ATM Activation by Oxidative Stress

See allHide authors and affiliations

Science  22 Oct 2010:
Vol. 330, Issue 6003, pp. 517-521
DOI: 10.1126/science.1192912


The ataxia-telangiectasia mutated (ATM) protein kinase is activated by DNA double-strand breaks (DSBs) through the Mre11-Rad50-Nbs1 (MRN) DNA repair complex and orchestrates signaling cascades that initiate the DNA damage response. Cells lacking ATM are also hypersensitive to insults other than DSBs, particularly oxidative stress. We show that oxidation of ATM directly induces ATM activation in the absence of DNA DSBs and the MRN complex. The oxidized form of ATM is a disulfide–cross-linked dimer, and mutation of a critical cysteine residue involved in disulfide bond formation specifically blocked activation through the oxidation pathway. Identification of this pathway explains observations of ATM activation under conditions of oxidative stress and shows that ATM is an important sensor of reactive oxygen species in human cells.

Patients with ataxia-telangiectasia (A-T) lack functional A-T mutated (ATM) protein and exhibit a pleiotropic phenotype that includes cerebellar ataxia, immunodeficiency, premature aging, and a high incidence of lymphoma (1). These defects may be connected through dysfunctional control of reactive oxygen species (ROS), indicated by observations that mammalian cells lacking ATM exhibit high concentrations of ROS and hypersensitivity to agents that induce oxidative stress (2). Lymphoma incidence and the loss of hematopoietic stem cells that occurs in mice lacking ATM can be suppressed by antioxidants (35), indicating an important role for ATM in regulating cellular defenses against redox stress. ATM is activated in response to changes in intracellular redox status (610), but it has not been clear what the initiating event is in these cases or how this relates to Mre11-Rad50-Nbs1 (MRN)–mediated ATM activation that is dependent on DNA double-strand breaks (DSBs).

To address these questions, we used primary human fibroblasts and induced oxidative stress with H2O2 or DSBs with bleomycin (11). Autophosphorylation of ATM on Ser1981, phosphorylation of the tumor suppressor p53 on Ser15, and phosphorylation of the protein kinase Chk2 on Thr68 all occurred in response to H2O2 or to bleomycin treatment (Fig. 1, A to C). Phosphorylation of histone H2AX (γ-H2AX), a marker for DNA DSBs, only occurred after bleomycin treatment (Fig. 1D), indicating that ATM activation induced by H2O2 can occur in the absence of DNA damage. Like histone H2AX, the heterochromatin protein Kap1 was phosphorylated at Ser824 after exposure of cells to bleomycin but not after H2O2 treatment (Fig. 1E), indicating that only a subset of ATM targets that are phosphorylated during the DNA damage response are also phosphorylated during oxidative stress. Treatment of cells with an ATM-specific inhibitor (Ku-55933) blocked phosphorylation of ATM, p53, and Chk2 during subsequent exposure to H2O2 (Fig. 1, F to H), confirming that they are in fact ATM-dependent. Similar results were also observed in cultured human embryonic kidney (HEK) 293T cells (fig. S1). Our results show a correlation between oxidative stress and phosphorylation of substrates that are not stably associated with DNA (p53 and Chk2), whereas DNA-associated substrates (H2AX and Kap1) are specifically targeted in response to DNA damage when ATM is recruited to DNA by MRN (12, 13).

Fig. 1

Activation of ATM by H2O2 in the absence of DNA DSBs. (A to E) Human primary fibroblasts (GM08399B) were treated with 250 μM H2O2 or 10 μg/ml bleomycin for 30 min. Proteins from cell lysates were analyzed by SDS-PAGE and Western blotting for phospho-ATM Ser1981, phospho-p53 Ser15, phospho-Chk2 Thr68, phospho-histone H2AX Ser140, phospho-Kap1 Ser824, or the nonphosphorylated proteins as indicated. (F to H) Fibroblasts were treated with the ATM inhibitor Ku-55933 for 1 hour before the addition of H2O2 and assayed for ATM phosphorylation events as in (A). (I to L) ATLD fibroblasts expressing GFP [(I) and (J)] or wild-type Mre11 [(K) and (L)] were treated with H2O2 as indicated (25, 50, or 100 μM). Phospho-p53 Ser15 and phospho-Chk2 Thr68 induced by H2O2 were compared with untreated and ATM inhibitor–treated samples and analyzed for ATM phosphorylation events as in (A).

Patients with ataxia-telangiectasia–like disorder (ATLD) express a mutated form of Mre11 and are impaired in activation of ATM through DSBs (12, 14). To test for the role of MRN in ATM activation by oxidative stress, we used fibroblasts established from an ATLD patient and complemented with either green fluorescent protein (GFP) as a negative control or wild-type Mre11. Expression of wild-type Mre11 in these cells restores normal amounts of functional MRN complex to the nucleus (15). p53 and Chk2 were both phosphorylated by ATM in ATLD cells and in Mre11-complemented cells that were exposed to H2O2 (Fig. 1, I to L). Thus, MRN is not essential for activation of ATM by oxidation, and ATM activation through this pathway may be separate from the DNA damage response machinery.

Dimeric ATM purified from mammalian cells is completely inactive but is strongly activated by addition of recombinant MRN complex and DNA ends in a purified protein system in vitro (13). We tested whether oxidative stress could also activate ATM in vitro. The addition of H2O2 to purified dimeric ATM (fig. S2) stimulated its activity toward a p53 substrate [the N terminus of p53 fused to glutathione S-transferase (GST)] to an extent similar to those observed with MRN and DNA (Fig. 2, A and B). ATM activation by H2O2 in vitro was completely inhibited by the reducing reagent N-acetyl-cysteine (NAC), whereas NAC had little or no effect on the stimulation by MRN and DNA (Fig. 2B). Autophosphorylation of ATM appeared not to be essential for H2O2-mediated activation of ATM, because full activity was observed with the Ser1981 → Ala1981 (S1981A) autophosphorylation site mutant (Fig. 2C). We store recombinant ATM in 1 mM dithiothreitol (DTT) because it exhibits spontaneous activation if purified and stored in the absence of reducing agents (fig. S3), so the effective concentration of H2O2 in the in vitro reactions is lower than cited here.

Fig. 2

Activation of purified ATM by oxidizing agents in vitro. (A) Kinase assays were performed with recombinant dimeric ATM and GST-p53 (amino acids 1 to 100) with H2O2 as indicated (0.0625, 0.125, 0.25, 0.5, and 1 mM). Phosphorylation by ATM was detected by SDS-PAGE and Western blotting with an antibody to p53 phospho-Ser15. (B) Kinase assays were performed as in (A) with 0.81 mM H2O2 but also with MRN and linear DNA as described (13) or with NAC (27, 83, or 250 μM). (C) Kinase assays as in (A) but with wild-type or S1981A dimeric ATM and H2O2 (0.81, 2.4, and 7.3 mM). (D) Biotinylated ATM was bound to streptavidin beads and incubated with GST-p53 in the presence or absence of H2O2 (0.27 mM). The amount of GST-p53 substrate bound was quantified with Western blotting using an antibody to GST. The average of three independent experiments is shown, with error bars indicating standard deviation. (E) ATM oxidized with H2O2 in vitro was analyzed by SDS-PAGE in the presence or absence of betamercaptoethanol (BME) and analyzed by Western blotting with an antibody to ATM. M and D indicate the positions of a 350-kD monomer and a 700-kD dimer, respectively. (F) ATM oxidized with H2O2 in human 293 cells was immunoprecipitated and analyzed as in (E) with ATM and phospho-ATM Ser1981 antibodies. (G) ATM oxidized in vitro was separated by glycerol gradient and visualized by Western blotting in comparison with untreated ATM. Positions of monomer and dimer ATM relative to molecular weight standards are shown. (H) Kinase assays with dimeric ATM in vitro as in (B) with 0.81 mM H2O2 except that NAC (0.25 or 0.5 mM) was added before incubation with substrate. (I) Kinase assays as in (A) except with H2O2 (0.09, 0.27, 0.8, 2.4, 7.3, or 22 mM) and diamide (0.9, 2.7, 8.3, 25, 75, or 225 μM). (J) Human cells stably expressing wild-type ATM were treated with 0.5 mM diamide for 30 min and analyzed by Western blotting for phospho-ATM Ser1981, phospho-p53 Ser15, ATM, or p53.

Association of the MRN complex with ATM results in an increase in the affinity of ATM for its substrates (13). To test the mechanism of H2O2 activation, we investigated substrate binding by immobilizing ATM on magnetic beads, incubating the beads with a substrate (GST-p53), and quantifying the amount of substrate bound. Oxidation of ATM by H2O2 increased the affinity of ATM for GST-p53 (Fig. 2D). H2O2-treated ATM also bound ATP more efficiently (fig. S4). Both results indicate that conformational changes occur in ATM upon oxidation.

In mammalian cells, ATM is initially an inactive, noncovalently associated dimer but converts to an active monomer upon DNA damage (16). With H2O2 treatment in vitro, however, ATM instead formed covalent dimers in denaturing SDS–polyacrylamide gel electrophoresis (PAGE) gels that were sensitive to reducing agents (Fig. 2E). A similar pattern of ATM dimer formation was also seen with ATM immunoprecipitated from human cells exposed to H2O2 and probed with antibodies directed against ATM or ATM phosphorylated at Ser1981 (Fig. 2F). The autophosphorylated ATM seen after H2O2 treatment was exclusively in the dimer state, indicating that the activated ATM did not undergo a dimer-to-monomer transition, as it does in response to DNA damage (16). Glycerol gradient ultracentrifugation of purified ATM treated with H2O2 in vitro confirmed that oxidized ATM migrated with the apparent molecular size of an ATM dimer (Fig. 2G).

These results led us to test whether one or more intermolecular disulfide bonds form between ATM monomers during oxidative stress and whether these are important for the activation of ATM. Disulfide bonds are reversible by reducing agents, whereas some other forms of oxidized cysteine are not. We therefore exposed purified ATM to H2O2, then incubated it with NAC, and then measured activity. Activation was strongly inhibited by NAC (Fig. 2H), consistent with disulfide bonds being required for activation by H2O2.

Disulfide bond formation can also be catalyzed by agents that do not generate ROS. For instance, the thiol oxidant diamide can act as a hydrogen acceptor for reactive thiol groups and promotes formation of disulfide bonds between closely opposed cysteine residues (17). Diamide does not form oxygen radicals and is effective as a thiol oxidant under hypoxic conditions (18, 19). ATM-mediated phosphorylation of p53 in vitro was stimulated by diamide (Fig. 2I). ATM autophosphorylation and phosphorylation of p53 on Ser15 were both increased after treatment of human cells with diamide (Fig. 2J), indicating that disulfide bond formation is very likely an essential component of ATM activation by oxidative stress.

To generate an ATM mutant deficient in the oxidation pathway, we made several mutations in cysteine residues in the ATM protein on the basis of conservation, association with cancer or A-T, and mass spectrometry–based analysis of ATM disulfide bonds (table S1), but mutation of these sites did not block ATM activation by oxidation. However, mutation of the only cysteine in the FRAP/ATM/TRRAP C-terminal (FATC) domain, Cys2991, resulted in an ATM mutant that was fully activated by MRN and DNA in vitro but not by H2O2 (Fig. 3A). Cys2991 is located close to the kinase domain of ATM and is conserved among all terrestrial vertebrates, but is not present in ATM from zebrafish or lower eukaryotes (fig. S5).

Fig. 3

Requirement of Cys2991 and the C terminus of the FATC domain for ATM activation by oxidative stress. (A) Kinase assays as in Fig. 2, A and B, with the wild-type (0.27 and 0.81 mM H2O2) and the C2991L ATM mutant (0.27, 0.81, and 2.4 mM H2O2). (B) ATM with (top) or without (bottom) H2O2 was analyzed for disulfide bond formation at the Cys2991 site as described in the text and (11). The 3181.49 peak shows NEM modification of the Cys2991-containing peptide in the treated sample; the control is modified by propionamide. (C) Kinase assays as in (A) except with ATM wild-type homodimer (WT-WT), C2991L-WT heterodimer (WT-CL), or C2991L homodimer (CL-CL) proteins with H2O2 (0.27 or 0.81 mM for WT and 0.27, 0.81, or 2.43 mM for WT-CL and CL-CL). (D) Kinase assays as in (A) except with the ATM mutant R3047X. (E) ATM dimerization assays as in Fig. 2E, except with 5 mM MnCl2 or H2O2 (0.27 mM). M and D indicate positions of ATM monomer and dimer, respectively. (F) Kinase assays as in (A) except with the C2991L ATM mutant, MnCl2 (5 mM), and 0.25 mM NAC.

To determine whether Cys2991 is involved in disulfide bond formation during ATM oxidation, we treated ATM in vitro with H2O2, separated the oxidized ATM by SDS-PAGE, blocked all reactive thiol groups with iodoacetamide, then reduced the thiols in disulfides with Tris(2-carboxyethyl)phosphine (TCEP), a disulfide-specific reducing agent. The thiols reduced by TCEP were modified with N-ethyl maleimide (NEM), and the protein was digested with trypsin. Analysis of the tryptic peptides by mass spectrometry showed that the peptide containing Cys2991 reacted with acrylamide in the gel to generate a propionamide adduct in the absence of oxidation, consistent with this residue being highly reactive. With oxidation treatment, the peptide shows addition of NEM, confirming that Cys2991 does form a disulfide bond during ATM oxidation (Fig. 3B and table S2).

The Cys2991 →Leu 2991 (C2991L) mutant still formed a disulfide–cross-linked dimer during H2O2 exposure, similar to the wild-type protein, even though it is not active (fig. S6), consistent with our observation that oxidized ATM contains at least several disulfide bonds (table S1). To determine whether Cys2991 is required in both subunits of the dimer, we prepared heterodimers of ATM containing one wild-type subunit and one C2991L subunit by using two different epitope tags and sequential affinity purification (20). The C2991L–wild-type heterodimeric ATM, like the C2991L homodimeric form, showed a normal response to MRN and DNA but was not activated by H2O2 in vitro (Fig. 3C). This result suggests that the disulfide bond formed at Cys2991 is intermolecular.

Deletion of the last 10 amino acids of ATM results in a mutant protein (R3047X) that causes A-T; however, patients with this mutation have been classified as A-T variants because they do not exhibit immunodeficiency and show reduced radiosensitivity compared with most A-T patients (2123). The recombinant R3047X ATM exhibits properties similar to that of the C2991L mutant in that MRN/DNA stimulation is normal but activation by oxidation is completely deficient (Fig. 3D). This region of the FATC domain has also been suggested to be a peroxisome targeting signal, and ATM has been reported to localize with the peroxisome (24).

MnCl2 has been widely used to activate immunoprecipitated ATM in vitro in the absence of MRN (2529), but the mechanism of this activation is not understood. We found that MnCl2 induced ATM cross-linking in vitro and that MnCl2-dependent activation was inhibited by the reducing reagent NAC (Fig. 3, E and F). The ATM C2991L mutant showed reduced activation by manganese compared with that of wild-type ATM (Fig. 3F). Thus, MnCl2 appears to primarily activate ATM through the oxidation pathway. This may be due to trace amounts of Mn3+ in preparations of MnCl2 (30).

To further investigate the functional effects of the C2991L mutation, we used lymphoblasts from an A-T patient lacking functional ATM and stably complemented these cells with either wild-type or C2991L alleles of ATM under the control of an inducible promoter (fig. S7). After induction, cells were exposed to a low concentration of H2O2 or to camptothecin, a topoisomerase poison that induces DNA breaks. Consistent with the results with purified proteins in vitro, wild-type ATM responded to both H2O2 and camptothecin, whereas the mutant allele only responded to camptothecin treatment (Fig. 4, A and B). To determine whether these phosphorylation events affected cell survival, we monitored both groups of cells for apoptosis, which is induced by ROS or DNA damage in lymphocytes (31). The cells expressing wild-type ATM showed a strong apoptotic response to both H2O2 and camptothecin (as measured by a fluorescent caspase 3 substrate assay), whereas cells expressing the mutant allele only underwent caspase activation in response to camptothecin (Fig. 4, C and D). Confirmation of these results was also obtained by using propidium iodide staining and annexin V to measure cleavage of nuclear DNA and loss of membrane integrity during apoptosis (figs. S8 and S9).

Fig. 4

Oxidative activation of ATM in human cells is blocked by the C2991L and R3047X mutations. (A and B) AT1-ABR human lymphocytes inducibly expressing WT or C2991L alleles of ATM were treated with H2O2 (25 μM) or camptothecin (10 μg/ml) as indicated. Cell lysates were analyzed by SDS-PAGE and Western blotting for phospho-ATM Ser1981, phospho-p53 Ser15, or the nonphosphorylated proteins as indicated. (C and D) AT1-ABR cells expressing WT or C2991L ATM were treated with H2O2 (25 μM) or camptothecin (5 μg/ml). The increase of caspase 3 activity during apoptosis was measured by using the fluorescent indicator PhiPhiLux-G2D2 (OncoImmunin, Incorporated, Gaithersburg, MD). (E) WT or R3047X A-T lymphocytes were treated with H2O2 (12.5 μM) or camptothecin (10 μg/ml). Cell lysates were analyzed by SDS-PAGE and Western blotting for phospho-ATM Ser1981, phospho-Chk2 Thr68, or the nonphosphorylated proteins as indicated.

Considering that C2991L and wild-type heterodimers are inactive in vitro (Fig. 3C), we also overexpressed either the C2991L or a wild-type allele of ATM in cells expressing endogenous wild-type ATM (fig. S10). The cells were treated with bleomycin or H2O2, and phosphorylation of p53 on Ser15 and Chk2 on Thr68 was quantified. Cells overexpressing wild-type ATM showed higher phosphorylation of both p53 and Chk2 in response to H2O2 compared with cells overexpressing C2991L ATM; thus, ectopic expression of the C2991L mutant acted as a dominant negative and inhibited the oxidative activation of wild-type ATM in human cells.

Immortalized lymphoblasts derived from an A-T patient expressing the R3047X ATM allele were also analyzed for responses to H2O2 and DNA damage, which showed that R3047X ATM failed to autophosphorylate ATM or phosphorylate Chk2 after exposure to H2O2 but showed a response to camptothecin that was similar to that seen in cells expressing wild-type ATM (Fig. 4 and fig. S10). It is difficult to make quantitative comparisons because the WT and mutant cells were derived from different individuals. The small decrease in responsiveness of the mutant cells to campothecin could mean that the reduced response to H2O2 also reflects a deficit in response to DSBs, but we interpret the result to show a specific deficit in the oxidative response of the R3047X mutant.

We identified and characterized a pathway of ATM activation that is separate from the previously defined pathway that depends on MRN and DNA ends. ATM appears to act as a redox sensor in human cells, and, given the large number of substrates identified as ATM targets after DNA damage (32), ATM may similarly regulate global cellular responses to oxidative stress. The observation that the R3047X mutation generates an ataxia phenotype in A-T patients but retains normal activation in response to DNA damage suggests that most of the clinical manifestations of A-T may result from an inability to effectively regulate ROS, an observation that has important consequences for A-T treatment strategies.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

References and Notes

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We are grateful to members of the Paull laboratory as well as to M. Weitzman for cell lines; to D. Johnson, G. Georgiou, J. Huibregtse, and K. Dalby for discussion; and to C. Walker for communication of data in press. This work was supported by NIH grant CA132813 (Paull laboratory), NHMRC 569591 (Lavin laboratory), and NIEHS 007784 (ICMB mass spectrometry core facility).

Stay Connected to Science

Navigate This Article