Inhibition of Glycogen Synthase Kinase 3β Attenuates Neurocognitive Dysfunction Resulting from Cranial Irradiation

There are now more than 10 million cancer survivors in the United States. With these numbers, chronic sequelae that result from cancer therapy have become a major health care problem. Although radiation therapy of the brain has improved cancer cure rates, learning disorders and memory deficits are a common consequence of this therapy. Here we show that glycogen synthase kinase 3β (GSK-3β) is required for radiation-induced hippocampal neuronal apoptosis and subsequent neurocognitive decline. Inhibition of GSK-3β either by small molecules (SB216763 or SB415286) or by ectopic expression of kinase-inactive GSK-3β before irradiation significantly attenuated radiation-induced apoptosis in hippocampal neurons. GSK-3β inhibition with SB216763 or SB415286 also decreased apoptosis in the subgranular zone of the hippocampus in irradiated mice, leading to improved cognitive function in irradiated animals. Studies of the molecular mechanisms of the cytoprotective effect showed that GSK-3β activity in hippocampal neurons was not significantly altered by radiation, pointing to the indirect involvement of this enzyme in radiation-induced apoptosis. At the same time, radiation led to increased accumulation of p53, whereas inhibition of the basal level of GSK-3β activity before radiation prevented p53 accumulation, suggesting a possible mechanism of cytoprotection by GSK-3β inhibitors. These findings identify GSK-3β signaling as a key regulator of radiation-induced damage in hippocampal neurons and suggest that GSK-3β inhibitors may have a therapeutic role in protecting both pediatric and adult cancer patients and may help to improve quality of life in cancer survivors. [Cancer Res 2008;68(14):5859–68]

Cranial irradiation is essential for the management of brain tumors. Many of these patients will exhibit neurocognitive deficits that involve impaired learning and memory. After cranial irradiation, cognitive decline is observed in both children (1, 2) and adults (3, 4) and is a result of hippocampal dysfunction. In fact, the severity of the cognitive deterioration seems to depend on the radiation dosage delivered to the medial temporal lobes including hippocampus (5). Hippocampal damage leads to defects in learning, memory, spatial navigation, visual motor processing, quantitative skills, and attention (6). The complex pathogenesis of radiation-induced cognitive deficit involves diminished neurogenesis by affecting neuronal progenitor cells that contribute new neurons to the hippocampus throughout life (7).

Although there is presently no pharmacologic prophylaxis of radiation-induced cognitive deficit, recent studies show that lithium exhibits neuroprotection against various neurologic insults (8, 9). One of possible neuroprotective mechanisms of lithium lies in its ability to inhibit glycogen synthase kinase 3β (GSK-3β), acting both by displacing magnesium and by indirectly inhibiting an activating phosphatase (10). Lithium also inhibits inositol monophosphate, leading to depletion of endogenous inositol in the cell and thereby reducing synthesis of inositol 1,4,5-triphosphate (11). GSK-3β belongs to a family of GSK-3, a multifunctional serine/threonine kinase (12) that has been implicated in multiple biological processes including embryonic development, cell differentiation, apoptosis, and insulin response (13, 14). GSK-3β is ubiquitously expressed in eukaryotic cells and is highly enriched in the brain and has been implicated in central nervous system dysfunctions like Alzheimer's disease (15), schizophrenia (16), dopamine-associated behaviors (17), bipolar disorders (18), and Parkinson's disease (19). Several kinases including Akt can attenuate GSK-3β enzymatic activity by phosphorylating the NH₂-terminal serine, Ser⁹ (13). Among a number of downstream targets, β-catenin is phosphorylated by GSK-3β and then degraded through the ubiquitin-proteasome system (20). Inhibition of GSK-3β activity therefore leads to stabilization and accumulation of β-catenin in the cytosol, which is shuttled into the nucleus where it functions to regulate gene expression. GSK-3β is also involved in cell cycle regulation through the phosphorylation of cyclin D1, which results in the rapid proteolytic turnover of cyclin D1 protein (21). In the present study, we therefore measured β-catenin and cyclin D1, which served as biomarkers of GSK-3β activity.

In our recent studies, we have shown that lithium prevents radiation-induced apoptosis in hippocampal neurons, leading to improved cognitive performance of irradiated mice. One of the molecular mechanisms of lithium neuroprotection is the inhibition of GSK-3β (9). In the present study, we show that GSK-3β signaling is a key regulator of radiation-induced apoptosis leading to neurocognitive deficits. Thus, pharmacologic disruption of GSK-3β activity may present a new approach to the protection of the brain during cranial irradiation.

Cell cultures and treatment. Mouse hippocampal neuronal HT-22 cells were obtained from David Schubert (The Salk Institute, La Jolla, CA) and maintained in DMEM with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Life Technologies). Human medulloblastoma Daoy was obtained from American Type Culture Collection and propagated in Eagle's MEM with 10% FBS, 2 mmol/L l-glutamine, 1.5 g/L sodium bicarbonate, 0.1 mmol/L nonessential amino acids, and 1.0 mmol/L sodium pyruvate. Human glioma D54 and mouse glioma GL261 cell lines were obtained from Dr. Yancie Gillespie (University of Alabama-Birmingham, Birmingham, AL) and maintained in DMEM with Nutrient Mixture F-12 1:1, 10% FBS, 1% sodium pyruvate, and 1% penicillin/streptomycin (Life Technologies). All cells were grown in a 5% CO₂ incubator at 37°C. For the radiation of cells, Mark I ¹³⁷Cs irradiator (J.L. Shepherd and Associates) was used delivering 1.84 Gy/min. Turntable ensured that the radiation was equally distributed.

Chemicals. SB415286 {3-[(3-chloro-4-hydroxyphenyl)amino]-4-(2-nitrophenyl)-1H-pyrrole-2,5-dione} and SB216763 [3-(2,4-dichlorophenyl)-4-(1-methyl-1H-indol-3-yl)-1H-pyrrole-2,5-dione] were purchased from Tocris Biosciences. AR-A014418 [N-(4-methoxybenzyl)-N′-(5-nitro-1,3-thiazol-2-yl)urea] and BIO [(2′Z,3′E)-6-bromoindirubin-3′-oxime] were purchased from EMD Biosciences (Calbiochem). All other chemicals including lithium chloride (LiCl) and camptothecin were purchased from Sigma.

Cloning, transfection, and cell sorting. Mouse wild-type GSK-3β (GSK-3β-WT) and kinase-inactive GSK-3β (GSK-3β-KI) cloned into pcDNA3 were obtained from David C. Seldin (Boston University Medical Center, Boston, MA; ref. 22) and subcloned into pPRIG, a bicistronic construct obtained from Patrick Martin (Université de Nice, Nice, France; ref. 23); pPRIG allows expression of any cDNA in the first cistron while keeping a high level of expression from its internal ribosome entry site–dependent second cistron encoding enhanced green fluorescent protein (eGFP). PCR was used to amplify the 1.3-kb fragment using a sense oligonucleotide (5′-GGGGTACATCGATCATGGCCTACCCATA-3′) and an antisense oligonucleotide (5′-GGTCTAGAGCGGCCGCGGGCTGTTCAGG-3′) by cloning into ClaI and NotI sites of pPRIG. The mutations were confirmed by sequencing the plasmids. HT-22 cells were seeded at 30% to 40% confluency and transfected the following day with 5 to 10 μg of different expression vectors and Lipofectamine 2000 (Invitrogen). Cells were analyzed between 24 and 48 h posttransfection for GFP expression. The cells were trypsinized 48 h later, resuspended in HBSS, and sorted for GFP-positive cells using the BD FACS Aria cell sorter. The GFP-positive cells were further used for clonogenic survival assays, apoptosis assays, and immunoblotting.

Clonogenic survival. Colony-forming assay was done as previously described (9). Briefly, calculated numbers of cells were plated to enable normalization for plating efficiencies. Cells were allowed to attach for 5 h and then irradiated with 0, 2, 4, 6, or 8 Gy. After 7- to 10-day incubation, plates were fixed with 70% ethanol and stained with 1% methylene blue. Colonies consisting of >50 cells were counted under a microscope. The survival fractions were calculated as (number of colonies / number of cells plated) / (number of colonies for corresponding control / number of cells plated).

Apoptosis assays for cultured cells. Apoptosis was determined by 4′,6-diamidino-2-phenylindole (DAPI) staining. The treated cells were washed with PBS, fixed in 4% paraformaldehyde at room temperature for 10 min, and stained with 5 μg/mL of DAPI at room temperature for 10 min. The nuclear morphology was observed with an Olympus BX60 fluorescent microscope equipped with Retiga 2000R digital camera. Apoptosis was quantified by scoring the percentage of cells with apoptotic nuclear morphology at the single-cell level. Condensed or fragmented nuclei were scored as apoptotic; average percentage of apoptotic cells (±SE) was calculated in five to seven randomly selected high-power fields.

Alternatively, cell death was determined by Annexin V/propidium iodide staining with the Apoptosis Detection Kit (BD PharMingen). Briefly, aliquots of 10⁵ treated cells were incubated with Annexin V/propidium iodide for 15 min at room temperature. The cells were then analyzed by flow cytometry using a two-color fluorescence-activated cell sorting analysis (BD LSR II). For each treatment, the average fold increase of apoptotic cells over control (±SE) was calculated.

Immunoblot analysis. Total protein was extracted from treated cells using M-PER mammalian protein extraction reagent (Pierce). Protein concentration was quantified with bicinchoninic acid reagent (Pierce). Protein extracts (40 μg) were subjected to Western immunoblot analysis with antibodies for the detection of phospho-GSK-3β^Ser9 and Tyr²¹⁶, GSK-3β, phospho-Akt^Ser473, Akt, β-catenin, cyclin D1 (all from Cell Signaling Technologies), and p53 (EMD Chemicals, Inc.); antibody to actin (Sigma) was used to evaluate protein loading in each lane. Immunoblots were developed with the Western Lightning Chemiluminescence Plus detection system (Perkin-Elmer) according to the manufacturer's protocol.

Mice and treatment. All animal procedures were approved by the Vanderbilt University Institutional Animal Care and Use Committee. Mice were housed up to five per cage on a 12-h light/dark cycle (lights on at 0600 h). Food (Purina Rodent Chow) and water were provided ad libitum. Timed pregnant C57/BL/6J mice were obtained from The Jackson Laboratory. Indicated doses of SB216763 and SB415286, dissolved in DMSO and LiCl dissolved in PBS, were given to mouse pups via i.p. injection on postnatal day 14 daily for 3 consecutive days. DMSO or PBS was injected to vehicle-control mice. After this, mice were anesthetized, restrained, and treated with the indicated doses of cranial irradiation using a Therapax DXT 300 X-ray machine (Pantak) delivering 2.04 Gy/min at 80 kVP. For histologic stainings, mice were sacrificed 24 h after irradiation by cervical dislocation under isoflurane anesthesia. For Morris water maze studies, mice were tested 8 wk postirradiation. Behavioral testing was done during the light part of the cycle.

Histochemistry. After treatment, mouse brain was removed and placed in 10% paraformaldehyde solution for 24 h. The frontal lobes were removed before embedding in paraffin. Sections were cut by microtome until the anterior hippocampus was visualized. Five-micrometer-thick coronal sections were then taken and placed on Superfrost Gold Plus slides (Erie Scientific). Tissue sections were stained with the DeadEnd Colorimetric terminal deoxyribonucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) System (Promega) and counterstained with H&E in the standard fashion. For the TUNEL experiments, at least three animals were used in each experimental group. The superior curvature of the hippocampus was used for quantification of apoptosis in all cases. The subgranular zone was identified as a 2- to 3-cell-wide layer adjacent to the granule cell layer, facing the hilus. TUNEL-positive neurons were counted under a light microscope (400×). At least three fields were counted per animal. The average number of TUNEL-positive cells per high-power field (±SE) was calculated.

Immunohistochemical staining for β-catenin. Five-micrometer-thick sections of mouse brain were prepared as described above and placed on charged slides. After paraffin removal, the sections were rehydrated and placed in heated Target Retrieval Solution (DakoCytomation) for 20 min. Endogenous peroxidase activity was inactivated with 0.03% hydrogen peroxide followed by a 1% bovine serum albumin solution for blocking interfering protein interactions. The sections were incubated with rabbit anti–β-catenin antibody (Sigma Aldrich; 1:1,000 dilution) for 30 min. The Dako Envision+ System, rabbit/horseradish peroxidase (DakoCytomation) was applied for 30 min. Visible results were produced with 3,3′-diaminobenzidine. The tissue sections were counterstained with Mayer's hematoxylin and observed under a light microscope (400×).

Morris water maze. Spatial learning and memory were evaluated in mice using the hidden-platform water maze paradigm. The water maze pool was 92 cm in diameter and filled with water made opaque with nontoxic tempura paint. An acrylic platform (10 × 10 cm) was submerged ∼0.5 cm below the surface of the water to allow relief from swimming. Sessions were captured by an overhead camera and analyzed in real time using an NIH Image Macro program on a Macintosh computer. Parameters measured included latency (time to reach platform), swimming velocity, and path length. Mice were trained to locate a clearly marked platform (days 1–5) for the visible platform experiments, the location of the platform with the start location of the mouse changed for each trial. There were four possible platform locations, one in each quadrant of the pool. The animals were given 2 d off before beginning the hidden platform experiments. In the hidden platform experiments, the platform location was fixed throughout the experiment, but starting positions of mice were different on each trial. Each trial continued for a maximum of 60 s or until the mouse reached the platform, whichever occurred first. If a mouse did not reach the platform in 60 s, it was captured by the experimenter and placed gently on the platform for 15 s. A trial started by placing a mouse into the water facing the wall of the pool in one of four possible start locations around the pool circumference. Once the mouse was placed in the water, the camera was immediately activated, and when the trial ended, the mouse was returned to its cage. Trials were done in a massed fashion; in other words, all mice underwent the first trial of the day before undergoing a second trial. Data were analyzed immediately following each set of trials, and once each group of mice reached the criterion of 8 s average latency, they were given a probe test 24 h after their last training session. In the probe trial, the platform was removed and mice were allowed 60 s to explore the water maze. The amount of time spent in each quadrant of the pool and platform location crossings were measured for the 60-s probe trial.

Statistical analyses. The mean and SE of each treatment group were calculated for all experiments. The number of samples is indicated in the description of each experiment. Statistical analysis was done with Kruskal-Wallis ANOVA. All pairwise comparison procedures including calculation of P value were done using the Student-Newman-Keuls method. P < 0.05 was considered significant.

GSK-3β-KI increases clonogenic survival and protects HT-22 neurons from radiation-induced apoptosis. Treatment with lithium effectively protects irradiated hippocampal neurons from apoptosis and improves cognitive performance in irradiated mice; however, the mechanism is not well understood (9). Lithium inhibits GSK-3β, which is associated with various degenerative neurologic disorders (24). To elucidate the mechanism of neuroprotection, we expressed a GSK-3β-WT and a GSK-3β-KI mutant in HT-22 hippocampal neurons using bicistronic vector pPRIG (23). The eGFP-positive cells were sorted and used for immunoblot, clonogenic survival, and apoptotic assays (Fig. 1A). As a first step to validate the GSK-3β constructs, we transfected HT-22 cells with 5 and 10 μg of plasmid DNA per 100-mm dish. Immunoblot analysis revealed a significant increase of GSK-3β protein level in cells transfected with 10 μg of GSK-3β-WT DNA or GSK-3β-KI DNA (Fig. 1B). To determine the role of GSK-3β in radiation-induced neuronal damage, we manipulated its levels in HT-22 cells and assessed the expression of β-catenin as an indicator of GSK-3β signaling. HT-22 cells were transfected with 10 μg of GSK-3β-WT, GSK-3β-KI, or PRIG vector alone and irradiated with 3 Gy. Cells transfected with GSK-3β-KI showed increased stabilization of β-catenin compared with transfection with GSK-3β-WT and PRIG vector alone (Fig. 1C).

We used clonogenic cell survival assay to study the effect of manipulation of GSK-3β on cell viability and survival. Disruption of GSK-3β activity by transiently transfecting cells with GSK-3β-KI significantly protected HT-22 neurons from cell death induced by irradiation as compared with cells overexpressing either the GSK-3β-WT or vector alone (Fig. 2A). The calculated D0 was 1.95 Gy for the dose-response curves representing all three transfected cell cultures (PRIG, PRIG-GSK-3β-WT, and PRIG-GSK-3β-KI). However, the calculated extrapolation number n was 2.5 for the dose-response curves representing PRIG and PRIG-GSK-3β-WT, and n = 4.5 for the dose-response curve representing PRIG-GSK-3β-KI. Because n represents the ability to accumulate and repair sublethal damage, this analysis indicates that inhibition of GSK-3β activity increases the repair capacity by ∼1.8-fold.

To determine the mechanisms of this increased survival, the transfected cells were irradiated with 3 Gy and, 24 hours later, stained with either the fluorescent DNA-binding agent DAPI or Annexin V and propidium iodide (Fig. 2B and C). After DAPI staining, cells were evaluated for the typical morphologic features of apoptosis, (i.e., condensed or fragmented nuclei at 24 hours after treatment; Fig. 2B,, arrows). Morphologic analysis of HT-22 neurons showed that transfection with GSK-3β-WT induced mild apoptosis (6.1% of total number of cells) as compared with vector (0.96%) or GSK-3β-KI (2.3%; Fig. 2B). Treatment with 3 Gy caused a significant increase in apoptosis in cells transfected with PRIG (18.5%) or GSK-3β-WT (14%), whereas transfection with GSK-3β-KI protected HT-22 neurons by decreasing radiation-induced apoptosis to 2.76% (Fig. 2B). Similar effects were observed in the analysis of Annexin V– and propidium iodide–stained cells (Fig. 2C). Transfection with GSK-3β-WT induced 2.2-fold higher apoptosis than with GSK-3β-KI or PRIG (Fig. 2C). Treatment with 3 Gy caused a significant increase in apoptosis in both PRIG- and GSK-3β-WT–transfected cells, 2- and 2.7-fold, respectively, as compared with sham-irradiated PRIG-transfected cells (Fig. 2C). However, transfection with GSK-3β-KI protected HT-22 neurons from radiation-induced apoptosis (Fig. 2C; 1.1-fold over the PRIG-transfected control). Taken together, these data show that ectopic expression of GSK-3β-KI protects HT-22 neurons from radiation-induced apoptosis.

GSK-3β inhibitors stabilize β-catenin and cyclin D1 in irradiated HT-22 cells. To characterize GSK-3β inhibition using chemical inhibitors, we studied four small-molecule inhibitors of GSK-3β. SB-216763 and SB-415286 are structurally distinct maleimides that inhibit GSK-3α/β in vitro with K_i of 9 and 31 nmol/L, respectively, in an ATP-competitive manner (25). The thiazole AR-A014418 and indirubin BIO inhibit GSK-3α/β with an IC₅₀ of 104±27 nmol/L (26) and 83 nmol/L (27), respectively, in an ATP-competitive manner. We studied the protein expression of cyclin D1 and β-catenin to monitor GSK-3β activity in HT-22 hippocampal neurons. Because active GSK-3β phosphorylates these proteins and promotes their degradation, the increased accumulation of β-catenin and cyclin D1 correlates with the decreased activity of GSK-3β and vice versa. All four inhibitors were observed to increase accumulation of β-catenin in the cytosol in a dose-dependent manner (Supplementary Fig. S1A). The maximal effective concentrations of GSK-3β inhibitors SB415286, SB216763, AR-AO14418, and BIO were 25, 10, 1, and 1 μmol/L, respectively (Supplementary Fig. S1A). The optimal time of treatment for SB415286 or SB216763 was 16 hours (Supplementary Fig. S1B). To analyze the effects of radiation on the GSK-3β-dependent pathway, HT-22 cells were treated with 10 μmol/L SB216763, 25 μmol/L SB415286, 1 μmol/L AR-AO14418, and 1 μmol/L BIO for 16 hours or with 3 mmol/L lithium for 7 days. The cells were lysed 6 hours after irradiation (3 Gy) and the lysates were studied by immunoblot analysis. Phosphorylation of GSK-3β at Ser⁹ (inhibitory) or at Tyr²¹⁶ (activating) and accumulation of β-catenin were not significantly changed in irradiated HT-22 neurons as compared with corresponding control cells, suggesting that radiation alone did not affect GSK-3β activity (Fig. 3). Treatment of HT-22 cells with each of the four GSK-3β inhibitors increased β-catenin protein level compared with their respective controls (Fig. 3). However, only pretreatment with SB415286 and lithium prevented radiation-induced decrease in accumulation of cyclin D1 (Fig. 3). In addition, pretreatment of irradiated HT-22 cells with SB216763 and SB415286 attenuated the phosphorylation of active form of GSK-3β^Tyr216 (Fig. 3). Based on these observations, we chose to study these two inhibitors in biological assays. Moreover, AR-A014418 cannot be used to study GSK-3β inhibition in the central nervous system because it lacks bioavailability within the brain (28).

GSK-3β inhibitors enhance clonogenic survival of irradiated hippocampal neurons. To determine the role of GSK-3β in cell viability, we first examined the effects of GSK-3β inhibitors on clonogenic cell survival in irradiated HT-22 hippocampal neurons. Pretreatment of HT-22 with 10 μmol/L SB216763 or 25 μmol/L SB415286 for 16 hours before irradiation significantly increased cell survival as compared with cells treated with radiation alone (Fig. 4A). Radioprotection was not observed in glioma cell lines D54 (human) and GL-261 (mouse) or human medulloblastoma cell line Daoy (Fig. 4A).

Apoptosis in irradiated HT-22 neurons is attenuated by GSK-3β inhibitors. Growing evidence supports the view that GSK-3β activation (29) and nuclear translocation (30) contribute to neuronal apoptosis. To elucidate the mechanism of increased survival in hippocampal neurons, we examined the nuclear morphology of dying cells by DAPI staining (Fig. 4B). Irradiated HT-22 cells pretreated with GSK-3β inhibitors showed a protective effect with the numbers of apoptotic cells reduced to 5.34% (SB216763), 5.58% (SB415286), and 5.36% (lithium), as compared with 27.8% in DMSO-pretreated irradiated cells (Fig. 4B). To further confirm and evaluate the induction of apoptosis, we analyzed treated cells by using Annexin V-FITC and propidium iodide staining in flow cytometry assay (Fig. 4C). This assay was applied only for the cells treated with SB415286 because SB216763 showed the artifact of fluorescence similar to that of FITC. In the absence of GSK-3β inhibitors, significantly more Annexin V–positive cells (>55%) were observed in response to irradiation as compared with cells treated with either SB415286 (32%) or lithium (25%) before irradiation (Fig. 4C). We also determined the effect of radiation dose on the level of apoptosis in HT-22 cells by using various doses of irradiation. A dose-dependent increase in Annexin V–positive cells was observed in HT-22 cells (Supplementary Fig. S2). Pretreatment of HT-22 cells with SB415286 (25 μmol/L) or lithium (3 mmol/L) protected the cells from radiation-induced apoptosis even at higher doses (up to 15 Gy), with SB415286 being more effective than lithium as compared with DMSO (Supplementary Fig. S2; 24% and 28% for SB415286, 30% and 44% for lithium, and 76% and 83% for DMSO, at 10 and 15 Gy, respectively).

Because radiation alone did not affect GSK-3β activity in HT-22 neurons (Fig. 3), the molecular mechanism of the observed protection from radiation-induced apoptosis by GSK-3β inhibitors is not direct. Studies of the molecular mechanism of radiation cytotoxicity in HT-22 neurons showed accumulation of p53 (Fig. 4D), one of the main mechanisms of radiation-induced apoptosis (31). However, inhibition of basal level of GSK-3β activity before irradiation prevented radiation-induced p53 accumulation, possibly leading to cytoprotection (Fig. 4D; ref. 32).

GSK-3β inhibitors stabilize β-catenin in hippocampus of irradiated mice. To determine whether small-molecule inhibitors of GSK-3β penetrate mouse brain and indeed lead to stable inhibition of GSK-3β activity as compared with lithium, we treated 1-week-old mice with a single dose of 0.6 mg/kg SB216763, 1.0 mg/kg SB415286, or 40 mg/kg lithium followed by a single dose of 7 Gy of cranial irradiation. Hippocampal sections were prepared 24 hours after irradiation and analyzed for increased immunologic staining for β-catenin, an indicator of GSK-3β inhibition. Sections from control and irradiated animals showed a very low level of β-catenin (Fig. 5A). However, treatment with each of the three GSK-3β inhibitors led to dramatically increased accumulation of β-catenin that remained elevated following irradiation (Fig. 5A), confirming significant inhibition of GSK-3β activity in mouse hippocampus.

GSK-3β inhibitors protect hippocampal neurons in vivo from radiation-induced apoptosis. Radiation-induced learning and memory deficits in animal models are correlated with an increase in apoptosis within hippocampal neurons (33). The severity of the impairment depends on the dose delivered to the hippocampus, the volume irradiated, and the age of the subject (2). To determine the effect of radiation dose on radiation-induced apoptosis in the mouse hippocampus, we treated 1-week-old mouse pups with the recommended dose of 0.6 mg/kg SB216763 or 1.0 mg/kg SB415286 (34) to various fractionation schedules of radiation doses of 3 Gy and a single dose of 7 Gy. Apoptosis was evaluated by comparing the average number of TUNEL-positive cells per high-power field (TPC) within the subgranular zone of the hippocampus (Supplementary Fig. S3). A single dose of 7 Gy induced more apoptosis (75 TPC) than 4 (9 TPC), 5 (47 TPC), or 6 (56 TPC) fractions of 3 Gy (Supplementary Fig. S3). To determine the effective inhibitor dose, we treated 2-week-old mice with single doses of 0.6 mg/kg SB216763 or 1.0 mg/kg SB415286 and a 10-fold higher dose of 6.0 mg/kg SB216763 or 10.0 mg/kg SB415286 and subjected them to a single radiation dose of 7 Gy (Supplementary Fig. S4). Both the recommended doses of SB216763 (30 TPC) or SB415286 (26 TPC) and a 10-fold higher dose of SB216763 (23 TPC) or SB415286 (33 TPC) protected the hippocampal neurons from radiation-induced damage with similar efficacy (Supplementary Fig. S4).

To determine whether GSK-3β inhibitors regulate cell survival and apoptosis in the subgranular zone of the hippocampus in comparison with lithium, we pretreated 1-week-old mice with a single dose of 0.6 mg/kg SB216763, 1.0 mg/kg SB415286, or 40 mg/kg lithium and treated them with a single dose of 7 Gy of cranial irradiation. Hippocampal sections were analyzed for apoptosis within the subgranular zone by TUNEL staining (Fig. 5B and C). Mice pretreated with SB216763 and SB415286 showed significantly less TUNEL-positive neurons (<15 TPC; Fig. 5B) as compared with control (55 TPC; Fig. 5B). This protective effect of the GSK-3β inhibitors was comparable to that of lithium (10 TPC; Fig. 5B). Similar results were obtained in a stereologic approach when TUNEL-positive cells were counted in five tissue sections prepared throughout the hippocampus of treated mice at the various cutting deepness (Fig. 5C; <25 TPC for SB216763 pretreatment and <35 TPC for SB415286 and lithium pretreatment, as compared with <205 TPC for radiation alone).

Prophylaxis with small-molecule GSK-3β inhibitors improves cognitive function in irradiated mice. Cranial irradiation in newborn rodents is associated with severe spatial navigation deficits in the Morris water maze (35). These radiation-induced deficits can be accounted for by neuronal precursor cell dysfunction in the subgranular zone of hippocampus (7). Spatial learning and memory were evaluated in C57/BL6 mice using hidden platform Morris water maze paradigm (36). The mice were treated with 0.6 mg/kg SB216763 or 1.0 mg/kg SB415286 and irradiated with a single dose of 7 Gy to the cranium, which was sufficient to induce apoptosis in the subgranular zone (Fig. 5; Supplementary Fig. S3). Corresponding sham-irradiated controls were included for comparison. Mice in all six treatment groups learned the Morris water maze task equally well as indicated by the similar latencies in reaching the platform and time spent by circling the perimeter as well as by the decreasing trend in both latencies and time in the perimeter over the course of 7 days of training (Supplementary Fig. S5). Mice also had similar swim speeds (data not shown). In the probe trials, the irradiated mice treated with SB216763 or SB415286 spent significantly more time in the target quadrant as compared with mice treated with DMSO (Fig. 6A; 30 versus 16 seconds), indicating intact memory for platform location in mice receiving prophylactic GSK-3β inhibitors. Similarly, irradiated mice treated with SB216763 or SB415286 were more likely to swim over the former platform location (Fig. 6B,, crossings) as compared with equivalent locations in the other three quadrants (Fig. 6B,, pseudo-crossings). On average, the irradiated mice pretreated with SB216763 crossed 8 times and mice pretreated with SB415286 crossed 10 times, compared with mice treated with DMSO control, which crossed the platform an average of 3 times in a 60-second session (Fig. 6B). The observed protection was not gender dependent, showing similar effects in Morris water maze for female and male mice in test parameters, latencies, and crossings (Supplementary Fig. S6).

Radiation therapy is a curative treatment for children with medulloblastoma and in some leukemias. Cranial irradiation, however, results in memory and learning deficits in children (37). Age dependence of radiation-induced brain injury has relegated the need to delay radiotherapy in infants (38). Even children treated with low-dose cranial irradiation have significantly reduced employment rates as compared with children treated with chemotherapy alone (39).

Pathophysiologic studies of radiation-induced neurocognitive deficit have identified apoptosis in the subgranular zone of the hippocampus, which in part contributes to deficits in memory and spatial navigation. Studies of lithium as a neuroprotective agent in irradiated hippocampal neurons showed that lithium improves cognitive performances in irradiated mice (9). This finding led to a phase I clinical trial that shows safety in patients treated with lithium during whole-brain irradiation as treatment for brain metastases.⁴

The protective mechanism of lithium could be pleiotropic, including, but not limited to, inhibition of GSK-3β (9). Here we show that GSK-3β is the key regulator of radiation-induced apoptosis and a potential molecular target for the development of protective agents to attenuate radiation-induced neurotoxicity. In the present study, the kinase-inactive genetic construct and small molecule inhibitors of GSK-3β decreased apoptosis in irradiated hippocampal neurons and increased their survival.

GSK-3β signaling. GSK-3β is a multifunctional kinase that has a role in various signaling pathways that regulate cell fate. GSK-3β is involved in Wnt signaling, protein synthesis, glycogen metabolism, and apoptosis. We studied small-molecule GSK-3β inhibitors SB415286, SB216763, AR-AO14418, and BIO. We evaluated GSK-3β activity based on the protein stabilization and accumulation of its downstream targets, β-catenin and cyclin D1, as well as on the level of activating phosphorylation GSK-3β^Tyr216. Although all four inhibitors tested at the optimal dose and time of treatment have shown stabilization of β-catenin, only pretreatment of irradiated HT-22 cells with SB216763 and SB415286 caused decreased phosphorylation of active form of GSK-3β^Tyr216. GSK-3β is predominantly present in the cytoplasm but also localizes to the nucleus and mitochondria, where it is highly active as compared with cytoplasmic GSK-3β (40). Nuclear GSK-3β is particularly interesting due to the fact that it can influence many signaling pathways and regulate the expression of many genes. The nuclear level of GSK-3β fluctuates due to intercellular responses and is highest in S phase of the cell cycle. These high levels of GSK-3β facilitate phosphorylation of nuclear cyclin D1. Cyclin D1 is required to integrate extracellular signaling during G₁ phase to regulate DNA replication and mitosis (21). Treatment with SB415286 prevented radiation-induced degradation of cyclin D1 in irradiated neurons similar to that observed with lithium. Taken together, inhibition of GSK-3β by small molecule inhibitors SB216763 or SB415286 helps to stabilize β-catenin and cyclin D1. Therefore, this novel GSK-3β–dependent mechanism of neuroprotection may regulate the cellular response of the hippocampal neurons to irradiation.

Regulation of apoptosis by GSK-3β. GSK-3β has been shown to induce apoptosis in a wide variety of conditions including DNA damage (32), hypoxia (41), and stress of the endoplasmic reticulum (42). In cell culture studies, apoptosis is either attenuated or prevented by inhibition of GSK-3β in primary neurons (43) and HT-22 hippocampal neurons (9). HT-22 cells treated with SB216763 or SB415286 showed increased survival from radiation damage compared with irradiated control cells. This protection was observed only in hippocampal neurons and was not observed in the cancer lines GL261, D54, and Daoy. We have determined that this selective protection is related to the role of GSK-3β in proapoptotic signaling. Cancer cells typically do not undergo apoptosis in response to ionizing radiation, in contrast to hippocampal neurons from the subgranular zone. By inhibiting GSK-3β, we prevent radiation-induced apoptosis, which is minimal in irradiated cancer cells. This differential response to radiation provides a means to improve the therapeutic effect of cranial irradiation.

GSK-3β promotes apoptosis by inhibiting prosurvival transcription factors, such as cyclic AMP–responsive element binding protein and heat shock factor 1 (14), and facilitating proapoptotic transcription factors, such as p53 (32). We have shown that the activity of GSK-3β in hippocampal neurons was not significantly altered by radiation, suggesting indirect involvement of this enzyme in radiation-induced apoptosis. On the other hand, dependence of radiation-induced apoptosis on p53 function and interaction between GSK-3β and p53 had been reported (31, 32). In our current study, the p53 level was significantly increased in irradiated HT-22 neurons, suggesting a p53-dependent mechanism of radiation-induced apoptosis in these cells. Treatment of HT-22 neurons with GSK-3β inhibitors before radiation led to inhibition of the basal level of GSK-3β activity, which was sufficient to prevent radiation-induced p53 stabilization and cytoprotection.

GSK-3β is required for cell viability. Disruption of murine GSK-3β gene results in embryonic lethality due to tumor necrosis factor-α–induced hepatocyte apoptosis during mid-gestation (44). GSK-3β is required in the early development of the hippocampus and has a key role in neuronal polarity (45). We observed that C57/BL6 mouse pups treated with 0.6 mg/kg SB216763 or 1.0 mg/kg SB415286 showed less apoptosis (TUNEL-positive cells) as compared with control mice when subjected to 7-Gy radiation. This radioprotective effect could be due to the inactivation of GSK-3β in the brain leading to reduced apoptosis. In direct contrast, overexpressing GSK-3β in the brain decreased levels of β-catenin and hyperphosphorylation of tau leading to neuronal stress and apoptosis (46).

Dominant negative GSK-3β transgenic mice displayed improved performances in the Morris water maze when crossed with amyloid precursor protein transgenic mice (47). This was attributed to the preservation of the dendritic structure in the frontal cortex and hippocampus and decreased tau phosphorylation. Further, GSK-3β was implicated to decrease the levels of amyloid precursor protein phosphorylation, which resulted in decreased amyloid-β production (47). It has been shown that overexpression of GSK-3β impairs spatial learning (15). In our study, C57/BL6 mice treated with GSK-3β inhibitors were able to overcome the deficits resulting from cranial irradiation. The irradiated mice treated with SB216763 or SB415286 showed improved performance in learning and located the hidden water platform much better as compared with the irradiated controls. These improved performances could be due to the protective effects of SB216763 or SB415286 by inhibiting GSK-3β, leading to attenuation of radiation-induced apoptosis. Maintaining appropriate levels of GSK-3β activity in the cell is critical because too little or too much GSK-3β activity can promote cell death in certain conditions. The long-term effect of cranial irradiation on 2-week-old mouse pups is most analogous to the response in the pediatric brain. It is possible that the response of the adult brain, particularly the aged brain, may well be different because neurogenesis is age dependent and proposed treatment may not be equally effective in the aged individual as compared with children. Therefore, GSK-3β is a molecule target for the development of small molecule inhibitors that prevent deleterious neurocognitive consequences of cranial irradiation in pediatric patients and improve the quality of life in childhood cancer survivors.

D.E. Hallahan and E.M. Yazlovitskaya: Patent pending. D.K. Thotala disclosed no potential conflicts of interest.

Grant support: Grant from the Witmer Foundation, the Ingram Charitable Foundation, the Vanderbilt-Ingram Cancer Center, and NIH grants R01-CA125757, R01-CA89674, and R01-CA125656.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Patrick Martin for the bicistronic construct, Dr. David C. Seldin for the mouse GSK-3β constructs, and Drs. Michael McDonald and John Alison at Vanderbilt Murine Neurobehavioral Laboratory and Allie Fu for help with behavioral studies.