BDNF-mediated signal transduction is modulated by GSK3β and mood stabilizing agents

Authors


Address correspondence and reprint requests to Xiaohua Li, Department of Psychiatry and Behavioral Neurobiology, The University of Alabama at Birmingham, 1075 Sparks Center, 1720 7th Avenue South, Birmingham, AL 35294-00017, USA. E-mail: xili@uab.edu

Abstract

Brain-derived neurotrophic factor (BDNF) is a major neurotrophin in the brain and abnormal regulation of BDNF may contribute to the pathophysiology of mood disorders. In the present study, we examined if alterations in the activity of glycogen synthase kinase-3-beta (GSK3β) or treatment with mood stabilizers modulated BDNF-mediated signal transduction pathways in differentiated human neuroblastoma SH-SY5Y cells. BDNF increased the phosphorylation of the forkhead transcription factor FKHRL1 through activation of the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway, and the phosphorylation of the cyclic AMP response element binding protein (CREB) through activation of extracellular signal-regulated kinase1/2 (ERK1/2). BDNF also increased serine9-phosphorylation of GSK3β, which inhibits GSK3β activity. Overexpression of GSK3β did not affect BDNF-induced phosphorylation of Akt, ERK1/2, or FKHRL1, but abolished CREB phosphorylation induced by BDNF. This inhibition of BDNF-induced CREB phosphorylation in GSK3β-overexpressing SH-SY5Y cells was blocked by treatment with lithium. In contrast to lithium, sodium valproate and lamotrigine did not affect BDNF-mediated signaling, whereas carbamazepine induced a rapid and prolonged phosphorylation of ERK1/2 and CREB in the absence or the presence of BDNF. Therefore, increased GSK3β selectively attenuates BDNF-induced CREB phosphorylation, and lithium and carbamazepine can facilitate activation of CREB.

Abbreviations used:
BDNF

brain-derived neurotrophic factor

CREB

cyclic AMP response element binding protein

EGF

epidermal growth factor

ERK

extracellular signal-regulated kinase

FKHRL1

forkhead L1 transcription factor

GSK3β

glycogen synthase kinase-3-beta

IGF-1

insulin-like growth factor-1

PI3K

phosphatidylinositol 3-kinase.

Brain-derived neurotrophic factor (BDNF), a major neurotrophin in the brain, is an important regulator of neuronal function, differentiation, and survival. BDNF binds to the TrkB receptor to initiate multiple signaling cascades, including the phosphatidylinositol 3-kinase (PI3K)/Akt pathway and the extracellular signal-regulated kinase 1/2 (ERK1/2) pathway (Encinas et al. 1999). Both the PI3K/Akt and the ERK1/2 signaling pathways lead to the regulation of transcription factors to control gene expression. Among the transcription factors subject to regulation by growth factor signaling, FKHRL1, a member of the pro-apoptotic forkhead transcription factor family, is phosphorylated by Akt and this modification maintains FKHRL1 in an inactive state (Brunet et al. 1999; Rena et al. 1999). Although several growth factors have been found to induce FKHRL1 phosphorylation (Kuo et al. 1996; Rena et al. 1999), its regulation by BDNF has not been reported. Cyclic AMP response element binding protein (CREB) is a transcription factor regulated by multiple signal transduction pathways, including ERK1/2 and Akt in different cell types (Xing et al. 1998; Pugazhenthi et al. 2000). CREB has been reported to be a key mediator of BDNF-induced gene expression and cell survival (Finkbeiner 2000; Finkbeiner et al. 1997).

BDNF may be an important neurotrophic factor involved in mood disorders (Altar 1999; Duman et al. 2000). This is indicated by the findings that chronic psychosocial stress, an animal model for depression, decreases the expression of BDNF (Smith et al. 1995), and centrally administered BDNF has antidepressant effects in animal models of depression (Siuciak et al. 1997). Furthermore, chronic antidepressant treatment not only increases the expression of BDNF and its receptor TrkB, but also blocks the down-regulation of BDNF in response to stress (Nibuya et al. 1995; Siuciak et al. 1997; Chen et al. 2001b). These effects of antidepressants appear to be mediated by the cyclic AMP signal transduction system, which increases the expression of BDNF (Duman et al. 1997). Thus, although some therapeutic agents facilitate upstream events in the BDNF signaling pathway, it remains to be determined if drugs used in mood disorders also affect the downstream signals generated by BDNF-activated receptors.

Glycogen synthase kinase-3-beta (GSK3β) is an important intracellular regulatory protein that is subject to phosphorylation by growth factor-stimulated signaling pathways. GSK3β is inhibited upon serine9-phosphorylation by growth factor receptor-induced activation of Akt (Cross et al. 1995). GSK3β is a protein kinase that has important regulatory effects in neural plasticity and survival. For example, increased GSK3β activity impairs activation of CREB (Grimes and Jope 2001a) and promotes cell death (Pap and Cooper 1998; Bijur et al. 2000; Li et al. 2002a). Foulstone et al. (1999) recently reported that BDNF, like insulin and epidermal growth factor (EGF), can inhibit GSK3β through increased serine9-phosphorylation in cerebellar granule cells. However, whether GSK3β has a regulatory role in BDNF-mediated signaling activities, such as BDNF-induced activation of transcription factors, is not known.

The mood stabilizer lithium is a direct inhibitor of GSK3β (Klein and Melton 1996) and lithium has been shown to be neuroprotective through the inhibition of GSK3β (Bijur et al. 2000). Lithium inhibits the activity of GSK3β by reversible binding to GSK3β, a different mechanism from that of BDNF which inhibits GSK3β by increasing serine9-phosphorylation. In addition to lithium, our previous study showed that sodium valproate and lamotrigine, but not carbamazepine, suppressed GSK3β-facilitated cellular apoptosis (Li et al. 2002a), although the mechanisms of these actions remain to be identified. This raises the question of whether lithium and possibly other mood stabilizers may have a role in facilitating BDNF-mediated signaling. Therefore, in the present study, we identified signaling pathways activated by BDNF, including phosphorylation of FKHRL1 and CREB in differentiated SH-SY5Y human neuroblastoma cells, and examined the effects of increased GSK3β activity and of mood stabilizers on BDNF-mediated signaling.

Materials and methods

Cell culture

Human neuroblastoma SH-SY5Y cells were grown on Corning 100-mm tissue culture dishes (Corning, NY, USA) in continuous culture RPMI-1640 medium (Cellgro, Herndon, VA, USA) containing 10% horse serum (Life Technologies, Gaithersburg, MD, USA), 5% fetal clone II (Hyclone, Logan, UT, USA), 2 mm l-glutamine, 100 U/mL penicillin, and 100 µg/mL streptomycin (Life Technologies), and were maintained at 37°C in 95% air and 5% CO2. Stably transfected SH-SY5Y cells overexpressing HA-tagged GSK3β (Bijur et al. 2000) were maintained in medium containing 100 µg/mL G418 (Alexis Biochemicals, San Diego, CA, USA). For most experiments, cells were plated at a density of 5 × 104 cells per 100-mm dish and differentiated to a neuronal phenotype (Kaplan et al. 1993) by maintaining cells in a low serum (5%) medium containing 10 µm l all-trans-retinoic acid for 6 days. On the seventh day, serum was withdrawn, and cells were maintained in serum-free medium for 5 h before further treatment or harvest.

Immunoblotting

Cells were washed twice with phosphate-buffered saline and were lysed with lysis buffer containing 10 mm l Tris–HCl, pH 7.4, 150 mm l NaCl, 1 mm l EDTA, 1 mm l EGTA, 0.5% NP-40, 10 µg/mL leupeptin, 10 µg/mL aprotinin, 5 µg/mL pepstatin, 1 mm l phenylmethanesulfonyl fluoride, 1 mm l sodium vanadate, 100 nm l okadaic acid, and 0.1 mm lβ-glycerophosphate. The lysates were collected in microcentrifuge tubes, sonicated for 10 s, and centrifuged at 20 800 g for 15 min. Protein concentrations in the supernatants were determined using the Bradford protein assay (Bradford 1976) and the lysates were stored at − 80°C until used for immunoblotting.

Cell lysates were mixed with Laemmli sample buffer [2% sodium dodecyl sulfate (SDS)] and placed in a boiling water bath for 5 min. Proteins were separated in 10% SDS–polyacrylamide gels, and the proteins were transferred to nitrocellulose. Blots were probed with antibodies to Akt, phospho-serine473-Akt, ERK1/2, phospho-ERK1/2, phospho-serine9-GSK3β, CREB, phospho-serine133-CREB (Cell Signaling Technology, Beverly, MA, USA), FKHRL1, phospho-threonine32-FKHRL1, and phospho-serine253-FKHRL1 (Upstate Biotech, Lake Placid, NY, USA).

To detect TrkB receptors, cells were washed twice with phosphate-buffered saline and were collected in TE buffer containing 20 mm l Tris–HCl, pH 7.4, 2 mm l EDTA, 10 µg/mL leupeptin, 10 µg/mL aprotinin, 0.1 mm l phenylmethanesulfonyl fluoride, and 1 mm l sodium vanadate. Cells were homogenized and centrifuged twice at 500 g for 5 min. The supernatants were centrifuged at 100 000 g at 4°C for 1 h to separate the cytosol and membrane fractions. The pellet containing the membrane fraction was resuspended in lysis buffer and used for immunoblotting with anti-TrkB (Santa Cruz Biotechnology, Santa Cruz, CA, USA).

Results

BDNF-mediated signal transduction in retinoic acid-differentiated human neuroblastoma SH-SY5Y cells

Differentiation of SH-SY5Y cells with retinoic acid can induce expression of the TrkB receptor which is responsive to BDNF stimulation (Kaplan et al. 1993). To measure BDNF-activation of signaling pathways in SH-SY5Y cells, the expression of TrkB receptors was measured after cells were treated with 10 µm l all-trans-retinoic acid (Fig. 1a). The level of TrkB receptors increased during differentiation with retinoic acid for 0–6 days. Parallel measurements of BDNF-mediated phosphorylation of Akt and ERK1/2, components of the two major BDNF signaling pathways, are shown in Fig. 1(b). Retinoic acid itself increased the levels of Akt and ERK1/2 after treatment for 4–6 days, and also increased the phosphorylation of Akt and ERK1/2 after 2 days of treatment, which remained stable at this higher level from 2 to 6 days of treatment. BDNF (5 ng/mL) did not significantly increase the phosphorylation of Akt or ERK1/2 until 4 days of treatment with retinoic acid. Therefore, the following experiments were conducted after cells were differentiated with retinoic acid for 6 days to obtain optimal BDNF-induced signaling.

Figure 1.

Retinoic acid-induced TrkB receptor expression and BDNF signaling in human neuroblastoma SH-SY5Y cells. SH-SY5Y cells were differentiated with 10 µm l retinoic acid (RA) for the indicated length of time (0–6 days) and then incubated in serum-free media for 5 h. BDNF (5 ng /mL) was added to the media 5 min before cells were washed with phosphate-buffered saline and whole cell extracts were prepared. Electrophoresis was conducted using 20 µg protein of whole cell lysates, and samples were immunoblotted for (a) TrkB and (b) phospho-serine473-Akt, total Akt, phospho-ERK1/2, and total ERK1/2. Each immunoblot shown is representative of three experiments with similar results.

A time course of BDNF-induced responses in differentiated SH-SY5Y cells showed that BDNF induced a rapid but transient increase of Akt phosphorylation, which peaked at 5 min and returned to baseline after 45 min of treatment (Fig. 2a). BDNF also increased ERK1/2 phosphorylation rapidly, with the maximal stimulation occurring at 5–10 min of treatment, followed by a gradual decline in phosphorylated ERK1/2, which remained above baseline through 60 min of treatment.

Figure 2.

Time course of BDNF-induced phosphorylation of Akt, ERK1/2, FKHRL1, and CREB. SH-SY5Y cells were differentiated with retinoic acid (10 µm l) for 6 days and then incubated in serum-free media for 5 h. BDNF (5 ng/mL) was added to the media for 5–60 min before cells were collected for protein extraction. Electrophoresis was conducted using 20 µg protein of whole cell lysates, and samples were immunoblotted for (a) phospho-serine473-Akt and phospho-ERK1/2, (b) phospho-threonine32-FKHRL1, phospho-serine254-FKHRL1, total FKHRL1, phospho-serine133-CREB, and total CREB using antibodies specific for the indicated proteins. Each immunoblot shown is representative of three experiments with similar results.

To identify transcription factors responsive to BDNF in SH-SY5Y cells, BDNF-induced phosphorylation of FKHRL1 and CREB were measured because FKHRL1 is known to be a substrate of Akt (Brunet et al. 1999; Rena et al. 1999), and CREB can be activated by phosphorylation of serine133 by either ERK or Akt in different cell types (Xing et al. 1998; Pugazhenthi et al. 2000). As shown in Fig. 2(b), BDNF robustly increased both threonine32- and serine253-phosphorylation of FKHRL1. This response was transient, with peak phosphorylation evident at 5 min, and the phosphorylation level returned to near baseline after 60 min of treatment. This indicates that BDNF, like insulin-like growth factor-1 (IGF-1) and EGF (Kuo et al. 1996; Rena et al. 1999), can increase the phosphorylation of the FKHRL1 transcription factor. In contrast, BDNF only moderately increased serine133-phosphorylation of CREB in differentiated SH-SY5Y cells, but the response lasted longer compared to the phosphorylation of FKHRL1. BDNF-induced CREB phosphorylation was evident at 5 min, it reached a peak at 30 min, and it remained elevated after 60 min of BDNF treatment.

To identify the specific signaling pathways that mediate BDNF-induced phosphorylation of FKHRL1 and CREB, LY249002, an inhibitor of PI3K which is upstream of Akt, and PD98059, which inhibits activation of ERK1/2, were applied in differentiated SH-SY5Y cells before cells were treated with BDNF (Fig. 3). LY249002 completely blocked BDNF-induced phosphorylation of Akt and FKHRL1, but had no effect on BDNF-induced phosphorylation of ERK1/2 and CREB. In contrast, PD98059 completely blocked BDNF-induced phosphorylation of ERK1/2 and CREB, but had no effect on BDNF-induced phosphorylation of Akt and FKHRL1. These results clearly identify the signaling pathways that mediate BDNF-induced phosphorylation of FKHRL1 and CREB in SH-SY5Y cells, with threonine/serine phosphorylation of FKHRL1 mediated by PI3K/Akt signaling, whereas the ERK signaling pathway mediates serine133-phosphorylation of CREB.

Figure 3.

Signaling pathways mediating BDNF-induced phosphorylation of FKHRL1 and CREB. SH-SY5Y cells were differentiated with retinoic acid (10 µm l) for 6 days and then incubated in serum-free media for 5 h. Cells were treated with LY249002 (20 µm l) or PD98059 (50 µm l) for 30 min and then with BDNF (5 ng /mL) for 5 min. Protein extracts were immunoblotted for phospho-serine473-Akt, phospho-ERK1/2, phospho-threonine32-FKHRL1, phospho-serine254-FKHRL1, and phospho-serine133-CREB. Each immunoblot shown is representative of three experiments with similar results.

GSK3β modulation of BDNF-mediated signaling in SH-SY5Y cells

GSK3β is known to be inhibited by phosphorylation of serine9 mediated by Akt, which can be induced by growth factors (Cross et al. 1995). However, the regulation of GSK3β by BDNF in neuroblastoma cells has not been reported previously. To determine if BDNF modulates GSK3β in differentiated SH-SY5Y cells, serine9-phosphorylation of GSK3β was measured after cells were treated with BDNF for 5–60 min (Fig. 4). Treatment with BDNF increased phospho-serine9-GSK3β, with a peak increase occurring at 5 min, followed by a gradual reduction which still remained higher than baseline after 60 min of BDNF treatment. This result is consistent with results in cerebellar granule cells in which BDNF, like insulin and EGF, induced serine9-phosphorylation of GSK3β (Foulstone et al. 1999).

Figure 4.

BDNF-induced serine9-phosphorylation of GSK3β. SH-SY5Y cells were differentiated with retinoic acid (10 µm l) for 6 days and then incubated in serum-free media for 5 h. Cells were treated with BDNF (5 ng/mL) for 5–60 min, and protein extracts were immunoblotted for phospho-serine9-GSK3β. The immunoblot shown is representative of three experiments with similar results.

Because BDNF increased serine9-phosphorylation and therefore inhibited GSK3β in SH-SY5Y cells, we further tested whether increased GSK3β activity is inhibitory for signaling processes downstream from BDNF stimulation. As there is no available activator of GSK3β to increase its activity in vivo, the effect of increased GSK3β activity on BDNF-mediated signaling was measured in HA-GSK3β stable-transfected SH-SY5Y cells. In these cells, the level of expression and activity of GSK3β is three- to fourfold of that found in wild-type SH-SY5Y cells (Bijur et al. 2000). In GSK3β-overexpressing cells, retinoic acid induced TrkB expression, although the level of expression was slightly lower than in control SH-SY5Y cells (Fig. 5a). BDNF (5 ng/mL) increased the phosphorylation of Akt and ERK1/2 in GSK3β-overexpressing cells (Fig. 5b), with a temporal pattern similar to that found in control SH-SY5Y cells. BDNF rapidly and transiently increased Akt phosphorylation, whereas phosphorylation of ERK1/2 was induced rapidly and lasted up to 60 min after BDNF treatment.

Figure 5.

Retinoic acid-induced TrkB receptor expression and BDNF signaling in GSK3β-overexpressing SH-SY5Y cells. (a) TrkB expression in control SH-SY5Y (SH) and GSK3β-overexpressing SH-SY5Y (GSK) cells in the absence or the presence of retinoic acid (RA) for 6 days. (b) Time course of BDNF-induced phosphorylation of Akt and ERK1/2 (left panel) and total Akt and ERK1/2 (right panel) in GSK3β-overexpressing SH-SY5Y cells. Cells were treated with retinoic acid for 6 days, incubated in serum-free media for 5 h, and then treated with BDNF (5 ng/mL) for the indicated time. Protein extracts were immunoblotted for phospho-Akt, Akt, phospho-ERK1/2, and ERK1/2. Each immunoblot shown is representative of three experiments with similar results.

To test if BDNF-induced phosphorylation of transcription factors is affected by increased GSK3β activity, BDNF-induced phosphorylation of FKHRL1 and CREB was measured in GSK3β-overexpressing cells. As shown in Fig. 6, BDNF increased the phosphorylation of FKHRL1 in GSK3β-overexpressing cells, with a similar temporal pattern and intensity as found in control SH-SY5Y cells. However, BDNF did not significantly increase CREB phosphorylation in GSK3β-overexpressing cells. Therefore, increased GSK3β had little effect on BDNF–induced phosphorylation of FKHRL1 but impaired the moderate stimulation of CREB phosphorylation induced by BDNF.

Figure 6.

BDNF-induced phosphorylation of FKHRL1 and CREB in GSK3β-overexpressing SH-SY5Y cells. Cells were differentiated with retinoic acid (10 µm l) for 6 days and then incubated in serum-free media for 5 h. Cells were then treated with BDNF (5 ng/mL) for 0, 5, or 30 min. Protein extracts were immunoblotted for phospho-threonine32-FKHRL1, phospho-serine254-FKHRL1, and phospho-serine133-CREB. Each immunoblot shown is representative of three experiments with similar results.

The effect of lithium on BDNF-mediated signaling pathways

To determine if lithium affects BDNF signaling, BDNF-induced protein phosphorylation in the presence or the absence of lithium (20 mm l for 1 h) was compared in control and GSK3β-overexpressing SH-SY5Y cells (Fig. 7). In control SH-SY5Y cells, lithium itself had no effect on the phosphorylation of Akt or FKHRL1, whereas it induced a small but significant increase in the phosphorylation of ERK1/2 and CREB (Fig. 7a,b). Lithium had no significant effect on BDNF-induced phosphorylation of Akt, FKHRL1, ERK1/2, or CREB in control SH-SY5Y cells. In GSK3β-overexpressing cells, lithium itself did not significantly affect the phosphorylation of Akt, FKHRL1, ERK1/2, or CREB (Fig. 7a,c). Similar to results in control SH-SY5Y cells, lithium had no effect on BDNF-induced phosphorylation of Akt and FKHRL1. In contrast, lithium significantly increased BDNF-induced phosphorylation of ERK1/2 and CREB.

Figure 7.

The effect of lithium on BDNF signaling in control (a,b) and GSK3β-overexpressing (a,c) SH-SY5Y cells. Cells were differentiated with retinoic acid (10 µm l) for 6 days and then incubated in serum-free media for 5 h. Cells were incubated in the presence or the absence of lithium (20 mm l) for 1 h and then with BDNF (5 ng/mL) for 5 min. (a) Representative immunoblots of phospho-serine473-Akt, phospho-threonine32-FKHRL1, phospho-serine254-FKHRL1, phospho-ERK1/2, and phospho-serine133-CREB. (b,c) Immunoreactive bands were analyzed by densitometer. Means ± SE, n = three experiments. In (c), each of the three experiments is from a different clone that overexpresses GSK3β (Bijur et al. 2000). *p < 0.05 (m lva) compared with untreated cells. **p < 0.05 (m lva) compared with BDNF-treated cells.

Using a high concentration of lithium (20 mm l) to elicit the maximal effect, the results shown in Fig. 7 indicated that lithium facilitated BDNF-induced phosphorylation of ERK1/2 and CREB in GSK3β-overexpressing cells. We then examined the concentration of lithium required to potentiate the phosphorylation of ERK1/2 and CREB induced by a lower concentration of BDNF (1 ng/mL) in GSK3β-overexpressing cells (Fig. 8). A low concentration of BDNF (1 ng/mL) had minimal effects on ERK1/2 and CREB phosphorylation in GSK3β-overexpressing cells, and lithium dose-dependently (1–10 mm l) potentiated the phosphorylation of ERK1/2 and CREB induced by 1 ng/mL BDNF, with the potentiation reaching a significant increase on CREB phosphorylation. This lithium concentration-dependent increase reflects the lithium dose-dependent inhibition of GSK3β that has been reported previously (Klein and Melton 1996; Grimes and Jope 2001a).

Figure 8.

The effect of lithium on BDNF-induced phosphorylation of ERK1/2 and CREB in GSK3β-overexpressing SH-SY5Y cells. Cells were differentiated with retinoic acid (10 µm lva) for 6 days and then incubated in serum-free media for 5 h. Cells were treated with the indicated concentrations of lithium for 1 h and then with a low concentration of BDNF (1 ng/mL) for 5 min. (a) Phospho-ERK1/2 and (b) phospho-serine133-CREB were detected by immunoblots and immunoreactive bands were analyzed by densitometer. Means ± SE, n = three experiments. *p < 0.05 (m lva) when compared with control (no lithium).

The effect of other mood stabilizers on Akt and ERK1/2 phosphorylation

The effects of anticonvulsant mood stabilizers, including sodium valproate, carbamazepine, and lamotrigine, on Akt and ERK1/2 phosphorylation in the absence and the presence of BDNF were examined in both control and GSK3β-overexpressing SH-SY5Y cells (Fig. 9a). Sodium valproate (10 mm l) itself had little effect on Akt and ERK1/2 phosphorylation and it did not affect the stimulatory effect of BDNF in both control and GSK3β-overexpressing cells. Cells treated with a dose range of sodium valproate from 0.5 mm l to 10 mm l for 1 h did not change Akt and ERK1/2 phosphorylation in the absence or the presence of BDNF (data not shown). Lamotrigine (300 µm l) also had no significant effect on Akt and ERK1/2 phosphorylation.

Figure 9.

The differential effects of sodium valproate, carbamazepine, and lamotrigine on BDNF-induced phosphorylation of Akt, ERK1/2 and CREB in control and GSK3β-overexpressing SH-SY5Y cells. Cells were differentiated with retinoic acid (10 µm lva) for 6 days and then incubated in serum-free media for 5 h. These cells received no further treatment (control), or were treated with sodium valproate (valproate, 10 mm lva), carbamazepine (90 µm lva), lamotrigine (300 µm lva), or EtOH (vehicle) for 1 h, and then with or without BDNF (5 ng/mL) for 5 min. (a) Phospho-Akt and phospho-ERK1/2 in control (SH) and GSK3β-overexpressing (GSK) cells. (b) Time course of the carbamazepine-induced increase in ERK1/2 phosphorylation in control SH-SY5Y cells. (c) The effect of PD98059 on the carbamazepine-induced increase in the phosphorylation of ERK1/2 and CREB in control SH-SY5Y cells. Each immunoblot shown is representative of three experiments with similar results.

Unlike sodium valproate or lamotrigine, carbamazepine had differential effects on Akt and ERK1/2 phosphorylation. Carbamazepine (90 µm l) increased basal ERK1/2 phosphorylation and enhanced BDNF-induced phosphorylation of ERK1/2 in both control and GSK3β-overexpressing cells. In contrast, it had no effect on Akt phosphorylation in the absence or the presence of BDNF. A time course of carbamazepine treatment showed that the induction of ERK1/2 phosphorylation occurred rapidly at 5 min and continued for 90 min of treatment (Fig. 9b). The effect of carbamazepine on ERK1/2 phosphorylation was completely blocked by PD98059 (Fig. 9c), indicating that the effect of carbamazepine was mediated by the ERK signaling pathway. As expected, carbamazepine also increased CREB phosphorylation (Fig. 9c) but had no effect on FKHRL1 phosphorylation (data not shown), which further supports the selective effect of carbamazepine on the ERK1/2 signaling pathway.

Discussion

The overall goals of this study were to determine if signaling systems activated by BDNF are regulated by GSK3β or mood stabilizers. Human neuroblastoma SH-SY5Y cells were used to study BDNF-mediated signaling because these cells express increased levels of TrkB receptors after differentiation with retinoic acid (Kaplan et al. 1993). Increased expression of TrkB receptors was evident between 2 and 6 days of differentiation, whereas the BDNF-induced phosphorylation of Akt and ERK1/2 was maximal after treatment with retinoic acid for 4–6 days. Therefore, we conducted experiments after cells were differentiated for 6 days with retinoic acid.

A comparison of BDNF-induced phosphorylation of Akt and ERK1/2 showed that BDNF induced a rapid but transient phosphorylation of Akt whereas phosphorylation of ERK1/2 induced by BDNF was prolonged. These differences likely reflect unique functions of each pathway. For example, it has been reported that Akt mediates the promotion of cell survival induced by BDNF and ERK1/2 mediates cell differentiation induced by BDNF (Encinas et al. 1999).

Because regulation of gene expression by transcription factors plays an important role in neural plasticity and survival, we examined BDNF-induced phosphorylation of FKHRL1 and CREB to determine which were linked to the PI3K/Akt and ERK signaling pathways in SH-SY5Y cells. FKHRL1 is known to be a pro-apoptotic transcription factor which is maintained inactive when phosphorylated and is activated by dephosphorylation (Brunet et al. 1999). Growth factors, such as IGF-1 and EGF, have been found to phosphorylate and inactivate FKHRL1 through activation of the PI3K/Akt signaling pathway. We found in the present study that FKHRL1 was phosphorylated by BDNF with a time course similar to BDNF-induced Akt phosphorylation, and the PI3K inhibitor LY249002 blocked BDNF-induced phosphorylation of FKHRL1. This indicates that FKHRL1 can be inhibited by BDNF-induced phosphorylation through the PI3K/Akt signaling pathway, an action that likely plays a role in the promotion of cell survival by BDNF.

BDNF and other growth factors are known to activate CREB by phosphorylation of serine133, a modification that is regulated by multiple signal transduction mechanisms. Although originally named due to its activation by cyclic AMP-dependent protein kinase (Yamamoto et al. 1988), both ERK and PI3K/Akt have been reported to mediate growth factor-induced CREB phosphorylation and activation (Xing et al. 1998; Pugazhenthi et al. 2000; Mehrhof et al. 2001). In SH-SY5Y cells, ERK1/2 was the main signaling pathway mediating BDNF-induced CREB phosphorylation because PD98059, which blocks activation of ERK1/2, completely blocked BDNF-induced CREB phosphorylation, whereas the PI3K inhibitor LY249002 had no effect on CREB phosphorylation. Compared to the rapid and transient phosphorylation of FKHRL1 induced by BDNF, CREB phosphorylation induced by BDNF was prolonged, which was similar to the increase in ERK1/2 phosphorylation stimulated by BDNF. These results suggest that prolonged phosphorylation of ERK1/2 may be required for sufficient activation of CREB by BDNF.

GSK3β is a substrate of Akt which phosphorylates and inhibits GSK3β, and thus, GSK3β is subject to inhibitory regulation by growth factors that activate the PI3K/Akt signaling pathway (Grimes and Jope 2001b). In the present study, we found that BDNF increased serine9-phosphorylation of GSK3β, a result consistent with the findings of Foulstone et al. (1999) in cerebellar granule cells. Because serine9-phosphorylation of GSK3β is the predominant mechanism used to inhibit its activity, we hypothesized that BDNF-mediated signaling might be impaired in conditions where there is insufficient inhibition of GSK3β. Therefore, we examined the effect of increased GSK3β on BDNF-induced phosphorylation in both the Akt and ERK1/2 signaling pathways using SH-SY5Y cells overexpressing active GSK3β by three- to fourfold (Bijur et al. 2000). Increased GSK3β activity had no effect on BDNF-induced Akt and FKHRL1 phosphorylation, indicating that GSK3β is not inhibitory for these responses to BDNF. In contrast, increased GSK3β activity suppressed BDNF-induced CREB phosphorylation without affecting ERK1/2 phosphorylation. This indicates that BDNF-induced inhibition of GSK3β normally may play an important role in the stimulation of CREB by BDNF. This result is consistent with the recent finding that GSK3β suppresses CREB activity (Bullock and Habener 1998; Grimes and Jope 2001a). This finding is especially interesting because of the evidence that BDNF plays an important role in the development and the treatment of mood disorders (Altar 1999; Duman et al. 2000), and CREB appears to be a key target of antidepressant treatment (Thome et al. 2000; Chen et al. 2001a). Our present findings suggest that GSK3β is a negative regulator of BDNF and CREB functioning.

Lithium is a mood stabilizer used in the treatment of bipolar disorder although its mechanism of action as a therapeutic agent is not clear. Lithium is a direct inhibitor of GSK3β (Klein and Melton 1996) and it inhibits the activity of GSK3β by reversibly binding to GSK3β, a different mechanism from BDNF, which inhibits GSK3β by increasing serine9-phosphorylation. We hypothesized that inhibition of GSK3β by lithium would facilitate BDNF-mediated signaling. Interestingly, we found that lithium selectively facilitated BDNF-induced phosphorylation of ERK1/2 and CREB, but had no effect on Akt or FKHRL1 phosphorylation induced by BDNF. The effect of lithium was especially significant in GSK3β-overexpressing SH-SY5Y cells, indicating that adequate inhibition of GSK3β is important for the maintenance of optimal activation of CREB by BDNF.

Similar to lithium, the mechanisms of the therapeutic effects of a group of anticonvulsant mood stabilizers, including sodium valproate, carbamazepine, and lamotrigine, also are unknown. Numerous studies have demonstrated that mood stabilizers influence multiple signal transduction mechanisms (Li et al. 2002b). We previously found that sodium valproate and lamotrigine, but not carbamazepine, suppressed GSK3β-facilitated apoptosis, an effect similar to lithium (Li et al. 2002a). In this study, we found that carbamazepine, but not sodium valproate or lamotrigine, rapidly increased the phosphorylation of ERK1/2 and CREB. The effect of carbamazepine was prolonged and was independent of BDNF. This effect of carbamazepine was completely blocked by PD98059. These findings indicate that carbamazepine activated the signaling pathway upstream of ERK leading to increased ERK1/2 and CREB phosphorylation. Thus both lithium and carbamazepine facilitated signaling leading to activation of CREB, but their mechanisms of action differed. Although treatment of differentiated SH-SY5Y cells with sodium valproate for 1 h did not affect Akt or ERK1/2 phosphorylation, a previous report by Yuan et al. (2001) showed that prolonged treatment with sodium valproate for 24 h increased ERK1/2 phosphorylation in undifferentiated SH-SY5Y cells. Their finding and our present results indicate that ERK signaling may be a common target of several mood stabilizers which regulate ERK signaling through different mechanisms.

In summary, the results of this study demonstrated that GSK3β is an inhibitory regulator of BDNF-induced signaling, and that this was selective for the pathway leading to phosphorylation of CREB, while BDNF-signaling leading to phosphorylation of FKHRL1 was independent of GSK3β. The results also showed that lithium and carbamazepine facilitate activation of CREB through different mechanisms, with the effect of lithium likely attributable to its inhibition of GSK3β, while carbamazepine activated signaling upstream of ERK leading to increased activation of CREB. The selective facilitation of ERK1/2 and CREB activation by two mood stabilizers further supports the potential importance of BDNF and CREB in the development of mood disorders and their potential role as therapeutic targets for antidepressants and mood stabilizers.

Acknowledgements

The authors thank Ling Song for help with many of the initial experiments in this study. Supported by NIH grant MH38752.

Ancillary