Direct, activating interaction between glycogen synthase kinase-3β and p53 after DNA damage

Methods

Results

To study responses to DNA damage we used camptothecin, a topoisomerase I inhibitor that is well known to cause DNA damage, which subsequently activates p53-mediated responses. Camptothecin treatment of human neuroblastoma SH-SY5Y cells caused time-dependent (Fig. 1A) and concentration-dependent (Fig. 1B) increases in the levels of p53 and p53-regulated protein p21^WAF1/CIP1. Subsequent to p53 accumulation there was increased proteolysis of the caspase-3 substrate PARP and increased activity of caspase-3, which is consistent with previous reports that DNA damage leads to p53-induced activation of the caspase-3-mediated apoptosis signaling cascade (4–6).

To test whether GSK3β was influenced by DNA damage caused by camptothecin treatment, GSK3β was immunoprecipitated from camptothecin-treated cells, and its activity was assessed by measuring the phosphorylation of recombinant tau, a well characterized substrate of GSK3β (15). These measurements revealed that camptothecin treatment resulted in a large and prolonged increase in the activity of GSK3β (Fig. 1C). Because GSK3β is located in both the cytosolic and nuclear compartments of cells, and some apoptotic stimuli increase nuclear levels of GSK3β (13), these compartments were fractionated from control and camptothecin-treated cells to examine which pool of GSK3β was activated after DNA damage. Surprisingly, there was no nuclear translocation of GSK3β after camptothecin treatment, because both the nuclear and cytosolic levels of GSK3β remained unchanged, but there was an exclusive activation of nuclear, not cytosolic, GSK3β (Fig. 1D). Nuclear GSK3β was activated within 1 h of camptothecin treatment and remained highly activated for at least 5 h (Fig. 1D). Treatment with another agent that causes DNA damage, etoposide, also increased p53 levels and activated GSK3β in the nucleus (Fig. 1E). These findings reveal a previously unrecognized effect of DNA damage, activation of GSK3β in the nucleus.

We examined the mechanisms underlying DNA damage-induced activation of GSK3β, first focusing on potential changes in the phosphorylation state of GSK3β. GSK3β is inhibited by phosphorylation of Ser-9 and activated by phosphorylation of Tyr-216 (7, 8, 16, 17). However, after camptothecin treatment there were no changes in the levels of nuclear phospho-Ser-9-GSK3β (Fig. 2A) or phosphotyrosine-GSK3β (data not shown). GSK3β activity also can be regulated by association with other proteins (7, 8). For example, axin facilitates GSK3β-mediated phosphorylation of β-catenin, and the inhibitory GBP inhibits GSK3β activity (18–20). Therefore, we tested whether there might be a direct, regulatory association of GSK3β with p53 that accumulated in the nucleus after DNA damage by using coimmunoprecipitation measurements. Immunoblots revealed that p53 coimmunoprecipitated with GSK3β from nuclear fractions prepared from cells treated with camptothecin (Fig. 2B). Similarly, camptothecin treatment resulted in the coimmunoprecipitation of GSK3β with p53 (data not shown). Thus, DNA damage increased nuclear p53 levels and activated nuclear GSK3β, and p53 formed a stable complex with GSK3β in the nucleus.

We tested whether p53, rather than other signals generated by DNA damage, accounted for activation of GSK3β. p53 transiently transfected into p53-null H1299 lung carcinoma cells (14) coimmunoprecipitated with GSK3β, and expression of p53 was sufficient to increase the activity of nuclear GSK3β (Fig. 2C). In contrast, mutant p53 (R175H), known to be functionally impaired, associated with but did not activate GSK3β. This result indicated that the DNA damage-induced activation of GSK3β resulted from a regulatory association with p53. The association of p53 with GSK3β, taken in conjunction with the lack of changes in GSK3β phosphorylation, raised the possibility that p53 may cause the activation of GSK3β directly. To test this hypothesis, the activity of recombinant GSK3β was measured in vitro in the absence and presence of recombinant p53. Remarkably, these measurements demonstrated that p53 directly increased the activity of recombinant GSK3β, because GSK3β activity increased in a p53 concentration-dependent manner (Fig. 2D). Because GSK3β phosphorylates substrates that are either unprimed, such as tau, or primed by previous phosphorylation (7, 8), we examined recombinant GSK3β-mediated phosphorylation of a primed substrate, phosphoglycogen synthase peptide-2, which also exhibited activation by recombinant p53 (n = 3; P < 0.05 compared with GSK3β activity in the absence of p53; data not shown). Thus, these experiments have shown that p53 is capable of directly binding to GSK3β, and this direct association increases the activity of GSK3β.

In the Wnt pathway, scaffolding regulatory proteins control the specificity of GSK3β-mediated signaling (7, 8). Therefore, we reasoned that association with p53 might direct the actions of GSK3β to processes involved in p53-induced signaling. To examine this hypothesis, several methods were used to inhibit GSK3β activity to test whether this modulated p53 mediated responses after DNA damage. Pretreatment with lithium, a selective inhibitor of GSK3β (21–24), or a structurally dissimilar inhibitor of GSK3β, sodium valproate (25), reduced the camptothecin-induced increases in p21, PARP proteolysis (Fig. 3A), and caspase-3 activity (data not shown). Lithium and sodium valproate pretreatment also attenuated etoposide-induced increases in p21 levels and PARP proteolysis (Fig. 3B), and stable expression of a dominant negative mutant of GSK3β (26) also attenuated camptothecin-induced PARP proteolysis (Fig. 3C). Thus, three treatments that reduce the activity of GSK3β attenuated signaling responses to p53. However, these conditions all reduce GSK3β activity throughout the cell. Therefore, we devised a method to selectively inhibit nuclear GSK3β, the pool that is activated by p53. For this purpose, the inhibitory GBP (19, 26) was stably expressed by using a vector with a nuclear localization sequence. Immunoblots confirmed the exclusive nuclear localization of expressed GBP (data not shown). Expression of nuclear GBP did not alter camptothecin-induced increases in p53, but effectively attenuated elevations in the level of p21, and PARP proteolysis induced by camptothecin or etoposide even when a higher concentration (5 μM) of camptothecin was tested (Fig. 3D), confirming that nuclear GSK3β promotes DNA damage-induced, p53-mediated responses. Thus, not only is nuclear GSK3β activated by association with p53 after treatment with DNA-damaging agents, but activated nuclear GSK3β contributes substantially to the signaling activities induced by p53.

To test whether p53-activated nuclear GSK3β affected its interactions with other substrates, the levels of β-catenin were measured because active GSK3β promotes the degradation of β-catenin (7, 8). Treatment with camptothecin reduced cytosolic β-catenin but not membrane-bound β-catenin, which is sequestered from accessibility by GSK3β, and this effect was blocked by inhibition of GSK3β with lithium (Fig. 3E).

Discussion

This identification of a direct, activating interaction between p53 and GSK3β reveals new mechanisms that regulate the actions of each of these proteins, provides a mechanistic explanation for several previously reported findings, and raises potentially new therapeutic strategies. Although several studies have reported that GSK3β is involved in apoptotic signaling, in most cases it was inferred from overexpression or inhibition of GSK3β. When endogenous GSK3β was found to be activated during apoptosis, the mechanism was attributed to changes in the phosphorylation state of GSK3β (7–12). The observed activation of GSK3β in response to DNA damage, which previously was not known to activate GSK3β, is unique because the phosphorylation state of GSK3β was unaltered and only nuclear GSK3β was activated, although the majority of cellular GSK3β is in the cytosol (13). It is now evident that these unique features are caused by the mechanism found to activate GSK3β after DNA damage, the association with p53, and the predominantly nuclear localization of p53. It is intriguing to note that this interaction provides a mechanism for p53 to down-regulate several survival-promoting transcription factors that are known to be phosphorylated by GSK3β in the nucleus (reviewed in ref. 8), the inhibition of which would promote p53's activation of the apoptotic program. Previously GSK3β was known to be regulated by protein complex formation, because axin and axin-related proteins can facilitate the phosphorylation of β-catenin by GSK3β (7, 8, 18, 20), and other GSK3β-binding proteins such as GBP can inhibit this process (7, 8, 15, 19, 20). However, the regulation of GSK3β by protein complexes was known only to occur in the context of the Wnt/wingless signaling pathway. Thus, activation of GSK3β by binding to p53 reveals that protein complexes outside the Wnt/wingless pathway also regulate the activity of GSK3β, which raises the possibility that additional proteins bind and regulate GSK3β independently of Wnt/wingless signaling. This interaction likely accounts for the findings that (i) expression of dominant negative p53 blocked apoptosis induced by overexpression of GSK3β (10), (ii) activation of Akt (protein kinase B), which inhibits GSK3β, is inhibitory for p53-mediated transcriptional and apoptotic activities (27–29), and (iii) activated p53 down-regulates β-catenin (30), because GSK3β promotes proteolysis of β-catenin (7). Furthermore, the broad cell survival-promoting effects of the GSK3β inhibitor lithium (8, 31, 32) may be attributable partly to attenuated p53 function (32, 33) consequential to GSK3β inhibition such as in ischemic brain, where lithium is protective (31). The revelation that GSK3β modulates p53 function presents opportunities for regulating the effects of p53. For example, chemotherapeutic p53-mediated signaling could be enhanced by activating GSK3β, and p53-mediated neuronal loss may be attenuated by inhibitors of GSK3β, although potential proliferative/cancerous effects must be assessed. Thus, these findings reveal mechanisms involved in DNA damage-induced intracellular signaling and raise possibilities for pharmacologically modulating cellular responses to DNA damage.