Reports

Molecular Linkage Between the Kinase ATM and NF-κB Signaling in Response to Genotoxic Stimuli

See allHide authors and affiliations

Science  24 Feb 2006:
Vol. 311, Issue 5764, pp. 1141-1146
DOI: 10.1126/science.1121513

Abstract

The transcription factor NF-κB modulates apoptotic responses induced by genotoxic stress. We show that NF-κB essential modulator (NEMO), the regulatory subunit of IκB kinase (IKK) (which phosphorylates the NF-κB inhibitor IκB), associates with activated ataxia telangiectasia mutated (ATM) after the induction of DNA double-strand breaks. ATM phosphorylates serine-85 of NEMO to promote its ubiquitin-dependent nuclear export. ATM is also exported in a NEMO-dependent manner to the cytoplasm, where it associates with and causes the activation of IKK in a manner dependent on another IKK regulator, a protein rich in glutamate, leucine, lysine, and serine (ELKS). Thus, regulated nuclear shuttling of NEMO links two signaling kinases, ATM and IKK, to activate NF-κB by genotoxic signals.

The NF-κB family of transcription factors is an important point of convergence for many signal transduction pathways, which regulate genes that are critical for processes such as development, innate and adaptive immune responses, cell migration, and apoptosis (1). NF-κB regulates apoptosis and is an attractive target for anticancer drug development (2). Inactive NF-κB is cytoplasmically localized because of its association with inhibitor proteins, such as IκBα. Distinct signaling cascades induced by stimuli such as tumor necrosis factor α (TNFα), bacterial lipopolysaccharide (LPS), and genotoxic agents elicit NF-κB–dependent transcription by activating the cytoplasmic IKK complex—composed of two catalytic subunits, IKKα (also called IKK1) and IKKβ (also called IKK2)—and a regulatory subunit, NEMO (also called IKKγ). Another important regulatory component of the IKK complex is a protein called ELKS (3). Upon activation, IKK phosphorylates IκB to promote its ubiquitin-dependent degradation. Liberated NF-κB then migrates into the nucleus and induces transcription of target genes.

ATM is a nuclear protein kinase that regulates apoptosis and cell cycle checkpoint responses after DNA double-strand breaks (DSBs) (4, 5). ATM is crucial for NF-κB activation by multiple DNA-damaging anticancer agents, including ionizing radiation (IR), the topoisomerase I inhibitor camptothecin (CPT), and the topoisomerase II inhibitors etoposide (VP16) and adriamycin/doxorubicin (69). NF-κB activation by CPT and VP16, but not by TNFα or LPS, requires a modification of free NEMO by small ubiquitin-like modifier 1 (SUMO-1) to promote its nuclear localization, which is then followed by its ubiquitination and activation of cytoplasmic IKK (6). Because ATM appeared critical for the latter modification of NEMO, we considered the possibility that ATM regulates NEMO activity by direct phosphorylation. ATM specifically phosphorylates serine or threonine residues followed by glutamine (Ser-Gln or Thr-Gln sites) (10). We identified five Ser-Gln motifs—Ser8, Ser85, Ser156, Ser364, and Ser383—that are candidate phosphorylation sites for ATM in the human NEMO sequence (fig. S1A).

To determine whether any of these Ser-Gln sites are required for NF-κB activation by DNA-damaging agents in vivo, we stably introduced Myc-tagged NEMO mutants with each of these serine residues substituted with alanine into the NEMO-deficient murine pre–B cell line 1.3E2 (fig. S1, B and C). NF-κB activation by multiple DSB-inducing agents or by LPS was defective in 1.3E2 cells but was restored by stable expression of wild-type NEMO (Fig. 1, A to C). NEMO with Ser85→Ala (NEMO-S85A) was the only mutant that completely failed to permit NF-κB activation by VP16 and CPT, whereas it fully restored activation by LPS (Fig. 1, B and C, and fig. S1C). Western blot analysis of IκBα indicated that this defect was associated with the loss of IκBα degradation in response to these agents (Fig. 1B), a finding consistent with the lack of DSB-inducible activation of IKK (Fig. 1D). The NF-κB activation defect in NEMO-S85A cells correlated with the loss of VP16-dependent transcription of the IκBα gene, a well-established NF-κB transcriptional target (Fig. 1E). Additionally, substitution of Gln86 to either alanine or asparagine caused NF-κB activation defects in response to DNA damaging agents that were indistinguishable from those of the NEMO-S85A mutant (Fig. 1F). Thus, Ser85 and Gln86 of NEMO are each required for DSB-inducible NF-κB activation but have no apparent role in the LPS signaling pathway.

Fig. 1.

Requirement of Ser85 of NEMO for NF-κB activation by DSB inducers. (A) NF-κB activation in NEMO-deficient 1.3E2 cells, or in those reconstituted with NEMO-WT, treated and analyzed by electrophoretic mobility shift assay (EMSA) (23). Dox, doxorubicin; Oct-1, octamer binding protein 1. (B) NF-κB activation and IκBα degradation in NEMO-WT and NEMO-S85A cells treated with LPS (L), VP16 (V), or CPT (C) were analyzed by EMSA and by anti-IκBα and anti-Myc Western blotting. The arrowhead points to IκBα protein, and the star indicates a possible IκBα degradation product. (C) Amounts of NEMO and actin from whole-cell lysates of 70Z/3, NEMO-WT, and NEMO-S85A cells were analyzed by Western blotting with antibodies (α) to NEMO and actin. Protein loading amount in lane 4 is two times the protein in lane 1. Exo., exogenous; Endo., endogenous. (D) IKK activation in NEMO-WT and NEMO-S85A cells treated with indicated agents were analyzed by immune complex kinase assay (23). KA, kinase assay. (E) IκBα mRNA levels in HEK293 cells as measured by reverse transcription polymerase chain reaction analysis. (F) NF-κB activation and Myc-tagged NEMO protein in NEMO with Gln86→Asn (Q86N) and NEMO with Gln86→Ala (Q86A) cells as analyzed by EMSA and Western blottting.

To determine whether ATM could phosphorylate NEMO on Ser85, we generated glutathione-S-transferase (GST)–fusion NEMO wild type (WT), NEMO-5A (in which all five serines are mutated to alanine), and NEMO-S85A recombinant proteins and used them in an in vitro ATM immunocomplex kinase assay (11). Ser85-dependent phosphorylation of NEMO was evident (fig. S2A). We also generated rabbit polyclonal antibodies that specifically recognized the phosphorylated Ser85 residue of NEMO (anti-pS85-NEMO). Anti-pS85-NEMO detected VP16-inducible and phosphatase-sensitive modification of NEMO-WT, but not of NEMO-85A, expressed in human embryonic kidney (HEK) 293 cells (Fig. 2A). CPT, IR, and doxorubicin also induced Ser85 phosphorylation of NEMO, but TNFα did not (Fig. 2B). Endogenous NEMO underwent inducible Ser85 phosphorylation, which was prevented by ATM-specific RNA interference (RNAi) (Fig. 2C). Moreover, AT cell lines (AT3LA and AT59) failed to induce NEMO Ser85 phosphorylation upon exposure to VP16 (Fig. 2D). Mutation of Ser85 blocked VP16-inducible incorporation of ortho-32P into NEMO in vivo (fig. S2B), suggesting that Ser85 is the predominant site of DSB-inducible phosphorylation.

Fig. 2.

Phosphorylation on Ser85 of NEMO by ATM. (A) Phosphorylation of NEMO-WT or NEMO-S85A transfected in HEK293 cells after VP16 treatment. NEMO proteins were immunoprecipitated (IP) with an antibody to Myc then analyzed by Western blotting (WB) with antibodies (α) to phospho-S85 (pS85)–NEMO and NEMO. The sample in lane 4 was treated with λ-phosphatase (λ-PPase) for 30 min after immunoprecipitation. (B) Induction of Ser85 phosphorylation of NEMO in HEK293 cells stably expressing NEMO-WT in response to CPT, IR, and doxorubicin (Dox), but not TNFα. (C) ATM siRNA prevents phospho-S85-NEMO detection in HEK293 cells. Whole-cell extracts were analyzed with antibodies to ATM, ATR, and DNA-PKcs. (D) Ser85 phosphorylation of NEMO is detectable in AT WT cells (721 and L40) but not in AT lymphoblast cells (AT3LA and AT59). (E) Coimmunoprecipitation of endogenous NEMO with an antibody to ATM in HEK293 cells. (F) Coimmunoprecipitation of endogenous ATM, but not ATR and DNA-PKcs, with an antibody to NEMO in HEK293 cells. NEMO-associated ATM is reactive to the antibody to phospho-S1981 ATM. (G) NF-κB activation in normal hPBL-T cells treated with TNFα or VP16 as analyzed by EMSA. (H) Coimmunoprecipitation of endogenous ATM, IKKα, and IKKβ with an antibody to NEMO in normal hPBL-T cells. (I) Induction of Ser85 phosphorylation of NEMO in normal hPBL-T cells. Black circle to the right indicates phosphorylated NEMO.

Consistent with the notion that ATM is a direct NEMO kinase, endogenous NEMO and ATM could be coimmunoprecipitated in a VP16-inducible manner (Fig. 2, E and F). NEMO-associated ATM interacted with a phospho-Ser1981 antibody (pS1981) that detects activated ATM (12). VP16-inducible interaction of NEMO was not detected with other ATM family members, DNA-dependent protein kinase catalytic subunit (PKcs), or ATM-Rad3-related (ATR) (Fig. 2F). Thus, ATM appears to selectively associate with NEMO to promote NF-κB activation under these conditions. The interaction of ATM and NEMO was also detected in normal human peripheral blood (hPBL)–T cells, which correlated with Ser85 phosphorylation of NEMO and NF-κB activation in these normal human cells (Fig. 2, G to I).

NEMO undergoes sequential SUMO-1 and ubiquitin modifications in response to treatment of cells with VP16 (6). Ubiquitination (Fig. 3A) but not SUMOylation (fig. S3A) of NEMO was decreased in cells expressing the NEMO-S85A mutant. Because the migration of ubiquitinated NEMO during gel electrophoresis was consistent with mono-ubiquitination, we hypothesized that phosphorylation of NEMO on Ser85 was required for its mono-ubiquitination. Therefore, we tested whether direct fusion of a ubiquitin moiety to NEMO-S85A could bypass the requirement of Ser85 phosphorylation for activation of IKK and NF-κB, a test that has been done previously to study transactivation domain function (13). To this end, we generated 1.3E2 cells stably expressing NEMO-S85A with an N-terminal ubiquitin tag (Ub-S85A) that lacked the C-terminal diglycine residues of ubiquitin and was joined to the first methionine of NEMO (fig. S3B). This modified NEMO did not appear to be rapidly degraded by the ubiquitin-fusion degradation pathway (14). Ub-S85A NEMO supported activation of IKK (Fig. 3C) and NF-κB (Fig. 3B) in response to VP16 and CPT treatments, whereas NEMO-S85A with a similarly fused N-terminal SUMO-1 tag did not (fig. S3, B and C). Ubiquitin fusion also reversed the NF-κB activation defects of NEMO-S85A in response to both IR and doxorubicin (Fig. 3, D and E). Furthermore, clonogenic survival assays demonstrated that NEMO-S85A cells were more radiation sensitive than were NEMO-WT cells, and this effect of the NEMO-S85A mutant was also reversed by ubiquitin fusion (Fig. 3F). Thus, ATM-dependent phosphorylation and ubiquitination of NEMO was also essential for a NF-κB–dependent cell survival response.

Fig. 3.

Ubiquitination and nuclear export of NEMO in response to DSB. (A) Ubiquitination of NEMO in HEK293 cells that stably express NEMO-WT or NEMO-S85A and that are transiently transfected with hemagglutinin (HA)–ubiquitin. Cell extracts were subjected to immunoprecipitation (IP) with an antibody (α) to Myc followed by Western blotting with antibodies to HA or Myc. Whole-cell lysates were also probed with the antibody to HA. The arrowhead indicates mono-ubiquitinated NEMO. IgG, immunoglobulin heavy chain; n.s., nonspecific band. (B) NF-κB activation and NEMO levels in NEMO-WT and Ub-S85A cells [treated with LPS (L), VP16 (V), or CPT (C)] as analyzed by EMSA and Western blotting, respectively. (C) IKK activation in NEMO-WT and Ub-S85A cells, as analyzed in Fig. 1D. (D and E) NF-κB activation in NEMO-WT, NEMO-S85A, and Ub-S85A cells after (D) IR treatment and (E) doxorubicin (Dox) treatment. Gy, grays. (F) Increased sensitivity to radiation of cells lacking NEMO is reversed by NEMO-WT and Ub-S85A but not by NEMO-S85A. 1.3E2, NEMO-WT, NEMO-S85A, and Ub-S85A cells were exposed to indicated doses of IR and analyzed for clonogenic survival assay (23). The average + SD of three experiments is plotted for each condition. Asterisks indicate that P < 0.05 in both an ANOVA and a Tukey's test.

Mono-ubiquitination can regulate protein targeting, such as endocytic trafficking and sorting (15), and was also implicated in nuclear export (16). We thus tested whether NEMO-S85A had a nuclear export defect that could be reversed by ubiquitin fusion. Nuclear immunostaining of NEMO-WT increased up to 120 min after VP16 treatment of cells immobilized on a glass chamber and then declined by 180 min (Fig. 4A and fig. S4, A to D). NF-κB activation, as detected by nuclear staining of p65, peaked at 180 min under these conditions (fig. S4, C and D). The peak levels of NEMO and p65 nuclear staining varied from ∼20 to 40% of exposed cells in different experiments, which correlated with the overall magnitude of NF-κB–dependent survival that we observed in this cell system (Fig. 3F). Nuclear staining of NEMO-S85A after VP16 treatment occurred with comparable kinetics to those of NEMO-WT but remained at the peak levels even 240 min after stimulation (Fig. 4A). This nuclear export defect of NEMO-S85A was bypassed by ubiquitin fusion (Fig. 4A).

Fig. 4.

Mono-ubiquitination of NEMO promotes assembly of an ATM:IKK complex in the cytoplasm. (A) A graph representing the percentage of cells showing nuclear staining of NEMO in NEMO-WT, NEMO-S85A, and Ub-S85A cells treated with VP16 for indicated times and measured by immunofluorescence with the use of an antibody (α) to Myc (23). The average + SD of triplicate studies is plotted for each condition. Asterisks indicate that P < 0.05 in both an ANOVA and a Tukey's test. (B) Coimmunoprecipitation of endogenous IKKβ with an antibody to ATM in NEMO-WT, but not in NEMO-S85A cells, after VP16 exposure. IP, immunoprecipitation; WB, Western blotting. (C) Coimmunoprecipitation of endogenous activated ATM with an antibody to IKKβ in HEK293 cells. (D) Detection of nuclear NEMO and ATM in the cytoplasm of HEK293 activated cells after VP16 exposure for indicated times. (E) NEMO siRNA prevents the appearance of cytoplasmic activated ATM in HEK293 cells. In the pS1981ATM panel, lanes 1 to 6 were exposed longer than lanes 7 to 12 by three times to reveal the lack of activated ATM in lanes 4 to 6 even after a long exposure. (F) IKK activity present in cytoplasmic ATM complexes in NEMO-WT and NEMO-S85A cells treated with VP16. Average ± SD is shown for each condition (n = 3). (G) Coimmunoprecipitation of endogenous ATM with Ub-S85A after VP16 exposure.

Polyubiquitination of certain proteins in the NF-κB signaling pathway, including NEMO, is proposed to provide a scaffold to assemble IKK signaling complexes to promote activation of IKK and NF-κB by immune and inflammatory stimuli (17). To probe the potential role of ubiquitin in the assembly of an IKK signaling complex in the DSB-dependent signaling pathway, we screened ATM and NEMO immunoprecipitates for the presence of other signaling proteins by immunoblotting. Anti-ATM precipitates prepared from NEMO-WT cells, but not those prepared from NEMO-S85A cells, contained IKKβ in a VP16-inducible manner (Fig. 4B). ATM was also detected in anti-IKKβ immunoprecipitates (Fig. 4C). Because IKKβ remains cytosolic in the NF-κB activation pathway induced by genotoxic stress (6), these results led us to test whether a fraction of ATM exits the nucleus upon DNA damage. Subcellular fractionation studies demonstrated that a small fraction of activated ATM was exported from the nucleus in a NEMO-dependent manner (Fig. 4, D and E). ATM immunoprecipitates prepared from cytoplasmic extracts demonstrated VP16-inducible presence of IKK activity (Fig. 4F). We used control immunoglobulin G (IgG) and a mutant GST-IκBα (Ser-Ser/Ala-Ala) substrate containing alanine substitutions at both IKK sites to demonstrate the specificity of IKK activity associated with cytoplasmic ATM. The relatively low amounts of IKK activity derived from ATM immunoprecipitates may be due to inefficiency of immunoprecipitation with antibodies to ATM. However, IKK activity was absent in anti-ATM immunoprecipitates prepared from NEMO-S85A cells. Moreover, this defect of NEMO-S85A to promote ATM-IKK interaction was reversed by ubiquitin fusion (Fig. 4G). Thus, the nuclear export function of NEMO was necessary to promote the ATM-IKK association and IKK activation in the cytoplasm. Moreover, NF-κB activation in Ub-S85A cells was completely blocked by the ATM inhibitors wortmannin and caffeine (fig. S4E). These findings indicate that ATM has another function in the cytoplasm: mediating the activation of IKK and NF-κB.

Previous studies implicated several cytoplasmic proteins in NF-κB activation after genotoxic stress, including receptor interacting protein 1 (RIP1) (18) and p90 ribosome S6 kinase (RSK1) (19). Reduction of RIP1 and RSK1 protein levels by RNAi had a small inhibitory effect on NF-κB activation by VP16. In contrast, RNAi reduction of ELKS, a newly discovered component of the cytoplasmic IKK complex (3), caused a severe NF-κB activation defect similar to that of NEMO reduction (Fig. 5A). The inhibitory effect of ELKS RNAi on NF-κB activation could not be attributed to a defect in Ser85 phosphorylation of NEMO or ATM-IKK interaction (Fig. 5, B and C). IKK activation was, however, strongly inhibited by ELKS RNAi (Fig. 5D). Interestingly, ATM associated with ELKS in a VP16-dependent manner (Fig. 5E). These findings suggest that ELKS is a downstream regulator that is essential for ATM-dependent IKK activation in response to DSBs.

Fig. 5.

Requirement of ELKS for IKK activation downstream of ATM:IKK complex formation. (A) Activation of NF-κB in response to VP16 in the presence of siRNA against NEMO, ELKS, RIP1, or p90RSK1. Western blot analyses of corresponding proteins are also shown. (B) Effects of ELKS depletion on Ser85 phosphorylation of NEMO. Asterisk indicates NEMO. (C) Effects of ELKS depletion on interaction of ATM and IKK. shELKS, small hairpin ELKS stable cell clone. (D) Effects of ELKS depletion on IKK activation. Average ± SD for each condition is shown (n = 3). (E) Association of ATM with ELKS. HEK293 cells were transfected with a Myc-ELKS-His construct and either left untreated or treated with VP16. Whole-cell lysates were immunoprecipitated (IP) with control IgG or antibodies (α) to ATM or Myc and analyzed by Western blotting with antibodies as indicated. (F) A model depicting the role of NEMO modifications, ATM-dependent events, and the point at which ELKS regulates activation of IKK and NF-κB in response to DSB inducers. Asterisks indicate activated ATM.

Our results collectively demonstrate that ATM phosphorylates Ser85 of NEMO in response to genotoxic stress and that this event is required for mono-ubiquitination of NEMO. This modification is essential for nuclear export of NEMO and ATM and for their subsequent interaction with the catalytic IKK subunit in the cytoplasm. Although potential cytosolic function for ATM in the regulation of β-adaptin and phosphorylation of the eukaryotic initiation factor 4E–binding protein 1 have been previously reported (20, 21), we demonstrate a previously unrecognized mechanism for stimulus-dependent nuclear export of ATM. In addition, our data demonstrate that ELKS is a critical component of DNA damage-induced IKK activation, acting downstream of cytosolic ATM-IKK complex formation. Because ATM also associates with ELKS upon genotoxic stress induction, we propose a model in which a cytosolic signaling complex containing NEMO, ATM, IKK catalytic subunits, and ELKS is assembled in response to genotoxic stress to mediate NF-κB activation (Fig. 5F). This model contrasts with those proposed for NF-κB activation in response to immune and inflammatory signals where Lys63-linked polyubiquitination plays a critical role in IKK activation (22). Despite the differences in the mechanisms and types of ubiquitination involved, our model also suggests that the role of ubiquitin in assembling an IKK signaling complex is a conserved strategy that has evolved to regulate NF-κB in response to DNA damage and immune and inflammatory signals.

Supporting Online Material

www.sciencemag.org/cgi/content/full/311/5764/1141/DC1

Materials and Methods

Figs. S1 to S4

References

References and Notes

View Abstract

Navigate This Article