p53 Induces NF-κB Activation by an IκB Kinase-independent Mechanism Involving Phosphorylation of p65 by Ribosomal S6 Kinase 1*

Apoptosis induced by p53 has been proposed to involve activation of the transcription factor NF-κB. Here we describe the novel molecular mechanism through which p53 and DNA-damaging agents activate NF-κB. NF-κB induction by p53 does not occur through classical activation of the IκB kinases and degradation of IκBα. Rather, p53 expression stimulates the serine/threonine kinase ribosomal S6 kinase 1 (RSK1), which in turn phosphorylates the p65 subunit of NF-κB. The lower affinity of RSK1-phosphorylated p65 for its negative regulator, IκBα, decreases IκBα-mediated nuclear export of shuttling forms of NF-κB, thereby promoting the binding and action of NF-κB on cognate κB enhancers. These findings highlight a rather unusual pathway of NF-κB activation, which is utilized by the p53 tumor suppressor.

Apoptosis, or programmed cell death, is a key cellular mechanism that participates in the regulation of normal growth, development, and homeostasis of organisms. Apoptosis also likely plays a key role in halting the growth of tumor cells early in their development. In this regard, the p53 tumor suppressor acts by inducing apoptosis and/or cell cycle arrest. Because of its central role as a negative regulator of cell growth, p53 activity is tightly regulated, principally at the level of protein stability (1). By various mechanisms, DNA damage, hypoxia, and oncogene activation lengthen the intracellular half-life of p53, resulting in its accumulation and increased action in the nucleus. The most important inhibitory regulator of p53 is Mdm2, a ubiquitin ligase that both targets p53 for degradation in the 26 S proteasome (2,3) and directly inhibits p53-mediated transcriptional activity (4). The p53-mediated apoptotic response at least partially depends on the induction of NF-B (5). However, the mechanism by which p53 activates NF-B is unknown.
The NF-〉/Rel family of transcription factors regulates the expression of various genes involved in growth, differentiation, development, apoptosis, and inflammation. These factors function as either homo-or heterodimers with p50/p65 corresponding to the prototypical NF-〉 complex (6,7). In resting cells, NF-〉 is largely sequestered in the cytoplasm by its associa-tion with the I〉␣ inhibitor. However, more recent studies indicate that this complex shuttles into and out of the nucleus but remains transcriptionally inactive because of its association with the I〉␣ protein (8). In response to many stimuli including proinflammatory cytokines, nuclear NF-〉 expression is induced by activation of the I〉 kinases (IKKs), 1 which in turn phosphorylate two amino-terminal serines in IB␣ (9,10). This phosphorylation reaction targets I〉␣ for ubiquitindependent degradation by the proteasome (11). The liberated NF-〉 complex translocates into the nucleus where it binds to 〉 enhancers in various target genes and activates transcription. The transcriptional activity of NF-〉 is additionally controlled by various post-translational modifications, including phosphorylation (12,13) and acetylation of the p65 subunit (14,15).
Elimination of NF-B activity by disruption of either the p65 (16) or the Ikk2 (Ikk␤) (17)(18)(19) gene in mice is associated with embryonic death because of fulminant tumor necrosis factor (TNF)-␣-induced apoptosis of hepatocytes. Thus, in this and other situations (20 -22), NF-〉 functions as a potent antiapoptotic factor. Under different circumstances, however, NF-〉 activation promotes a proapoptotic response (23)(24)(25)(26)(27)(28)(29)(30)(31). In addition to differences conferred by the type of cell involved, the function of NF-〉 as an antiapoptotic or proapoptotic factor likely depends on the nature of the inducing stimulus. Indeed, in the same cells, NF-〉 antagonizes TNF-␣-activated apoptosis yet promotes H 2 O 2 -induced apoptosis (32,33). The timing and duration of NF-〉 activity relative to the death stimulus may play a key role in determining which of these cellular responses occurs (27,34).
The tumor suppressor p53 is one of the most important cellular mediators of apoptosis. Interestingly, many proapoptotic stimuli, including DNA-damaging agents, hypoxia, oxygen radicals, and ionizing irradiation, activate both p53 and NF-〉. Ryan et al. (5) suggested a link between these responses, showing that p53 expression directly activates NF-B and that this NF-〉 activity is required for p53-induced apoptosis. However, the molecular mechanism by which p53 activates NF-B remains unclear. We now describe a series of studies dissecting the novel mechanism by which p53 activates NF-〉.
Transfections and Reporter Gene Assay-Cells were plated in 60-mm dishes and transfected with the indicated expression vectors together with Bor p53-dependent reporter luciferase genes with FuGENE 6 transfection reagent (Roche Applied Science). In each experiment, cells were cotransfected with a ␤-galactosidase reporter plasmid to permit normalization of differences in transfection efficiency occurring in the individual cultures. Cell extracts were prepared 18 h later or as indicated, and luciferase activity was determined with the luciferase assay system (Promega, Madison, WI). ␤-Galactosidase activity was measured with the Galacto-Light ␤-galactosidase assay (Tropix, Bedford, MA).
Immunoprecipitation, Immunoblotting, and Kinase Assays-Cells were washed with phosphate-buffered saline and lysed in buffer containing 50 mM Hepes, pH 7.2, 250 mM NaCl, 0.5% Nonidet P-40, 5 mM EDTA, 1 mM dithiothreitol, 1 mM NaF, 0.1 mM Na 3 VO 4 , and protease inhibitor mixture (Roche Applied Science) for 30 min on ice. Precleared lysates (15,000 ϫ g, 3 min) were immunoprecipitated for 3 h with the indicated antibodies, and the immunoprecipitates were collected using protein A-conjugated agarose (Roche Applied Science). Immunoprecipitated proteins were separated by SDS-PAGE and visualized by immunoblotting with the indicated antibodies.
When used in in vitro kinase assays, the immunoprecipitates were washed in kinase buffer containing 20 mM Hepes, pH 7.2, 10 mM MgCl 2 , 4 mM NaF, 2 mM MnCl 2 , 1 mM dithiothreitol, and 0.1 mM Na 3 VO 4 . Recombinant active RSK1 (0.1 g) (Upstate Biotechnology) was employed instead of immunoprecipitates in some assays. [␥-32 P]ATP (20 Ci, specific activity 6,000 Ci/mol) and 1 g of a substrate were added to the kinase reactions followed by incubation for 30 min at 30°C. The reaction products were analyzed by SDS-PAGE and autoradiography. Fractionation of cell lysates was performed by fast protein liquid chromatography (FPLC) using a Superose 6 molecular sieving column (Amersham Biosciences).
Electrophoretic Mobility Shift Assays-EMSA was performed as described (39). In brief, cells in 6-well plates were washed in phosphatebuffered saline, harvested, and resuspended in 0.4 ml of buffer A (10 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol). After 10 min, 23 l of Nonidet P-40 was added, and the sample was vortexed for 2 s. Nuclei were separated from cytosolic fractions by centrifugation at 13,000 ϫ g for 10 s. To isolate nuclear proteins, pelleted nuclei were resuspended in 25 l of buffer B (20 mM Hepes, pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, and 0.1 mM phenylmethylsulfonyl fluoride) at 4°C. After 30 min, lysates were centrifuged at 13,000 ϫ g for 30 s, and supernatants containing nuclear proteins were collected. Nuclear proteins were analyzed in binding reactions performed with radiolabeled consensus B enhancer oligonucleotides (Promega) and resolved on nondenaturing 5% polyacrylamide gels.
In Vitro Binding Assays-GST-IB␣ or His-p65 (0.5 g) were phosphorylated by recombinant active RSK1 as described above, except that 1 M unlabeled ATP was used in the kinase reaction. GST-IB␣ or phosphorylated GST-IB␣ was bound to glutathione-Sepharose 4B beads (Amersham Biosciences), and the beads were washed with lysis buffer. His-p65 or phosphorylated His-p65 (0.5 g) was added to the beads and incubated for 5 min at 4°C. The beads were washed in lysis buffer, and the binding of p65 or phosphorylated p65 to GST-IB␣ was analyzed by immunoblotting.
Small Interfering RNA-mediated Gene Silencing-Control small interfering RNA (siRNA) was purchased from Dharmacon Research (Lafayette, CO). siRNA for RSK1 was designed according to Elbashir et al. (40) and prepared by Dharmacon Research. Transfections of siRNA were performed using Oligofectamine (Invitrogen). After 24 h, the cells were replated and transfected with a second dose of siRNA, together with B-dependent reporter gene and expression vectors for p53 and p65. Alternatively, the cells were activated as indicated and analyzed for reporter gene activity. The amount of RSK1 in each of the cell extracts was determined by immunoblotting with anti-RSK1 antibodies.

RESULTS
Transcriptionally Inactive Mutants of p53 Effectively Induce NF-B-To assess the role of the individual domains within p53 in the activation of NF-〉, we used Saos-2 cells, which do not express endogenous p53. We transfected Saos-2 cells with p53 mutants characterized previously (Fig. 1A) and analyzed the ability of these mutants to activate an NF-〉-dependent reporter gene. Wild-type p53 activated both p53-and NF-〉dependent reporter genes (Fig. 1B). The amino-terminal region of p53 contains a transcriptional activation domain. Leu-22 and Trp-23 within this region are required for the binding of transcriptional coactivators and for the transcriptional activation of p53-dependent genes (41). Although the compound mutant p53(L22Q,W23S) failed to activate the p53-dependent reporter gene (Fig. 1B), this p53 mutant strongly activated an NF-〉dependent reporter gene. Indeed, the p53(L22Q,W23S) mutant induced a severalfold increase in NF-〉 activity compared with wild-type p53. Thus, the ability of p53 to activate expression of p53-dependent target genes is apparently not required for p53 activation of NF-〉.
p53 sequence-specific transcriptional activity determined by the central part of the protein is preserved when the aminoterminal transcriptional activation domain of p53 (amino acids 1-80) is replaced with the transactivation domain from the herpes simplex virus VP16 protein (VP16-p53-(80 -393)) (37). Despite effective p53-dependent gene expression, this fusion protein displayed a nearly complete loss of NF-〉-inducing activity (Fig. 1B). This finding suggests that the amino-terminal region of p53 is required for NF-〉 activation, although this requirement is distinct from the role played by this aminoterminal region in the activation of p53 target genes.
p53 contains five copies of the motif PXXP located between residues 61 and 94. These proline-rich motifs form recognition sites for SH3 domains, and their deletion impairs p53-induced apoptosis (36). Consistent with results described previously (36), a p53 mutant lacking this proline-rich region (p53(⌬PP)) retained p53-induced transcriptional activity. However, the p53(⌬PP) mutant exhibited a severely reduced ability to activate NF-〉 reporter gene activity (Fig. 1B). These results further confirm that p53-mediated activations of p53-dependent and NF-〉-dependent target genes are separable biological responses.
Next, we analyzed p53-induced NF-〉 DNA binding in EM-SAs. Consistent with our prior functional studies, transient expression of both wild-type p53 and the transcriptionally inactive p53(L22Q,W23S) mutant induced NF-〉 binding activity (Fig. 1C). Furthermore, the p53(L22Q,W23S) mutant induced higher levels of NF-〉 binding than wild-type p53. The NF-〉-binding complex induced by wild-type p53 and p53(L22Q,W23S) consisted principally of p50/p65 heterodimers, based on supershifting studies with antibodies specific for different members of the Rel family of proteins (data not shown). An NF-B complex with identical mobility was activated similarly by other NF-〉 inducers, including the NF-〉-inducing kinase and TNF-␣ (Fig. 1C).
p53 Activation of NF-B Does Not Involve the Degradation of IB␣-The ability of p53 to activate NF-〉 without inducing the transcription of p53-dependent genes suggests that p53 directly engages a signaling pathway that leads to NF-〉 induction. In resting cells, NF-〉 is sequestered chiefly in the cytoplasm through binding to its inhibitor, I〉␣. Translocation of NF-B into the nucleus and engagement of 〉 enhancers generally result from signal-induced phosphorylation of IB␣, followed by ubiquitylation and degradation of this inhibitor by the 26 S proteasome.
To investigate whether p53 activation of nuclear NF-B involves I〉␣ degradation, we used Saos-2 cells in which p53 expression can be induced by the addition of doxycycline (Saos-2-Tet-On-p53) (5). Resting Saos-2-Tet-On-p53 cells express very low levels of p53. After incubation of the cells with doxycycline, p53 expression increased markedly between 4 and 18 h ( Fig. 2A), mimicking activation of p53 by protein stabilization. As in the transient transfection system (Fig. 1C), p53 expression induced NF-B DNA binding activity; however, this response was not evident until 12-18 h (Fig. 2B), suggesting that the kinetics of p53 activation of NF-B are quite slow. The p53-induced NF-B response persisted for the entire 48 h of the experiment (data not shown). During this time, the intracellular levels of I〉␣ did not change (Fig. 2C), nor was there evidence for marked I〉␣ phosphorylation on serines 32 and 36 (Fig. 2D). Conversely, treatment of these cells with TNF-␣ both induced I〉␣ degradation (Fig. 2C) and increased the levels of phosphorylated I〉␣ (Fig.  2D). Expression of p53 also did not alter the cellular content of the I〉␣ homologs I〉␤ and I〉⑀ (data not shown). Together, these findings indicate that p53 activation of NF-B occurs slowly and does not involve IB␣ degradation.
To characterize further the p53-induced pathway of NF-B activation, we next tested the effects of the proteasome inhibitor, MG132, which effectively blocks NF-B activation upon TNF-␣ treatment by preventing the degradation of phosphorylated and ubiquitylated I〉␣. As expected, treatment with MG132 promoted accumulation of phosphorylated IB␣ in the cytoplasm after TNF-␣ stimulation. Pretreatment of cells with MG132 for 2 or 8 h markedly increased the amount of the more FIG. 1. Induction of NF-〉 activity by various p53 mutants. A, schematic representation of p53 mutants studied. B, Saos cells were transfected with 2.5 g of the indicated p53 expression vectors and either a 〉 enhancer (left panel) or p53 enhancer-dependent luciferase reporter cassettes (right panel). A ␤-galactosidase reporter plasmid was cotransfected to permit normalization of differences in transfection efficiency. Cells were harvested 18 h later and assayed for reporter gene activity. The graphs show fold induction of normalized luciferase activities compared with control. Error bars indicate the standard deviations derived from triplicate determinations. WT, wild-type. C, Saos cells were transfected with the indicated expression vectors (1 or 2.5 g of DNA/dish) and harvested 18 h after transfection. Control-transfected cells were incubated with TNF-␣ for 15 min before harvesting. Nuclear extracts were prepared and analyzed for binding to 32 P-radiolabeled 〉 enhancer oligonucleotide in EMSA followed by autoradiography. NIK, NF-B-inducing kinase. slowly migrating, phosphorylated form of I〉␣ in cells activated with TNF-␣. In contrast, no changes in the amount of phosphorylated I〉␣ were observed in cells expressing p53 (Fig. 2E). As expected, MG132 treatment efficiently blocked TNF-␣-induced translocation of p65 into the nucleus (Fig. 2F). In contrast, MG132 had no effect on p53-induced accumulation of p65 in the nucleus (Fig. 2F). Together, these findings suggest that the activation of NF-〉 by p53 does not involve an action of the 26 S proteasome.
p53 Activates NF-B Independently of the IB Kinases: Possible Role of RSK1-The signalsome, a multimolecular complex containing IKK1 and IKK2 and the noncatalytic adaptor molecule NEMO/IKK␥, mediates phosphorylation of I〉␣ in response to most NF-〉-inducing stimuli (9,10). The absence of I〉␣ phosphorylation and degradation during activation of NF-B by p53 prompted us to evaluate whether p53 activates NF-〉 in the absence of a functional signalsome. For these studies, we used MEF lacking both IKK1 and IKK2 (IKK1/2 Ϫ/Ϫ ). As expected, TNF-␣ induced NF-B activity in wild-type MEF but not in the IKK1/2 Ϫ/Ϫ MEF (Fig. 3A). Conversely, NF-B was effectively induced by p53 in the IKK1/2 Ϫ/Ϫ MEF. This finding indicates that IKK1 and IKK2 are not required for the activation of NF-B by p53 and further strengthens the notion that a distinct "nonclassical" pathway of NF-B induction is involved. Surprisingly, p53 did not activate NF-B in wild-type MEF. This binding is likely explained by the fact that transfection of wild-type MEF with p53 expression vectors consistently results in lower levels of p53 than found in IKK1/2 Ϫ/Ϫ MEF (Fig. 3B). In contrast to the IKK1/2 Ϫ/Ϫ MEF, wild-type MEF express substantially higher levels of the NF-Binducible Mdm2 protein, a ubiquitin ligase that targets p53 for degradation (42). Thus, accelerated degradation of p53 in the wildtype MEF likely explains the lower levels of p53 expression and the lack of NF-B induction in these cells.
Next, we analyzed the activation status of the IKKs in the presence of p53. Signalsomes isolated from p53-and TNF-␣-activated Saos-2-Tet-On-p53 cells were subjected to in vitro kinase assays using GST-I〉␣ as a substrate. In contrast to the readily detectable IKK activity observed after TNF-␣ treatment, induction of p53 or stimulation with phorbol 12-myristate 13-acetate (PMA) did not lead to activation of the IKKs (Fig. 3C). However, consistent with an earlier report (5), expression of p53 or treatment with PMA did activate the serine/ threonine protein kinase RSK1 (Fig. 3C). In contrast, TNF-␣ did not induce increased RSK1 activity. These results further confirm that IKK activation is not involved in p53-mediated activation of NF-B but instead raise the possibility that RSK1 may play a key role in the p53 pathway leading to the induction of NF-B.
This possibility was strengthened by the finding that RSK1 (Fig. 3D, upper panel) but not IKK1 and IKK2 (Fig. 3D, lower panel) coimmunoprecipitated with p53. In addition, immunoprecipitates of RSK1 ( Fig. 3D; note the large amounts of RSK1) did not contain IKK1. In reverse immunoprecipitates of signalsomes performed with anti-NEMO antibodies, no RSK1 was detected.
To characterize these molecular complexes in more detail, we size fractionated components present in the cell lysates employing FPLC on a Superose 6 column. As reported previously (9,10), the IKK signalsome eluted as a large complex with an apparent molecular mass of 700,000 -900,000 (Fig. 3E). In contrast, RSK1 was detected in smaller complexes of 80,000 -300,000, in both resting (Fig. 3E) and p53-expressing cells (not shown). p53 was found in fractions distributed over a broad molecular mass range (60,000 -600,000) which overlapped with both RSK1 and the IKKs. These results are consistent with the notion that subpopulations of RSK1 and p53 associate in a complex that is distinct in size from the signalsome.
RSK1 Is Required for Induction of NF-B by p53-Although p53 assembles with RSK1 in vivo and p53 expression is associated with RSK1 activation, it remained unclear whether Saos-2-Tet-On-p53 cells were incubated with doxycycline (Dox) to activate expression of p53 or with TNF-␣ for the indicated times. A, whole cell extracts were prepared, and the expression of p53 was analyzed by immunoblotting (WB) with an antibody specific for p53. Note increasing levels of p53 expression between 4 and 18 h. B, nuclear extracts were prepared and examined for binding to 32 P-radiolabeled 〉 enhancer oligonucleotide by EMSA followed by autoradiography. Note the delayed appearance of NF-B DNA binding activity at 18 h in Saos cells induced to express p53. C, the content of I〉␣ was determined by immunoblotting of whole cell extracts with an antibody specifically reacting with IB␣. Note that TNF-␣ induced depletion of IB␣ within 15 min, whereas IB␣ levels did not change over the entire time course of p53 induction. D, phosphorylation of I〉␣ was analyzed by immunoblotting of the whole cell extracts with an antibody specific for phosphorylated serines 32 and 36 of IB␣. Note induction of phospho-IB␣ (P-IB␣) by TNF-␣ but not by p53. E, to inhibit degradation of IB␣ by the proteasome, cells were treated with the inhibitor MG132 for the last 2 or 8 h of the incubation with doxycycline. Alternatively, TNF-␣ was added for the last 8 or 15 min of incubation with MG132. Whole cell extracts were prepared, and the content of I〉␣ was determined by immunoblotting with an antibody specific to IB␣. Positions of IB␣ and the more slowly migrating phosphorylated IB␣ are indicated. F, Saos cells were treated with MG132 similarly as in E. Nuclear extracts were analyzed for p65 by immunoblotting. The position of a nonspecific band is indicated (N.S.). Note that MG132 impairs nuclear translocation of p65 induced by TNF-␣ but not by p53.
RSK1 is actually required for p53 induction of NF-B. To assess this possibility, we employed siRNA to "knock down" RSK1 expression in Saos-2 cells followed by evaluation of p53-induced activation of a B-dependent luciferase reporter gene in these cells. Immunoblotting analysis revealed diminished expression of RSK1 in Saos-2 cells transfected with RSK1-specific siRNA, whereas transfection of a control siRNA produced no inhibitory effect on RSK1 expression (Fig. 4A). Transfection of these siRNAs did not alter the expression of a control cellular protein, ␣-tubulin. When functional responses were assessed, we observed that depletion of RSK1 with siRNA reduced p53mediated activation of the 〉-luciferase reporter gene (Fig. 4B) but not the NF-〉 response induced by p65 or human T-cell lymphotrophic virus, type I, Tax (not shown). These findings support a central role for RSK1 in p53-induced activation of NF-〉.
Inducible Phosphorylation of p65 by RSK1-Phosphorylation of nuclear p65 plays a key role in promoting the full NF-〉 transcriptional response (12,13,43). RSK1 also modulates the activities of various transcription factors by phosphorylation (44 -47). Because p53 expression increased both the kinase activity of RSK1 (Fig. 3C) and the transcriptional activity of NF-〉, we examined whether RSK1 targets p65 for phosphorylation. RSK1 was immunoprecipitated from control cells, p53-expressing cells, or cells activated with PMA. Immunoprecipitates were subjected to in vitro kinase assay employing the baculovirus-derived His epitope-tagged p65 (His-p65) as a substrate. In addition to RSK1 autophosphorylation, we observed phosphorylation of His-p65 by RSK1 stimulated by either p53 (Dox) or PMA (Fig. 5A). This phosphorylation of p65 by RSK1 appeared specific because immunoprecipitation of other kinases (RSK2 and protein kinase C␣ and C) prepared from identically stimulated cells did not lead to phosphorylation of His-p65 (data not shown). In addition, activated RSK1 phosphorylated GST epitope-tagged p65 isolated from Escherichia coli or T7 epitopetagged p65 isolated from mammalian cells but failed to phosphorylate His-IKK2 or GST-p53 (data not shown).
To identify the residues in p65 that are targeted for phosphorylation by RSK1, we performed in vitro kinase assays using GST fusion proteins containing various fragments of p65 as substrates. RSK1 effectively phosphorylated the carboxylterminal fragment of p65 present in GST-p65-(361-551) but Immunoblotting with an antibody specific for lamin A was performed to verify equal loading. C, Saos-2-Tet-On-p53 cells were incubated with doxycycline (Dox) for various times to induce expression of p53 or with TNF-␣ and PMA, and the whole cell extracts were prepared. Signalsomes were isolated by immunoprecipitation with anti-NEMO antibodies. The activities of coimmunoprecipitated IKK1/2 were analyzed by an in vitro kinase assay with GST-I〉␣ as the substrate. Equal IKK1 content in the NEMO immunoprecipitates was verified by immunoblotting of the immunoprecipitates with antibody specific for IKK1. RSK1 activity was determined by immunoblotting analysis of the cell extracts with an antibody that specifically reacts with phosphorylated RSK1. The amount of RSK1 present in each condition was assessed by immunoblotting with antibody specific for RSK1. The p53 content in the whole cell extracts was monitored by immunoblotting with antibody specific for p53. D, Saos-2-Tet-On-p53 cells were incubated with doxycycline for 24 h to induce expression of p53, and whole cell extracts were prepared and adjusted to equal protein concentration. Equal amounts of protein were verified by immunoblotting of the whole cell extracts (lysate). p53, RSK1, and NEMO were immunoprecipitated (IP) with specific antibodies, and the immunoprecipitates were analyzed by immunoblotting with antibodies reacting with RSK1 and IKK1. Note the coimmunoprecipitation of small quantities of RSK1 with p53. E, cell lysates were fractionated by FPLC using a Superose 6 column. The amounts of IKK1, p53, and RSK1 in the individual fractions were determined by immunoblotting with antibodies specific for each of these proteins. not the amino-terminal p65 fragment present in GST-p65-(1-181) nor the intervening p65 fragment present in GST-p65-(182-360) (Fig. 5B). This finding suggests that RSK1 targets p65 for phosphorylation at one or more residues located within the carboxyl-terminal domain (amino acids 361-551).
The carboxyl-terminal domain of p65 is phosphorylated on serine 529 by casein kinase II (13), and on serine 536 by IKK2 (48). To characterize further the phosphorylation of p65 by RSK1, we used a well characterized polyclonal antibody that specifically reacts with phosphoserine 536 on p65. This ␣-phosphoserine 536 antibody reacted with phosphorylated His-p65 but not with His-p65(536A) when these substrates were added to the in vitro kinase assay in the presence of recombinant IKK2 (data not shown). Next, we performed an in vitro kinase assay with recombinant RSK1 and His-p65 as the substrate. The ␣-phosphoserine 536 antibody effectively recognized a phosphorylated product in this reaction, suggesting that RSK1 directly phosphorylates serine 536 of p65 in vitro (Fig. 6A). To analyze p65 phosphorylation in vivo, Saos-2-Tet-On-p53 cells were incubated with doxycycline to induce p53 expression or were activated with TNF-␣, and the immunoprecipitated p65 proteins were analyzed by immunoblotting. In both cases, the ␣-phosphoserine 536 antibody reacted with the immunoprecipitated p65 (Fig. 6B). These findings indicate that p53-activated RSK1 mediates phosphorylation of p65 on at least serine 536 in vivo. NF-B Activation by DNA-damaging Agents Requires p53 and RSK1-Topoisomerase inhibitors doxorubicin and etoposide are DNA-damaging agents that induce DNA breaks, thereby stabilizing and activating p53. These agents have also been demonstrated to induce NF-B (20, 49, 50). An intriguing question was whether in our experimental system these DNAdamaging agents require p53 to induce NF-B. To test this possibility, we compared parental Saos-2 cells lacking p53 to Saos-2-Tet-On-p53 cells, which, in the absence of doxycycline, express very low amounts of p53 reminiscent of the levels found in physiological situations. Both doxorubicin and etoposide effectively induced NF-B activity in Saos-2-Tet-On-p53 cells but not in Saos-2 cells (Fig. 7A). These findings implicate p53 in the activation of NF-B by these DNA-damaging agents.
As we have shown earlier, p53 expression activates RSK1 (Fig. 3C), and RSK1 activity is required for NF-B activation (Fig. 4). Similarly, doxorubicin and etoposide activated RSK1 in Saos-2-Tet-On-p53 cells but not in Saos-2 cells (Fig. 7B). This p53-dependent activation of RSK1 was completely blocked by pretreatment of the cells with the MEK1 inhibitor UO126. MEK1 corresponds to an upstream component of the mitogenactivated protein kinase pathway that leads to the activation of RSK1. Thus, p53 is essential for up-regulation of the MEK1/ RSK1 pathway and subsequent NF-B activation. These findings also suggest that p53 binding to RSK1 is likely not a sufficient signal for RSK1 activation because the U0126 results indicate that activation of the MEK1 is required for the response.
To test further whether RSK1 is required for NF-B activation by these DNA-damaging agents, we used RSK1 siRNA to deplete endogenous RSK1 in Saos-2-Tet-On-p53 cells. We then tested activation of the B-luciferase reporter. Expression of the RSK1 siRNA but not control siRNA effectively prevented activation of luciferase in response to doxorubicin and etoposide. In contrast, TNF-␣-mediated activation of the NF-B response was not affected markedly by these siRNAs (Fig. 7C). We conclude that NF-B activation occurring in response to doxorubicin and etoposide requires RSK1 like the p53 response described previously.
DNA-damaging Agents Induce RSK1-mediated Phosphorylation of p65-Next we examined whether these DNA-damag- ing agents induce phosphorylation of p65 on serine 536 as observed for p53. Saos-2-Tet-On-p53 cells were stimulated with doxorubicin and etoposide followed by the fractionation of cytoplasmic and nuclear proteins. Stimulation with doxorubicin and etoposide resulted in phosphorylation of nuclear but not cytoplasmic p65 on serine 536 (Fig. 8A). In contrast, TNF-␣ stimulation revealed phosphorylation of both cytoplasmic and nuclear forms of p65 on serine 536. This result is consistent with prior reports demonstrating that activated RSK1 moves from the cytoplasm into the nucleus (51). In contrast, TNF-␣mediated activation of IKK2 promotes phosphorylation of cytoplasmic forms of p65 that upon IB␣ degradation translocate into the nucleus. Interestingly, treatment with doxorubicin and etoposide resulted in the appearance of a protein doublet recognized by ␣-phosphoserine 536 antibody, raising the possibility that phosphorylation also occurs at another site in p65.
Doxorubicin-and etoposide-induced phosphorylation of p65 on serine 536 was completely blocked by pretreatment of cells with the MEK1 inhibitor UO126 (Fig. 8A). Earlier we showed that UO126 abolished doxorubicin-and etoposide-induced p53dependent RSK1 activation (Fig. 7B). To test whether RSK1 is specifically required for phosphorylation of p65 on serine 536, we knocked down RSK1 expression in Saos-2-Tet-On-p53 cells by treatment with RSK1 siRNA. This treatment resulted in a substantial decrease of both p65 accumulation in nucleus and p65 phosphorylation on serine 536 after activation with doxorubicin and etoposide. In contrast, TNF-␣-induced phosphorylation of p65 remained unchanged (Fig. 8B). Therefore, RSK1 is essential for doxorubicin-and etoposide-induced phosphorylation of serine 536.
Phosphorylation of p65 by RSK1 Decreases Its Affinity for IB␣-I〉␣ functions as a key negative regulator of NF-〉 activity, acting in both the cytoplasm and nucleus. Because p53 expression does not induce cytoplasmic degradation of I〉␣, we considered the possibility that p53 acts by altering the basal nucleocytoplasmic shuttling properties of the NF-〉⅐I〉␣ complex favoring nuclear retention of NF-B.
In addition to our finding that RSK1 phosphorylates nuclear p65, prior studies have shown that this kinase also phosphorylates I〉␣ on serine 32 (52,53). Thus, we investigated whether these post-translational modifications might negatively influence the interaction of these two proteins, leading to a slow mechanism of NF-B activation. Recombinant His-p65 and GST-I〉␣ proteins were phosphorylated with RSK1 in vitro, mixed, and tested for assembly by pull-down of GST-I〉␣ with glutathione-agarose. Phosphorylation of IB␣ by RSK1 had no apparent effect on its binding to p65 (Fig. 9). However, phosphorylation of p65 by RSK1 markedly decreased its binding to both unphosphorylated and phosphorylated forms of I〉␣. These results suggest that p53-induced, RSK1-mediated phosphorylation of p65 decreases its affinity for I〉␣, thereby providing a possible mechanism for the enhanced nuclear localization and increased DNA binding of NF-B.

DISCUSSION
Activation of NF-〉 by p53 has been proposed to play a key role in p53-induced apoptosis (5). Conversely, the classical, TNF-␣-activated NF-B pathway based on the IKK-triggered, proteasome-mediated degradation of IB␣ often promotes antiapoptotic effects in cells. Our studies now reveal that p53 induces NF-B activation by a novel mechanism. Occurring independently of the activation of p53-dependent gene expression, p53 induces phosphorylation of nuclear p65 on serine 536 through the action of the protein kinase RSK1. RSK1-phosphorylated p65 exhibits a lower affinity for its negative regulator, IB␣. This impaired assembly of p65 with I〉␣ may lead to a slow increase in the steady-state levels of nuclear p65 as a result of the spontaneous shuttling of this complex into the nucleus coupled with reduced I〉␣-dependent nuclear export of p65 (Fig. 10). We further demonstrate that genotoxic stress inducers such as doxorubicin and etoposide activate NF-B though this p53-dependent pathway involving RSK1 activation and phosphorylation of p65. This nonclassical pathway of NF-B activation by p53 likely accounts for the slow tempo of the response and could contribute to the activation of a distinct set of genes that participate in the apoptotic response.
We provide several lines of evidence demonstrating that the classical IKK/I〉 signaling pathway is not involved in p53induced activation of NF-〉. First, we do not detect degradation or phosphorylation of I〉␣ in p53-expressing cells under conditions where NF-〉 was effectively induced (Fig. 2, C, D,  and E). Second, small molecule inhibition of the proteolytic function of the 26 S proteasome does not impair p53-induced accumulation of p65 in the nucleus (Fig. 2F). Third, p53 expression does not lead to activation of the IKKs (Fig. 3C), and furthermore, p53 induction of NF-〉 occurs in MEF where both the IKK1 and IKK2 genes had been disrupted (Fig. 3A).
Ryan et al. (5) suggested that p53-induced apoptosis depends on I〉␣ degradation based on the fact that overexpression of I〉␣ superrepressor blocked the p53 response. However, these authors did not address directly the role of I〉␣ degradation in p53-induced activation of NF-〉. It is possible that the binding of the IB␣ superrepressor to p65 may not be affected adversely by p65 phosphorylation. Alternatively, phosphorylation of serine 32 in IB␣ may play a role in the impaired assembly of IB␣, which was not detected in our in vitro interaction assay.
Various nonclassical NF-〉 activation pathways have been described. For example, the TNF-␣-activated protease calpain degrades I〉␣ independently of both IKKs and the ubiquitin/ proteasome pathway (54). In addition, induction of NF-〉 by short wavelength ultraviolet light does not require activation of IKKs or phosphorylation of I〉␣ on serines 32 and 36 (55,56). These findings are similar to those described here except that UV-C light induced modest, ubiquitin-dependent degradation of I〉␣ by the proteasome.
Nevertheless, it is well documented that NF-〉 can be activated in the absence of I〉␣, I〉␤, or I〉⑀ degradation (57). Various conditions, including hypoxia, reoxygenation, treatment with the tyrosine phosphatase inhibitor pervanadate (58,59), activation by nerve growth factor (60), H 2 O 2 (61), and exposure to silica (62), induce NF-〉 through tyrosine phosphorylation of I〉␣, which promotes its physical dissociation from NF-〉. However, we did not observe tyrosine phosphorylation of I〉␣ or degradation of I〉␤ and I〉⑀ as a result of p53 expression (data not shown). Like p53, all of these stimuli function as "slow" inducers of NF-〉, displaying kinetics that are quite prolonged compared with TNF-␣. Interestingly, in addition to NF-B, p53 is also activated during these stress conditions. It is therefore conceivable that under such conditions, NF-〉 is activated by the p53-triggered pathway we describe here.
The observation that p53 expression activates NF-〉 without engaging the cytoplasmic components of the classical NF-〉 inductive pathway directed our attention to the potential role of nuclear events. NF-〉⅐I〉 complexes shuttle steadily between cytoplasm and nucleus (63). DNA damage induced by genotoxic drugs such as doxorubicin and etoposide lead to the accumulation of p53 in the nucleus (64). We found that DNA-damaging agents activate RSK1, a downstream target of the MEK1/2 mitogen-activated protein kinases. We further show that RSK1 activation is required for the induction of NF-B by these genotoxic agents or by p53. When activated, RSK1 is known to translocate into the nucleus (51). Consistent  8. RSK1 is required for phosphorylation of p65 on serine 536. A, Saos-2-Tet-On-p53 cells were pretreated with UO126 where indicated. Next, cells were activated with doxorubicin, etoposide, or TNF-␣. Nuclear and cytoplasmic extracts were prepared, and samples were immunoblotted (WB) with the anti-phosphoserine 536 antibody (upper panels). Equivalent protein loading was verified by immunoblotting with antibody specific for a ␣-tubulin (cytoplasmatic fractions) and lamin A (nuclear fractions). B, Saos-2-Tet-On-p53 cells were transfected with the indicated siRNAs and activated 3 days later with doxorubicin, etoposide, or TNF-␣. Nuclear extracts were prepared and analyzed by immunoblotting with anti-phosphoserine 536 antibody or a p65 antibody. The RSK1 expression was assessed by immunoblotting. Equal loading was verified by immunoblotting with an antibody to lamin.
with this finding, we show that activated RSK1 phosphorylates nuclear but not cytoplasmic forms of p65 on serine 536. We further demonstrate that phosphorylation of p65 by RSK1 reduces its binding to I〉␣. This results in an uncoupling of the NF-〉⅐I〉 shuttling cycle, leading to the steady accumulation of free p50/p65 heterodimers in the nucleus. These NF-B complexes bind to DNA and induce target gene expression. Higashitsuji et al. (65) have recently identified HSCO (Hepatoma Subtracted-cDNA library Clone One), a nuclear-cytoplasmic shuttling protein commonly overexpressed in hepatocellular carcinomas. HSCO inhibits apoptosis induced by DNA-damaging agents. This effect appears to occur as a consequence of the ability of HSCO to bind to p65 and to accelerate nuclear export of NF-B back into the cytoplasm. This finding highlights how modulation of the NF-B⅐IB␣ nucleocytoplasmic shuttling cycle can serve as a mechanism for controlling NF-B activity and subsequently apoptosis.
It remains unclear whether RSK1-mediated phosphorylation of p65, like phosphorylation by protein kinase A (12) and casein kinase II (13), affects the DNA binding and transcriptional activity of p65. Protein kinase A and mitogen-and stressinduced kinase 1 (MSK1) phosphorylate serine 276 of p65 in response to interleukin-1, lipopolysaccharide, and TNF-␣, respectively (12,43). Serine 529 is phosphorylated by casein kinase II in response to TNF-␣ (13, 66), whereas Ser-536 is phosphorylated by IKKs (48). We now demonstrate that p53activated RSK1 leads to phosphorylation of p65 on at least Ser-536 in vivo (Fig. 6B). Our preliminary results indicate that mutation of serine 536 does not completely abrogate RSK1induced phosphorylation of p65 in vitro, 2 suggesting that additional phosphoacceptors may be targeted by RSK1. In this regard, we observed two forms of p65 phosphorylated at serine 536 after treatment with doxorubicin and etoposide (Fig. 8). Thus, additional residues within the carboxyl-terminal fragment of p65 are likely targeted by RSK1. The different patterns of p65 phosphorylation induced by different stimuli could contribute to the activation of a different subset of cellular genes and thus to different physiological responses. This effect could be mediated through differential association/dissociation of p65 with various transcriptional coactivators/corepressors (14,15,67,68). However, given the limited complexity of these phosphorylations, it seems unlikely to represent the only mechanism by which specificity of the NF-〉 response is achieved. The final transcriptional response likely depends on the composition of the inducible NF-〉 complexes, the availability of other transcription factors, and a complete set of modifications, including phosphorylation, acetylation, ubiquitylation, and methylation involving both the relevant transcription factors and the histones present in the surrounding chromatin environment.
The amino-terminal domain of p53 is critically required for NF-〉 activation because replacement of this region with an unrelated transactivation domain (fusion protein VP16-p53-(80 -393)) eliminates transcription of the NF-〉-dependent reporter gene (Fig. 1B). Furthermore, NF-〉 activity is reduced significantly when the proline-rich region of p53 that forms recognition sites for SH3 motifs is deleted. Consistent with a role for NF-〉 induction in the p53-mediated apoptosis, the proline-rich region of p53 is also required for p53-induced apoptosis (36,69). Our finding that transcriptionally inactive forms of p53 can induce NF-〉 suggests that p53 may interact directly with components of the NF-〉 signaling machinery. In this regard, prior studies have implicated the interplay of p53 with other cellular proteins in the proapoptotic response (70 -72). We have observed that p53 associates with RSK1 in a 100 -300-kDa complex (Fig. 3D). However, the fact that RSK1 sequence does not contain a classical SH3 motif makes it less likely that the assembly of these proteins occurs through the proline-rich motif.
RSK1 is known to activate several transcription factors by phosphorylation, including cAMP-response element-binding protein (46), c-Fos (44), SRE (45), and ER81 (47). These phosphorylation reactions occur in the nucleus and are facilitated by the formation of the ternary complex among the individual 2 J. Bohuslav, unpublished observations. FIG. 9. RSK1 phosphorylation of p65 decreases its interaction with IB␣. His-p65 and GST-I〉␣ were phosphorylated in vitro with recombinant active RSK1. GST-I〉␣ and phosphorylated GST-I〉␣ were bound to glutathione-agarose beads and incubated with His-p65 or phosphorylated His-p65. After unbound His-p65 was removed by washing, the binding of His-p65 to I〉␣ was analyzed by immunoblotting (WB) with antibody to p65. Comparable amounts of GST-I〉␣ in the binding reactions were confirmed by immunoblotting with antibodies specific for GST (lower panel).

FIG. 10. A model for the activation of NF-B by the p53 tumor suppressor.
p53 induces RSK1 activation, which in turn promotes phosphorylation of serine 536 in p65 as well as other carboxylterminal sites in p65. These nuclear forms of p65 stem from the basal nucleocytoplasmic shuttling of the NF-B⅐IB␣ complex. RSK1-mediated phosphorylation of p65 decreases the affinity of p65 for IB␣, resulting in reduced nuclear export and an increase in DNA binding of NF-B. Thus, p53 activation of NF-B results from the disruption of the normal nucleocytoplasmic shuttling of NF-B⅐IB␣ complexes. This mechanism likely accounts for the slow tempo of NF-B activation induced by p53 and is consistent with the lack of involvement of IKK activation and IB␣ phosphorylation and degradation in the p53 induction of NF-B. transcription factors, RSK1, and p300/CBP. p65, RSK1, and p53 all associate with p300/CBP (67,73,74), and we have observed an interaction between p53 and RSK1. In addition, the amino-terminal domain of p53 that is involved in CBP binding is also required for NF-〉 induction by p53 (Fig. 1B). Therefore, p300/CBP might act as a scaffold protein that binds both p65 and the p53/RSK1 complex, leading to the phosphorylation of p65 by activated RSK1. However, because the majority of RSK1 appears to reside in complexes of 100 -300 kDa (Fig. 3E), either only a very small fraction of the cellular p53/ RSK1 is assembled with p300/CBP and p65 or this complex is only formed in a transient manner. Further, the ability of the MEK1 inhibitor to block p53 activation of RSK1 suggests that upstream signaling in the mitogen-activated protein kinase pathway is also key for p53-mediated activation of RSK1.
In conclusion, we describe the molecular mechanism by which p53 activates NF-〉. Instead of the classical antiapoptotic IKK⅐I〉␣-dependent pathway, this mechanism involves RSK1 activation, which in turn leads to p65 phosphorylation in the nucleus on serine 536 and possibly other residues culminating in diminished assembly of phosphorylated p65 with I〉␣ (Fig. 8). DNA-damaging agents also activate NF-B through this p53-and RSK1-dependent pathway. In addition to the enhancement of DNA binding, the specific pattern of p65 phosphorylation by RSK1 together with other potential modifications may induce a set of genes that executes the final p53-induced proapoptotic response. Indeed, several NF-〉-induced proapoptotic genes have been identified, including TRAIL (24), CD95 (APO-1/Fas) ligand, DR4 and DR5 (29,75), JunB (76,77), c-myc (78,79), and TGF-␤-inducible early gene (77,80). Because the NF-B pathway has attracted great interest as a potential target for the development of small molecule inhibitors with immunomodulatory and anti-inflammatory effects, it is essential to develop a more complete understanding of the contrasting effects of NF-B in the proand antiapoptotic signaling pathways.