Address correspondence and reprint requests to Dr. H. K. Manji at Laboratory of Molecular Pathophysiology, Wayne State University School of Medicine, UHC 9B, 4201 St. Antoine Boulevard, Detroit, MI 48201, U.S.A.
Abstract : Valproic acid (VPA) is a potent broad-spectrum anti-epileptic with demonstrated efficacy in the treatment of bipolar affective disorder. It has previously been demonstrated that both VPA and lithium increase activator protein-1 (AP-1) DNA binding activity, but the mechanisms underlying these effects have not been elucidated. However, it is known that phosphorylation of c-jun by glycogen synthase kinase (GSK)-3β inhibits AP-1 DNA binding activity, and lithium has recently been demonstrated to inhibit GSK-3β. These results suggest that lithium may increase AP-1 DNA binding activity by inhibiting GSK-3β. In the present study, we sought to determine if VPA, like lithium, regulates GSK-3. We have found that VPA concentration-dependently inhibits both GSK-3α and -3β, with significant effects observed at concentrations of VPA similar to those attained clinically. Incubation of intact human neuroblastoma SH-SY5Y cells with VPA results in an increase in the subsequent in vitro recombinant GSK-3β-mediated 32P incorporation into two putative GSK-3 substrates (~85 and 200 kDa), compatible with inhibition of endogenous GSK-3β by VPA. Consistent with GSK-3β inhibition, incubation of SH-SY5Y cells with VPA results in a significant time-dependent increase in both cytosolic and nuclear β-catenin levels. GSK-3β plays a critical role in the CNS by regulating various cytoskeletal processes as well as long-term nuclear events and is a common target for both lithium and VPA ; inhibition of GSK-3β in the CNS may thus underlie some of the long-term therapeutic effects of mood-stabilizing agents.
Valproic acid (VPA), a simple branched fatty acid anticonvulsant, has demonstrated efficacy in the treatment of bipolar affective disorder (BD ; manic-depressive illness). Because the therapeutic effects of mood stabilizers like VPA are only observed after chronic administration (Bowden et al., 1994), they have been postulated to involve alterations of gene expression in critical neuronal circuits (Jope and Williams, 1994 ; Manji et al., 1995 ; Hyman and Nestler, 1996). In our endeavors to elucidate the molecular mechanisms by which mood-stabilizing agents may bring about their therapeutic effects, we have been concurrently investigating the effects of VPA and lithium. These are two structurally highly dissimilar agents, and although they likely do not exert their therapeutic effects by precisely the same mechanisms, identifying targets that they regulate in concert may provide important clues about the molecular mechanisms underlying mood stabilization. We previously examined VPA's effects on the DNA binding activity of activator protein-1 (AP-1) in cultured cells in vitro using gel mobility-shift assays (verified by both competition assays with unlabeled consensus and mutant AP-1 binding oligonucleotide and supershift assays with antibodies against c-Jun) (Chen et al., 1997). Incubation of rat C6 glioma cells with VPA resulted in one- to twofold increases in the DNA binding activity of AP-1 transcription factors in both a time- and concentration-dependent manner that occurred within the drug's therapeutic range. Similar VPA-induced increases in AP-1 DNA binding activity have also been observed in human SH-SY5Y neuro-blastoma cells and, importantly, in rat brain during in vivo treatment (Asghari et al., 1998 ; G. Chen and H. K. Manji, presented at the annual meeting of the American College of Neuropsychopharmacology, Waikola Village, Hawaii, 1997). Moreover, lithium has also recently been demonstrated to increase AP-1 DNA binding activity in cultured neurons and in rat brain during chronic in vivo treatment (Ozaki and Chuang, 1997Asghari et al., 1998). These data indicate that lithium and VPA, at concentrations similar to those attained clinically, exert effects on the AP-1 family of transcription factors ; however, the mechanisms underlying these effects have not been fully elucidated.
Recently, a hitherto completely unexpected target for the action of lithium has been identified. Lithium, at therapeutically relevant concentrations, has been demonstrated to inhibit glycogen synthase kinase (GSK)-3β (Klein and Melton, 1996). GSK-3β, which is regulated by protein kinase B and protein kinase C (Goode et al., 1992 ; Cross et al., 1995), phosphorylates c-jun at three sites adjacent to the DNA binding domain, thereby reducing AP-1 DNA binding (Lin et al., 1993 ; Stambolic et al., 1996). It has therefore been postulated that the lithium-induced increases in AP-1 DNA binding activity may be mediated, in part, via inhibition of GSK-3β (Ozaki and Chuang, 1997 ; Yuan et al., 1998 ; Jope, 1999). As VPA also increases AP-1 DNA binding activity, we have undertaken the present study to determine if VPA, like lithium, also regulates GSK-3.
MATERIALS AND METHODS
GSK-3 in vitro activity assay
Recombinant and purified GSK-3α and GSK-3β were from New England BioLabs (Beverly, MA, U.S.A.) and Upstate Biotechnology (Lake Placid, NY, U.S.A.). CREB phosphopeptide (a GSK-3 substrate) was from New England BioLabs. The reaction mixture (20 μl) contained 30 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 0.1 γg/μl bovine serum albumin, 2 μg of CREB phosphopeptides, 100 μM Mg·ATP, 0.5 μCi of [32P] ATP, 0.5 U of GSK-3α or -3β, and different concentrations of MgCl2 (0-40 mM). The reaction was carried out at room temperature for 20 min and stopped by addition of 75 mM H3PO4. 32P-incorporated CREB phosphopeptides were collected using Whatman P81 cation exchange chromatography paper, and the radioactivity was measured using a scintillation counter.
Effects of VPA on GSK-3α and GSK-3β
The reactions were carried out as above, with varying concentrations of VPA (0-6 mM), in the absence or presence of additional MgCl2 (20 mM). Additional studies were conducted in the presence of both VPA (0.6 mM) and lithium (1 mM), concentrations that are approximately those attained clinically in the treatment of BD.
Effects of VPA on GSK-3 and β-catenin levels in intact SY5Y cells
Cell culture studies were conducted largely as previously described (Chen et al., 1997). For the back-phosphorylation studies, SH-SY5Y cells were treated with 0.6 mM sodium valproate for 1 day, and then nuclear fractions were prepared. Five micrograms of nuclear protein was heated at 95°C for 3 min to inactivate endogenous protein kinases and phosphatases and then incubated with 1 U of GSK-3β at 37°C for 20 min in 20 μl of reaction mixture containing 20 mM Tris-HCl (pH 7.5), 10 mM MgCl2, 5 mM dithiothreitol, 250 μM ATP, and 1 μCi of [32P]ATP. The incubation was stopped by addition of 5 × Laemmli buffer, and the mixture was heated at 95°C for 3 min. The sample was then subjected to 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). For the immunoblotting studies, cells were treated with VPA for varying intervals (12 h-6 days). Cytosolic and nuclear fractions were prepared, and western blotting was performed with monoclonal anti-β-catenin antibodies (1 : 500 ; Transduction Laboratories, Lexington, KY, U.S.A.), using previously described methods (Chen et al., 1998). The autoradiogram was analyzed with an imaging system equipped with NIH Image 1.55 software.
RESULTS AND DISCUSSION
Effects of VPA on GSK-3 activity in vitro
Mg2+ potently increased the activity of both GSK-3α and GSK-3β, with maximal stimulation of both being observed with 20 mM MgCl2 (Fig. 1a). Assays were conducted in the presence of 150 μM Mg2+ unless otherwise specified. VPA dose-dependently inhibited the activity of GSK-3 isoforms (Fig. 1b and c). In the absence of additional Mg2+, 0.6 mM VPA produced a significant 54% reduction in the activity of GSK-3α (Fig. 1b). In the presence of 20 mM Mg2+, 0.6 mM VPA produced a smaller but significant 25% reduction in the activity of GSK-3α (Fig. 1c). Similarly, in the absence of additional Mg2+, 0.6 mM VPA produced a marked (86%) inhibition of GSK-3β activity (Fig. 1b) ; in the presence of 20 mM Mg2+, 0.6 mM VPA produced a smaller but significant reduction in the activity of GSK-3β activity (~22% reduction ; Fig. 1c).
Because many BD patients refractory to either VPA or lithium alone are often coadministered both treatments (Freeman and Stoll, 1998), we sought to investigate if these two agents had additive effects. Addition of lithium (1.0 mM) resulted in an additional ~35% reduction in the activity of GSK-3α and GSK-3β in the presence of low (150 μM) Mg2+ and an additional ~20% reduction in the activity of GSK-3α and GSK-3β in the presence of high (20 mM) Mg2+ (Fig. 1d). These additive effects of lithium and VPA on GSK-3 suggest that the two drugs may exert their effects at different sites, but additional studies will be necessary to establish this definitively. To determine if all mood stabilizers inhibit GSK-3, we investigated the effects of carbamazepine. Carbamazepine at concentrations ranging from 5 to 500 μM (plasma levels during clinical treatment are ~40 μM) was completely without effect on GSK-3α or GSK-3β activity (data not shown).
Effects of VPA on endogenous GSK-3β and β-catenin levels in intact SH-SY5Y cells
To determine if inhibitory effects of VPA are also observed on endogenous GSK-3β in intact human cells of neuronal origin, a back-phosphorylation assay (Guitart and Nestler, 1992 ; Atkins et al., 1997) was used. Incubation of human neuroblastoma SH-SY5Y cells with 0.6 mM VPA for 24 h (Fig. 2a and b) resulted in a significant increase in the subsequent in vitro recombinant GSK-3β-mediated 32P incorporation into two putative GSK-3 substrates (~85 and 200 kDa). These results are consistent with an inhibition of endogenous GSK-3β activity by VPA in intact SY5Y cells, resulting in the availability of additional substrate phosphorylation sites for the subsequent 32P incorporation mediated by the recombinant enzyme. To investigate in vivo GSK-3β inhibition more directly, we examined VPA's effects on a major GSK-3 substrate, β-catenin. It has been clearly established that GSK-3β regulates β-catenin levels (Yost et al., 1996 ; Ikeda et al., 1997 ; Larabell et al., 1997). Inhibition of GSK-3β results in β-catenin accumulation, likely due to a decrease in the rate of β-catenin protein degradation ; monitoring β-catenin levels has thus been used as a means of monitoring in vivo GSK-3β activity (Hedgepeth et al., 1997). As would be expected from in vivo inhibition of GSK-3β, VPA produced a marked increase in both cytosolic and nuclear β-catenin levels in a time-dependent fashion (Fig. 2c and d) ; clearly additional mechanisms may also be operative.
In conclusion, we have demonstrated that VPA, like lithium, exerts significant inhibitory effects on the activity of GSK-3β both directly in vitro and also on endogenous GSK-3β in intact human neuroblastoma SY5Y cells. Significant inhibitory effects are clearly observed at concentrations of VPA approximating those attained clinically in the treatment of BD. Furthermore, addition of lithium at therapeutic concentrations results in additive inhibitory effects to that of VPA. At present, the precise therapeutic relevance of these novel findings is unclear. It is noteworthy that GSK-3β is known to play a critical role in regulating long-term events in the CNS by modulating various cytoskeletal processes via its effects on tau, neurofilaments, and synapsin I ; also, GSK-3β may regulate gene expression in the CNS via phosphorylation of c-jun, nuclear translocation of β-catenin, and nuclear export of NF-ATc (Klein and Melton, 1996 ; Wagner et al., 1996 ; Yost et al., 1996 ; Ikeda et al., 1997 ; Lucas and Salinas, 1997). Thus, inhibition of GSK-3β in the CNS by VPA and lithium may underlie some of the long-term therapeutic effects of these mood-stabilizing agents and is worthy of further study.
The editorial assistance of Ms. Celia Knobelsdorf and the support of the Theodore and Vada Stanley Foundation and the Joseph Young Senior Foundation are gratefully acknowledged.