Lithium Suppresses Astrogliogenesis by Neural Stem and Progenitor Cells by Inhibiting STAT3 Pathway Independently of Glycogen Synthase Kinase 3 Beta

Transplanted neural stem and progenitor cells (NSCs) produce mostly astrocytes in injured spinal cords. Lithium stimulates neurogenesis by inhibiting GSK3b (glycogen synthetase kinase 3-beta) and increasing WNT/beta catenin. Lithium suppresses astrogliogenesis but the mechanisms were unclear. We cultured NSCs from subventricular zone of neonatal rats and showed that lithium reduced NSC production of astrocytes as well as proliferation of glia restricted progenitor (GRP) cells. Lithium strongly inhibited STAT3 (signal transducer and activator of transcription 3) activation, a messenger system known to promote astrogliogenesis and cancer. Lithium abolished STAT3 activation and astrogliogenesis induced by a STAT3 agonist AICAR (5-aminoimidazole-4-carboxamide 1-beta-D-ribofuranoside), suggesting that lithium suppresses astrogliogenesis by inhibiting STAT3. GSK3β inhibition either by a specific GSK3β inhibitor SB216763 or overexpression of GID5-6 (GSK3β Interaction Domain aa380 to 404) did not suppress astrogliogenesis and GRP proliferation. GSK3β inhibition also did not suppress STAT3 activation. Together, these results indicate that lithium inhibits astrogliogenesis through non-GSK3β-mediated inhibition of STAT. Lithium may increase efficacy of NSC transplants by increasing neurogenesis and reducing astrogliogenesis. Our results also may explain the strong safety record of lithium treatment of manic depression. Millions of people take high-dose (>1 gram/day) lithium carbonate for a lifetime. GSK3b inhibition increases WNT/beta catenin, associated with colon and other cancers. STAT3 inhibition may reduce risk for cancer.

Long used to treat bipolar depression and hematopoietic disorders [14], lithium stimulates NSCs neurogenesis in the hippocampus [15] and subventricular zone [16], causing sustained increases of gray matter volume in patients [17,18,19,20]. Lithium also stimulates transplanted NSCs to produce more neurons [21] as well as axonal growth in injured spinal cord [22,23]. Other glycogen synthetase kinase (GSK) blockers mimic these lithium effects on neurogenesis and regeneration.
Recent study shows lithium inhibits GSK3b and invokes downstream effects on NSCs development. It increases betacatenin accumulation [24], which combines with WNT to stimulate NSC proliferation and neurogenesis. RNAi inhibition of beta-catenin abolishes these lithium-induced effects [25]. Beside the effect on stimulating NSCs proliferation and neurogenesis, lithium is also found reducing astrogliogenesis by NSCs [26], but the mechanisms underlay remains an enigma.
Lithium inhibits multiple messenger systems [27,28], including the JAK/STAT3 pathway [29] known to stimulate astrocytosis [30]. We therefore studied the effects of lithium and other GSK3b blockers on astrogliogenesis by NSCs isolated from neonatal rat brains. Both lithium and another GSK3b inhibitor SB216763 stimulated neurogenesis but only lithium suppressed astrogliogenesis by NSCs. In addition, analysis of restricted progenitor cell proliferation revealed that both lithium and SB216763 promotes neuronal restricted progenitor (NRP) cell proliferation, but only lithium inhibited the proliferation of GRPs. Further investigation showed that lithium not only strongly inhibited STAT3 activation, but also abolished the effect of a STAT3 agonist AICAR on inducing STAT3 activation and astrogliogenesis, indicating that lithium suppresses astrogliogenesis through inhibiting STAT3. Nevertheless, neither specific GSK3b inhibitor SB216763 nor molecular blockade of GSK3b with GID5-6 overexpression inhibited astrogliogenesis or STAT3 activation induced by serum or AICAR, These results together indicate that lithium inhibits astrogliogenesis through a non-GSK3b-mediated inhibition of STAT3.

Neural Stem Cells and Progenitor Cells
Growing NSC in growth media expectedly produced heterogeneous cultures of cells that expressed neuronal, astrocytic, and oligodendroglial markers. After 7 days in growth in serum-free media (DMEM with bFGF and EGF), NSCs proliferated and congregated in loose colonies that expressed nestin ( Figure 1A), an intermediate filament protein present in NSC and progenitor cells [37]. The cells often formed neurospheres. After dissociation, replating, and growth in 10 ng/ml bFGF and EGF for 24 hours, the cells were unipolar or bipolar with short processes ( Figure 1B) and almost all (9760.85%) expressed nestin ( Figure 1C). Very few cells (1.0060.43%) expressed GFAP, characteristic of mature astrocytes ( Figure. 1D). Likewise, only 1.0060.81% expressed Tuj1 ( Figure. 1E), a neuronal marker. Only 0.5060.21% expressed GalC ( Figure 1F), a major myelin galactosphingolipid and oligodendroglial marker. Many cells expressed A2B5 or PSA-NCAM, presumptive markers [38,39] for GRPs and neuronal restricted progenitors (NRPs) respectively comprising 39.063.03% ( Figure 1G) and 16.064.58% ( Figure 1H) of the cultures.
When transferred to neurobasal media containing B27 (NB27, Invitrogen), many of the cells began to show mature neuronal, astrocytic, and oligodendroglial markers. After 7 days in NB27 media, 32.161.4% of the cells expressed the neuronal marker Tuj1 and 42.861.9% expressed the astrocytic marker GFAP ( Figure 2C1). The cells expressing Tuj1 or GFAP also had respectively the morphology of neurons and astrocytes. In general, astrocytes outnumbered neurons.
In summary, NSC cultures produced progenitor and differentiated cells. In serum-free growth media, the cells proliferated and formed loose colonies of nestin-expressing cells and neurospheres. Many cells expressed A2B5 or PSA-NCAM, presumptive markers for glial-restricted and neural-restricted progenitor cells. Relatively few cells expressed mature neuronal, astrocytic, and oligodendroglial markers. However, when grown for 7 days in NB27 medium, NSCs produce mostly GFAP+ astrocytes and Tuj1+ neurons, the former more than the latter.

GSK3b Inhibition Promotes NSCs Proliferation in NB27 Medium
Lithium and other GSK3b inhibitors stimulate growth of NSC and progenitor cells [21,25,40]. We therefore examined the effects of lithium and the GSK3b inhibitor SB216763 on NPC's grown in NB27 medium. Even in the absence of added growth factors, NSCs and NPC's continued to proliferate in NB27 ( Figure 2A) and SB216763 was more potent than lithium in stimulating proliferation in NB27 without growth factors. Lithium (1 mM) increased total cell number by 1.1 fold after 7 days. In contrast, SB216763 (10 mM) markedly stimulated NSC proliferation, increasing total cell number by 1.4 fold (p,0.05 compared to lithium-treated cultures) after 7 days ( Figure 2B).

Lithium but not SB216763 Suppresses Astrogliogenesis
After 7 days in NB27, many cells began showing mature astrocytic marker GFAP and neuronal marker Tuj1. We stained and counted the cells, estimating the total numbers and percentage of cells of each lineage. While both lithium and SB216763 enhanced neurogenesis, only lithium suppressed astrogliogenesis. Lithium significantly increased the number of Tuj1 positive cells by 1.2860.03 fold in 0.5 mM LiCl (P,0.05), 1.4760.06 fold in 1.0 mM LiCl (P,0.05), 1.3360.03 fold in 3.0 mM LiCl (P,0.05), 0.6660.02 fold in 5.0 mM LiCl (P,0.05), 2.2560.07 fold in 10 mM SB216763 (P,0.05) from control level ( Figure 2C3). Conversely, lithium reduced the GFAP-positive cells number from control level to 0.7560.05 fold in 0.5 mM LiCl (P,0.05), 0.6460.02 in 1.0 mM LiCl (P,0.05), 0.5560.02 in 3.0 mM LiCl (P,0.05), 0.5460.03 in 5.0 mM LiCl (P,0.05). Conversely, SB216763 treatment did not reduced astrocytes number (0.9060.06 fold versus Control, P.0.05, Figure 2C2). We also found the similar tendency with S100beta immunostaining ( Figure S1), which is another astrocytes marker. These results indicate that lithium and SB216763 exert different effects on astrogliogenesis.
Western blots confirmed the decrease of GFAP and increase of Tuj1 protein in the lithium-treated cultures ( Figure 2D). Compared to control untreated cultures, 5 mM LiCl treatment reduced GFAP expression by 50% (p,0.01) and LiCl treatment increased Tuj1 expression, particularly at 1 mM. Lithium had a higher doseresponse curve for suppressing astrogliogenesis than neurogenesis. The inhibitory effects of lithium on astrocytic production continued to increase up to 5 mM while the stimulatory effects of lithium on neurons peaked at 1 mM.
Lithium suppression of astrogliogenesis was most prominent at lithium concentrations $3 mM. At these doses, lithium may be toxic to cells, hence reducing the number of astrocytes by causing apoptosis [41] rather than by inhibiting astrogliogenesis. We therefore studied the effects of lithium and SB216763 on apoptosis in the cultures, using the TUNEL assay [42,43]. Neither 10 mM SB216763 nor 1-3 mM LiCl increased apoptosis of cells cultured in NB27 medium for 7 days. However, higher (5 mM) concentrations of LiCl nearly doubled the number of TUNEL-positive cells ( Figure 2E). While lithium toxicity may partly explain the decline in astrocytes at 5 mM, it cannot account for the reduction of astrocyte count at 3 mM. we observed that lithium and SB2 induced neurite spreading and branching, which may duo to their inhibition on GSK3b [44].
In summary, analysis of cell counts of Tuj1-and GFAPexpressing cells revealed that, while both lithium and SB216763 stimulated neurogenesis, only lithium suppressed astrogliogenesis. SB216763 increased the number of neurons but did not reduce the number of astrocytes. Lithium increased the number of neurons and reduced the number of astrocytes. Higher concentrations of lithium (3 mM) were required to suppress astrogliogenesis than to stimulate neurogenesis (1 mM). Although lithium increased apoptosis of astrocytes at 5 mM, it did not do so at 3 mM, suggesting that the reduction of astrocytes at 3 mM was not due to lithium-induced apoptosis.

Lithium and SB216763 Effects on GRP Proliferation
Both lithium and SB216763 reduced the percentage of GRPs, as measured by A2B5 immunostaining. The percentage of A2B5+ cells fell from 3961.6% in control cultures to 2561.7% after 2 day in 3 mM LiCl culture. SB216763 also reduced the percent of A2B5+ cells but not as much as lithium ( Figure 3A). Combining the percentage and cell count data ( Figure 2A) indicated that the actual number of GRPs increased with time in untreated and SB216763-treated cultures. Lithium-treated cultures however, showed little or no increase in A2B5+ cell counts, suggesting that lithium inhibits proliferation or production of A2B5+ cells.
To confirm this hypothesis, we stained the cells for Ki-67, a nuclear and nucleolar protein that increases with somatic cell proliferation [45,46]. Lithium dramatically reduced the Ki-67 positive fraction of A2B5+ cells to 38%, compared to 68% in control untreated cultures. In SB216763-treated cultures, Ki-67 positive fraction was 64%, not significantly different (p.0.05) from control untreated cultures ( Figure 3B).
In summary, lithium (3 mM) reduced both the percentage and the number of GRPs but GSK3b blocker SB216763 did not. Lithium markedly reduced the fraction of A2B5+ cells that stained for Ki-67, a nuclear marker that reflects cell division, from 68% in untreated control cultures to 38%. SB216763, however did not significantly alter the fraction of Ki-67 labeled A2B5+ cells, confirming that lithium but not SB216763 suppressed proliferation of GRPs.

Lithium and SB216763 Effects on NRP Proliferation
To assess whether lithium and SB216763 stimulated proliferation of NRP cells, we assessed the effects of these drugs on PSA-NCAM expressing cells. Most Tuj1+ cells co-localized with PSA/   NCAM after 5 days of differentiation ( Figure S3). Both lithium and SB216763 significantly increased percentages of PSA-NCAM+ cells in 5-day NB27 cultures, i.e. 9% in control cultures compared to 34% in 1 mM LiCl and 41% in 10 mM SB216763-treated cultures ( Figure 3C). Analysis of the cell counts indicates that both drugs significantly increased the number of NRPs in the culture.
Double staining for Ki-67 and PSA-NCAM revealed that both lithium and SB216763 robustly increased the Ki-67 fraction of PSA-NCAM+ cells, i.e. from 14% in control cultures to 51% in 1 mM LiCl and 64% in SB216763-treated cultures ( Figure 3D). This suggests that both lithium and SB216763 enhanced production or proliferation of NRPs in the cultures.
In summary, GSK3b blockade by lithium or SB216763 stimulated production of more neurons. Double staining for Ki-67 and PSA-NCAM revealed that both drugs enhanced production or proliferation of NRPs.

Lithium Effects on JAK/STAT3 Activation and Astrogliogenesis
JAK/STAT3 regulates astrocytic production by NSCs. Molecular suppression of STAT3 gene expression or pharmacological inhibition of STAT3 activity markedly reduces astrogliogenesis [30,47,48]. One recent study [29] reported that very high concentrations of LiCl (20 mM) blocked STAT3 activation induced by lipopolysaccharide (LPS) or interferon in astrocytes.
Since our experiments showed that lower LiCl concentrations (3 mM) suppressed astrocytosis, we were interested to know whether this concentration of lithium would block STAT3 activation induced by gentler stimuli.
In summary, both serum and AICAR stimulate astrocytosis by activating STAT3. We confirmed that adding 0.5% serum increased P-Tyr705-STAT3 to 176 of baseline (pre-treatment) levels, associated with increased astrocytosis and GFAP at 24 hours. Adding 3 mM LiCl reduced P-Tyr705-STAT3 to 46 of baseline and prevented the astrocytosis. The STAT3 agonist AICAR likewise activated STAT3 and increased astrocytes and expression of P-Tyr705-STAT3. Applying 3 mM LiCl to the culture dramatically reduced the number of cells expressing GFAP in control and AICAR-treated cultures. These data indicate that lithium blocks STAT3 activation and prevents astrocytosis.

Effect of SB216763 on JAK/STAT3 Activation and Astrogliogenesis
Before assessing the effects of the GSK3b blocker SB216763 on STAT3 activation and astrocytosis, we verified that lithium and SB216763 blocked GSK3b mediated phosphorylation of betacatenin, a widely used assay of GSK3b activity [34,36,52]. As shown in Figure 5A, 30 minutes of treatment with 5-20 mM of LiCl significantly reduced phosphorylated beta-catenin (p-betacatenin). SB216763 likewise inhibited formation of p-beta-catenin.
In summary, pharmacological blockade by SB216763 did not block STAT3 activation, manifested by no differences of P-Tyr705-STAT3 levels induced by serum or AICAR. We confirmed that SB216763 blocked GSK3b mediated phosphorylation of beta-catenin and is approximately 1000 times more potent than lithium. Increasing the dose of SB216763 to 20 mM did not block STAT3 either. Another GSK3b blocker SB415286 did not prevent the STAT3 activation by serum. SB216763 also did not block AICAR-induced increase in GFAP. In contrast, lithium blocked the AICAR-induced rise in P-Tyr705-STAT3 and reduction of GFAP.
In summary, transfection and overexpression of GID5-6 effectively inhibited GSK3b activity and stimulated proliferation of NPC but did not stop inhibition STAT3 phosphorylation or GFAP production. Thus, lithium inhibits STAT3 activation and astrogliogenesis through a mechanism not involving GSK3b.

Discussion
Wexler, et al. [25] previously reported that lithium stimulates hippocampal neurogenesis by inhibiting GSK3b and elevating beta-catenin. Our experiments confirmed that both lithium and the GSK3b blocker SB216763 stimulated neurogenesis in NSC cultures grown in NB27 medium, increasing both the proportion and number of cells that express PSA-NCAM, as well as the production of Tuj1, as determined by Tuj1 single and BrdU/Tuj1 double staining ( Figure S2). Lithium also reduced the proportion  . Specific GSK3b blockade has no effect on STAT3 activation and astrogliogenesis. A. Lithium and GSK3b blocker SB216763 inhibit beta catenin phosphorylation (p-beta-Catenin). NSCs were treated with SB216763 (SB2, 10 mM) and LiCl (0, 5, 10, 20 mM) for 30 min. The GSK3b activity was assessed by detection of p-beta-Catenin. B1. SB216763 had no effect on serum-induced STAT3 activation. NSCs were cultured in NB27 medium with 0.5% FBS in the presence of 10 mM SB216763 for the indicated time. B2. Serum increased p-STAT3 over time and SB216763 did not change this curve. Data are expressed as mean 6 sem, averaged from three independent experiments and normalized to control values (n = 3, * P,0.05 vs. control, # P,0.05 vs. SB2 treatment group, one way ANOVA with Dunnett's post-test). C. STAT3 activation on NSCs treated with lithium and specific GSK3b inhibitors SB216763 and SB415286. NSCs were treated with LiCl, SB216763, SB415286 and STAT3 inhibitor Stattic at indicated concentrations for 24 h. D. STAT3 activation on NSCs incubated with 1 mM AICAR for 24 h with or without a 45-minute pretreatment of LiCl (3/ 5 mM) or SB216763 (SB2, 10 mM). E. GFAP expression on NSCs stimulated with AICAR for 3 days in the presence or absence of lithium. doi:10.1371/journal.pone.0023341.g005 Figure 6. GSK3b inhibition by GID 5-6 does not mimic lithium effect. NSCs were transfected by electrophoresis with liposomes containing DNA to make Myc-labelled GID5-6, which binds GSK3b and prevents its docking to the cytoplasmic protein axin and phosphorylating beta-catenin. GID5-6/LP is an ineffective analog of GID5-6. A. Transfection efficiency was assessed by immunostaining the cells for Myc (green) after 24 h. Most of the cells were nestin+ (red). B. The effect of GID 5-6 transfection on GSK3b activity. GSK3b activity was assessed by immunoblotting for GSK3b phosphorylated at Ser9 (Ser-P-GSK3b). C. The effect of GID 5-6 transfection on STAT3 activation on NSCs incubated with 0.5% FBS for 24 hours. D. GID5-6 transfection increased cell numbers by 1.2 fold versus GID5-6 LP transfection group as measured by CyQUANT Assay. E. Neither GID5-6 nor GID5-6 LP affected the number of GFAP expressing cells. F. GID5-6 transfection had no effect on number of GFAP-expressing cells. Data were expressed as mean 6 sem obtained from three independent experiments (n = 3, * p,0.05 vs. control, t -test). G. GID5-6 transfection did not affect GFAP level on NSCs but lithium (3 mM) markedly reduced GFAP level. doi:10.1371/journal.pone.0023341.g006 and number of cells expressing A2B5, as well as cells expressing the mature glial marker GFAP.
Several investigators have noted these inhibitory effects of lithium on glial cells [26,53], our further investigation showed that lithium prevented increases in the number of A2B5+ and GFAP+ cells in NSC cultures but SB216763 did not. In lithium-treated cultures, counts of A2B5+ and GFAP+ cells did not increase as much as in untreated cultures. In SB216763-treated cultures, the number of A2B5+ and GFAP+ cells increased and did not differ from untreated cultures. This is the first evidence suggesting that lithium suppressed astrogliogenesis may not through non-GSK mechanisms.
We hypothesized that lithium blocks phosphorylation of STAT3, a messenger system known to stimulate astrogliogenesis. To test this hypothesis, we measured P-Tyr705-STAT3 as an indicator of STAT3 activation. Adding 0.5% serum or the specific STAT3 agonist AICAR rapidly increased P-Tyr705-STAT3 protein and GFAP levels in NSC cultures. Lithium blocked this P-Tyr705-STAT3 and GFAP increase with the same doseresponse as it inhibited astrogliogenesis. Neither SB216763 nor GID5-6, a highly specific molecular blocker of GSK3b blocked induced P-Tyr705-STAT3 or GFAP increases. Together these results provide convincing evidence that lithium inhibits astrogliogenesis in NSC cultures by preventing STAT3 phosphorylation through non-GSK3b mechanisms.
In contrast, GSK3b inhibition stimulates neural progenitor cells to proliferate. Both lithium and SB216763 markedly increased the fraction of Ki-67+ cells amongst PSA-NCAM+ cells but not A2B5+ cells. Ki-67 is a marker of nucleolar and nuclear proteins expressed by dividing or recently divided cells. In control untreated cultures, only 14% of PSA-NCAM+ cells labeled for Ki-67 compared to 51% in 1 mM lithium-treated cultures and 64% in 10 mM SB216763-treated cultures.
Lithium clearly inhibits STAT3 in NSC cultures. Beurel & Jope [29,54,55] had earlier reported that STAT3 activation depends on GSK3b in astrocytes and microglia. They found that 20 mM lithium and other drugs that blocked GSK3b and suppressed STAT3 activation induced by lipopolysaccharide (LPS) and interferon-gamma in mouse primary astrocytes and microglia. Like Beurel & Jope, we found that lithium inhibits STAT3. However, unlike Beurel and Jope, we found that SB216763 did not block serum-or AICAR-activation of STAT3. We thus chose to test another and more specific GSK3b blocker, i.e. GID5-6, to see if it would inhibit serum-or AICAR activation of STAT3. We speculate this discrepancy might be due to the different culture condition and the dominance of regulating pathways among different cell types.
The cytoplasmic protein axin plays a critical role in GSK function [36]. In order for GSK3b to phosphorylate (inactivate) beta-catenin, both molecules must bind to axin. GID5-6 is the part of axin that specifically binds GSK3b. While overexpression of full length axin will cause more inactivation of beta-catenin, expression of GID5-6 should inhibit GSK3b and prevent betacatenin phosphorylation. We confirmed that expression of GID 5-6 blocked GSK3b activity and phosphorylation of beta catenin in NSCs. However, GID 5-6 did not affect serum-or AICARinduced STAT3 activation or astrogliogenesis. These results indicate that specific blockade of GSK3b does not prevent STAT3 activation by serum or AICAR.
Thus, our data indicate that GSK3b blockade does not necessarily inhibit STAT3 activation in NSC cultures. While GSK3b may play an important role activating STAT3 in astrocytes and microglia stimulated by LPS and interferon gamma [29,54,55], GSK3b does not seem to do so in NSC cultures stimulated by gentler STAT3 agonists. The effect of lithium on STAT3 and astrogliogenesis appears to be mediated by non-GSK mechanisms in A2B5+ NSC stimulated by 0.5% serum and AICAR. Lithium may affect STAT3 directly or indirectly.
We hope that our study will direct attention towards lithium's effects on the JAK (Janus kinase) and STAT3 pathway. This pathway not only stimulates astrogliogenesis [30,48,51,62,63,64] but also microglial activation [65,66]. Lithium inhibition of STAT3 would explain the dramatic reduction of activated microglia and macrophage due to lithium treatment of NSC transplanted into spinal cord [21].
STAT3 inhibition may explain lithium's remarkable lack of carcinogenicity. Lithium inhibition of GSK3b increases WNT/ beta-catenin, known to be associated with cancer [67,68,69]. Yet, millions of people have taken lithium for their lifetime without reports of increased cancer. In fact, lithium reduces formation of some tumors [29,70,71,72,73]. JAK/STAT3 activation also increases SOCS (suppressors of cytokine signalling), abnormalities of which cause cancer [74,75]. By inhibiting STAT3, lithium should reduce SOCS levels.
Our finding that lithium inhibits astrogliogenesis at 3 mM should be of interest for those seeking to grow neurons from NSC. At 1 mM, lithium stimulates neurogenesis without inhibiting astrogliogenesis. However, at 3 mM, lithium strongly stimulates neurogenesis and inhibits astrogliogenesis at the same time, without increasing apoptosis. Growing NSC in 3 mM lithium should produce predominantly neuronal cultures while growing them in 1 mM lithium or specific GSK3b blockers will allow astrocytes to grow. To inhibit astrogliogenesis, higher doses of lithium should be used.
Lithium is an attractive therapy for CNS regeneration. It is safe and robustly stimulates proliferation of endogenous [14] and transplanted neural stem cells [21,40], as well as axonal regeneration [22,23]. It increases brain concentrations of neurotrophins [14,76,77,78,79]. We have now shown that lithium suppresses astrogliogenesis by inhibiting STAT3, an effect that other specific GSK3b blockers seem to lack. At 3 mM concentrations, lithium thus may prevent or retard gliosis after brain and spinal cord injury.
In conclusion, lithium stimulates neurogenesis and suppresses astrogliogenesis by NSCs. We hypothesized that lithium blocks STAT3, which induces astrogliogenesis and microglial activation. Lithium, SB216763, and GID5-6 all inhibited GSK3b, prevented inactivation of beta-catenin, and stimulated neurogenesis. However, only lithium blocked STAT3 activation and astrogliogenesis induced by 0.5% serum or the STAT3 agonist AICAR, these findings indicate that lithium blocked STAT3 activation through non-GSK3b mechanisms. Lithium inhibition of STAT3 not only explains why lithium suppresses astrogliogenesis and microglial activation but also may explain the low carcinogenicity of lithium in clinical use.

Materials and Methods
For the purposes of this article, we use the term ''neural stem/ progenitor cells'' to refer to cells isolated from the subventricular zone of rats. When placed in growth media with epidermal growth factor (EGF) or fibroblast growth factor (FGF), these cells proliferated and produced neural progenitor cells (NPCs) expressing A2B5 and PSA-NCAM, respectively markers for glialrestricted or neuronal-restricted precursors. When placed in neurobasal media with B27 (NB27, Invitrogen), the cells differentiated to express mature neuronal or astroglial markers, respectively Tuj1 and GFAP. We used the following methods to prepare and identify NSC, to sort the cells, to assess proliferation and apoptosis, to quantify GSK3b and STAT3 activation, and to transfect cells with GID5-6 to block GSK3b.

NSC Preparation and Treatments
We isolated NSCs from neonatal Fischer 344 rats. The Animal Care and Facilities Committee at Rutgers University approved all animal procedures (Protocol: Rat Breeding Colony, NO. 99-032). Newborn rats (P0 or P1) were anesthetized with isoflurane (5%) and decapitated. Under sterile and ice-cold conditions, we removed the brain, dissected out the lateral wall of the lateral ventricle, and dissociated the tissue by gentle trituration with firepolished Pasteur pipettes [31].
The cells grew in a 37uC humidified 5% CO 2 incubator. We added growth factors every day, changed media every 2 days, and passaged the cells after 7 days. We designated first and second passage cells as P1 and P2, and used only P1 or P2 NSCs in this study. After passage, the cells were cultured in plates or cover slips coated with poly-L-lysine (0.01%, Sigma Aldrich, St. Louis, USA) and laminin (10 mg/ml, Invitrogen), placed in NSC culture media for 1-2 days, and then transferred to basic neurobasal medium plus B27 (NB27) for differentiation assays. Lithium chloride (Sigma Aldrich, St. Louis, USA) was dissolved in Milli-Q water, AICAR (Sigma Aldrich, St. Louis, USA) and SB216763 (10 mM, Tocris Bioscience, Ellisville, USA) were dissolved in dimethyl sulfoxide (DMSO, Sigma Aldrich, St.Louis, USA). We added lithium chloride and other drugs to the culture medium after the passage and then assessed the cells by immunocytochemistry or Western blots 2 to 7 days later, depending on the outcome measure used.

Immunocytochemistry
For immunocytochemistry, we cultured cells on 12-mm poly-Llysine/laminin coated glass cover slips at a density of 2610 4 cells/ cover slip. At specified time points, we immersed the cells for 5-20 minutes in 4% phosphate-buffered (0.01 M phosphate) paraformaldehyde, then blocked non-specific binding sites with 10% normal goat serum (Vector) and solubilized lipids with 0.3% Triton-X-100 for 1 hour. For surface antigen A2B5 or PSA-NCAM labeling, the cells were fixed in 4% paraformaldehyde for only 5 minutes and no Triton-X-100 was added to the blocking or washing solution. The fixation, blocking, and solubilization were done at room temperature (RT).
To label cells, we incubated the cells overnight in primary antibodies diluted in phosphate buffered (0.01 M phosphate) solution containing 3% normal goat serum and 0.3% Triton-X-100. Table 1 lists the primary antibodies and their sources. After washing with phosphate-buffered saline (PBS), we applied secondary antibodies corresponding to the primary antibody. The secondary antibodies were conjugated to fluorescent Alexa 568 or 488 (1:400, Molecular Probes) or rhodamine (TRITC) or FITC-conjugated goat anti-mouse (1:200, Jackson ImmunoResearch, goat anti-mouse IgM, m-chain specific). We stained nuclei with Hoechst 33342 (Molecular Probes, 5 mg/ml) and mounted the cover slips with Gelmount.
We photographed fluorescent images with a Zeiss Axiovert 200 M epifluorescent microscope or a Zeiss LSM510 confocal microscope, using 206 magnification and 0.4 aperture, an AxioCam camera, and Axiovision 4.6 software (Carl Zeiss, Germany). Cell percentage counts were based on numbers of nuclei and situation of the nuclei in cells expressing specific markers. The actual cells number of each lineage was determined by the combination with the percentage counts and the total number of cells in each treatment, as determined by proliferation assay.

Proliferation and Apoptosis Assays
To assess proliferation, we seeded NPC's on poly-L-lysine/ laminin coated 96-well plates with an initial density of 10 4 cells/ well. The cells were cultured in NSC medium for 24 hours and then in NB27 medium supplemented with lithium and other drugs. At planned time points, we froze the cells at 270uC overnight, thawed the cells, and quantified DNA using Cy-QUANT (Invitrogen) Assay and a Fluoroskan Ascent microplate reader (Thermo Labsystem, US.) with excitation wavelength at 485610 nm and emission detection wavelength at 530612.5 nm. We verified that the DNA signal relates linearly to cell number according to a standard curve provided by the manufacturer. The data are expressed in mean 6 standard error of mean (mean 6 sem).
To assess apoptosis, we used terminal deoxynucleotidyl transferase dUTP nick end labeling (DeadEnd TM Fluorometric TUNEL System, Promega). After incubating the cells in NB27 medium supplemented with various treatments for different times, we used 4% paraformaldehyde in PBS to fix the cells for 25 min at RT, permeabilized the cells with 0.1% Triton X-100 and 0.1% sodium citrate for 5 min, washed them several times with PBS, transferred the cells to an equilibration buffer (200 mM potassium cacodylate, 25 mM Tris-HCl, 0.2 mM DTT, 0.25 mg/ml BSA, 2.5 mM cobalt chloride, pH 6.6) for 10 min, and then immersed the slides for 60 min at 37uC in a reaction buffer containing 0.3 U/ml terminal deoxynucleotidyl transferase and 50 mM FITC-fluorescein-12-dUTP. To stop the reaction, we immersed
We transferred the gels to PVDF (polyvinylidene difluoride) membranes that had been soaked in methanol for 1 min and equilibrated in transfer buffer containing 25 mM Tris base, 190 mM glycine, 20% methanol, and 0.005% SDS [32]. After blocking with Tris buffered saline (TBS) buffer containing 0.1% Tween (TBST) and 5% non-fat dry milk (TBST+5% milk) for 1 hour at RT, we applied primary antibodies in TBST+5% milk overnight at 4uC, washed with TBST, applied horseradish peroxidase (HRP) conjugated secondary antibody in PBS+5% milk for 1 hour, imaged the bound antibodies using ECL plus Western Blotting Detection System (GE Healthcare, UK), analyzed with Kodak Molecular Imaging Software (v4.4.4), and normalized the phosphorylated protein to non-phosphorylated total protein. Table 2 lists the antibodies used.
For immunocytochemistry, we grew 3610 4 cells on cover slips. To assess proliferation, we seeded 10 4 cells per well and counted before and after 7 days in culture. For protein analyses, we grew the cells in 6-well plates at 5610 5 cells/well. After transferring to NB27 media, we assessed the cells at 24 hours for GSK3b and STAT3 phosphorylation.

Statistical Analyses
Data in the figures represent mean6sem and n indicates the number of experiments. We used the Student's unpaired t-test to compare two groups and analysis of variance (ANOVA) to compare multiple groups, followed by Dunnett's posthoc test to compare pairs of groups. A p-value of ,0.05 indicates significance. Figure S1 Lithium but not SB216763 suppresses S100b expression. NSCs were grown for 7 days in NB27 containing LiCl (0.5, 1.0, 3.0 mM) or SB216763 (10 mM) and then stained for S100b (red); nuclei were stained with Hoechst 33342 (blue). The photomicrographs (A1) show representative fields from each treatment group (Control, 3 mM LiCl and SB21763). The graphs show actual number counts of S100b+ cells (A2), normalized to untreated control counts. Lithium reduced the S100b+ cells number by 0.7260.08 fold in 0.5 mM LiCl (P,0.05) compared to control, 0.5160.09 in 1.0 mM LiCl (P,0.05), 0.4460.11 in 3.0 mM LiCl (P,0.05). In contrast, SB216763 treatment did not reduce astrocyte number (1.160.09 fold versus Control, P.0.05). Data are expressed as mean 6 sem from three independent experiments (n = 3, * denotes P,0.05 vs. control, one way ANOVA with Dunnett's post-test). (TIF) Figure S2 Both lithium and SB216763 stimulate neurogenesis by NSCs. NSCs were differentiated in NB27 medium containing LiCl (1.0 mM) or SB216763 (10 mM) for 7-8 days. BrdU (10 mM, Sigma) was added to cultures 2 days before fixation. Data are expressed as mean 6 sem from three independent experiments (n = 3, * denotes P,0.05 vs. control, one way ANOVA with Dunnett's post-test).

Supporting Information
(TIF) Figure S3 The co-localization of PSA/NCAM and Tuj1 after 5 days of differentiation. NSCs were grown for 5 days in NB27 containing LiCl (0.5, 1.0, 3.0, 5 mM) or SB216763 (10 mM) and then stained for Tuj1 (red) and PSA/NCAM (green), nuclei were stained with Hoechst 33342 (blue). Most Tuj1+ cells co-localized with PSA/NCAM after 5 days. The photomicrographs (A1) show representative fields of the co-localization of PSA/NCAM and Tuj1 from each treatment group (control, 1 mM LiCl, SB216763). The graphs show the percentage of Tuj1 and PSA/NCAM double positive cells out of total cells (A2). Data are expressed as mean 6 sem from three independent experiments (n = 3, * denotes P,0.05 vs. control, one way ANOVA with Dunnett's post-test). (TIF)