GSK3β Overexpression in Dentate Gyrus Neural Precursor Cells Expands the Progenitor Pool and Enhances Memory Skills*

In restricted areas of the adult brain, like the subgranular zone of the dentate gyrus (DG), there is continuous production of new neurons. This process, named adult neurogenesis, is involved in important cognitive functions such as memory and learning. It requires the presence of newborn neurons that arise from neuronal stem cells, which divide and differentiate through successive stages in adulthood. In this work, we demonstrate that overexpression of glycogen synthase kinase (GSK) 3β in neural precursor cells (NPCs) using the glial fibrillary acidic protein promoter during DG development produces an increase in the neurogenic process, increasing NPCs numbers. Moreover, the transgenic mice show higher DG volume and increased number of mature granule neurons. In an attempt to compensate for these alterations, glial fibrillary acidic protein/GSK3β-overexpressing mice show increased levels of Dkk1 and sFRP3, two inhibitors of the Wnt-frizzled complex. We have also found behavioral differences between wild type and transgenic mice, indicating a higher rating in memory tasks for GSK3β-overexpressing mice compared with wild type mice. These data indicate that GSK3β is a crucial kinase in NPC physiology and suggest that this molecule plays a key role in the correct development of DG and adult neurogenesis in this region.

In restricted areas of the adult brain, like the subgranular zone of the dentate gyrus (DG), there is continuous production of new neurons. This process, named adult neurogenesis, is involved in important cognitive functions such as memory and learning. It requires the presence of newborn neurons that arise from neuronal stem cells, which divide and differentiate through successive stages in adulthood. In this work, we demonstrate that overexpression of glycogen synthase kinase (GSK) 3␤ in neural precursor cells (NPCs) using the glial fibrillary acidic protein promoter during DG development produces an increase in the neurogenic process, increasing NPCs numbers. Moreover, the transgenic mice show higher DG volume and increased number of mature granule neurons. In an attempt to compensate for these alterations, glial fibrillary acidic protein/GSK3␤-overexpressing mice show increased levels of Dkk1 and sFRP3, two inhibitors of the Wnt-frizzled complex. We have also found behavioral differences between wild type and transgenic mice, indicating a higher rating in memory tasks for GSK3␤-overexpressing mice compared with wild type mice. These data indicate that GSK3␤ is a crucial kinase in NPC physiology and suggest that this molecule plays a key role in the correct development of DG and adult neurogenesis in this region.
Adult neurogenesis is the production of new neurons not only during development but throughout life. It has been described in many mammalian species, including human (1). This process mainly takes place in two brain regions: the subventricular zone at the lateral ventricle (2) and the subgranular zone (SGZ) 2 in the dentate gyrus (DG) at the hippocampus (3). In both regions there are neural stem cells that divide to reach, at the end of the process, a mature newborn neuron. These precursors have an astrocyte-like phenotype expressing the glial fibrillary acidic protein (GFAP) (4,5).
DG is a discrete structure that is postnatally developed (6). After birth, the first granule neurons are formed in the hippocampal neuroepithelium and move to form a DG primordium. Afterward, granule precursor cells, also from this area, migrate to the primordium. These mitotic cells form a germinal matrix that gives rise to an adult DG after 1 month of life (7,8). After finishing the formation, some neural precursor cells remain in the SGZ dividing and constantly generating newborn granule neurons (9), which connect with the preexistent trisynaptic circuit of the hippocampus (10). Newborn granule neurons have a major role in hippocampus-dependent memory and learning tasks and in anxiety and depression (11).
For these reasons, understanding neural precursor cell proliferation and differentiation and adult neurogenesis is an important goal. GSK3 is a kinase that has been postulated as an important player in this process (12)(13)(14)(15). GSK3 is a serine-threonine kinase with two highly conserved isoforms, ␣ and ␤. GSK3␤ has a pivotal role in neurodevelopment, regulating different processes such as neuronal polarity, axonal outgrowth, migration, apoptosis, neurotransmission, synaptic development, and plasticity (16,17).
Some evidence about the importance of GSK3 in neurogenesis comes from genetically modified models. In this sense, the loss of DISC1 (disrupted in schizophrenia 1), a GSK3 inhibitor by protein interaction, promotes a decrease in neural precursor cells (NPCs) proliferation and an increase in differentiation, which is overridden using GSK3 inhibitors (18). Furthermore, in double knock-in mice with constitutive active GSK3, by the mutation of the phosphorylated inhibitory serines 21 and 9 to alanines in both GSK3␣/␤ isoforms, a decrease in NPC proliferation was detected (12). The opposite phenotype is observed in knock-out embryos of both GSK3 isoforms in NPCs nestinϩ (13). However, some findings related to GSK3 functions in NPCs are controversial. For instance, the growth of the NPCs depends of the levels of GSK3 inhibition (19).
NPCs progressively generate more differentiated progeny, which eventually mature into granule neurons. Thus, the role of GSK3␤ in adult neurogenesis should be analyzed step by step. In previous reports, we found that an overexpression of the kinase in later stages of neurogenesis from DCX cells to mature neurons has negative consequences for the neurogenic process (14,20). However, the direct role of GSK3␤ overexpression in the first step of the process in vivo, just in NPCs, has not been addressed so far.
Here, we further studied the consequences of in vivo GSK3␤ overexpression in NPCs using the GFAP promoter on adult hippocampal neurogenesis, having in mind the possible effect during DG formation. An increase in the number of NPCs, as well as in the total number of mature granule neurons and DG volume, was observed. However, these effects are not only due to adult neurogenesis but are also attributable to the overexpression during DG development. This atypical environment is controlled and retained to avoid a higher increase in neurogenesis by high levels of Wnt inhibitor pathway. Overall, the alterations of all these parameters change the behavior of these GFAP/GSK3 transgenic mice, increasing their memory skills.

Animals
Mice were bred at the Centro de Biología Molecular "Severo Ochoa" animal facility, under standard laboratory conditions in accordance with European Community Guidelines and handled in accordance with European and local animal care protocols. The mice were housed 4 -5/cage with food and water available ad libitum and maintained in a temperature-controlled environment on a 12/12-h light/dark cycle with light onset at 8 a.m.
The studies in adult neurogenesis were done with GFAP/ GSK3␤ double transgenic, with all experimental procedures authorized by the Bioethics Committee of Centro de Biología Molecular Severo Ochoa (Universidad Autónoma de Madrid-Consejo Superior de Investigaciones Científicas, UAM-CSIC, Madrid, Spain). GFAP/GSK3␤ mice were obtained by crossing Bi-TetO ␤-gal GSK3␤ mice (21) (carrying the bidirectional Tetresponsive promoter followed by GSK3␤ and ␤-galactosidase cDNAs, one in each direction) with GFAP-tTa mice (Jackson Laboratory; B6.Cg-Tg(GFAP-tTa)110Pop/J no. 005964). WT animals resulting from crossing GFAP-tTa line with C57/BL6 mice were used as a control group. All experiments were conducted in animals 3.5 months old, except those shown in Fig. 7 and Table 1.

Tissue Processing
The mice were fully anesthetized with an intraperitoneal pentobarbital injection (Dolethal, 60 mg/kg body weight) and transcardially perfused with saline 0.9% followed by 4% paraformaldehyde in phosphate buffer. The brains were removed and postfixed overnight in 4% paraformaldehyde. For immunofluorescence experiments, 50-m-thick sagittal brain sections were obtained on a Leica VT1200S vibratome.
The animals used for biochemical analysis were fully anesthetized with an intraperitoneal pentobarbital injection and transcardially perfused just with saline. The brains were removed, and the hippocampi were quickly dissected on ice and frozen in liquid nitrogen.
When the immunofluorescence was perform to detect thymidine analogs, a preincubation of 30 min with 2 N HCl was performed. Then sections were gently washed in phosphate buffer following the described protocol.

Western Blotting Analysis
Extracts for Western blotting analysis were prepared by homogenizing the sample tissue in ice-cold extraction buffer consisting of 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 1 mM sodium orthovanadate, 1 mM EDTA, a protease inhibitor mixture COMPLETE TM (Roche), and 1 M okadaic acid. The samples were homogenized, and protein content was determined by the Bradford method. Twenty micrograms of total protein were electrophoresed on 10% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane (Schleicher & Schuell). Prior to antibody hybridation, membranes were blocked with 5% nonfat dried milk. The following primary antibodies were used: mouse anti-␤-galactosidase (1:5000 Promega, catalog no. Z378A, lot 18637309); rabbit anti-GFAP (1:1000 Promega, catalog no. G560A, lot 11173502); and mouse anti-␤-actin (1:5000 Sigma, catalog no. A5441, lot 014M4759). The membranes were incubated with the antibody at 4°C overnight in the same blocking solution. Secondary goat anti-mouse and anti-rabbit antibody (1:5000; Dako) conjugates with HRP and ECL detection reagents (Amersham Biosciences) were used for immunodetection.

Dentate Gyrus Volume
The volume of the dentate gyrus was estimated stereologically by applying the Cavalieri method to each series of Nisslstained sections, as described previously (23). DG areas were measured in each slice of the series using ImageJ software (version 1.33; National Institutes of Health, Bethesda, MD).

Cell Counting
The total number of precursor cells BLBPϩ, Sox2ϩ, immature neurons DCXϩ, and cells in mitosis in P14 mice (PHisHϩ) were calculated using the physical dissector method adapted for confocal microscopy as described elsewhere (23). Confocal stacks of images were obtained using a LSM710 Zeiss confocal microscope with a 63ϫ oil immersion objective. Four stacks of images were analyzed per animal.
Total number of positive cells for the markers IdU, CldU, and fractin were quantified under an inverted Axiovert200 Zeiss fluorescence microscope (ϫ40 oil immersion objective), using the optical dissector method (23). Briefly, using series composed of every eighth section, the cells labeled for each marker in every section were counted, and the total number of cells counted was then multiplied by 8 to obtain the total number of cells in the whole DG.
Colocalization studies of IdUϩ (12 h) with CldUϩ (2 h) analyzed a sufficient number of IdUϩ cells (1000 cells per genotype obtained from a representative number of animals from each group) and classifying them as negative or positive for the other marker. In the case of colocalization studies with IdUϩ (4 weeks), all cells positive for thymidine analog were analyzed for different maturation steps markers (DCX or Calbindin). Data are shown as percentages relative to the total number of IdUϩ cells. Total mature granule neurons was analyzed through the application of a physical dissector method developed for confocal microscopy (Zeiss LSM710) 100ϫ oil immersion objective.

Quantitative PCR
The hippocampus from one hemisphere of each animal was used to isolate total RNA with the QIAzol lysis reagent (Qiagen) and RNeasy mini kit (Qiagen). The samples were homogenized in TissueLyser II with stainless steel beads (5 mm). Reverse transcription was performed using the high capacity cDNA reverse transcription kit (Applied Biosystems) in a final RNA concentration of 20 ng/l. Quantitative PCR was developed on a LightCycler 480 (Roche Applied Science) (384 multiwell plates). 5 ng of cDNA/well was run in triplicate. A final volume of 10 l in each well was prepared with the following concentrations: primers (250 nM concentration), Universal Probe Library Probes (100 nM), and LightCycler 480 Probes Mastermix (2ϫ) (Roche Applied Science). Data from each gene was normalized by 18S ribosomal gene as control. The following Universal Probe Library gene-specific primers were used: DKK1 (forward, CCGGGAACTACTGCAAAAAT; reverse, CCAAGGTTTTCAATGATGCTT), sFRP3 (forward: CACC-GTCAATCTTTATACCACCT; reverse: TCAGCTATAGAG-CCTTCTACCAAGA), and Rn18s (forward: CGGCTACCAC-ATCCAAGGAA; reverse: GCTGGAATTACCGCGGCT) as endogenous control.

Behavioral Studies
Open Field-Major locomotor activity was measured in clear Plexiglas boxes measuring 43.2 cm ϫ 43.2 cm outfitted with photobeam detectors for monitoring horizontal and vertical activity. Activity levels were recorded with an activity monitor (MED Associates, St. Albans, VT). Locomotor activity data were collected via PC and analyzed with the MED Associates activity monitor data analysis software. The mice were placed in a corner of the open field apparatus and left to move freely. The variables recorded included: vertical counts, jumps, and average velocity (cm/s). The data were individually recorded for each animal for 15 min.
Fear Conditioning Test-A special system for fear conditioning (Panlab) was used. It consists of a box of sound attenuation. The floor, which detects any mouse movements, was made of stainless steel rods connected to a shock delivery apparatus. Inside the box, there was a loudspeaker to emit acoustic stimuli of known intensity, frequency, and duration to the experimental subjects. The apparatus was connected to a stimulus programming device to predetermine number, duration, and rate of sound and electric shock.
For conditioning phase, mouse was placed inside the conditioning apparatus. The 6-min protocol consisted of allowing the mouse 3 min for exploration. After this time, there were two cycles of acoustic stimuli lasting 30 s followed by 2 s of 0.8-mA electric foot shock; the cycles were separated by 1 min. As a control, animals from both genotypes were conditioned with the same protocol removing the electric shock after the acoustic stimuli. 24 h later, conditioned freezing was measured. The animals were again placed inside the conditioning apparatus and left there for 6 min. While they were there, neither electrical nor acoustic stimuli were administered. Freezing time was measured with a PC and was analyzed with Freezing v1.3.03 software. The control animals did not show freezing, demonstrating that neither the cage nor the acoustic stimulus produce an aversive conditioning.

Statistical Analysis
The data are presented as mean values Ϯ S.E. Statistical analyses of data were performed by applying a Student's t test for each statistical comparison. p Ͻ 0.05 values were considered significant. An Mann-Whitney U test was used when a nonparametric test was required. All statistics were analyzed using SPSS 17.0.1 software (SPSS, 1989; Apache Software Foundation).

Mouse Design to Overexpress GSK3␤ in Neural Precursors-
To analyze the role of GSK3␤ in NPCs, we have generated transgenic mice where GSK3␤ overexpression is driven by a GFAP promoter (GFAP/GSK3␤ mice; Fig. 1A). To this aim, we have used GSK3␤ transgenic mice (Bi-TetO GSK3␤), which were generated in our laboratory as previously described (21). Briefly, these transgenic mice carry a bidirectional TetO promoter (Bi-TetO) followed by GSK3␤ cDNA encoding in 5Ј a myc epitope in one direction and a ␤-galactosidase (␤-Gal) reporter gene fused with a nuclear location signal in the other (Fig. 1B). The GFAP/GSK3␤ mice derive from crossing homozygous Bi-TetO GSK3␤ and GFAP-tTa lines (Fig. 1A). The last one (24) carries a tTa under a truncated GFAP promoter, so the double transgenic progeny are expected to overexpress GSK3␤ in GFAP-positive cells, such as NPCs. GFAP/GSK3␤ transgenic mice were viable, had a normal lifespan, and grew normally. The percentage of each Bi-TetO promoter is formed by seven copies of the palindromic tet operator sequence containing two cytomegalovirus minimal promoter sequences (P), one in each end, in divergent orientations. This bidirectional promoter is followed by a GSK3␤ cDNA sequence (fused with a myc epitope at its 5Ј-end) in one direction and ␤-galactosidase (LacZ) sequence encoding a nuclear localization signal in the other. The minimal promoter is active, and therefore the transgene is expressed only when it has joined the transactivator. This allows muting the expression with tetracycline administration. C and D, Western blotting detection of transgene expression using ␤-gal antibody in hippocampus (C) and cortex (D) of GFAP/GSK3␤ mice and WT mice. Actin and GFAP has been used as a load control. Lines show molecular masses in kDa. E-G, ␤-Galactosidase immunofluorescence on sagittal brain sections from GFAP/GSK3␤ mice, showing DG (E), cortex (F), and subventricular zone (G). The white arrows in F show mature astrocytes positive for ␤-gal. H-K, triple immunofluorescence showing a ␤-gal-positive cell (H) which also express myc epitope corresponding to transgenic GSK3␤ (I). As expected, this cell is a neural precursor cell, positive for BLBP marker (J). K shows merge of all channels. DG, dentate gyrus; Hil, hilus; I-VI, cortical layers; LV, lateral ventricle; St, striatum; RMS, rostral migratory stream. genotype in the littermates was the expected Mendelian frequency of 50%.
Pattern of Transgene Expression-First, we checked the pattern of transgene expression in hippocampus and cortex by Western blotting and immunofluorescence ( Fig. 1, C, D, and E-G). When GFAP/GSK3␤ mice were analyzed by Western blotting, ␤-Gal reporter protein was detected in the double transgenic mice (GFAP/GSK3␤) in hippocampus (Fig. 1 C) and cortex (Fig. 1D). No ␤-gal expression was observed in wild type mice (Fig. 1, C and D). Probing protein extracts with anti-myc (which label the N-terminal end of transgenic kinase) and anti-GSK3␤ antibodies we did not observe any increase in GSK3␤ levels in the hippocampal or cortical samples, demonstrating that few cells show transgenic GSK3␤ expression (data not shown).
Immunofluorescence analysis of GFAP/GSK3␤ mice brain sections showed restricted region expression. More precisely, the expression of the transgene takes place in the two brain regions with adult neurogenesis: in DG (Fig. 1E) in the SGZ of the hippocampus where neurogenic precursors are located and the subventricular zone at the lateral ventricle (Fig. 1G). In cortex, a reduced expression was essentially taking place in astrocytes (Fig. 1F). No ␤-gal expression was detected in striatum or brainstem (not shown). To gain insight into which cell populations overexpress GSK3␤, we performed immunohistochemistry with anti-myc antibody. In the hippocampus, increased immunoreactivity for MYC-GSK3␤ was restricted to the DG and colabeled with BLBP (Fig. 1, H-K). Although ␤-gal-positive cells were observed in the cortex, no immunoreactivity for MYC-GSK-3␤ was observed.
GFAP/GSK3␤ Mice Overexpress GSK3␤ in NPCs of the DG-Taking into account that we are using a bidirectional promoter that allows the expression of GSK3␤ and the expression of a reporter gene, ␤-galactosidase, under the control of GFAP promoter, we first analyzed whether the reporter (and consequently transgenic GSK3␤) was expressed in NPCs in the dentate gyrus (DG) of the hippocampus. For this aim, we performed double immunofluorescence experiments against ␤-gal and GFAP protein (Fig. 2, A-E). Cells overexpressing GSK3␤ were in the correct SGZ location, and a characteristic GFAP apical process was found (white arrow). In addition, other cell types positive for GFAP were located in the hippocampus and also express the transgene (white arrowhead). To demonstrate that these cells are astrocytes, we used S100␤, a molecular marker of mature astrocytes that does not label radial astrocytes acting as neurogenic progenitors. Fig. 2 (F-K) shows triple immunolabeling for GFAP, ␤-galactosidase, and S100␤. White arrows show a NPCs located in the SGZ, positive for ␤-galactosidase and negative for S100␤ and with GFAP apical processes penetrating the GCL. The white arrowhead shows a double positive non-neurogenic astrocyte, and the thin white arrows show a mature astrocytes in the hilus and in the DG, which do not express the transgene. These data demonstrate that astrocytes do not form a homogeneous population of cells.
To unequivocally confirm the overexpression of GSK3␤ by NPCs, we perform double immunofluorescence with specific NPCs markers. Thus, brain lipid-binding protein (BLBP, a stage-specific marker for radial glia-like cell) (Fig. 2, L-P) and Sox2 (Fig. 2, Q-U) confirmed the expression of transgenic GSK3␤, detected by ␤-gal reporter, in adult hippocampal precursor cells.
Adult neurogenesis consists of different stages starting by the proliferation of type B cells (astrocytes) that then differentiate into DCXϩ cells and mature turning into functional neurons (11). Thus, we tested the expression of the transgene in the different stages that characterize the process (Fig. 3). Using Ki67 antibody, which labels dividing cells, we demonstrated that ␤-gal expression colocalizes with Ki67 (Fig. 3, A-C). This fact shows that GSK3␤ transgene expression remains in dividing cells in the SGZ. We also checked a more mature state in the neurogenic process. We used DCX antibody for neuroblast labeling. No colocalization between ␤-gal protein and DCX was seen (Fig. 3, D-F). Finally, we tested whether transgene was expressed in mature granule cells using NeuN antibody. Positive cells for ␤-gal antibody did not express NeuN marker (Fig.  3, G-I). These data suggest that exogenous GSK3␤ is expressed in dividing cells located in the SGZ and does not remain during the neuron's lifetime.

GFAP/GSK3␤ Mice Show an Increase in Neural Progenitor Cells, Mature Granule Cells, and Dentate Gyrus Volume-Hav-
ing demonstrated the expression of transgene in NPCs, we examined the physiological implication of this overexpression. First, we analyzed whether any variation exists in DG volume using the Cavalieri method. The quantification of the total volume showed a significant increase in DG volume of GFAP/ GSK3␤ compared with control mice (Fig. 4, A-C). To support this finding, we also measured the number of total mature granule neurons in the DG in each genotype. The quantification demonstrated a higher number of granule neurons in GFAP/ GSK3␤ mice compared with matched control (Fig. 4, D-F).
After observing an increase in DG volume and in the number of mature cells, we analyzed whether these changes could be produced by an alteration in NPCs caused by GSK3␤ overexpression. Using a physical dissector method, we quantified the amount of neural precursor by assessing the number of cells labeled with BLBP (Fig. 5, A and B) and with Sox2 (Fig. 5, C and  D). With both markers, we have shown an increase in the number of neural precursor cells in GSK3␤ transgenic mice compared with WT mice (Fig. 5, G and H). Then we checked what happened with the next differentiated stage of newborn neurons, characterized by expression of DCX (Fig. 5, E and F). The quantification of DCX positive cells showed, agreeing with previous data, an increase in the number of these cells in GFAP/ GSK3␤ (Fig. 5I). These data suggest that increases in the number of NPCs, mature cells, and dentate gyrus volume in transgenic mouse are due to the effect of GSK3␤ overexpression in vivo.
The Increase of Neural Progenitor Cells in GFAP/GSK3␤ Is Due to GSK3␤ Overexpression during Development-To go further in these findings, we tested the hypothesis of an increase in proliferation as a reason for the observed neural precursors increase. We carried out the evaluation of proliferation rate by immunolabeling using the antibody anti-phosphohistone H3, a mitotic marker. No statistical differences between the genotypes were noted in NPCs proliferation (data not shown). To rule out that the alteration was at the proliferation level, we carried out thymidine analog experiments. Using different thymidine analog incorporation, we checked the division cell cycle re-entry to discard differences in division speed. Thus, IdU was injected, and after 12 h, CldU was injected. The animals were sacrificed 2 h later, and the total number of double labeled NPCs was quantified (Fig. 6, A and B). No statistical differences were detected between WT and GFAP/GSK3␤ mice (Fig. 6C), suggesting that in transgenic mice, an increase in re-entry into the cell cycle does not constitute a likely hypothesis.
We also carried out other kinds of experiments with thymidine analogs at different time points to check possible differences in neuronal survival caused by GSK3␤ overexpression. We quantified new neurons at 24 h (Fig. 6, D and E), 1 week (Fig.  6, G and H), and 4 weeks of age. In the same manner, there were not statistical differences in survival at 24 h (Fig. 6F), 1 week (Fig. 6I), or 4 weeks (Fig. 6J) between new cells generated in GFAP/GSK3␤ or in WT mice. Then we analyzed the number of DCXϩ neuroblasts and the number of calbindinϩ immature neurons positive for IdU (Fig. 6, K and L). We did not find statistical differences between GFAP/GSK3␤ mice and WT mice in IduϩDCXϩ and IduϩCalbindinϩ, suggesting that in transgenic mice, alterations in differentiation/maturation stages do not constitute a likely hypothesis.
Then, to know whether a decrease in apoptosis might be related to the increased number of DG neurons, the death rate was evaluated by immunostaining with the anti-fractin antibody, which labels actin fragments cleaved by caspase-3 (Fig.  6M). As observed, the number of fractinϩ cells in the SGZ was not significantly higher in GSK3␤-overexpressing mice. This suggests that a decrease in apoptosis in NPCs in the adult does not likely happen in the SGZ.
Having in mind these results and the fact that the transgene is expressed in GFAP/GSK3␤ mice from the last stages of the embryonic period, we hypothesized that the changes observed could be produced for the effect of GSK3␤ overexpression in NPCs during DG formation. This structure is formed postnatally during the first month of age. Interestingly, there is a peak of expression of GSK3␤, during that period after birth, when development of DG is taking place (25,26). The overexpression of the kinase in GFAP/GSK3␤ mice in this period could produce an alteration of this process. To test this notion, we first tested in Nissl-stained sections from WT and GFAP/GSK3␤ mice whether the observed increase in adult DG volume (Fig. 4, A-C) occurred at the same grade in younger mice (Table 1).  Quantification revealed that in young animals (14-day-and 1-month-old mice), that increase was not observed. Then we quantified the number of NPCs in 14-day-old mice (Fig. 7).
Using a physical dissector method, we estimated the amount of neural precursor by assessing the number of cells labeled with BLBP (Fig. 7, A and B) and with Sox2 (Fig. 7, C and D). With both markers we observed a statistical increase in the number of neural precursor cells in GSK3␤ transgenic mice compared with WT mice (Fig. 7, G and H). Then we carried out the evaluation of proliferation rate by immunolabeling using the antibody anti-phospho histone H3, a mitotic marker (Fig. 7, E and F). Statistical differences between both genotypes were observed in NPCs proliferation (Fig. 7I). Thus, the increase in the starting pool of NPCs when DG   formation was taking place could explain why we detect an increase of NPCs in adult mice. However, this fact is not consistent with the lack of differences in the NPCs proliferation rate analyzed with division label antibodies (phospho histone H3) or with thymidine analog incorporation studies (Fig. 6). We reasoned that this could be due to the presence of external factors, which could slow down the proliferation in GFAP/GSK3␤ transgenic mice. (27)(28)(29), and recently the Wnt pathway inhibitor Dickkopf 1 (Dkk1) was proposed as an effector of that decrease (30) (Fig. 8A). On the other hand, during physical exercise, an increase in neurogenesis occurs (31,32), and this effect has been attributed to the decreased levels of another Wnt pathway inhibitor, secreted frizzled-related protein 3 (sFRP3) (33) (Fig.  8A). Thus, we tested the possibility that both secreted factors were altered in GFAP/GSK3␤ mice brain. Using quantitative PCR, we found a statistically significant increase in mRNA levels of these secreted factors in GFAP/GSK3␤ compared with control mice (Fig. 8B), suggesting that high levels of Wnt-frizzled complex inhibitors could act in a compensatory manner, decreasing the expected proliferation based in the increase of progenitor cells.

GFAP/GSK3␤ Transgenic Mice Present High Levels of Wnt Inhibitors-A decrease of neurogenesis is well stablished in the elderly
Behavioral Alteration in GFAP/GSK3␤ Mice-As we described above, GSK3␤ overexpression in NPCs produces a subsequent increase of neural progenitors in GFAP/GSK3␤ mice. To check whether that increase has any behavioral effect, we tested transgenic and control mice in different behavioral paradigms. First, we checked mice in a general locomotion and behavioral test, such as open field test. After 15 min in the arena, the analysis of different parameters registered by the apparatus showed no differences in the general exploratory behavior of transgenic mice respect to WT. In this sense, GSK3␤-overexpressing mice showed the same values in different parameters such as the total amount of vertical counts and jumps and the same average velocity (Fig. 9A).
Taking into account that GSK3␤ overexpression in NPCs produces an alteration in hippocampal neurogenesis, we challenged the animals in a hippocampus-dependent task, like a fear conditioning test. After a training day, the percentage of immobility or freezing was indicative of ambient remembering dependent on the hippocampus. The data showed a significant increase in freezing time of GFAP/GSK3␤-overexpressing mice compared with WT mice (Fig. 9B). Thus, in a hippocampus-dependent memory test, behavioral differences were observed.

Discussion
The development of the brain requires the decision of precursors to proliferate or differentiate. Different structures are made at different times during development. Adult neurogenesis takes place in the DG, a postnatal developed structure. At approximately P4, a DG primordium can be visualized, and at that time an increase in the proliferation of neuronal cells occurs. At day 14, the shape is that of a mature DG but smaller, and the maturation finished at approximately P21. During that stage, changes in GSK3␤ takes place. Thus, an increase in GSK3␤ levels during this period of intense neurite outgrowth, but also cell division, is observed (25,26). When DG structure is finished, a decrease in GSK3␤ levels occurs, and proper adult neurogenesis takes place. In this process, astrocyte-like stem cells start to proliferate. In this work, we have tested the effect of GSK3␤ overexpression specifically in those astrocyte-like cells.
GSK3␤ Overexpression in NPCs-Inhibition of the GSK3 has been associated with the maintenance of the pluripotent state in embryonic stem cells (34,35). Thus, it has been suggested that GSK3 acts as a kinase of integration because of its essential role in Wnt3 signaling (36). The Wnt transduction pathway involves inhibition of GSK3␤, which causes an increase in the levels of ␤-catenin and consequent activation of promoters implicated in proliferation and maintenance of pluripotent stem cells present in the SGZ. The importance of that route in this structure is reflected by the fact that the hippocampus is missing in a Wnt-3a deficient mouse (37). In good correlation, overexpression of ␤-catenin (38), as well as knock-out of GSK3 in neural progenitors (13), results in a massive hyperproliferation. However, other studies suggest that Wnt decreases the number neurospheres and leads to neuronal differentiation (39 -42). To further complicate the scenario, the response of neural progenitors appears to depend on the levels of GSK3. Thus, the cellular decision of these cells to differentiate or to split depends largely on levels of GSK3 activity (19). Most of the studies described above have been developed either in cell cultures or in animal models based primarily on the generation of animals deficient in some of the proteins involved in the transduction of Wnt. However, taking into account that NPCs progressively generate more differentiated progeny that eventually mature into granule neurons, the role of GSK3␤ in adult neurogenesis should be analyzed step by step. In previous reports, we found that the overexpression of the kinase in later stages of neurogenesis from DCX cells to mature neurons has negative consequences for the neurogenic process (14,20). However, the direct role of GSK3␤ overexpression in the first step of the process in vivo, just in NPCs, has not been addressed so far.
To this end, we have expressed GSK3␤ gene under a GFAP promoter. GFAP is not only expressed in NPC; it can be found in many types of glial cells. However, in GFAP/GSK3␤ mice, a particular pattern of GFAP expression is observed because of the use of a truncated promoter. The promoter consists of a 2.2-kb human GFAP promoter adjacent to the starting codon of the protein (43). This fragment seems to determine the expres-sion in a restricted subset of cells, agreeing with other models generated with this GFAP promoter fragment (24,44). In this sense, full-length GFAP promoter has different regulator regions far from starting codon that regulate protein expression in other cell type like Schwann cells. These regions are missing in GFAP/GSK3␤ mice, and in fact no peripheral nervous system expression is seen in our model (24). Moreover, the response of GFAP transcription in a reactive gliosis is regulated at the promoter level. A fragment missed in GFAP-tTa promoter and located 13.2 kb upstream of the starting codon is responsible of a stronger gliosis response (45). This could be the reason, because transgene expression is not observed in the glial scar generated after stereotaxic injection (data not shown). Related to mature astrocytes, no wide expression of the transgene in them is detected. However, some dispersed labeling (S100␤ϩ) can be seen corresponding to some mature astrocytes.
Expression of endogenous GFAP starts in mouse in the last days of the embryonic period (embryonic days 17 and 18) when glial cells exchange the expression of vimentin for GFAP. The GFAP expression increases from birth, reaching the highest levels between days 8 and 10 (46). GFAP/GSK3␤ transgenic mice show an increase of NPCs present in the immature DG (14 days after birth). These data suggest that increasing levels of GSK3␤ produce an expansion of neural precursor cells pool during DG development. That expansion likely is responsible for the increase in NPCs and in DG volume observed in the adult mice.
Thus, the main findings observed in GFAP/GSK3␤ mice, a greater number of neural progenitors BLBP-and Sox2-positive together with a greater volume of the DG accompanied by an increase in granular neurons, seem to suggest that an increase of GSK3 activity is important in the generation of new neurons and, more important, for a better performance in a hippocampus-dependent task such as a fear conditioning test. Interestingly, our results mirror, to some extent, what is happening in the development of the DG. This structure presents a fundamentally postnatal development, and it is in that period when GSK3 levels reach higher values in the central nervous system (25,26). In fact, GSK3␤ enzyme, but not the ␣ isoform, has a peak at 2 weeks, a period of intense neurite outgrowth, but also cell division, at least in the DG.
Possible Mechanisms for GSK3␤ Regulation of NPCs in SGZ-Taking into account all previous data, a likely mechanism by which GSK3␤ could regulate neurogenesis in GFAP/GSK3␤ mice would be by controlling levels of the molecules that are involved in neurogenesis, mainly ␤-catenin in the Wnt pathway (for a revision see Ref. 47). This assumption rests on the finding that during the asymmetric cell division, ␤-catenin phosphorylated by GSK3␤ is mainly inherited by one daughter cell (48). We hypothesized that increased levels of GSK3␤ in NPCs could produce the lack of asymmetrical protein distribution during precursor division, leading to progenitor pool expansion. It has been reported that Wnt proteins are necessary to maintain pluripotency and self-renewal of stem cells (47)(48)(49). During precursor cell division, the asymmetrical distribution of proteins of Wnt pathway determines which daughter cell will remain as a precursor cell (50). Therefore, it is possible that during the asymmetric division of neuronal progenitors, the two daughter cells have different levels of GSK3. Thus, in the asymmetric division that takes place in neural progenitors, daughter cells with lower GSK3 activity accumulate ␤-catenin (and perhaps other proproliferative proteins) and therefore continue to be stem cells. Conversely, the daughter cell with greater GSK3 activity eliminates pro-proliferative proteins and differentiates. In GFAP/GSK3␤ mice, the high levels of GSK3␤ overexpression obtained in NPCs could mask the polarized distribution of Wnt proteins. As a result, in both daughter cells high levels of GSK3 will exist, ␤ avoiding the effect of this asymmetrical distribution in precursors division. Thus, the division of a GFAP stem cell will produce two more linage-restricted cells (rapid amplifying cells) characterized by Sox2 and BLBP markers. This fact could explain the reported increased number of NPCs in GSK3␤ transgenic mice, although some experiments will be necessary to really confirm this hypothesis.
Another possible hypothesis that could explain the observed increase of NPCs in adult transgenic mice would be an increase in precursor cells proliferation. More proliferation events would be expected if more precursor cells are present, but no differences in mitosis rate were detected in GFAP/GSK3␤overexpressing mice. Interestingly, an increase in apoptosis was also not observed. The explanation for this phenomenon could be the increased levels of Dkk1 and sFRP3, two Wnt-Frizzled inhibitors (30,33). These molecules inactivate the Wnt pathway with the subsequent increase in ␤-catenin degradation and reduction of the expression of proneurogenic genes. GFAP/ GSK3␤ mice show high mRNA levels of Dkk1 and sFRP3, in a likely attempt to keep the enhanced pool of NPCs controlled and avoid and excess of neurogenesis.
In summary, in this work we have analyzed the consequences of GSK3␤ overexpression in NPCs. We have found increases in NPCs and DG volume and an improvement in memory skills. In previous studies, it was demonstrated that overexpression of GSK3␤ under CamkII promoter (14,20), a late promoter that is expressed in immature neurons, results in cell death, memory impairment and a decrease in DG volume Thus, it seems that during neurogenesis, changes in the levels of GSK3␤ occur, and they are needed to carry out the process in a proper way. Accordingly, to maintain the pluripotency state a decrease in GSK3 activity is necessary, whereas an increase in GSK3 activity seems to be necessary during asymmetric division of neuronal precursors to initiate differentiation.