction with TCF (23).
Endothelial cells of the adult vasculature are characterized by a quiescent state, which is controlled in part by the organization of cadherin and catenin complexes at cell-cell contacts (24). However, in response to injury or cytokine stimulation, tyrosine phosphorylation of ␤-catenin and dissociation from the cadherin complex is associated with EC migration and proliferation (24). Conversely, sequestration of ␤-catenin by the endothelium-specific VE-cadherin in the junctional complex inhibits EC migration and proliferation (25). ␤-Catenin cleavage by caspase-3 and dissociation from the complex is also an early event in the EC onset to apoptosis (26). Truncation of the ␤-catenin binding domain of VE-cadherin induces EC apoptosis, indicating that ␤-catenin may have a survival role in EC (27).
Increased cytoplasmic levels of ␤-catenin have been reported in EC during the neovascularization process occurring after myocardial infarction (28). The expression of different components of the Wnt signaling pathway in vascular cells is also altered in a rat model of restenosis (29) or in vitro in response to mitogens (29) or cell passage (30). EC proliferation occurs during angiogenesis and various vascular diseases (31). Cell cycle progression in EC is inhibited, as in other mammalian cells, by two key regulators: the cyclin-dependent kinase inhibitor p21 Cip and the tumor suppressor p53 (32). Increased expression of p21 Cip and p53 is observed in atherosclerotic lesions (31). Stabilization of p53 and induction of p21 Cip by a variety of injuries, stresses, or inhibitors of proliferation lead to cell cycle arrest either at the G 1 /S or G 2 /M transitions (32).
Lithium has been shown to affect cell proliferation either positively (33) or negatively (34), depending on the cell type. Lithium induces proliferation of erythroid progenitor cells (35) and terminal differentiation of WEHI-3B myelomonocytic leukemia cells (36). Wnt5a and Wnt10b are also able to promote proliferation of hematopoietic progenitor cells (37). Lithium, similarly to Wnt7A, has been shown to extend neurite outgrowth and to induce differentiation of cerebellar granule neurons (38). Several recent reports suggest that lithium may exert its therapeutic effects through neuroprotection against various stresses: ischemia, glutamate excitotoxicity, radiation, and heat shock (39 -42). Such survival effects of lithium on neuronal cells seem to be mediated in part by the activation of the survival Akt/protein kinase B pathway (40) as well as by the inactivation of GSK-3␤ activity (42).
Taken together, it appears that lithium as well as Wnt and ␤-catenin signaling may have various effects on cells depending on their differentiated state and whether they are immortalized or transformed. In the current report, we have determined that in primary EC lithium induces stabilization of both ␤-catenin and the tumor suppressor p53. This precedes cell cycle arrest in G 2 /M phase. Thus, our results reveal a novel mechanism that may account for the developmental and neuroprotective effects of lithium.

MATERIALS AND METHODS
Cells, Treatments, and Reagents-Bovine aortic EC (BAEC) were isolated as described previously (43) and grown in Dulbecco's modified Eagle's medium/F-12 containing 5% fetal bovine serum and used at passages 3-8. At least four different isolates of BAEC were used in the present study. Mouse fibroblast (12)1/CA cells carrying a p53-responsive ␤-galactosidase reporter construct have been described (44). The mouse embryonic fibroblasts MEFp53 ϩ/ϩ , MEFp53 ϩ/Ϫ , and MEFp53 Ϫ/Ϫ were derived from parental, heterozygous, and p53 knockout mice, respectively, and were kindly provided by Dr. Tyler Jacks (45). All of the mouse fibroblast cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Life Technologies, Inc.). Cell irradiation with 25 J/m 2 UV light for 10 s was used as positive control for p53 activation and was conducted as previously described (44). Actinomycin D was from Sigma.

Measurement of [ 3 H]Thymidine Incorporation and [ 3 H]Leucine
Incorporation-Cells were grown in 24-well culture clusters for 24 h prior to treatment with LiCl, or NaCl as control, with the indicated doses and times. In the last 2 h of treatment, 1 Ci/ml [methyl-3 H]thymidine (specific radioactivity 6.7 Ci/mmol; ICN) was added. Thereafter, the cells were washed rapidly with ice-cold 5% trichloroacetic acid and then incubated on ice for 20 min in 5% trichloroacetic acid. After two washes with ice-cold PBS, the cells were solubilized with 0.25 N NaOH and transferred to counting vials containing 5 ml of liquid scintillant (ICN). All results represent triplicate samples and three independent experiments. For the determination of [ 3 H]leucine incorporation, 24 h after seeding, the cells were treated with various doses of NaCl or LiCl for 24 h prior to the addition of 0.5 Ci/ml of [ 3 H]leucine (specific radioactivity 132 Ci/mmol; ICN) followed by a 24-h incubation at 37°C. Thereafter, the cells were washed and solubilized as described above. The experiment was carried out with triplicate samples and was repeated twice.
Cell Cycle Analysis-The cell cycle distribution was determined by flow cytometric analysis of propidium iodide-labeled cells. Briefly, BAEC were plated in six-well culture clusters 24 h prior to treatment with either NaCl or LiCl for the indicated doses and times. After trypsinization, cells were collected, washed twice with ice-cold PBS, and fixed in ice-cold 70% ethanol. Cells were then washed twice with ice-cold PBS, resuspended in PBS containing 100 units/ml RNase A, incubated at 37°C for 30 min, stained with propidium iodide (20 g/ml), and analyzed by FACScan (Becton Dickinson).
Senescence-associated ␤-Galactosidase Assay-BAEC were grown on glass coverslips in six-well culture clusters 24 h prior to cell treatment either with NaCl or LiCl. Four days post-treatment, BAEC were stained for senescence-associated ␤-galactosidase (SA-␤-galactosidase) activity at pH 6 for 16 h, as previously described (46). Phase-contrast and color photomicrographs of stained cells were taken.
Northern Blot Analysis-Total RNA was isolated by the acid guanidine thiocyanate/phenol/chloroform method using Trizol (Life Technologies). RNA (10 -20 g) was size-fractionated in a 1% agarose gel containing 6% formaldehyde in MOPS buffer and then transferred to Nytran nylon membrane (Schleicher and Schuell) using 10ϫ SSC (1.5 M NaCl and 0.15 M sodium citrate). The filters were hybridized for 12-16 h at 65°C with 32 P-labeled DNA probes (10 6 Ci/ml) in Church and Guilbert buffer. DNA probes were labeled with [␣-32 P]dCTP (specific activity 3000 Ci/mmol; ICN) using the oligoprimer labeling kit (Amersham Pharmacia Biotech). After hybridization, the membranes were washed under standard conditions and autoradiographed using X-Omat MR films (Eastman Kodak Co.). Radioactive mRNA signals were quantified by PhosphorImager (Molecular Dynamics, Inc.). Each experiment was repeated 2-3 times.
Cell Fractionation and Extracts-Whole cell extracts were prepared as follows. After incubation and treatments, cells were washed with PBS and lysed in 10 mM Hepes, pH 7.6, 1.5 mM MgCl 2 , 1 mM EDTA, 1% Triton X-100, 10 mM NaF, 2 mM NaVO 3 , 1 mM dithiothreitol supplemented with a mixture of protease inhibitors (Roche Molecular Biochemicals). For cell fractionation into cytosolic and membrane extracts, cells were washed in PBS, collected, and disrupted in 10 mM Hepes, pH 7.6, 1.5 mM MgCl 2 , 1 mM EDTA, 10 mM NaF, 2 mM NaVO 3 , 1 mM dithiothreitol, and protease inhibitors by repeated passage through a 23-gauge needle. After removal of nuclei by centrifugation at 800 ϫ g for 5 min, the supernatant was centrifuged at high speed (100,000 ϫ g for 30 min) to separate the membrane (pellet) and the cytosolic (supernatant) fractions. Nuclear extracts were prepared as previously described (50).
Immunoblot Analysis-Protein (30 g) was subjected to SDS-polyacrylamide gel electrophoresis (10%) and then transferred to polyvinylidene difluoride membrane (Millipore Corp.). Loading of equal amounts of protein was verified by staining with Ponceau-S (Sigma). Filters were incubated for 2 h at room temperature in 5% nonfat milk blocking mix (Bio-Rad) in 50 mM Tris-HCl, pH 7.6, 0.15 M NaCl, 0.05% Tween 20, followed by 12-16 h at 4°C with either mouse monoclonal anti-p53 antibody (PAb240) or rabbit polyclonal anti-p27 Kip antibodies (Santa Cruz) or rabbit polyclonal anti-␤-catenin (Sigma) in the same buffer. After 3 washes, blots were incubated with the appropriate secondary antibody conjugated to horseradish peroxidase Upstate Biotechnologies, Inc., Lake Placid, NY, developed using the enhanced chemiluminescence substrate (Amersham Pharmacia Biotech), and exposed to Kodak X-Omat-MR film. The polyclonal antibodies specific for p21 Cip used in this study were directly conjugated to horseradish peroxidase (Santa Cruz Biotechnology, Inc.).
Transient Transfection and Luciferase Assay-BAEC were grown in 12-well culture clusters, and 24 h later at 60% confluency, cells were transfected for 2 h in serum-free medium with 5 l of Exgen 500 (Euromedex) and with 0.5 g of the following cesium chloride-purified luciferase constructs: the TOPflash and FOPflash constructs containing, respectively, wild-type or mutated TCF binding sites upstream of the thymidine kinase minimal promoter (14); the p21 promoter luciferase constructs Ϫ2300-p21L, Ϫ210-p21L, and -94-p21L previously described (48); the p53-CA-luciferase reporter construct containing p53 consensus A sites and HSP70 minimal promoter (51); and the promoterless pGL3-basic (Promega). 5 ng of pRSV-␤-galactosidase was cotransfected to monitor the efficiency of transfection. Normal medium was then added for 22 h prior to treatment with NaCl, LiCl, or UV light as positive control for the indicated times. Cells were lysed 30 -48 h after transfection, and the luciferase activity assays and ␤-galactosidase activity assays were performed accordingly to the manufacturers' instructions (Promega and Tropix, respectively) and monitored using a Dynex luminometer. Statistical Analysis-Data are presented as the mean Ϯ S.D. of at least three independent experiments unless designated otherwise. Statistical analysis was performed using Student's t test, and a value of p Ͻ 0.05 was considered to be significant.

Lithium Induces ␤-Catenin Stabilization in BAEC without
Inducing an Active ␤-Catenin-TCF Pathway-We investigated the effect of lithium on the ␤-catenin-TCF pathway in primary BAEC by using two standard assays: stabilization and nuclear accumulation of ␤-catenin and transcriptional activation of a luciferase reporter driven by TCF-responsive elements. The epithelial HEK293 cells were used as positive control, since they are responsive to both lithium and Wnt signaling (52). As shown in Fig. 1A, lithium treatment for 12 h in BAEC increased the cytosolic pool of ␤-catenin, similar to the case in epithelial HEK293 cells, a known responsive cell line for Wnt signaling. A slight increase of ␤-catenin nuclear pool was also observed in both cell lines. To test whether this ␤-catenin stabilization was followed by an increase in the ␤-catenin-TCF complex activity, BAEC and HEK293 were transfected with either the TOPflash or FOPflash constructs (14). 24 h posttransfection, the cells were treated for 12 h with either 5 mM NaCl as control or with 5 mM LiCl. A 4-fold increase in luciferase activity was observed in HEK293 cells transfected with TOPflash and treated with lithium but not with the FOPflash construct containing mutated TCF sites (Fig. 1B). However, under the same conditions, lithium did not increase the activity of the TOPflash construct in BAEC. This lack of effect in BAEC was also observed with higher doses of LiCl (20 mM) (not shown). Therefore, in contrast to HEK293, lithium induced in BAEC stabilization of ␤-catenin without inducing an active ␤-catenin-TCF complex.
Lithium Inhibits Proliferation of Primary BAEC-The effect of lithium on BAEC proliferation was examined by measuring the incorporation of [ 3 H]thymidine into DNA. Subconfluent and asynchronous BAEC were incubated with various concentrations of LiCl or with NaCl as a control for the indicated times. Cells were pulse-labeled with [ 3 H]thymidine in the last 2 h of the incubation. As shown in Fig. 2A, the amount of [ 3 H]thymidine incorporation in lithium-treated cells (5 mM) was only 49.7 Ϯ 7.9% at 12 h and 19.5 Ϯ 7.9% at 24 h as compared with sodium-treated cells (p Ͻ 0.05). This effect was also observed at lower concentration of lithium (3 mM) with an inhibition of 32 Ϯ 5.8 and 43 Ϯ 7.9% after, respectively, 24-and 48-h treatment (p Ͻ 0.05). Cell viability was monitored by measuring the rate of protein synthesis (incorporation of [ 3 H]leucine into proteins) and the number of cells was determined in a parallel experiment. As shown in Fig. 2B, the number of BAEC was decreased by 38 Ϯ 2.5 and 55 Ϯ 2.45% after 48-h treatment of a 70% confluent culture with 5 and 10 mM lithium, respectively. After correcting for cell number, the protein synthetic rate was found to be unchanged after 48 h of lithium treatment with doses up to 10 mM (Fig. 2C). These results indicated that lithium inhibits BAEC proliferation without affecting their viability.
Lithium Induces a G 2 /M Cell Cycle Arrest in BAEC-Inhibition of cell proliferation can occur through activation of several possible checkpoints during cell cycle progression. To determine the effects of lithium on the cell cycle, subconfluent and asynchronous BAEC were treated with LiCl for various times, and cell cycle profiles were monitored by flow cytometric analysis of DNA content ( Fig. 3A and Table I). The number of cells with a 4N DNA content was increased at 16 and 24 h in a dose-dependent manner by lithium ( Fig. 3A and Table I) 4.1% of cells in the G 2 /M phase of the cell cycle as compared with 20.1 Ϯ 5.6% for 5 mM LiCl-treated cells or with 10% for the control cells (p Ͻ 0.05). Since lithium is known also to inhibit the inositol 1,4,5-trisphosphate breakdown in myoinositol, we investigated whether the effect of lithium occurred by inhibiting the phosphoinositol turnover. The addition of 10 mM myoinositol, the limiting factor in the inositol pathway turnover, did not affect the ability of lithium to induce a cell cycle arrest in BAEC (Fig. 3A), thus ruling out that the cell cycle arrest induced by lithium occurs via depletion of the inositol pathway. The percentage of cells with less than 2N DNA content, characteristic of apoptotic cells, was not modified by lithium treatment (not shown). The accumulation of cells in G 2 /M phase persisted over 4 days (Table I). After 24 h of treatment of BAEC, lithium removal resulted in an increase of cells in the S phase (22 Ϯ 6.4%) in 24 h similar to the percentage observed in control cells (29.6 Ϯ 6.1%) (p Ͼ 0.1). After 4 days of treatment, only a slight increase in S phase cells was observed after lithium removal (p Ͻ 0.05) ( Table I). To confirm the loss of the reversibility of lithium-induced cell cycle arrest, BAEC pretreated with sodium or lithium for 3 days were seeded at 5 ϫ 10 5 cell density in the absence of lithium. After 2 days of culture, the number of lithium-pretreated cells was reduced by about 30 -40% as compared with sodium-pretreated cells, whereas there was no significant difference at 24 h, indicating that the adhesion ability of lithium-pretreated cells was not impaired (Fig. 3B). Therefore, these results were also indicating that lithium-induced cell cycle arrest in BAEC occurred in the absence of cytotoxicity.
Lithium Induces a Senescence-like Phenotype in BAEC-Morphological changes in BAEC were observed after lithium treatment with an increased size and a more spread and flattened appearance (Fig. 4A). No rounded cells, characteristic of apoptosis, were observed. The number of multinucleated cells was slighty increased, but after cell nuclei staining with 4Ј,6-diamidino-2-phenylindole (DAPI), the most striking features of lithium-treated BAEC were enlarged nuclei and uncondensed chromatin, characteristic of interphase nuclei (Fig. 4B). Altogether, these observations were consistent with cell cycle arrest occurring in early G 2 . These morphological changes and the irreversibility of the cell cycle arrest (Fig. 3B and Table I) had similarity to the senescent cells (53). Therefore, we analyzed lithium-treated BAEC for expression of the SA-␤-galactosidase marker (46). The ␤-galactosidase staining was more intense in lithium-treated cells as compared with control cells (Fig. 4C). After counting the positive cells among 100 cells total, a 4-fold increase in SA-␤-galactosidase-positive cells after 5 days of lithium treatment was observed as compared with sodium-treated cells.
Lithium Up-regulates the Expression of p21 Cip in BAEC-To further characterize the effects of lithium on BAEC, we investigated the expression of genes involved in the control of cell cycle progression. As shown in Fig. 5A, lithium treatment induced a time-dependent increase in the cyclin-dependent kinase inhibitor p21 Cip mRNA levels, with a maximum 4-fold increase after treatment with 5 mM LiCl for 8 h. This increase was sustained up to 48 h as detected by Northern blot. Meanwhile, the expression of p27 Kip , another cyclin-dependent kinase inhibitor, was slightly down-regulated (1.5-fold decrease at 4 and 8 h). The expression of cyclin D1, a target gene of the ␤-catenin-TCF pathway (16), was biphasic with a slight increase of 1.5-fold at 4 and 8 h of lithium treatment followed by a down-regulation at 24 and 48 h as compared with control cells. The mRNA levels of the tumor suppressor p53 was not significantly modified by lithium treatment (Fig. 5A). Up-regulation of p21 Cip expression was also confirmed at the protein level by Western blot analysis. A time-dependent accumulation

FIG. 4. Lithium induces a senescent-like morphology in BAEC.
A, phase-contrast photomicrograph (ϫ 10 magnification) of live BAEC treated for 5 days with either 5 mM LiCl or 5 mM NaCl. B, fluorescence photomicrograph (ϫ 20 magnification) of BAEC treated with either 5 mM NaCl or 5 mM LiCl for 24 h and then fixed and stained with ␤-catenin (green) and with 4Ј,6-diamidino-2-phenylindole (DAPI) for nuclei visualization (blue). C, phase-contrast and color photograph (ϫ 20 magnification) of BAEC stained for the SA-␤-galactosidase activity (blue). BAEC were grown on coverslips and at 70% confluence were treated with either 5 mM LiCl or 5 mM NaCl for 5 days. SA-␤-galactosidase staining was done at pH 6. of p21 Cip protein was observed in whole cell extracts of BAEC treated with 5 or 10 mM LiCl as compared with control-treated cells (Fig. 5B). On the other hand, the expression of p27 Kip protein was down-regulated by lithium.
Expression of p21 Cip Is Transcriptionally Regulated by Lithium-We investigated next whether an increase of p21 Cip gene transcription or an increase of p21 Cip mRNA stability could account for lithium-induced p21 Cip mRNA expression in BAEC. The p21 Cip mRNA stability was determined in BAEC after exposure to 5 mM NaCl or 5 mM LiCl for 6 h followed by an incubation with actinomycin D, an inhibitor of RNA synthesis, for various times. A similar half-life of about 3.5 h was determined for p21 Cip mRNA both in control and lithium-treated cells, thus ruling out a post-transcriptional effect (Fig. 6A). The transcriptional activation of p21 Cip promoter by lithium was investigated using the luciferase reporter driven by different fragments of the human p21 Cip promoter (48) for which binding sites for various transcription factors have been delineated. These results are represented schematically in Fig. 6B. After 12 h of treatment with 10 mM LiCl, no significant changes in luciferase activities were observed with the smaller fragments, Ϫ291/ϩ16 and Ϫ94/ϩ16, of the p21 Cip promoter (Fig. 6B). In contrast, a 1.7-fold induction of luciferase activity was observed with the Ϫ2300/ϩ16 fragment of p21 Cip promoter, which contains two binding sites for p53 (p Ͻ 00.5) (Fig. 6B). Altogether, these results showed that lithium regulated the transcription of p21 Cip gene and suggested that p53, a potent inducer of p21 Cip expression (54), may be involved in this process.
Lithium Activates and Stabilizes Endogenous p53-We next examined whether the transcriptional activity of p53 could be increased by lithium. First, the mouse (12)1/CA fibroblast carrying consensus response elements for p53, controlling expression of the ␤-galactosidase reporter gene (44), were treated either with NaCl, LiCl, or UV light as a positive control. Whole cell extracts were prepared at various times after treatment, ␤-galactosidase activity was determined, and expression of p21 Cip was followed by immunoblotting. Maximal effects were observed at 16 h after lithium addition, with a reproducible 2.5-fold increase in ␤-galactosidase activity as compared with sodium-treated cells (Fig. 7A). UV light irradiation induced ␤-galactosidase activity by about 15-fold, as previously reported (44). A similar accumulation of p21 Cip protein was observed both in lithium-treated and UV-irradiated (12)1/CA cells (Fig. 7A).
The activation of p53 in response to lithium treatment was confirmed in BAEC by transient transfection of a luciferase reporter construct driven by p53-responsive elements (p53-CA-L) (51). After, 24 h of recovery, the transfected cells were treated with NaCl, LiCl, or UV irradiation for the indicated times, and luciferase activity of the cell extracts was determined. As shown in Fig. 7B, lithium treatment increased p53dependent transcription of the luciferase reporter with a maximal effect of 2.3-fold increase at 9 h of treatment. BAEC irradiated with UV light displayed only a slight increase of 1.5fold in the luciferase activity both at 6 and 9 h, although the accumulation of p21 Cip protein was stronger in UV-irradiated cells as compared with lithium-treated cells (Fig. 7B). However, lithium was able to induce the accumulation of p53 protein after 12 h of treatment similarly to UV light irradiation, as detected by immunoblotting using the PAb240 monoclonal antibody (Fig.  7B). Therefore, these results showed that lithium induced both activation and accumulation of endogenous p53 in BAEC.
Lithium-induced p21 Cip Expression Is Dependent on p53-To determine if p53 mediated the induction of p21 Cip expression by lithium, we investigated the effects of lithium in the mouse embryonic fibroblast MEFp53 ϩ/ϩ , MEFp53 ϩ/Ϫ , and MEFp53 Ϫ/Ϫ , cell lines, which are wild-type, heterozygous, and null for p53 alleles, respectively (45). MEFp53 cells were treated for 12 h with NaCl, LiCl, or UV light as a positive control, and cell extracts were prepared and analyzed for p21 Cip expression by immunoblotting. Lithium, like UV irradiation, was able to induce p21 Cip accumulation in MEF wild-type and heterozygous for the p53 alleles, but not in MEFp53 null cells (Fig. 8). These results suggest that lithium induced a p53-dependent pathway in various primary cells including BAEC.

DISCUSSION
In this report, we have shown that in primary BAEC, lithium, an inhibitor of GSK-3␤ and an activator of the Wntsignaling pathway, induced stabilization and nuclear translocation of ␤-catenin without inducing its transcriptional activity. Moreover, lithium induced in BAEC cell cycle arrest in the G 2 /M phase associated with the induction of p21 Cip expression in a p53-dependent pathway. Lithium-induced cell cycle arrest was not accompanied by signs of apoptosis or loss of cell viability but rather with the appearance of a senescent-like phenotype. These effects are not restricted to BAEC but occur also in other primary cells: the mouse embryonic fibroblast and bovine vascular smooth muscle cells (data not shown). From these findings, we propose that ␤-catenin stabilization and activation of p53 by lithium may account for its protective or survival effects reported in cells that were challenged by various stresses: radiation, ischemia, staurosporine, and glutamate excitotoxicity (39 -42).
Although lithium induced ␤-catenin stabilization in BAEC, it failed to induce ␤-catenin transactivation functions as assessed FIG. 5. Lithium increases p21 Cip expression. A, Northern blot analysis of cell cycle-regulated genes in BAEC after lithium treatment. Subconfluent BAEC were incubated with NaCl (Ϫ) or LiCl (ϩ) for the indicated times. Total RNAs were isolated and then analyzed by Northern blot for the expression of cyclin D1, p21 Cip , p27 Kip , and p53 mRNAs. Intensities of the radioactive bands were quantified by a PhosphorImager. Normalization of the RNA amounts was achieved using the expression of the ribosomal protein L32 (rpL32) mRNA as standard. B, Western blot analysis of p21 Cip and p27 Kip protein expression in BAEC after lithium treatment. Whole cell extracts from BAEC treated with 10 mM NaCl (0) or with LiCl for the indicated doses and times were analyzed by immunoblotting with specific antibodies for p21 Cip and p27 Kip .
with TCF-responsive luciferase reporter constructs (Fig. 1), whereas the epithelial cell line HEK293 demonstrated not only similar levels of ␤-catenin stabilization and nuclear localization ( Fig. 1) but also an increase of transcriptional activity as previously described (52). Similar observations have been made by others with the Jurkat T-cell line in comparison with the mouse epithelial cell line C57MG (55). We did not detect any activity of the ␤-catenin-TCF complex in primary human umbilical vein EC (HUVEC) or in the immortalized microvascular EC cell line HMEC-1 (data not shown), although all of these EC expressed TCF4. 2 In agreement with this absence of ␤-catenin-TCF transcriptional activity in BAEC, we did not observe a significant increase in cyclin D1 expression, a known TCF target gene (16), in response to lithium (Fig. 5). Wright et al. (56) have shown that primary mouse brain microvascular EC in response to 2 C. Mao, unpublished results.  (12)1/CA harboring a ␤-galactosidase reporter construct driven by p53-responsive elements (CA) was either treated with NaCl (Ϫ) or with 5 and 10 mM LiCl or irradiated with UV light for the indicated times. After cell lysis, ␤-galactosidase activities were measured, and p21 Cip expression was analyzed by immunoblotting. B, lithium-induced p53 activation and accumulation in BAEC. BAEC were transfected with 0.5 g of p53-CA luciferase reporter construct (p53-CA-L) or with the luciferase reporter driven by the thymidine kinase minimal promoter (FOP). pRSV-␤-galactosidase vector was co-transfected for normalization of transfection efficiency. 24 h post-transfection, the cells were treated with 10 mM NaCl (C) or with 10 mM LiCl (Li) or were UV-irradiated (UV) for 6 and 12 h, and cell lysates were prepared. Luciferase activities were measured and normalized to ␤-galactosidase activities. The results are mean Ϯ S.D. of the normalized luciferase activity obtained for each condition in four independent experiments performed in duplicate. Expression of p53 and p21 Cip was analyzed in parallel by immunoblotting.
Wnt1 display ␤-catenin-TCF transcriptional activity. Therefore, it is conceivable that additional modifications of ␤-catenin and/or TCF factors that are not induced by lithium in EC must occur to get full activity. Indeed, both inactivation of GSK-3␤ and activation of protein kinase C seem to be required for maximal activation of the ␤-catenin-TCF complex in HEK293 cells (52).
Lithium induced cell cycle arrest in G 2 /M without affecting cell viability in BAEC (Figs. 2 and 3). In addition to inhibiting GSK-3␤, lithium is also known to inhibit the inositol monophosphatase 1, leading to inhibition of inositol phosphate dependent pathways via depletion of the cellular myoinositol pool (8). However, the addition of myoinositol did not prevent or rescue the inhibitory effects of lithium on cell cycle progression (Fig. 3A). Therefore, the lithium effects on cell cycle, like its developmental defects in Xenopus (8), do not correlate with the inhibition of inositol phosphate pathways. Recently, Smits et al. (34) have observed a similar G 2 /M cell cycle arrest in various transformed (P19 embryonal carcinoma, U2OS osteosarcoma, and SK-N-MC neuroepithelioma) or immortalized (NIH3T3) cell lines after lithium treatment. They have shown that the activity of the cyclin B-Cdc2 complex, required for the entry into mitosis, was impaired after lithium treatment due to sustained phosphorylation of Cdc2 on tyrosine residue 15 (34). Our results show that lithium is able to induce p21 Cip expression, which is also a signal for cell cycle arrest in G 2 (Fig. 5). Therefore, both mechanisms of G 2 arrest are probably induced by lithium with induction of p21 Cip responsible for the sustained arrest that we have observed (Table I and Fig. 3B) (57). Although inhibition of cell proliferation by ␤-catenin signaling has not been described to date, the reported induction of cell apoptosis occurring after overexpression of ␤-catenin (21)(22)(23) was probably preceded by cell cycle arrest. It will be very informative to determine whether stabilization of ␤-catenin also induces a cell cycle arrest in G 2 .
Here, we have demonstrated that lithium is inducing stabilization and activation of p53 (Fig. 7) and that p21 Cip induction is mediated by p53 (Fig. 8). Stabilization and activation of p53 is mediated mainly through specific phosphorylation induced by a variety of stimuli, which can be grouped in three classes: DNA-damaging agents such as UV light, agents that affect microtubules, and inappropriate spatio-temporal expression of factors involved in cell cycle regulation (58). We have observed a more spread morphology of BAEC after lithium treatment, indicating that changes in the organization of the cytoskeleton are occurring (Fig. 4). Changes in the microtubule network have been described in response to Wnt7A (28,36) as well as in response to lithium and to dvl1, an upstream Wnt-signaling component mediating GSK3␤ inactivation (59). Therefore, an increase in microtubule stability may be the signal for cell cycle arrest in the G 2 phase, where disorganization of the microtubule must occur prior to formation of the mitotic spindle and cell rounding. Alternatively, ␤-catenin may be the signal for cell cycle arrest. Indeed, Orford et al. (20) have shown that the nuclear localization of ␤-catenin was cell cycle regulated in the epithelial Madin-Darby canine kidney cells with a peak during the S phase. Damalas et al. (60) have reported the accumulation of p53 in mouse fibroblasts NIH3T3 overexpressing a stable form of ␤-catenin (S37A). It is thus possible that a sustained retention of ␤-catenin in the nucleus during G 2 or G 2 /M transition can signal for p53 induction and cell cycle arrest. On the other hand, stabilization of ␤-catenin and p53 induced by lithium may occur in parallel through a similar mechanism. Further analyses are required to distinguish between these hypothesizes.
Some features of BAEC treated with lithium are associated with replicative senescence (46,53): adoption of a flat and enlarged cell shape (Fig. 4), the appearance of SA-␤-galactosidase activity (Fig. 4), and reduced growth ability (Fig. 3) accompanied by the accumulation of p21 Cip and p53 proteins (Figs. 5 and 7). Interestingly, changes in cell shape and appearance of multinucleated cells have been noted for renal mesenchymal cells constitutively expressing Wnt3A (61), which is similar to our observations with lithium-treated BAEC. Overexpression of p21 Cip and p53 can cause premature senescence in low passage fibroblasts (62,63). Therefore, further investigations are needed to determine whether lithium and abnormal activation of Wnt signaling may increase the rate of cell senescence.
Cell cycle-arrested or senescent cells display a reduced sensitivity to stress or DNA-damaging agents (53). The observed decrease in the apoptotic rate appears to be mediated by an activation of p53 and p53-dependent up-regulation of genes involved in DNA repair and cell survival. Once this survival state of the cell is established, p53 expression is dispensable (53). In addition, ␤-catenin may also mediate cell survival. Indeed, previous reports have shown a correlation between decreases in ␤-catenin levels and decreases in cell survival in response to anticancer drugs or stresses (21,22,64). The mechanisms of lithium action during bipolar therapy are not clearly understood, although recently a possible survival effect of lithium was proposed (1). Indeed, several studies have shown that lithium protects neuronal cells from undergoing apoptosis in response to a variety of insults, such as ischemia (40) or glutamate excitotoxicity (39). Activation of the phosphatidylinositol 3-kinase/AKT pathway (39) and inactivation of GSK-3␤ (42) have been implicated in this protection, the effects of phosphatidylinositol 3-kinase/AKT pathway probably being mediated in part by inactivation of GSK-3␤ (42). Down-regulation of p53 and Bax, a proapoptotic factor, in conjunction with an up-regulation of the antiapoptotic Bcl2 have also been proposed to mediate the protective effects of lithium in cerebellar granule cells (41). Such effects were only observed after a long term treatment of 7 days with lithium and, thus, are not in contradiction to our observations, since down-regulation of p53 is often observed after its activation (58). This down-regulation is due in part to an increase of MDM2 expression and of p53 targeting for degradation (58). Our results, in addition to strengthening the hypothesis of a cell survival effect of lithium in various cell types, suggest that stabilization of both p53 and ␤-catenin may contribute to this effect.
Multinucleated and enlarged EC have been described in the aging vasculature but also in vascular lesions such as atherosclerosis and postangioplasty restenosis (65). Increases in p53 expression in EC as well as in smooth muscle cells and macrophages are associated with atherosclerotic lesions (32,65). This increase appears to be part of the protective mechanisms that cells develop to limit the extent of proliferation in response to injury (33,58). Indeed, in models of arterial injury, prolifera- tion of vascular smooth muscle cells is inhibited by transfer of wild-type p53 with an absence of apoptosis (66), and conversely, abnormal proliferation of vascular smooth muscle cells is observed after transfer of p53 antisense oligonucleotide (67). Similar observations have been made in the double knockout mice p53 Ϫ/Ϫ and ApoE Ϫ/Ϫ , where an acceleration of the atherosclerotic process is occurring as compared with ApoE Ϫ/Ϫ mice (68). This acceleration is mediated by an increase of cell proliferation and decrease of cell death. Therefore, lithium treatment, by inducing p53 and inhibiting cell proliferation in the absence of cell apoptosis, may also have a protective effect in vascular diseases. Although a direct role of the ␤-catenin or GSK-3␤ pathways has not been studied to date in the development of vascular diseases or in vascular cell survival, it is worth noting that the inactivation of GSK-3␤ and stabilization of ␤-catenin can occur in response to nitric oxide (69), a potent regulator of vascular functions. Therefore, lithium may be a good therapeutic agent in vascular diseases involving vascular cell proliferation, such as restenosis after coronary angioplasty.