ATM Activation by DNA Double-Strand Breaks Through the Mre11-Rad50-Nbs1 Complex

See allHide authors and affiliations

Science  22 Apr 2005:
Vol. 308, Issue 5721, pp. 551-554
DOI: 10.1126/science.1108297

This article has a correction. Please see:


The ataxia-telangiectasia mutated (ATM) kinase signals the presence of DNA double-strand breaks in mammalian cells by phosphorylating proteins that initiate cell-cycle arrest, apoptosis, and DNA repair. We show that the Mre11-Rad50-Nbs1 (MRN) complex acts as a double-strand break sensor for ATM and recruits ATM to broken DNA molecules. Inactive ATM dimers were activated in vitro with DNA in the presence of MRN, leading to phosphorylation of the downstream cellular targets p53 and Chk2. ATM autophosphorylation was not required for monomerization of ATM by MRN. The unwinding of DNA ends by MRN was essential for ATM stimulation, which is consistent with the central role of single-stranded DNA as an evolutionarily conserved signal for DNA damage.

In mammalian cells, DNA double-strand breaks trigger activation of the ataxia-telangiectasia mutated (ATM) protein kinase, which phosphorylates downstream targets that initiate cell-cycle arrest, DNA repair, or apoptosis. Several of these targets, including p53, Chk2, Brca1, and H2AX, function as tumor suppressors in vivo, and the phosphorylation of these factors is critical for their function after DNA damage.

The Nbs1 (nibrin) protein is also a substrate for ATM, and abrogation of Nbs1 phosphorylation inhibits checkpoint signaling during the S phase (chromosome replication) of the cell cycle (1). Nbs1 is part of the Mre11-Rad50-Nbs1 (MRN) complex, which is essential for DNA double-strand–break repair and genomic stability. Cells from patients with Nijmegen breakage syndrome (NBS) or ataxia telangiectasia–like disorder (ATLD) express mutant forms of the Nbs1 or Mre11 protein, respectively, and exhibit decreased levels of ATM substrate phosphorylation, particularly on Chk2 (25) and Smc1 (6, 7), despite the presence of wild-type ATM. Thus, the MRN complex may not only be a downstream effector of ATM but also may function in activating ATM to initiate phosphorylation of cellular substrates.

MRN stimulated ATM activity in vitro toward p53, Chk2, and histone H2AX in a kinase assay with purified recombinant components (8). MRN and ATM associated through multiple protein-protein interactions, and MRN contributed to ATM kinase activity by increasing the affinity of ATM for its substrates. In this in vitro assay, however, there was no effect of DNA on ATM, either with or without MRN (Fig. 1A).

Fig. 1.

Requirement of DNA and MRN for activity of dimeric ATM. (A) Kinase assays contained 0.125 nM monomeric ATM, 7.5 nM MRN, 7.5 nM MR, 50 nM GST-p53 (amino acids 1 to 102), and 10 ng of linear DNA. Western blots were probed with antibody to phosphoserine 15 of p53. (B) Coomassie-stained SDS–polyacrylamide gel electrophoresis (SDS-PAGE) of MRN, MR(S1202R)N, MR, Mre11, Flag-tagged monomeric ATM (monomer), Flag-ATM–HA-ATM dimeric complex (dimer), and S1981A dimeric complex (dimer S1981A). (C) Glycerol gradient sedimentation analysis of monomeric ATM, dimeric ATM, and S1981A dimeric ATM, with positions of molecular size markers as indicated. ATM concentrations are shown relative to the most concentrated fraction within each gradient. (D) Kinase assays with 0.2 nM dimeric ATM, 3.6 nM MRN, 50 nM GST-p53 substrate, and linear DNA, probed with antibody to phosphoserine 15 of p53. (E) Kinase assays with 0.2 nM dimeric ATM, 4.8 nM MRN, 200 nM GST-Chk2 substrate, and linear DNA, probed with antibody to phosphothreonine 68 of Chk2. (F) Kinase assays as in (D) with indicated amounts of MRN or MR (1.8 and 3.6 nM, respectively). (G) Kinase assays as in (D) with indicated amounts of GST-p53. Anti-GST Western blots are shown in fig. S1.

ATM exists in vivo as an inactive multimer that dissociates into active monomers after DNA damage or other forms of cellular stress (9). Thus, the ATM we studied previously may have been monomeric, either already present as a monomer in cells or converted into monomers during purification. To study the multimeric form of ATM specifically, we transfected human 293T cells with two ATM expression constructs encoding Flag- and hemagglutinin (HA)–epitope–tagged ATM, and we modified our purification procedure to preserve multimeric interactions. Sequential purification with antibodies directed against the Flag and HA epitopes yielded ATM complexes (Fig. 1B). Glycerol gradient sedimentation analysis of complexes containing both epitope-tagged forms of ATM showed that the majority of this protein fractionated as a dimer, whereas the previously purified form of ATM fractionated as a monomer (Fig. 1C).

Unlike monomeric ATM, dimeric ATM required both the MRN complex and DNA for activity. Dimeric ATM was tested for kinase activity with a glutathione S-transferase (GST) fusion protein containing residues 1 to 102 of p53, and phosphorylation was detected with a phosphospecific antibody directed against p53 protein phosphorylated on serine 15 (S15). Minimal activity was seen with ATM alone, or with ATM with MRN, whereas substantial phosphorylation was seen with ATM, MRN, and linear DNA incubated together in the kinase reaction (Fig. 1D). Similar results were seen with dimeric ATM and a GST-Chk2 substrate when an antibody that is specific for phosphothreonine 68 was used (Fig. 1E). In each case, MRN increased the amount of phosphorylated product by only two- to fivefold, whereas DNA plus MRN yielded an increase in phosphorylated product of 80- to 200-fold over ATM alone. The addition of DNA to dimeric ATM in the absence of MRN did not stimulate kinase activity (Fig. 1, D and E). Dimeric ATM also required the complete MRN complex and was not stimulated by MR complex, which lacks Nbs1 (Fig. 1F), unlike monomeric ATM, which is stimulated by both the MR and MRN (Fig. 1A). Identical results were observed with full-length Chk2 and p53 (fig. S2) (10).

The level of ATM stimulation by DNA was modulated by the substrate concentration (Fig. 1G and fig. S1). The increase in phosphorylation induced by DNA over that by MRN alone was greater than 60-fold with 12.5 nM GST-p53 substrate and 32-fold with 50 nM substrate, but only 18-fold with 200 nM substrate. Therefore, DNA and MRN may act on dimeric ATM by stimulating substrate recruitment, in similar fashion to how they act on monomeric ATM (8).

In vivo, ATM is activated by DNA double-strand breaks. To test whether DNA ends are required with dimeric ATM, we added closed circular plasmid DNA instead of linear DNA fragments. The circular DNA stimulated ATM fivefold in the presence of MRN, but when restriction enzymes were also included in the kinase reaction, the phosphorylation of p53 and Chk2 was increased by 13- to 25-fold (Fig. 2A). We did not observe any DNA sequence specificity (11); however, optimal stimulation of dimeric ATM required that the length of the DNA fragment be at least 1 to 2 kb (Fig. 2B).

Fig. 2.

Importance of DNA ends for ATM activation through MRN. (A) Kinase assays as in Fig. 1 are shown except with closed circular relaxed plasmid DNA (uncut plasmid DNA) instead of linear DNA fragments. One unit of Dra III, Bam HI, or Eco RV restriction enzyme was added as indicated. Each enzyme cuts a unique site in the plasmid. (B) Kinase assays with linear DNA of various lengths. (C) Biotinylated 415-bp and 2.3-kb DNA fragments were attached to streptavidin-coated magnetic beads and incubated with MRN and ATM. Associated MRN and ATM were detected by Western blotting. The input lane has one-10th of ATM and one-third of MRN used in the reaction. (D and E) Binding assays as in (C), except with MR or Mre11 compared to MRN.

To determine which proteins are required for DNA binding, we attached a 2.3-kb DNA fragment to magnetic beads through a biotin-streptavidin interaction. The DNA-bound beads were incubated with recombinant ATM and MRN, and we identified the bound proteins by Western blotting. MRN bound to the DNA-containing beads irrespective of the presence of ATM, whereas ATM was associated with the beads only when MRN was present (Fig. 2C). Thus, in vitro, MRN is required for the stable association of ATM with DNA. The MR complex, which lacks Nbs1, also stimulated ATM binding to DNA (Fig. 2D), but Mre11 alone did not (Fig. 2E), suggesting that the interaction between MR and ATM (8) is through the Rad50 component of the complex. Both MRN and MR recruited ATM to DNA (Fig. 2D), yet only MRN stimulated the kinase activity of ATM (Fig. 1F). Similarly, ATM bound equally well to 415–base pair (bp) and 2.3-kb DNA fragments in the presence of MRN (Fig. 2C), yet was only efficiently activated by the larger fragments (Fig. 2B). Thus, ATM recruitment to DNA does not appear to be sufficient for activation.

ATM autophosphorylation of Ser1981 (S1981) was shown to be required for the activation of ATM activity and for the damage-induced conversion of ATM dimers to monomers (9). A small amount of ATM S1981 autophosphorylation in the in vitro assay was detected with a phosphospecific antibody, and it increased threefold in the presence of MRN and DNA (Fig. 3A). To test whether autophosphorylation of ATM on S1981 is necessary for the DNA stimulation of dimeric ATM in vitro, we cotransfected Flag- and HA-tagged ATM S1981A mutant (where Ser1981 is replaced by Ala) expression constructs and purified complexes with antibodies to Flag (anti-Flag) and anti-HA (Fig. 1B). This S1981A protein also migrated as a dimer in the glycerol gradient (Fig. 1C). The S1981A dimeric ATM, as well as the wild-type dimeric protein, responded similarly to DNA and MRN, exhibiting greater than 200-fold stimulation of both p53 and Chk2 phosphorylation by DNA and MRN (Fig. 3, A and B). Thus, in vitro, S1981 autophosphorylation is not essential for MRN-dependent stimulation of dimeric ATM by DNA.

Fig. 3.

ATM autophosphorylation of S1981. (A and B) Kinase assays as in Fig. 1D, with S1981A dimeric ATM compared with wild-type dimeric ATM, with 3.6 nM MRN and linear DNA as indicated. (C) The Flag-ATM–HA-ATM dimer was bound to anti-Flag conjugated to agarose beads. Dissociation of ATM was monitored by Western blotting of the supernatant with anti-HA and antibodies to Flag, Rad50, and Nbs1. The input lane has one-third of the ATM and one-10th of the MRN used in the reaction. (D) Dissociation of the S1981A dimeric ATM complex as in (C).

The conversion of ATM dimers into monomers occurs in human cells after DNA damage and correlates with ATM activation (9). To assay the dimer-to-monomer transition in vitro, we bound the Flag-ATM–HA-ATM dimer preparation to anti-Flag conjugated to agarose beads. After washing the beads, we added MRN and DNA, isolated the beads again, and analyzed the supernatant to look for dissociated ATM proteins (Fig. 3C). Flag-ATM was not observed in the supernatant because it was still bound to the beads, but one-third of the total HA-ATM used in the reaction was found in the supernatant when MRN was added to the dimeric ATM. This dissociation of HA-tagged ATM from the Flag-tagged ATM on the beads was not dependent on DNA and occurred similarly with S1981A dimeric ATM (Fig. 3D). Neither Mre11 alone nor a nonspecific protein had any effect on ATM dimer dissociation (fig. S2).

The Mre11 protein exhibits manganese-dependent nuclease activity in vitro (12). However, the nuclease activity of Mre11 is not active in the kinase assays shown here, because all of the reactions were performed in magnesium only, conditions which do not allow Mre11 nuclease activity (12) but support ATM activity in the presence of MRN.

The MRN complex also exhibits DNA binding and DNA unwinding activities that are dependent on both adenine nucleotides and Nbs1 (13, 14). We tested the MR(S1202R)N mutant complex (where Ser1202 of Rad50 is replaced by Arg), which is specifically deficient in the adenosine triphosphate (ATP)–dependent functions of MRN (15), for stimulation of ATM activity. This mutant complex did not stimulate ATM activity (Fig. 4A), which indicates that at least one of the ATP-dependent activities of Rad50 is required for ATM stimulation. The deficiency of the mutant complex does not seem to be in DNA binding, because the MR(S1202R)N complex bound to DNA in equal amounts as wild-type MRN and also recruited ATM to DNA (Fig. 4B). The mutant complex also dissociated the ATM dimer similarly to the wild-type protein (fig. S2).

Fig. 4.

Requirement of Rad50 ATP binding and DNA unwinding for DNA-dependent stimulation of ATM. (A) Kinase assays as in Fig. 1D with MR(S1202R)N complexes or wild-type MRN assayed with GST-p53. (B) DNA binding assays as in Fig. 2C, except with MR(S1202R)N mutant complexes compared with wild-type MRN. (C) Kinase assays with MRN, ATM, and DNA containing either normal cut ends or closed hairpin ends. (D) Kinase assays comparing 2.7-kb linear DNA (paired ends) with 2.7-kb DNA containing 60 bp of heterology on one end (unpaired ends).

To test the importance of DNA unwinding for ATM stimulation, we prepared a DNA substrate with closed hairpins on each end. This substrate did not stimulate ATM activity, which indicates that opening of the DNA helix is required for MRN stimulation of dimeric ATM (Fig. 4C). We then prepared a substrate containing 60 noncomplementary base pairs at one end to mimic an unwound DNA molecule. This DNA (“unpaired ends”) complemented the MR(S1202R)N mutant for ATM stimulation (Fig. 4D); thus, the MRN-specific role of ATP in this reaction may be to stimulate DNA unwinding.

In this study, we reconstituted DNA damage signaling to ATM with recombinant purified components. The accumulating evidence indicates that DNA breaks are sensed directly by the MRN complex, which binds DNA, unwinds the ends, recruits ATM, and dissociates the ATM dimer. These results are consistent with recent studies in budding yeast, which show that the Mre11-Rad50-Xrs2 complex localizes to DNA breaks very rapidly in vivo (16, 17) and recruits the ATM homolog Tel1 to DNA breaks (18). However, our experiments do not recapitulate the requirement for ATM autophosphorylation that is observed in human cells (9). Effects of other factors, including protein phosphatases 5 (19) and 2A (20) and chromatin remodeling complexes, may be needed to reconstitute the effects of autophosphorylation that are seen in vivo. Our results in this study define in mechanistic detail the pathway for ATM activation after the formation of a DNA double-strand break and provide a biochemical foundation for the characterization of other factors that influence the activity of ATM in cells.

Supporting Online Material

Materials and Methods

Figs. S1 and S2


References and Notes

Stay Connected to Science

Navigate This Article