Glycogen Synthase Kinase-3β (GSK3β) Binds to and Promotes the Actions of p53

Abstract

The recent discovery of direct interactions between two important regulators of cell fate, the tumor suppressor p53 and glycogen synthase kinase-3β (GSK3β), led us to examine the mechanism and outcomes of this interaction. Two regions of p53 were identified that regulate its binding to GSK3β. Deletion of the p53 activation domain-1 (AD1), but not mutations that prevent MDM2 binding through the AD1 domain, enhanced GSK3β binding to p53, indicating that the AD1 domain interferes with p53 binding to GSK3β. Deletion of the p53 basic domain (BD) abrogated GSK3β binding, and a ten amino acid region within the C-terminal BD domain was identified as necessary for binding to GSK3β. GSK3β activity was not required for p53 binding, but inhibition of GSK3β stabilized the association, suggesting a transient interaction during which active GSK3β promotes actions of p53. This regulatory role of GSK3β was demonstrated by large reductions of p53-induced increases in the levels of MDM2, p21, and Bax when GSK3β was inhibited. Besides promoting p53-mediated transcription, GSK3β also contributed to mitochondrial p53 apoptotic signaling. After DNA damage, mitochondrial GSK3β co-immunoprecipitated with p53 and was activated, and inhibition of GSK3β blocked cytochrome c release and caspase-3 activation. Thus, GSK3β interacts with p53 in both the nucleus and mitochondria and promotes its actions at both sites.

The tumor suppressor protein, p53, is a transcription factor that is activated by a wide variety of cellular stresses, notably including conditions that cause DNA damage (1–3). Post-translation modifications, such as phosphorylation or acetylation, and interactions with other proteins are important regulators of the level and activity of the normally short-lived p53 (3, 4). For example, p53 binding by MDM2 enhances its degradation, while inhibition of MDM2 can result in stabilization and accumulation of p53. Activated p53 induces the expression of a variety of gene targets, including p21^waf1/cip1, MDM2, Bax, and others, leading to two well-defined cellular responses, cell cycle arrest, and apoptosis.

Another enzyme that can contribute to apoptotic signaling is glycogen synthase kinase-3β (GSK3β)¹ (5). GSK3β is a serine/threonine kinase that plays crucial roles in development, such as in the Wnt developmentally regulated signaling pathway (6). GSK3β is also a critical component in several receptor-coupled signaling pathways (7), such as those emanating from growth factor-stimulated receptors that activate the intermediary protein kinase Akt, which phosphorylates and inhibits GSK3β (8). Additionally, activated GSK3β has been shown to promote apoptotic signaling in a number of conditions (9–14). Especially noteworthy was the finding that apoptosis induced by the overexpression of GSK3β was p53-dependent (10). Furthermore, several apoptotic stimuli recently were found to cause the accumulation of GSK3β in the nucleus, co-localizing it with p53 (15). Most recently, nuclear GSK3β was found to directly associate with p53 after DNA damage, and evidence that GSK3β contributes to p53-induced signaling was reported (16). This evidence included the findings that p53-mediated increases in p21 protein levels and caspase-3 activation were blocked by lithium, a selective inhibitor of GSK3β (17), by expression of dominant-negative GSK3β, and by selective inhibition of nuclear GSK3β by means of nuclear-targeted expression of the inhibitory GSK3β-binding protein, GBP (16).

The present study was designed to further clarify the interactions between GSK3β and p53. The results identify two regions of p53 that regulate binding to GSK3β, demonstrate that inhibition of the activity of GSK3β stabilizes its association with p53, documents mitochondrial as well as nuclear association of GSK3β and p53, and show that GSK3β contributes to the transcriptional and apoptotic actions of p53.

EXPERIMENTAL PROCEDURES

Cell Culture

Human neuroblastoma SH-SY5Y and p53-null human lung carcinoma H1299 cells that express inducible wild-type or mutated forms of p53 were grown as described previously (12, 18). Subcellular fractions were prepared as described previously (15). Drugs used include lithium (Sigma), rottlerin (Tocris), and 2-thio(3-iodobenzyl)-5-(1-pyridyl)-[1,3,4]-oxadiazole (tip-oxadiazole; marketed as GSK-3 Inhibitor II; Calbiochem).

Immunoblot Analysis

For immunoblotting, cells were washed twice with phosphate-buffered saline, and were lysed with lysis buffer (20 mm Tris, pH 7.5, 150 mm NaCl, 2 mm EDTA, 2 mm EGTA, 1 mm sodium orthovanadate, 50 mm sodium fluoride, 100 μm phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, 10 μg/ml aprotinin, 5 μg/ml pepstatin A, 1 nm okadaic acid, and 0.2% Nonidet P-40). The cell lysates were collected in centrifuge tubes, sonicated, and centrifuged at 16,000 × g for 10 min at 4 °C. Protein concentrations were determined using the bicinchoninic acid method (Pierce). For immunoprecipitation, samples were incubated with the indicated primary antibody overnight at 4 °C with gentle agitation, followed by incubation with protein G- or protein A-Sepha-rose beads (Amersham Biosciences) for 1 h at 4 °C, and the immune complexes were washed three times. Samples were mixed with Laemmli sample buffer (2% SDS) and placed in a boiling water bath for 5 min. Proteins were resolved in SDS-polyacrylamide gels, transferred to nitrocellulose, and membranes were probed with antibodies to p53 (Upstate Biotechnology, Inc., Lake Placid, NY), GSK3β, p21, proteolyzed poly(ADP-ribose) polymerase (PARP) 85-kDa fragment (BD Pharmingen, San Diego, CA), anti-active casapse-3 (Cell Signaling, Beverly, MA), p53 (Ab-2), MDM2 (Oncogene, San Diego, CA), or HA tag (Covance, Berkeley, CA). Immunoblots were developed using horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit IgG, followed by detection with enhanced chemiluminescence. Protein bands were quantitated with a densitometer and each experiment was carried out two or more times.

GSK3β Activity

The activity of GSK3β was measured essentially as described previously (12). Briefly, GSK3β was immunoprecipitated and activity was measured by incubation in 30 μl of kinase buffer (20 mm Tris, pH 7.5, 5 mm MgCl₂, 1 mm dithiothreitol, 250 μM ATP, 1.4 μCi [γ-³²P]ATP; Amersham Biosciences), with 0.1 μg/μl recombinant tau protein (Panvera, Madison, WI) for 30 min at 30 °C. Laemmli sample buffer (25 μl) was added to stop the reaction. Samples were placed in a boiling water bath for 5 min, proteins were separated in 8% SDS-polyacrylamide gels, gels were vacuum-dried, exposed to a phosphoscreen overnight, and quantitated using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The efficiency of GSK3β immunoprecipitation was determined by immunoblotting for GSK3β.

Northern Analysis

Northern blot measurements were carried out as described previously (19). Total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. RNA was separated by electrophoresis in 1% agarose gels containing formaldehyde and transferred to nitrocellulose membranes. p21 or MDM2 cDNA (18) was random prime-labeled with [³²P]dCTP (Amersham Biosciences). Blots were hybridized with labeled probes at 42 °C for 18 h and then washed in two changes of 2× saline-sodium citrate and 0.1% SDS at 20 °C for 20 min and once in 1× saline-sodium citrate and 0.1% SDS at 55 °C for 10 min. Results were obtained using a PhosphorImager (Molecular Dynamics), and bands were quantitated using ImageQuant. All experiments were repeated two or more times.

RESULTS

p53 Domains Regulating Its Interaction with GSK3β

In order to identify the domains of p53 that regulate its interaction with GSK3β, we used p53-null H1299 cells that inducibly express HA-tagged wild-type or deletion mutants of p53, which have been described previously (18). The domains of p53 that were investigated are diagrammed in Fig. 1A. In addition to wild-type p53, we studied p53 with deletions of the activation domain 2 (ΔAD2) comprising residues 43–63, of the activation domain 1 (ΔAD1) comprising residues 1–42, of the proline-rich domain (ΔPRD) comprising residues 64 –92, and of the C-terminal basic domain (ΔBD) comprising residues 364 –393. For each of these, cells expressing wild-type and mutant p53 were examined in parallel. After 24 h of induced expression of p53, cells were harvested, nuclear fractions were prepared, the levels of nuclear p53 were measured by immunoblot analysis, and GSK3β was immunoprecipitated. This was followed by measurements of the levels of immunoprecipitated GSK3β and of co-immunoprecipitated p53 to identify which domains of p53 affected its association with GSK3β.

Fig. 1. Domains of p53 that regulate its association with GSK3β.

A, diagram of the domains of p53 that were studied. B, AD2: The expression of wild-type p53 and ΔAD2-p53 was induced in H1299 cells for 24 h, followed by preparation of nuclear extracts, immunoblotting p53, immunoprecipitating GSK3β, and immunoblotting the immunoprecipitated GSK3β and the co-immunoprecipitated p53. C, AD1: The expression of wild-type p53 and ΔAD1-p53 was induced for 24 h, followed by preparation of nuclear extracts, immunoblotting p53, immunoprecipitating GSK3β, and immunoblotting the immunoprecipitated GSK3β and the co-immunoprecipitated p53. D, AD1(−)-p53: The expression of wild-type p53 and p53 containing an intact AD1 domain but with two point mutations, (L22Q, W23S), called AD1(−)-p53, was induced for 24 h and compared with ΔAD1-p53. Full induction of the two constructs with AD1 modifications led to much higher protein levels than of wild-type p53. In order to adjust this, partial induction was used for the two constructs with: full withdrawal of doxycycline for wild-type p53, 1 ng/ml doxycycline for AD1(−)-p53, and 0.1 ng/ml doxycycline for ΔAD1-p53, to obtain levels of the p53 constructs that were similar to wild-type p53. This was followed by preparation of nuclear extracts, immunoblotting p53, immunoprecipitating GSK3β, and immunoblotting the immunoprecipitated GSK3β and the co-immunoprecipitated p53. E, PRD and BD: The expression of wild-type p53, ΔPRD-p53, and ΔBD-p53 was induced for 24 h, followed by preparation of nuclear extracts, immunoblotting p53, immunoprecipitating GSK3β, and immunoblotting the immunoprecipitated GSK3β and the co-immunoprecipitated p53. F, Δ(374 –393)-p53: The expression of wild-type p53 and Δ(374 –393)-p53 was induced for 24 h, followed by preparation of nuclear extracts, immunoblotting p53, immunoprecipitating GSK3β, and immunoblotting the immunoprecipitated GSK3β and the co-immunoprecipitated p53. G, diagram of the region of p53 required for binding to GSK3β.

Results obtained with ΔAD2-p53 and ΔAD1-p53 are shown in Fig. 1, B and C, respectively. The expression level of ΔAD2-p53 in the nucleus was nearly identical (89 ± 8%; mean ± S.E.; n = 4) to that of wild-type p53, with ΔAD2-p53 demonstrating slightly faster mobility than wild-type p53 due to the deletion of the AD2 domain (Fig. 1B). Additionally, equal amounts of nuclear GSK3β were immunoprecipitated from each cell line, and a similar amount of ΔAD2-p53 (76 ± 19%; n = 3) as wild-type p53 co-immunoprecipitated with the GSK3β (Fig. 1B). Thus, the AD2 domain of p53 does not participate in regulating binding to GSK3β.

Different results were obtained with ΔAD1-p53 (Fig. 1C). The levels of nuclear ΔAD1-p53 were higher than wild-type p53 (208 ± 53% of wild-type p53; p < 0.05; n = 4) because p53 stability is increased by deletion of the MDM2 binding site in the AD1 domain of p53. Most notably, although equivalent amounts of nuclear GSK3β were immunoprecipitated, greater amounts of ΔAD1-p53 compared with wild-type p53 co-immunoprecipitated with GSK3β (1381 ± 543% of wild-type p53; p < 0.05; n = 4). Thus, the amount of p53 associated with GSK3β normalized to the nuclear level of p53 was 6-fold greater for ΔAD1-p53 than wild-type p53 (6.1 ± 1.1 compared with the ratio with wild-type p53; p < 0.05; n = 4). These data show that deletion of the AD1 domain facilitates the association of GSK3β and p53, indicating that the AD1 region of p53 can interfere with the binding of GSK3β to p53. This inhibition could be due to interference of the AD1 domain of p53 with the binding of GSK3β, or due to impaired access of GSK3β to p53 by MDM2 bound to the AD1 domain of wild-type p53. To examine this question further, we studied the association of GSK3β with a p53 construct, AD1(−), containing an intact AD1 domain but with two point mutations (L22Q, W23S) which inhibit MDM2 binding to p53 (20). After induction the level of AD1(−)-p53 was extremely high, therefore expression was controlled by titration of the doxycycline concentration to obtain a level of AD1(−)-p53 similar to that of fully induced wild-type p53. Similarly, the expression of ΔAD1-p53 also was limited to match the level of wild-type p53. Using these methods, the expression levels of all p53 constructs were similar, as were the immunoprecipitated amounts of GSK3β (Fig. 1D). Upon examination of the co-immunoprecipitation of AD1(−)-p53 and wild-type p53 with GSK3β, these were found to be equivalent, whereas more ΔAD1-p53 (5-fold) co-immunoprecipitated (Fig. 1D). This result shows that eliminating the binding site for MDM2 was not sufficient to enhance the binding of p53 to GSK3β. Taken together, these results demonstrate that a functional AD1 domain of p53 has an inhibitory regulatory role in the association of p53 with GSK3β independently of the binding of MDM2.

Comparisons of the association of GSK3β with p53 also were made using ΔPRD-p53 and ΔBD-p53. The levels of nuclear p53, immunoprecipitated GSK3β, and co-immunoprecipitated p53, were equivalent in cells expressing ΔPRD-p53 and wild-type p53 (Fig. 1E). In contrast, experiments with the p53 construct with deletion of the basic domain demonstrated that this region was critical for the association of p53 with GSK3β. Unlike all of the other p53 constructs, no ΔBD-p53 was detected in immunoprecipitates of GSK3β although the levels of nuclear p53 and of immunoprecipitated GSK3β were similar using cells expressing ΔBD-p53 and wild-type p53 (Fig. 1E). This result indicated that the BD domain was necessary for the binding of GSK3β to p53. To examine this further, we used cells expressing p53 containing a partial deletion (residues 374 –393) of the BD domain (residues 364 –393). These measurements revealed that Δ(374 –393)-p53 co-immunoprecipitated with GSK3β equivalently with wild-type p53 (Fig. 1F). Taken together these results indicate that residues 364 –373 of p53, consisting of the amino acid sequence AHSSHLKSKK, are critical for the association of p53 with GSK3β (Fig. 1G).

Recent structural studies of GSK3β have demonstrated that a phosphorylated serine on a GSK3β substrate binds the “primed substrate binding pocket” of GSK3β to facilitate phosphorylation by GSK3β of a serine four amino acids removed from the phosphorylated serine (21, 22). Within this binding pocket of GSK3β is arginine-96, which is critical for binding to primed substrates, and mutated R96A-GSK3β is unable to bind primed substrates (22). The sequence of the GSK3β-binding region of p53 contains a potential “primed” phosphorylation site for GSK3β binding, SXXXS(p), if serine 371 of p53 is phosphorylated. Although this is unlikely since little post-translational modifications occur when p53 is overexpressed in unstressed cells, to test if p53 binds GSK3β in this manner, wild-type p53 and either wild-type-GSK3β or R96A-GSK3β (each with a HA tag to distinguish them from endogenous GSK3β) were expressed in H1299 cells. Subsequently, p53 was immunoprecipitated, and the co-immunoprecipitation of GSK3β with p53 was examined. The levels of expressed p53 and of the variants of GSK3β were essentially equivalent in the cells expressing wild-type GSK3β and R96A-GSK3β (Fig. 2A). Following immunoprecipitation of p53, the amount of co-immunoprecipitated GSK3β was the same for wild-type GSK3β and R96A-GSK3β (103 ± 12%; n = 3) (Fig. 2B). These results are consistent with hypothesis that the association of p53 with GSK3β is not mediated by the binding of pre-phosphorylated p53 to the phosphobinding pocket of GSK3β.

Fig. 2. Association of p53 with wild-type GSK3β and R96A-GSK3β.

H1299 cells were transiently transfected with wild-type, untagged p53, and either wild-type-GSK3β (WT) or R96A-GSK3β, each with an HA tag. A, immunoblots show equivalent levels of endogenous GSK3β, expressed GSK3β, and p53 in each sample. B, 24 h after transfection, p53 was immunoprecipitated (DO-1 antibody), and the levels of the immunoprecipitated p53 and the co-immunoprecipitated GSK3β were measured by immunoblot analyses which showed equivalent association of wild-type GSK3β and R96A-GSK3β with p53.

The binding of p53 to GSK3β previously was reported to increase the activity of GSK3β (16), but those experiments did not examine if this was a transient interaction temporally limited by the activity of GSK3β. To address this issue, cells were incubated with lithium, a selective inhibitor of GSK3β (17), to test if GSK3β activity influenced its association with p53. Following induction of wild-type p53 expression in H1299 cells, the association of p53 with endogenous GSK3β was examined by co-immunoprecipitation from nuclear fractions. Inhibition of GSK3β with 20 mm lithium did not inhibit the co-immunoprecipitation of p53 with GSK3β, but to the contrary increased the association (Fig. 3, A and B). These results indicate that GSK3β activity is not required for association with p53 and that inhibition of GSK3β stabilizes its association with p53.

Fig. 3. Influence of GSK3β activity on the association of p53 with GSK3β.

The expression of wild-type p53 was induced in H1299 cells for 24 h in the presence or absence of 20 mm lithium, a concentration capable of inhibiting GSK3β by more than 80% (17). Nuclear extracts were prepared, p53 levels were measured by immunoblot analysis, GSK3β was immunoprecipitated, and the levels of immunoprecipitated GSK3β and co-immunoprecipitated p53 were measured by immunoblotting. A, representative immunoblots, and B, quantitative values obtained from three independent experiments, showing that inhibition of GSK3β stabilized the association of GSK3β and p533.

Functional Regulation of p53 by GSK3β

Induction of p53 leads to increased expression of a large group of p53-regulated proteins (23), including regulators of p53 (e.g. MDM2), of the cell cycle (e.g. p21), and of cell death (e.g. Bax). To test if these outcomes of p53 activation show differential sensitivities to the influence of GSK3β, we examined the effects of the inhibitor of GSK3β, lithium, on a representative from each of these groups. In the absence of induced expression of p53, p53-null H1299 cells expressed nearly undetectable levels of MDM2 and p21, whereas a low basal expression of Bax was detected (Fig. 4A). Following induced expression of p53, large increases in each protein were evident. Inhibition of GSK3β by treatment with lithium greatly diminished the p53-induced increased levels of all three proteins, without altering the level of p53 (Fig. 4, A and B). The expression of MDM2 appeared to be slightly less sensitive to inhibition of GSK3β than the other two, as the p53-induced increase in MDM2 was inhibited by lithium by 52 ± 8%, whereas the p53-induced increases in p21 and Bax were inhibited by 84 ± 9% and 80 ± 4%, respectively. The large inhibition by lithium of increased p21 following the induced expression of p53 matches well our previous report that increases in p21 following camptothecin treatment were nearly completely blocked by inhibiting GSK3β with lithium, dominant-negative GSK3β, or expressed GBP, an inhibitory GSK3β-binding protein (16). In addition to inhibiting GSK3β, lithium also has moderate inhibitory effects on casein kinase 2 and mitogen-activated protein kinase-activated protein kinase 2 (24). To further verify that the effect of lithium was due to inhibition of GSK3β°, we directly compared the effects of lithium to those of two other small molecule protein kinase inhibitors. These include rottlerin, which is a potent inhibitor of GSK3β but also inhibits equally well mitogen-activated protein kinase-activated protein kinase 2, p38-regulated/activated kinase, and protein kinase A and is a weaker inhibitor of other kinases (24), and tip-oxadiazole, a potent inhibitor of GSK3β but currently lacking an extensive study of specificity (25). As shown in Fig. 4, C and D, the new results following lithium treatment match well with the initial results, as in the two groups of experiments lithium treatment reduced MDM2 levels by 52% (Fig. 4B) and 49% (Fig. 4D), p21 levels by 84 and 97%, and Bax levels by 80 and 70%, respectively. Inhibition of GSK3β by treatment with 5 μm tip-oxadiazole or 5 μm rottlerin reduced the p53-induced increased levels of MDM2 more than lithium, by 99 and 94%, respectively (Fig. 4, C and D). p21 levels were reduced by 99% by tip-oxadiazole, whereas rottlerin had a more variable and weaker effect, reducing p21 by 65%. Bax levels were reduced by 75% by rottlerin, similar to the effect of lithium, whereas tip-oxadiazole had less effect, causing a reduction of 48%. Thus, in each case all three inhibitors of GSK3β attenuated these responses by 48 to 99%, although there was some variation among the three inhibitors in the extent of inhibition which may be attributable their different specificities and magnitudes of inhibition of GSK3β. Altogether, these results are consistent with the proposal that GSK3β activity contributes to p53-induced expression of MDM2, p21, and Bax.

Fig. 4. Regulation of p53 actions by inhibition of GSK3β.

A, H1299 cells capable of inducibly expressing wild-type p53 were used to test if treatment with 20 mm lithium during the 24 h induction of p53 expression affected the levels of p53, MDM2, p21, or Bax, as indicated in representative immunoblots. B, the results from three experiments described in A were quantitated and are expressed as the percent of values obtained with induced expression of p53 in the absence of lithium. For quantitation of the effect of lithium on the level of Bax, basal Bax levels were subtracted prior to calculating the difference between Bax levels induced by p53 expression in the absence and presence of lithium. Means ± S.E.; n = 3; *, p < 0.05. C, wild-type p53 was inducibly expressed in H1299 cells to compare the effects of three inhibitors of GSK3β, 20 mm lithium (Li), 5 μm rottlerin (Rot), or 5 μm tip-oxadiazole (tip), on the levels of MDM2, p21, or Bax, as indicated in representative immunoblots. D, the results from three experiments described in C were quantitated and are expressed as the percent of values obtained with induced expression of p53 in the absence of any GSK3β inhibitor. Means ± S.E.; n = 3; *, p < 0.05;+, p = 0.06.

To further examine if GSK3β participates in regulating the expression of these proteins, northern blotting was carried out to test if inhibition of GSK3β by lithium blocked the increases in mRNA levels of p21 and MDM2 induced by p53. Induction of wild-type p53 in H1299 cells increased the p21 mRNA level, and inhibition of GSK3β with lithium reduced this response by 56 ± 7% (n = 3) (Fig. 5A). Lithium treatment also reduced the p53-induced increased level of MDM2 mRNA in H1299 cells by 47 ± 10% (n = 3) (Fig. 5B). Taken together, the measurements of protein and mRNA levels demonstrate that GSK3β activity is necessary for full transcriptional activity of p53.

Fig. 5. Increased p21 and MDM2 mRNA levels caused by p53 are blocked by inhibition of GSK3β with lithium.

A, induction of wild-type p53 for 24 h in H1299 cells increased the p21 mRNA level and co-treatment with 20 mm lithium inhibited this increase. B, induction of wild-type p53 for 24 h in H1299 cells increased the MDM2 mRNA level and co-treatment with 20 mm lithium inhibited this increase. β -Actin levels were unaffected by induced expression of p53 or treatment with lithium.

Transcriptional-independent, as well as -dependent, processes contribute to p53-mediated apoptosis. Evidence is accumulating that mitochondrial p53 is an important intermediate in transcriptional-independent apoptosis (26 –29). Our further studies examining mitochondrial events following DNA damage demonstrated that regulatory interactions of GSK3β with p53 are not limited to the nucleus. Previously we showed that inhibition of GSK3β by pharmacological and molecular methods attenuated camptothecin-induced, p53-mediated, caspase-3 activation and apoptosis in SH-SY5Y cells (16). As noted above, in addition to the nucleus, p53 also has been identified in mitochondria, where GSK3β also is known to reside (30). Therefore, we tested if the mitochondrial p53 level increased, and compared this with nuclear p53, following camptothecin treatment of SH-SY5Y cells. Treatment with 1 μm camptothecin caused a time-dependent increase in nuclear p53, soon followed by increased mitochondrial p53 (Fig. 6A). This was verified not to be due to contamination with cytosolic or nuclear proteins by immunoblotting for markers of all three fractions, including tubulin for cytosol, the transcription factor cyclic AMP response element-binding protein (CREB) for the nucleus, and cytochrome oxidase for mitochondria (Fig. 6B). Because previously we found that nuclear p53 associates with, and activates, nuclear GSK3β (16), we tested if these actions also occurred in mitochondria. This was found to be the case, as after camptothecin treatment mitochondrial p53 co-immunoprecipitated with GSK3β (Fig. 6C). Furthermore, the activity of mitochondrial GSK3β was increased, whereas the total level of mitochondrial GSK3β remained unaltered (Fig. 6D). Thus, similar signaling events occurred in both the nucleus and the mitochondria, including accumulation of p53, the association of p53 with GSK3β, and the activation of GSK3β.

Fig. 6. Inhibition of GSK3β protects SH-SY5Y cells from camptothecin toxicity.

A, camptothecin treatment (1 μm) of SH-SY5Y cells resulted in time-dependent increases in the nuclear and mitochondrial levels of p53. B, subcellular fractions were immunoblotted for tubulin as a cytosolic (Cyt) marker, CREB as a nuclear (Nuc) marker, and cytochrome oxidase (COX IV) as a mitochondrial (Mito) marker. C, with or without camptothecin (CT) treatment (1 μm; 3 h), mitochondria were prepared, GSK3β was immunoprecipitated, and immunoprecipitants were immunoblotted for GSK3β and p53. D, with or without camptothecin treatment (1 μm; 3 h), mitochondria were prepared, GSK3β was immunoprecipitated, and the activity and levels of immunoprecipitated GSK3β were measured. E, camptothecin treatment (1 μm; 3 h) increased cytosolic cytochrome c levels and this was blocked by inhibition GSK3β with 1 h pretreatments with 20 mm lithium (Li), 5 μm rottlerin (Rot), or 5 μm tip-oxadiazole (tip). F, camptothecin treatment (1 μm; 3 h) increased the levels of cleaved, active caspase-3 and cleaved PARP, and these were blocked by inhibition GSK3β with 1 h pretreatments with 20 mm lithium, 5 μm rottlerin, or 5 μm tip-oxadiazole.

To test if GSK3β contributed to the mitochondrial targeted apoptotic signaling, the effects of three inhibitors of GSK3β were examined on the release of mitochondrial cytochrome c to the cytosol, which leads to activation of caspases (31–33), and the activation of caspase-3. Treatment with camptothecin stimulated the release of cytochrome c into the cytosol (Fig. 6E). Inhibition of GSK3β by pretreatment with 20 mm lithium, 5 μm rottlerin, or 5 μm tip-oxadiazole blocked the camptothecin-induced release of cytochrome c. Furthermore, all three inhibitors of GSK3β attenuated the camptothecin-induced activation of caspase-3, as measured by the appearance of cleaved active caspase-3 bands and the proteolysis of the caspase-3 substrate PARP (Fig. 6F). These results indicate that GSK3β plays a role in p53-induced apoptotic signaling, and the blockade of signals leading to cytochrome c release upon inhibition of GSK3β likely accounts for the block of caspase-3 activation shown here and previously reported using several other methods to inhibit GSK3β (16).

DISCUSSION

Considering the importance of both p53 (1–3) and GSK3β (7) in cell fate determination, the recent finding that these two proteins interact synergistically (16) led us to address two aspects of this interaction in this investigation: what are some of the factors that control the association, and what are some of the consequences of this interaction. Regarding the first goal, we identified two domains of p53 that modulate its interaction with GSK3β. Secondly, both nuclear and mitochondrial interactions of GSK3β and p53 were identified. This information indicates that when the level of p53 is elevated it plays a role in directing the actions of GSK3β, and that GSK3β, in turn, contributes an important facilitating role in promoting the actions of p53, suggesting that these two proteins co-operate as partners in controlling cellular responses to DNA damage.

Two domains of p53, the BD and AD1 regions, had regulatory influences on its association with GSK3β. The BD in the C-terminal region of p53 was identified as necessary for the binding of GSK3β to p53. Specifically, amino acids 364 –373 of p53 were found to be obligatory for its association with GSK3β. Previous studies have documented that this basic domain of p53 is a region that regulates the function of p53, although the precise regulatory actions of the BD remain controversial (1, 34, 35). One function of the BD appears to be to facilitate the binding to GSK3β, which subsequently facilitates the transcriptional and apoptotic actions of p53, as indicated by the inhibitory effects of blocking GSK3β activity. It is of interest that other proteins that bind to the basic domain, such as the Werner syndrome protein (36), the Y-box binding protein YB-1 (37), and TFIIH subunits (38), also have been reported to facilitate actions of p53. Apparently the regulation of proteins associating with the BD region of p53 significantly modifies the actions of p53, likely causing protein-selective differences in the repertoire of signals derived from p53. Additionally, a possibly unique aspect of the association of p53 and GSK3β is that not only is p53 function enhanced, but the activity of GSK3β also is increased by the association of the two proteins (16). Furthermore, we found that the activity of GSK3β contributed to regulating the stability of the complex, as inhibition of GSK3β activity increased the association of p53 with GSK3β. Such an effect could be ascribed to a direct phosphorylating action of GSK3β on p53, or to an indirect action mediated by the GSK3β -induced phosphorylation of another protein interacting with p53 or GSK3β. In either case, it may indicate that the association of p53 and GSK3β is transient, with the activity of GSK3β influencing the duration of binding to p53. Finally, considering the many links between GSK3β and apoptosis (7), it is intriguing to note the correlation between the capacity of p53 constructs to bind GSK3β and their ability to induce cell death, as ΔBD-p53 is unable to bind GSK3β or to induce apoptosis (39).

The AD1 region is the other domain of p53 that had a regulatory influence on its association with GSK3β. The AD1 domain is the binding site for MDM2, which targets p53 for degradation. Thus, deletion or functional mutations of the AD1 domain, by eliminating MDM2 binding, lead to accumulation of p53. Deletion of the AD1 domain allowed significantly more association of GSK3β with p53 than occurred with wild-type p53. That this was unlikely due to elimination of MDM2 binding was evident from the results with the AD1(−)-p53 construct in which elimination of MDM2 binding, but retention of the AD1 region, did not alter the association of GSK3β with p53. Thus, the AD1 region, or another protein bound to this region, regulates the capacity of p53 to associate with GSK3β.

The mechanisms by which either p53 or GSK3β individually promotes apoptosis remain uncertain, thus the mechanisms accounting for apoptosis following the concerted action of these two molecules remain to be determined. Apoptosis caused by p53 can be directed by transcriptional-dependent and transcriptional-independent mechanisms. The former is clearly an action of nuclear p53, and there is accumulating evidence that the latter is a mitochondrial action of p53. For example, some apoptotic conditions cause translocation of p53 to the mitochondria and constructs of p53 with decreased ability to translocate to the mitochondria are weaker activators of apoptosis (26, 27, 29). More recently this association was further solidified by the finding that the Arg-72 polymorphic variant of p53 translocates to the mitochondria much better than the Pro-72 variant and is at least five times better at inducing apoptosis (28). Our findings suggest that both nuclear and mitochondrial interactions between p53 and GSK3β may be relevant in this capacity. This was indicated by the findings that inhibition of GSK3β was capable of interfering with p53-mediated apoptotic signaling at two levels, including a rapid mitochondrial action and a slower nuclear transcriptional action. Inhibition of GSK3β blocked cytochrome c release and caspase 3 activation, suggesting that mitochondrial GSK3β may contribute to the mitochondrial apoptotic signaling pathway. One possibility is that GSK3β acts in concert with mitochondrial p53, which bound mitochondrial GSK3β, to facilitate mechanisms causing cytochrome c release. Previous studies have reported that mitochondrial p53 contributed to apoptotic signaling upstream of mitochondrial cytochrome c release (26 –29). The association of p53 with GSK3β in the mitochondria, and the concomitant activation of GSK3β, represents the first enzymatic regulatory action of mitochondrial p53 that has been identified, although mitochondrial p53 has been shown to bind other proteins in mitochondria (26, 28). Thus, the association of p53 with GSK3β in mitochondria and the activation of mitochondrial GSK3β may contribute to the mitochondrial apoptotic action of p53, although the intermediate steps whereby GSK3β activity promotes cytochrome c release remain to be determined. We also note that although only the association of GSK3β with p53 was examined, the inhibitors used also block the activity of GSK3α so it is quite possible that both isoforms of GSK3 may contribute to the signaling activities reported here.

In the nucleus, GSK3β was found to co-localize with p53 following apoptotic stimuli-induced nuclear accumulation of GSK3β (15, 16). Clearly the nucleus represents an important localization for the interactions of p53 and GSK3β since the two proteins associate in the nucleus, p53 activates nuclear GSK3β, and inhibition of GSK3β reduces the expression of p53-regulated proteins. For example, the involvement of Bax mediated by transcriptional activation of p53 often contributes to apoptosis and was evident in H1299 cells upon induced expression of p53. Inhibition of GSK3β with lithium greatly abrogated this p53-induced Bax expression, likely by inhibiting the facilitation of GSK3β on the transcriptional activity of p53. Thus protection from apoptosis may be provided by inhibiting the facilitating effects of GSK3β both on the rapid mitochondrial and slower nuclear apoptotic responses to DNA damage.