Research ArticleDNA damage

ATM directs DNA damage responses and proteostasis via genetically separable pathways

See allHide authors and affiliations

Science Signaling  09 Jan 2018:
Vol. 11, Issue 512, eaan5598
DOI: 10.1126/scisignal.aan5598

The many roles of ATM

The kinase ATM is classically known for its role in coordinating the response to DNA damage. DNA damage is caused by various intracellular and extracellular stimuli, including oxidative stress and free radicals. Lee et al. found critical amino acid residues that enable ATM to coordinate a response to DNA damage that is independent of its response to oxidative stress. Activation of ATM by either pathway promoted mitochondrial function and autophagy, thus mediating cell survival through metabolic changes. ATM activation via oxidative stress additionally promoted the clearance of toxic protein aggregates. These findings expand the roles of ATM and suggest that the loss of ATM function, such as in the neurodegenerative disease ataxia telangiectasia (A-T), causes broader cellular stress than that limited to a defective DNA damage response.


The protein kinase ATM is a master regulator of the DNA damage response but also responds directly to oxidative stress. Loss of ATM causes ataxia telangiectasia, a neurodegenerative disorder with pleiotropic symptoms that include cerebellar dysfunction, cancer, diabetes, and premature aging. We genetically separated the activation of ATM by DNA damage from that by oxidative stress using separation-of-function mutations. We found that deficient activation of ATM by the Mre11-Rad50-Nbs1 complex and DNA double-strand breaks resulted in loss of cell viability, checkpoint activation, and DNA end resection in response to DNA damage. In contrast, loss of oxidative activation of ATM had minimal effects on DNA damage–related outcomes but blocked ATM-mediated initiation of checkpoint responses after oxidative stress and resulted in deficiencies in mitochondrial function and autophagy. In addition, expression of a variant ATM incapable of activation by oxidative stress resulted in widespread protein aggregation. These results indicate a direct relationship between the mechanism of ATM activation and its effects on cellular metabolism and DNA damage responses in human cells and implicate ATM in the control of protein homeostasis.


Ataxia telangiectasia (A-T) is a disorder characterized by progressive cerebellar degeneration, predisposition to lymphoid malignancies, and diabetes that is caused by the loss of the A-T mutated (ATM) kinase. Cells from A-T patients lack the ability to initiate DNA damage–induced checkpoints and are deficient in responses to DNA double-strand breaks (DSBs) (1). ATM-deficient cells also exhibit abnormalities in responses to other forms of cellular stress, including oxidation (2, 3), hypoxia (4), hyperthermia (5), and hypotonic stress (6).

ATM was initially characterized solely as a regulator of the DNA damage response, a rapid initiation of checkpoints and DNA repair that requires the Mre11/Rad50/Nbs1 (MRN) complex to recruit and activate ATM at sites of DSBs (1, 79). The importance of MRN in ATM activation is evident in the similar clinical phenotype of patients with A-T–like disorder (ATLD) or Nijmegen breakage syndrome (NBS), caused by hypomorphic mutations in the Mre11, Rad50, or Nbs1 genes (10, 11). The MRN complex localizes to sites of DSBs, recruits ATM through interactions with Nbs1 and Mre11/Rad50 (MR), facilitates the conversion of inactive dimeric forms of ATM into active monomeric forms, and promotes the stable binding of ATM substrates for efficient phosphorylation (8). We have also demonstrated that ATM can be activated by oxidative stress independently of MRN or DNA damage (3). In this pathway, multiple disulfide bonds are formed within the ATM dimer that induce an active conformation. The disulfide formed at C2991 is particularly important, because mutation of this residue blocks the oxidation-mediated activation of ATM without affecting MRN/DNA-mediated activation.

ATM deficiency has been linked for many years with observations of high levels of reactive oxygen species (ROS) and inability to respond appropriately to oxidative conditions (2). For instance, A-T patients exhibit increased oxidative damage to lipids and DNA (12) and lower levels of antioxidants in their blood plasma (13). A-T fibroblasts show increased sensitivity to hydrogen peroxide and nitric oxide donors (14, 15), also suggesting a high basal level of oxidative stress. ATM-deficient mice exhibit a loss of hematopoietic stem cells that is attributable to high ROS (16), and the incidence of T cell lymphomas in these mice is delayed and reduced by feeding with antioxidants (1721). Our previous observations that ATM-deficient cells expressing the C2991L (CL) allele of ATM or the A-T patient allele R3047X exhibit high amounts of ROS and are resistant to peroxide-induced apoptosis (3) suggest that the activation of ATM by oxidation is causally linked to regulation of global redox homeostasis and to the A-T neurodegeneration phenotype.

After DNA damage, ATM phosphorylates many proteins including histone H2AX, structural maintenance of chromosomes protein 1 (SMC1), KRAB-associated protein 1 (KAP1), checkpoint kinase 2 (CHK2), and the transcription factor p53. However, H2AX and KAP1 are not phosphorylated in the presence of oxidative stress (3), presumably because ATM is not recruited to DNA sites where these substrates are located. ATM activation by cell cycle arrest during mitosis does not also result in the phosphorylation of SMC1 or p53, substrates that are known to be phosphorylated after ionizing radiation (IR) (22). These observations suggest that ATM may activate specific downstream effectors in addition to common substrates depending on which cellular stress is present.

To delineate which activities of ATM are specific to its oxidation or DNA damage–related functions, it is essential to have separation-of-function alleles that affect each pathway specifically. The CL allele of ATM serves this purpose for the oxidative pathway as described above, but no mutants have been described that specifically alter the DNA damage–dependent pathway. Here, we identify such an allele and characterize the responses of ATM that rely on MRN-dependent activation of ATM. These results show that there are two distinct and independent pathways to activate ATM: one that operates through MRN that is induced by DSBs and regulates checkpoint functions and DNA repair, and one that is activated by oxidative stress and governs ROS abundance and protein homeostasis. Global phosphoproteomic analysis of human cells expressing these separation-of-function alleles indicates widespread changes with the loss of ATM activation by oxidative stress, largely due to deficiency in phosphorylation events normally catalyzed by the protein kinase CK2. Here, we found that CK2 aggregated in the absence of functional ATM and that global increases in protein aggregation were associated with the general lack of ATM activation by oxidative stress. These studies provide novel insights into the mechanisms of ATM activation and the DNA damage–independent functions of ATM in human cells and indicate the pleiotropic nature of the signaling deficiencies in ATM-deficient cells that stem from the loss of oxidation-induced activation.


The 2RA ATM mutant is deficient in activation by MRN and DNA

The mammalian target of rapamycin (mTOR) protein kinase, which is a member of the phosphatidylositol 3-kinase (PI3K)–related kinase (PIKK) family and is structurally related to ATM, is regulated by phosphatidic acid (PA) (23). Although ATM responds to different stimuli, we hypothesized that mutations that disrupt PA stimulation of mTOR may serve as a guide for the location of mutations in ATM that might disrupt its regulation. Therefore, we looked for residues in ATM located at a similar distance from the conserved kinase active site as the Arg2109 residue in mTOR that affects PA-mediated signaling (23). In ATM, there are two arginine residues (R2579 and R2580), which are analogous in their position to Arg2109 on mTOR and located at the end of the FRAP/ATM/TRRAP (FAT) domain in ATM (Fig. 1A). We mutated these sites to alanine (R2579A/R2580A, “2RA”) in the human ATM gene and expressed and purified the recombinant protein (fig. S1), as previously described (24). The mutant protein was tested in comparison to the wild-type protein in an in vitro assay using the N terminus of p53 fused to glutathione S-transferase (GST) as a substrate in which we monitored phosphorylation at Ser15 by quantitative Western blotting analysis. This comparison showed that the 2RA mutations resulted in the loss of ATM activation by MRN and DNA (Fig. 1, B and C). We also tested for activation by oxidation using hydrogen peroxide, which activates ATM through a DNA-independent mechanism (3). The 2RA ATM mutant still exhibited H2O2-mediated stimulation similar to that in the wild-type protein (Fig. 1D).

Fig. 1 Disruption of MRN- and DNA-dependent ATM activation by R2579A/R2580A mutations of ATM.

(A) Schematic diagram of ATM structure; mutations analyzed in this study are indicated. Features and domains within ATM consist of the nuclear localization signal (NLS), the FRAP/ATM/TRRAP (FAT), the kinase domain (PI3K), and the FAT C-terminal (FATC) domain. (B and C) Kinase assays with dimeric wild-type (WT) or R2579A/R2580A (2RA) ATM (1.35 nM), Mre11/Rad50/Nbs1 (MRN) (9.6 nM), glutathione S-transferase (GST)–p53 substrate (6.25 nM), and linear DNA (~140 nM) probed with antibody to phospho-Ser15 of p53. Data are means ± SD of three independent experiments, showing the fold change in phospho-signal with MRN/DNA relative to the reactions without MRN/DNA. (D) Kinase assays with dimeric WT or 2RA ATM (2.7 nM) and GST-p53 substrate (12.5 nM) in the presence of H2O2 (0.27, 0.81, and 2.4 mM). (E and F) Binding assays with WT or 2RA ATM and biotinylated MRN (E) or Mre11/Rad50 (MR) (F) (20 nM; incubated with 50 nM ATM and isolated with streptavidin-coated magnetic beads), as indicated. Bound proteins were visualized by Western blotting with antibodies against ATM, Nbs1, or Rad50. (G and H) Kinase assays as in (A) and (B) except with triple-mutant ATM (R2579A/R2580A/C2991L, “2RA + CL”). (I and J) Kinase assays as in (A) and (B) except using heterodimeric ATM (WT/2RA and 2RA/CL) as indicated.

Because ATM can interact with MRN through both Nbs1 and MR (25), we performed binding assays with ATM and biotinylated MRN or MR to investigate the mechanism of ATM inactivation in the 2RA mutant. The 2RA ATM mutant bound to MRN similarly to wild-type ATM but was deficient in binding to MR (Fig. 1, E and F). These results suggest that interaction of ATM and MR, in addition to interaction with Nbs1, is necessary to activate the kinase activity of ATM and that the 2RA mutations disrupt this interaction.

We have previously shown that mutation of a conserved cysteine residue in the FAT C-terminal (FATC) domain of ATM generates a separation-of-function mutant (CL) that cannot be activated by oxidative stress but shows normal activity in the presence of MRN and DNA (3). Here, we generated a combination mutant (“2RA + CL”) containing both the 2RA and CL mutations. As expected, the 2RA + CL ATM double mutant was completely deficient in both MRN/DNA- and H2O2-dependent ATM activation, as monitored by p53 phosphorylation in vitro (Fig. 1, G and H). To determine whether the 2RA mutations are dominant, we then coexpressed the 2RA mutant with either wild-type ATM or the CL mutant, using Flag and hemagglutinin (HA) tags on the proteins to purify heterodimeric ATM complexes of WT/2RA, 2RA/CL, and WT/WT purified in the same manner for comparison. The heterodimer combinations of wild-type and 2RA (WT/2RA) or 2RA and CL (2RA/CL) each showed partial activity in the presence of MRN and DNA in vitro compared to the combination of two wild-type ATM proteins (WT/WT), indicating that the 2RA mutations do not block the function of another ATM monomer in trans (Fig. 1I). This was expected, considering that ATM is initially in dimeric form but becomes monomeric in response to DNA damage (6). In contrast, the 2RA/CL ATM heterodimer showed a complete loss of activity in the presence of H2O2 (Fig. 1J), consistent with our previous finding that the C2991 residue must be intact in both subunits of the ATM dimer for activation by oxidation (3). In contrast, the 2RA/WT heterodimeric mutant did exhibit oxidative activation, albeit with lower efficiency than the wild-type complex, similar to our result with the 2RA homodimer complex (Fig. 1D). These results together confirm that two distinct mechanisms exist for the activation of ATM and can be genetically separated.

ATM is activated by two independent pathways in cells

To confirm the in vitro results with the 2RA mutant, we established an inducible expression system to test the effects of ATM mutations in human adherent cells. We used a human osteosarcoma cell line (U2OS) containing an FRT recombination site integrated into the genome adjacent to a cytomegalovirus (CMV) promoter blocked by two copies of the Tet operator (Flp-In T-REx system, Invitrogen). Vector only or the wild-type, 2RA, or CL alleles of ATM were transfected into the cells with a construct expressing Flp recombinase, and stable clones were generated. In each cell line, we depleted endogenous ATM with lentivirus containing ATM-specific short hairpin RNA (shRNA) before inducing ATM expression with doxycycline. The cells were then treated with DNA-damaging agents, including IR and camptothecin (CPT), and oxidizing agents, including hydrogen peroxide (H2O2) and sodium arsenite, to examine the effects of the ATM mutations on substrate phosphorylation in cells. We used arsenite here because work from other groups has indicated that arsenite stimulates checkpoint activation responses that are dependent on ATM (26) and that ATM is required for the survival of human cells exposed to arsenite, although arsenite does not detectably generate DNA DSBs (fig. S2) (27). Arsenite is thought to induce ROS through activation of NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase isoforms as well as through the mitochondria (28, 29) and can produce both superoxide and hydrogen peroxide in the cytoplasm. Arsenite is stable in serum-containing medium over long periods unlike exogenously added hydrogen peroxide (30, 31) and reportedly induces checkpoint responses mediated by caffeine-sensitive protein kinases (26). As expected, ATM depletion resulted in defective phosphorylation of known ATM substrates, including KAP1 and CHK2, after each treatment we used (Fig. 2).

Fig. 2 Separation-of-function mutations in ATM dictate responses to DNA damage and oxidative stress.

Human U2OS osteosarcoma cells were depleted for endogenous ATM using short hairpin RNA (shRNA) and induced to express various ATM alleles as indicated. (A) Cells were induced for ATM expression with doxycycline (10 ng/ml) and exposed to camptothecin (CPT) (10 μM) for 1 hour. ATM activity was examined using antibodies directed against phospho-Kap1 Ser824, and ATM levels were assessed with anti-ATM antibody. (B) U2OS cells were treated as in (A) except that expression was induced with doxycycline (1 μg/ml). ATM activity was assessed with antibodies directed against ATM, phospho-ATM Ser1981, Kap1, phospho-Kap1 Ser824, Chk2, and phospho-Chk2 Thr68 as indicated. (C) U2OS cells expressing shRNA against ATM (shATM) and either vector (CTRL) or various ATM alleles as in (B) were treated with 10 μM CPT for 1 hour, and the levels of phosphorylated Kap1 were determined, in comparison to total Kap1 protein, and normalized with the phosphorylated signal from WT ATM–expressing cells. NS, not significant. (D) U2OS cells were depleted of endogenous ATM and induced for recombinant ATM expression as in (B) but were exposed to IR (10 Gy) followed by 1-hour recovery. Phosphorylation was assessed as in (B). (E) U2OS cells were treated as in (B) but exposed to 100 μM H2O2 or arsenite for 1 hour in the presence of doxycycline (1 μg/ml). Phosphorylation was assessed as in (B). (F) U2OS cells expressing various ATM alleles as in (B) were treated with 100 μM H2O2 for 1 hour in serum-free medium, and the amount of phosphorylated Chk2 was quantitated in comparison to total Chk2 protein and normalized with the phosphorylated signal from WT ATM–expressing cells. Data are means ± SD of three independent experiments. *P < 0.05 and **P < 0.005.

Induction of ATM expression with low levels of doxycycline (10 ng/ml) resulted in approximately equivalent levels of recombinant ATM compared with the endogenous ATM protein in the U2OS cell line (Fig. 2A). Under these conditions, treatment of cells with CPT resulted in phosphorylation of KAP1 on Ser824, which is ATM-dependent (32). Comparison of the mutant alleles revealed that the wild-type and CL-expressing cell lines efficiently phosphorylate KAP1, whereas the 2RA-expressing line is completely deficient, similar to our results with purified ATM in vitro.

We also analyzed the responses of the cell lines with a higher amount of doxycycline induction (1 μg/ml), which yielded higher abundance of recombinant ATM, and found a similar deficit in KAP1 phosphorylation in the 2RA cell line (Fig. 2B). In three biological replicates, the fold change in KAP1 phosphorylation was reproducibly diminished in cells expressing the 2RA allele in comparison to cells expressing the wild-type allele (Fig. 2C and fig. S3). In contrast, the cells expressing the CL mutant showed wild-type levels of KAP1 phosphorylation.

Unlike CPT treatment, exposure of the cell lines to IR (10 Gy) resulted in equivalent phosphorylation of KAP1 and CHK2 by the 2RA ATM mutant in comparison to wild-type, although cells expressing the CL allele failed to fully restore phosphorylation of CHK2 (Fig. 2D). Because IR induces the formation of ROS and DNA DSBs, we considered the possibility that both modes of activation are operating when this form of DNA damage is induced. To test this idea directly, we compared the responses of the U2OS cell lines to two different forms of ROS induction. We have previously shown that treatment of cells with low levels of hydrogen peroxide induces ATM activation via the oxidation pathway independent of DNA damage (3). Here, we compared peroxide to an alternative method of inducing ROS with sodium arsenite (26) and found that both treatments induce ATM-dependent phosphorylation of CHK2 on Thr68 but not KAP1 phosphorylation (Fig. 2, E and F). In this case, only the cells expressing the CL allele fail to phosphorylate CHK2, consistent with our previous results. Notably, the peroxide and arsenite treatments used here did not induce DSBs, as measured by γ-H2AX phosphorylation (fig. S2). Although the 2RA ATM appears to be less efficient than the wild-type protein in responding to peroxide in vitro, the mutant exhibited wild-type or even higher levels of activity in response to the oxidative agents in cells.

MRN-dependent ATM activation is critical for cell survival after DNA damage but not oxidative stress

To understand the effects of the 2RA and CL ATM mutations on the responses of human cells after DNA damage and oxidative stress, we performed clonogenic cell survival assays after IR, CPT, or arsenite exposure. Cultures of cells with shRNA depletion of ATM showed significantly lower rates of survival after treatment with each of these agents compared to control shRNA-treated cells, and this sensitivity was partially recovered by expression of wild-type ATM (Fig. 3). After treatment of the U2OS cells expressing the ATM mutant alleles with the DNA-damaging agents IR or CPT, cells expressing CL ATM showed similar or higher survival rates compared to cells expressing wild-type ATM (Fig. 3, A and B). The cells expressing 2RA ATM showed significantly lower survival after exposure to these DNA-damaging agents compared to cells expressing wild-type ATM or CL ATM, although the survival was generally intermediate between ATM-depleted cells and those expressing wild-type ATM, suggesting that 2RA ATM has some residual activity or that there is some contribution of the oxidative pathway under these conditions (Fig. 3, A and B). For comparison, we also tested cells expressing the kinase-deficient D2889A (DA) mutant allele of ATM (33), which exhibited extremely poor survival of both IR and CPT (fig. S4), clearly worse than the ATM-depleted cells with no complementation, suggestive of a dominant-negative effect (34).

Fig. 3 ATM deficient in activation via MRN exhibits defects in survival of DNA damage and in DSB resection.

(A to C) U2OS cells depleted for endogenous ATM (shATM) or treated with a control shRNA (shCTRL) were induced to express vector only (CTRL) or various ATM alleles [doxycycline (1 μg/ml)] as indicated and were analyzed for cell survival after treatment with IR (A), CPT (B), or arsenite (C) as indicated. Data are means ± SE of three independent experiments. P values (inset) were assessed by Student’s t test. (D) DNA end resection at Asi SI–induced breaks in cells expressing various ATM alleles. Endogenous ATM in U2OS–ER–Asi SI cells was depleted by shRNA treatment, and cells were complemented with induced expression of ATM alleles [doxycycline (1 μg/ml)] as indicated, 3 days before treatment with 4-hydroxytamoxifen (600 nM for 4 hours), which induces nuclear translocation of the Asi SI enzyme (98). Harvested genomic DNA was either digested with restriction enzymes to distinguish between single-stranded DNA (ssDNA) and double-stranded DNA or mock-digested as described previously (99). Quantitation of single-stranded DNA intermediates generated by resection was performed by real-time polymerase chain reaction. Data are means ± SD of three experiments. *P < 0.05 and ***P < 0.0005. DSB, double-strand break; nt, nucleotides; DA, D2889A. (E and F) U2OS cells depleted of endogenous ATM with shRNA and inducibly expressing various ATM alleles [doxycycline (1 μg/ml)] were synchronized in G1–early S with aphidicolin (2 μg/ml; 17 hours). The intra-S cell cycle checkpoint was analyzed by quantification of the percentage of G2-M cells 17 hours after removal of aphidicolin and treatment with CPT (1 μM) (E) or arsenite (100 μM) (F) compared with untreated group, as indicated. Data are means ± SE of three independent experiments. **P < 0.005.

The cells were also treated with sodium arsenite to test for survival of oxidative stress. Whereas ATM-depleted cells were not as sensitive to this range of arsenite treatment compared to the DNA-damaging agents, it was clear that ATM-depleted cells had reduced survival compared to control shRNA-expressing cells (Fig. 3C). The cells expressing CL ATM showed a marked sensitivity to arsenite compared to ATM-depleted cells, a pattern opposite to that seen in response to IR and CPT. In contrast, both the 2RA and wild-type ATM alleles complemented the survival deficit of ATM-depleted cells after exposure to arsenite (Fig. 3C). These results are generally consistent with the in vitro results, indicating that 2RA ATM and CL ATM are blocked from activation after DSBs and oxidative stress, respectively (Figs. 1 and 2).

MRN-mediated ATM stimulation regulates DNA end resection

ATM activity is important for homologous recombination (35), which initiates with DNA end resection. To examine the effects of the ATM alleles on DNA end resection in human cells, we introduced an estrogen receptor fusion of AsiSI endonuclease (ER–AsiSI) to our U2OS Flp-In T-REx cells containing wild-type or mutant ATM alleles and used a quantitative polymerase chain reaction (qPCR)–based method to measure the levels of single-stranded DNA produced at two sites in the genome that exhibit high-efficiency Asi SI cleavage (36). Consistent with the critical role of ATM in DNA resection, depletion of ATM led to decreased DNA resection at both DSB sites (Fig. 3D). The resection was rescued by expression of wild-type ATM or CL ATM, but not of 2RA or DA kinase-deficient ATM, suggesting that MRN-dependent ATM activation is specifically required for DNA end resection after double-strand breakage.

Cell cycle checkpoints are disrupted in cells expressing ATM separation-of-function alleles

ATM activation is required for DNA damage–induced cell cycle checkpoints (11), and A-T cells also fail to exhibit checkpoint activation in response to ROS (37). We therefore investigated whether the 2RA and CL ATM mutants efficiently arrest the cell cycle after DNA damage or oxidative stress. We monitored the intra-S phase checkpoint, which has previously been shown to be dependent on ATM (38), by arresting cells in early S phase with a low amount of aphidicolin, adding CPT or arsenite when cells were released into S phase, and measuring the percentage of cells subsequently in G2-M using propidium iodide (PI) staining. ATM depletion generated a higher percentage of cells in G2-M after either CPT or arsenite treatment (Fig. 3, E and F), suggesting that the cell cycle checkpoint is not efficiently activated. The ratio of G2-M cells after CPT or arsenite treatment dropped significantly when wild-type ATM was expressed (Fig. 3, E and F). Expression of the CL ATM allele in ATM-depleted cells efficiently activated the cell cycle checkpoint after CPT treatment but not after arsenite treatment. As expected, 2RA-expressing cells showed an opposite pattern to that of CL-expressing cells, with efficient checkpoint activation after arsenite treatment but not with CPT-mediated DNA damage. Last, cells expressing DA ATM showed defects in cell cycle arrest after both CPT and arsenite treatment. These differences are not due to changes in the S phase populations of the cell lines expressing the various ATM alleles (fig. S5). These results further establish that there are two distinct mechanisms to activate ATM—DNA damage and oxidative stress—and indicate that this separation of function also applies to the activation of the intra-S phase checkpoint.

Stress-specific activation of ATM controls ROS, mitochondrial function, and autophagy

ATM-deficient cells have been reported to have higher levels of ROS and to have impaired mitochondrial function (2, 39, 40), although the underlying basis of the mitochondrial phenotype is not clear. We have also previously reported that ROS levels are higher in ATM-deficient lymphoblast cells overexpressing the CL allele of ATM compared to cells overexpressing the wild-type allele (3). Here, we confirmed that expression of the CL allele in U2OS cells depleted of ATM results in a higher amount of ROS compared to cells expressing the wild-type allele, as measured by the fluorescent reporter for total ROS levels, 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) (Fig. 4A). In contrast, cells expressing the 2RA allele showed wild-type ROS levels, indicating that in untreated human cells, ATM responds to oxidative stress to control ROS levels. Because many types of ROS exist in human cells, we sought to determine whether an excess of superoxide accounts for the increase measured by H2DCFDA. For this, we used the dihydroethidium (DHE) fluorescent reporter, which monitors superoxide, and found that cells expressing CL ATM showed significantly lower levels of superoxide than those expressing wild-type ATM (Fig. 4B). These results suggest that cells expressing CL ATM exhibit defects in redox homeostasis but do not accumulate superoxide, an observation generally consistent with a previous finding that overexpression of superoxide dismutase in ATM-deficient mice exacerbates, rather than alleviates, their radiation sensitivity and hematopoietic abnormalities (41). Expression of an inducible CL allele in the previously described human lymphoblast AT1ABR cell line (42) also resulted in this pattern of DHE staining (fig. S6).

Fig. 4 The mechanism of ATM activation determines the functional response of ATM in cell cycle checkpoint regulation and in ROS homeostasis.

(A and B) Reactive oxygen species (ROS) levels were measured in U2OS cells depleted of endogenous ATM with shRNA (Fig. 2) and inducibly expressing various ATM alleles [doxycycline (1 μg/ml)] using the general ROS indicator 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) (A) or a probe for superoxide levels, dihydroethidium (DHE) (B). Fluorescence was analyzed by flow cytometry using 10,000 cells per cell line and normalized with data from cells expressing WT ATM. (C) Cells expressing various ATM alleles as indicated were analyzed for membrane potential–dependent mitochondrial mass using MitoTracker Red staining, followed by analysis using flow cytometry with 10,000 cells per cell line and normalization with data from cells expressing WT ATM. Data are means ± SD of three independent experiments. *P < 0.05 and **P < 0.005. (D) U2OS Flp-In T-REx cells expressing WT, CL, or 2RA alleles were infected with retrovirus containing the mKeima mitochondria-targeted pH indicator, and pH changes in the mitochondria were measured by comparison of the emission when excited at 561 or 488 nm. The y axis shows the ratio of the emission values, normalized to the cells expressing the WT allele. Quantitation of the mean (± SD) emission ratios (561 nm/488 nm) each from four independent experiments with 10,000 cells is analyzed per cell line. ***P < 0.0005. (E) Acridine orange was used to stain acidic vesicular organelles (AVOs) in U2OS cells expressing WT, CL, or 2RA alleles, with either no treatment, CPT (5 μM), or arsenite treatment (100 μM) as indicated. The autophagy inhibitor spautin-1 (10 μM) was used to confirm that acridine orange reflects autophagy-dependent vesicles. (F) Acridine orange staining in U2OS cells expressing WT, CL, or 2RA alleles with either no treatment, CPT (5 μM), or arsenite treatment (100 μM) as indicated was quantified by fluorescence-activated cell sorting, showing here the fold increase in the percentage of cells with 585-nm emission relative to untreated cells. Data are means ± SD of three biological replicates with about 5000 cells measured per replicate. *P < 0.05. (G and H) U2OS cell lines expressing WT or mutant cell lines as indicated were treated with CPT (5 μM) or arsenite (100 μM) for 120 min, followed by 48-hour recovery, and then blotted for LC3-II and LC3-I. Data are means ± SD of three independent biological replicates, plotted as fold increase relative to the untreated controls. *P < 0.05, **P < 0.005.

Because a high amount of ROS is often associated with mitochondrial dysfunction and because the loss of mitochondrial integrity has been reported in ATM-deficient cells, we asked whether a failure in oxidative activation of ATM is responsible for this phenotype. To do this, we measured mitochondrial mass (dependent on membrane potential) with MitoTracker Red CMXRos (43). The mitochondrial membrane potential in cells expressing CL ATM was reduced in ATM-deficient cells compared to that in cells expressing wild-type ATM (Fig. 4C), and the cells expressing the 2RA showed an intermediate level. These results indicate that both activation pathways affect mitochondrial function, although the CL mutation blocking ATM activation by oxidative stress exhibited a more striking effect. Similar to the assays for ROS levels, the mitochondrial assays also showed a strong effect of the CL allele in the lymphoblast AT1ABR cell line (fig. S6). We also quantitated the amounts of carnitine and its derivatives, given that acylcarnitine metabolites are reportedly biomarkers of mitochondrial stress (44) and disruption of carnitine homeostasis results in mitochondrial dysfunction (45). Overall, levels of carnitine and its derivatives were significantly higher in A-T patient lymphoblast cells expressing CL ATM than in cells expressing wild-type ATM (fig. S6), consistent with the hypothesis that mitochondrial function in cells expressing CL ATM is impaired.

Last, we investigated whether turnover of mitochondria and other organelles is affected by the ATM separation-of-function alleles by measuring mitophagy and autophagy, respectively. We examined mitophagy in the cell lines by expressing a mitochondria-targeted pH indicator protein that responds to different wavelengths of light in a manner that depends on the pH of the environment (46, 47). The amount of lysosome-dependent mitochondrial turnover can thus be measured by the pH of the mitochondria-targeted probe. We tested the mKeima probe using the mTOR inhibitor rapamycin, which induces mitophagy (48), in comparison to wortmannin, which blocks mitophagy through inhibition of PI3K (49, 50). In wild-type U2OS cells, rapamycin increased the fold change in mKeima emission, whereas wortmannin reduced the ratio, as expected (fig. S7). To test the ATM alleles for effects on mitophagy, we expressed the mKeima sensor in cells treated with shRNA against ATM and expressing either the wild-type or CL alleles. We found that the ratio of emission with 561-nm excitation relative to emission with 488-nm excitation was significantly lower in cells expressing the CL allele compared to those expressing wild-type or 2RA ATM (Fig. 4D), indicating less efficient delivery of mitochondria to the lysosome.

In addition to mitophagy, ATM has also been implicated generally in the control of macroautophagy, the process by which proteins and organelles are degraded and recycled (5154). To determine whether either pathway of ATM activation affects control over macroautophagy, we used acridine orange to quantify acidic lysosomal vesicles in the U2OS cell lines expressing ATM alleles. This technique has previously been used to identify a role for ATM in Adriamycin-induced autophagy (54), which also confirmed the specificity of this technique for ATM-dependent autophagic flux using p62 and LC3 quantitation. Here, we found that both CPT and arsenite treatments increased the number of acidic vesicles in U2OS cells, and that this increase was largely blocked by the macroautophagy inhibitor spautin-1 (Fig. 4E) (55). Whereas all the cell lines exhibited an increase in acridine orange staining with stress, cells expressing the 2RA ATM allele showed significantly less change than untreated cells in response to either treatment (Fig. 4, E and F). In contrast, cells expressing the CL ATM allele were more proficient than those expressing wild-type ATM in their response to CPT but were deficient in their response to arsenite (Fig. 4, E and F). In addition, we analyzed cells depleted for ATM in the absence of mutant allele expression and also observed a deficiency in CPT- and arsenite-induced acidic vesicles (fig. S8).

To confirm this result using a different method, we also measured the amount of lipidated, membrane-bound LC3 (LC3-II) relative to soluble LC3 (LC3-I), which is a commonly used marker for autophagosome formation (56). Expression of the 2RA allele of ATM during CPT treatment reduced the amount of LC3-II to a similar level as seen in wild-type cells treated with spautin-1, an inhibitor of macroautophagy (55), whereas expression of the CL allele had an intermediate effect with this treatment (Fig. 4G). In contrast, the amount of LC3-II in response to treatment with arsenite was only reduced with expression of the CL allele, which was equivalent to the effect of spautin-1 in wild-type cells (Fig. 4H). Thus, the oxidative response of ATM is critical for inducing macroautophagy after oxidation.

ATM activation via oxidative stress is required for CK2 function

Our studies using ATM protein that is deficient in oxidative activation suggested that there are global effects on redox homeostasis and metabolism. To examine these effects in detail and to determine whether there are specific phosphorylation targets downstream of ATM for each activation pathway, we performed phosphoproteomic analysis using ATM-deficient AT1ABR patient lymphoblasts expressing wild-type, CL, 2RA, or 2RA + CL ATM alleles. Through this analysis, we identified 2694 phosphopeptides with a false discovery rate (FDR) of 1% (table S1). A histogram of the raw intensities from each cell line shows that the phosphoproteomes are grossly similar, as expected (Fig. 5A), and comparisons of the nonparametric Spearman’s rank correlation coefficient, ρ, also suggest that the phosphopeptide quantifications are grossly similar across all the cell lines (Fig. 5B). However, examination of the ρ values reveals that the cells expressing 2RA and CL ATM were the most dissimilar in phosphopeptide quantitation, whereas cells expressing CL and 2RA + CL ATM were the most similar to each other and to the uncomplemented AT1ABR cell line. Specifically, comparison of the CL/AT1ABR Spearman’s rank correlation coefficient with the 2RA/AT1ABR coefficient after Fisher z transformation yields a P value of <0.00001, as did comparisons of the CL/WT with the 2RA/WT data. Thus, cells expressing the wild-type or 2RA alleles were statistically very different from the parental A-T cell line compared to cells expressing the CL ATM alleles, and the expression of ATM CL results in fewer changes in the phosphoproteome compared to the ATM-deficient cell line.

Fig. 5 Phosphoproteomic analysis of A-T patient lymphoblast cells expressing an oxidation-deficient ATM mutant shows defects in global phosphorylation patterns.

(A) AT1ABR lymphoblast cells deficient in WT ATM were complemented with inducible expression of WT, CL, 2RA, or 2RA + CL ATM alleles. Phosphopeptides were enriched and analyzed by mass spectrometry (MS). A histogram of total phosphopeptide counts (log2 values) in each cell line is shown after normalizing raw values for each cell line to the parental AT1ABR cell line. (B) Pairwise comparisons of the phosphoproteomic raw data from AT1ABR cells with or without expression of ATM. The raw intensity of each phosphopeptide was analyzed in pairwise comparisons of each cell line, and Spearman’s rank correlation coefficient, ρ, was calculated for each pairwise comparison. (C) Hierarchical clustering of all the phosphopeptides. Phosphopeptide levels were normalized to the parental AT1ABR cell line. Each vertical line within a cell line is a phosphopeptide and varies from green to black to red to represent decreased, equal, or increased levels of the phosphopeptides compared to the parental AT1ABR cell line, respectively. (D) Phosphopeptides from the phosphoproteome or the C2991 Dependent Cluster [indicated by the yellow box in (C)] were analyzed with motif-x. Sequences contained six residues N-terminal and C-terminal of each phosphorylation event. Fold enrichment is shown relative to the abundance of motifs in the proteome. (E and F) Empirical cumulative distribution functions of the ratio of phosphopeptide intensities of cells expressing WT ATM or cells expressing the CL allele, comparing the phosphoproteome and the predicted ATM phosphoproteome (E) or the predicted CK2 phosphoproteome (F). (G and H) Results from Kolmogorov-Smirnov tests showing the predicted number of phosphopeptides in the data set for each kinase and the P value of the observed changes in the CL- and 2RA + CL–expressing cell lines relative to cells expressing the WT allele. The red line marks P = 0.05.

Analysis of the total phosphoproteome data set by hierarchical clustering revealed a group of 314 peptides significantly less phosphorylated in the parental AT1ABR cells and cells expressing CL ATM (both CL ATM and 2RA + CL ATM), hereafter referred to as the C2991 Dependent Cluster (Fig. 5C; in the heat map, all values were normalized to the uncomplemented AT1ABR cell line phosphopeptide signal). Examination of the sequences surrounding these phosphorylation sites revealed that few of these phosphopeptides contained an S/TQ motif, the canonical sequence for ATM targets, whereas the remaining sites showed sequences distinct from this motif. Because ATM was not the predominant kinase responsible for phosphorylating the proteins in this group, we performed motif-x analysis (57, 58) to extract motifs from the phosphoproteome and the C2991 Dependent Cluster to identify which kinase(s) might be responsible for phosphorylating each phosphosite. Eight motifs were extracted from the C2991 Dependent Cluster—tP, sPXK, sP, sDXE, sEXE, RXXs, sD, and sE—and the results were compared to the motifs from the entire phosphoproteome. The four motifs in which acidic residues followed the phosphorylated residue, which is typical of substrates for the protein kinase CK2, were significantly enriched in the C2991 Dependent Cluster (Fig. 5D).

To further characterize the kinases that could be phosphorylating the phosphopeptides identified in the C2991 Dependent Cluster, the ratios of intensities of phosphopeptides from cells expressing wild-type or CL ATM in the phosphoproteome and each kinase-specific predicted phosphoproteome were analyzed using the two-sample Kolmogorov-Smirnov (K-S) test, a nonparametric test to determine whether two empirical cumulative distribution functions (ECDFs) are from the same or separate distributions. The K-S statistic is the largest difference between two ECDFs. If the K-S statistic is above a critical value, then the null hypothesis that the samples are drawn from the same distribution is rejected, suggesting that predicted substrates of the kinase are either over- or underrepresented in the C2991 Dependent Cluster. We used Group-based Prediction System (GPS) (59) to predict the substrates for each kinase based on known substrates. These putative substrates were entered as the phosphoproteomes of each kinase, the ratios of the phosphopeptides in wild-type versus CL ATM–expressing cells were calculated, and the ECDFs of the kinase-specific phosphoproteomes and the full phosphoproteome were compared using the K-S test. From this analysis, the phosphorylation of substrates by the kinase CaMKII, CK2, and ATM was predicted to depend on oxidation-induced ATM activation, given that the ECDFs of the phosphoproteome and the kinase-specific phosphoproteomes were significantly different (Fig. 5, E to H). Because the WT/CL ratio of these phosphopeptides is shifted to the right, these phosphopeptides had lower abundances in cells expressing CL ATM. In contrast, the ECDF of CK1 and several other cellular kinases tightly overlaps the ECDF of the full phosphoproteome, suggesting that there is no difference between the CK1 phosphoproteome and the entire phosphoproteome as measured by the K-S test (Fig. 5H and fig. S9).

Functional ATM is required to maintain cellular ROS amounts and CK2 activity

Because a reduction in CK2 activity in AT1ABR cells expressing CL ATM was suggested by the phosphoproteomic analysis, we investigated the causes of this phenomenon. First, we compared the abundances of CK2 subunits—CK2α, CK2α′, and CK2β—in AT1ABR cells expressing wild-type and CL ATM. Cells expressing CL ATM had similar amounts of all three subunits compared to cells expressing wild-type ATM, suggesting that the reduction in predicted phosphorylation targets of CK2 is not due to the loss of protein (fig. S10). We then examined whether there was a change in CK2 activity in the absence of ATM function by immunoprecipitating CK2 complexes expressed in human embryonic kidney (HEK)–293T cells in the presence or absence of 10 μM KU-55933, an ATM inhibitor. We tested the immunoprecipitated CK2 in an in vitro assay and observed a reduction in CK2 activity with the enzyme isolated from ATM inhibitor–treated cells compared to the untreated cells (Fig. 6A). However, there was little difference in the specific activity of the kinase with or without ATM inhibitor when the kinase activity was normalized to the level of CK2β-V5 protein immunoprecipitated. We found that overall immunoprecipitated CK2β levels were about twofold lower in HEK-293T cells in the presence of ATM inhibitor than in untreated cells, although there were similar amounts of CK2β-V5 in the input lysates of the untreated and KU-55933–treated cells (Fig. 6B). This suggests that ATM inhibition reduces the efficiency of CK2β immunoprecipitation and corresponding levels of CK2 activity in the in vitro assay.

Fig. 6 Loss of ATM oxidative stress activation causes aggregation of CK2β.

(A) In vitro kinase assay with immunoprecipitated CK2. HEK-293T cells stably expressing V5-tagged CK2β were treated with 10 μM ATM inhibitor (KU-55933) or an equivalent amount of dimethyl sulfoxide (DMSO) for 16 hours, and CK2β was immunoprecipitated with magnetic anti-V5 beads. CK2 was incubated with 1.67 mM ATP, 5 fCi of [γ-32P]ATP, and 1.15 μg of GST-CK2 substrate for 1 hour, and 32P-labeled substrate was analyzed by PhosphorImager. (B) Normalized amount of immunoprecipitated (IP) CK2β-V5 from (A) with the levels of CK2β-V5 normalized to levels in mock-treated cells. Quantitation was performed on the LI-COR system using Image Studio version 4.0. *P < 0.05. (C to E) Distribution profile of CK2 subunits CK2α (C), CK2α′ (D), and CK2β (E) in HEK-293T cells after sucrose gradient sedimentation. HEK-293T cells expressing CK2α-V5, CK2α′-V5, or CK2β-V5 were treated with 10 μM KU-55933 or an equivalent amount of DMSO for 16 hours, harvested, and lysed in the absence of detergent. One milligram of lysate in a total volume of 500 μl of lysis buffer was added to the top of a sucrose gradient made with 1-ml layers of 50% to 5% sucrose in 5% increments. After ultracentrifugation, 500-μl fractions were collected and analyzed by Western blotting. (F and G) Distribution profile of CK2α (F) and CK2β (G) from the AT1ABR cells expressing WT or CL ATM alleles, analyzed by sucrose gradient sedimentation. Lysates were analyzed as in (C) to (E) except with induced AT1ABR cells expressing WT or CL alleles of ATM as indicated. (H) Sucrose gradient sedimentation pattern of V5-tagged CK2β stably expressed in U2OS cells treated with 10 μM KU-55933 or an equivalent amount of DMSO for 16 hours or treated with a combination of KU-55933 and 1 mM N-acetylcysteine (NAC), as indicated. (I) Analysis of detergent-resistant aggregates in U2OS cells treated with 10 μM KU-55933 or DMSO or a combination of KU-55933 and 1 mM NAC, as indicated. Aggregate fractions were isolated and compared with total lysate using Western blotting for stably expressed V5-tagged CK2β. (J) Analysis of detergent-resistant aggregates in U2OS Flp-In T-REx cells expressing WT, CL, or 2RA ATM treated with 25 μM arsenite as shown in (I). (K) Aggregate fractions were isolated from U2OS cells expressing shCTRL or ATM shRNA and probed for CK2β, ATM, and β-actin as indicated from two independent cultures. (L) Means ± SD of CK2β abundance in aggregates, quantified from three independent experiments [including the replicates shown in (K)]. **P < 0.005.

CK2 is composed of two catalytic α subunits (CK2α or CK2α′) bound to two β subunits (CK2β) in a heterotetramer, with larger oligomeric forms also observed (60). The CK2 enzyme is known to aggregate in response to oxidative stress (6164), with the CK2α′ specifically shown to be prone to aggregation (65). On the basis of this evidence, we hypothesized that CK2 aggregation may be the cause of reduced activity and reduced efficiency of immunoprecipitation from ATM inhibitor–treated cells. To test this possibility, we treated cells expressing V5-tagged CK2α, CK2α′, or CK2β with KU-55933 or dimethyl sulfoxide (DMSO) as a control and separated protein in the lysates using sucrose gradient sedimentation, followed by Western blotting. CK2α-V5 showed a similar distribution pattern in ATM inhibitor–treated cells and untreated cells (Fig. 6C). However, a subset of CK2α′-V5 and CK2β-V5 protein showed a shift to the denser fractions, and ATM inhibitor–treated cells showed a twofold increase in the level of protein observed in these fractions (Fig. 6, D and E). To examine whether this ATM inhibitor–triggered aggregation is related to the reduction of endogenous CK2 activity in AT1ABR cells and to use an inhibitor-independent approach, we compared the distribution profile of CK2α or CK2β from the AT1ABR cells expressing wild-type or CL ATM after sucrose gradient analysis. Although CK2α showed similar distributions in both cell lines (Fig. 6F), CK2β again showed a large (~10-fold) increase in the aggregated fraction in the cell line expressing the CL allele compared to the wild-type–expressing line (Fig. 6G). This is similar to the aggregation previously reported for CK2 and may provide an explanation for the reduction in overall CK2 activity in AT1ABR cells expressing the CL ATM allele.

Because the CL ATM mutant is impaired in oxidative activation by H2O2 (3) and human cells expressing CL ATM exhibited higher amounts of ROS (Fig. 4A and fig. S1), it is possible that the increase in ROS caused by ATM inhibition may cause CK2β aggregation. To test this hypothesis, we compared distribution profiles of CK2β expressed in U2OS cells treated with ATM inhibitor by itself or in the presence of the reducing agent N-acetylcysteine (NAC). As expected, CK2β showed a marked shift to the denser fraction, with 20% of the protein appearing in the aggregated form from cells treated with ATM inhibitor. However, in the presence of both ATM inhibitor and 1 mM NAC, the levels of CK2β aggregate were significantly reduced and similar to the levels in untreated cells (Fig. 6H).

To confirm this, we isolated aggregated proteins from cellular lysates by sequential detergent extractions and sedimentation analysis (66). The level of CK2β-V5 in the insoluble fraction containing protein aggregates was increased fourfold in the ATM inhibitor–treated cells in comparison to the untreated cells and was markedly reduced in cells treated with both ATM inhibitor and NAC (1 mM), although CK2β levels were identical in the total lysates (Fig. 6I). Similarly, expression of the CL mutant allele in U2OS cells also promoted aggregation of CK2β as measured by the increased presence of CK2β in the aggregate fraction (Fig. 6J). These results suggest that functional ATM is required for maintaining protein homeostasis and that the loss of this activity of ATM alters the solubility of the CK2 kinase and thus the subset of the phosphoproteome dependent on CK2 (Fig. 5).

Widespread protein aggregation occurs in cells lacking oxidative activation of ATM

Our observation of CK2β aggregation in cells expressing the CL allele of ATM suggested that other polypeptides may also be prone to aggregation in these cells. To identify these, we isolated the detergent-insoluble fraction as described above in U2OS cells expressing the wild-type, CL, or 2RA alleles of ATM and analyzed the proteins by mass spectrometry (MS). We also did this analysis with concurrent arsenite treatment to model conditions of oxidative stress. The results show that a small number of proteins are enriched in both CL- and 2RA-expressing cells compared to wild-type (about 40 per cell line; see also table S2) but that a large number of proteins (about 500) appear in the aggregate fraction of CL-expressing cells exposed to arsenite (Fig. 7A). The fold enrichment of these proteins relative to their levels in the cells expressing the wild-type allele is over 100-fold for some polypeptides (>1.5-fold shown in Fig. 7B). The levels of CL/WT enrichment for CK2β and CAMKIIδ kinases in the aggregate fraction are 9- and 150-fold, respectively (table S2). Gene ontology analysis of the proteins that appear in the aggregate fraction of the cells expressing the CL allele show a strong enrichment for nuclear proteins including the Replication Factor C (RFC) complex, DNA repair proteins (including Mre11 and Rad50), and RNA processing enzymes (Fig. 7C). The presence of Rad50 in the aggregate fraction was validated by Western blotting (fig. S11). The aggregated proteins isolated from the cells expressing the CL allele of ATM were also predicted to have a higher risk of aggregation, based on the TANGO and WALTZ algorithms that estimate aggregation propensity and amyloid-like structures (67, 68), and were significantly longer than average polypeptides in human cells (Fig. 7D). Overall, this analysis shows that expression of mutant ATM that cannot be activated by oxidative stress results in a global increase in protein aggregation, particularly in cells that are also exposed to low levels of ROS.

Fig. 7 Loss of ATM oxidative stress activation causes global protein aggregation.

(A) U2OS cells depleted for endogenous ATM and inducibly expressing WT, CL, or 2RA alleles with or without concurrent arsenite treatment (25 μM) were lysed, and detergent-resistant aggregate fractions were isolated and analyzed by MS using label-free quantification. Three biological replicates were analyzed for each cell line, and proteins enriched by ≥1.5-fold in cells expressing the CL or 2RA alleles compared to the level in cells expressing WT ATM are shown (only those with P < 0.05 by a t test). (B) Fold enrichment over WT for proteins identified in (A) for each target. (C) Gene ontology analysis of proteins enriched in aggregate fraction of cells expressing the CL allele with arsenite treatment (Panther Gene Ontology database). The analysis included Bonferroni correction for multiple testing; all results shown had P < 0.05. (D) Comparison of TANGO and WALTZ scores for 497 polypeptides isolated from U2OS cells expressing the CL allele with arsenite treatment in (A) compared to the entire proteome. Polypeptide length is shown in amino acids. n = 3 biological replicates per sample; P values by two-tailed t test. CI, confidence interval.


The ATM protein kinase responds to both DNA DSBs and oxidative stress. Here, we identified two adjacent arginine residues, Arg2579/2580, at the end of the FAT domain in ATM that are essential for ATM activation by MRN and DNA in vitro. Mutating these resides (creating a 2RA mutant) did not affect H2O2-dependent ATM activation, indicating that 2RA ATM is only blocked in MRN-dependent ATM activation, likely due to the deficiency in binding to MR that we have observed here. However, it is currently unknown whether the Arg2579/2580 region of ATM directly interacts with MR. Heat repeats 17/18 and 21/22 in Schizosaccharomyces pombe Tel1 (the ATM homolog), which are N-terminal to the location of Arg2579/2580 in human ATM, were identified as Nbs1-binding regions previously (69), but a detailed understanding of this interaction will likely require structural analysis.

Most of our functional characterization of the 2RA ATM allele was performed with a Tet-inducible system in human U2OS cells combined with shRNA-mediated depletion of the endogenous protein. This conditional system is preferable to analysis of A-T patient cells expressing mutant alleles because ATM-deficient cells are extremely sensitive to transfection and because ATM deficiency induces adaptive compensatory responses over time that are not fully understood and can confound analysis of mutant alleles (2). Using this inducible system, we found that survival of CPT-induced DNA damage, checkpoint activation, and resection of DNA ends were specifically impaired with expression of the 2RA ATM allele, suggesting that the activation of ATM via the MRN/DSB pathway is specifically responsible for these DNA damage–related outcomes.

In contrast to cells expressing the 2RA allele, cells expressing the CL allele of ATM (2, 3) fail to phosphorylate downstream substrates in response to low levels of peroxide or arsenite stress, a defect most apparent with phosphorylation of Chk2. Similarly, these cells failed to activate the intra-S phase checkpoint in response to arsenite exposure and showed poor survival in clonogenic assays after this treatment. Our results with arsenite in this work and our previous observations with hydrogen peroxide indicate that the oxidative pathway of ATM activation is specifically required for ATM responses to ROS in human cells.

With expression of the CL of ATM, we observe higher overall ROS levels compared to cells expressing the wild-type allele, whereas levels of superoxide are lower with expression of the mutant. This is generally consistent with our previous observations and a number of other groups indicating high levels of ROS when ATM activity is absent (2, 3). Our previous results with hydrogen peroxide in cells and in vitro are consistent with a higher level of this compound in ATM-deficient cells, which is also supported by a report that overexpression of mitochondria-targeted catalase (which removes hydrogen peroxide) reduces or even rescues hematopoiesis deficiencies, thymic lymphomas, and immune system function in mice lacking ATM (70). Catalase activity is lower in ATM-deficient human cells compared to normal controls (71) and in ATM−/− mice where the deficit was reported specifically in the cerebellum (72), an important detail given the Purkinje cell specificity of the A-T disorder in humans. The ATM protein has also been reported to be associated with peroxisomes (52, 73), the location of catalase activity in human cells, which is perhaps relevant to the overall deficit in catalase function in cells lacking ATM.

Mitochondrial dysfunction has previously been observed in ATM-deficient mouse thymocytes and human cells; specifically, an increase in total mitochondrial mass, a decrease in mitochondrial complex I activity, a loss of mitochondrial membrane potential, and a decrease in mitophagy were measured in the absence of ATM function (21, 39, 40). Our experiments with the CL ATM mutant indicate that a loss of oxidative activation of ATM does lead to lower levels of mitochondria with normal membrane potential and a deficiency in mitophagy. Last, levels of carnitine derivatives are significantly higher in cells expressing the ATM CL allele, indicating aberrant regulation or flux in fatty acid oxidation pathways, perhaps as a compensatory mechanism in response to oxidative stress. Carnitine derivatives are important for transport of long-chain fatty acids into the mitochondria and have been observed to act as neuroprotective agents (74). Treatment of lymphoblasts from A-T patients with l-carnitine reduced levels of chromosomal abnormalities after exposure to oxidative DNA damage (75). Together, it is clear that there are changes in mitochondrial function that occur with ATM inhibition or deletion and that the CL allele that lacks oxidative stress activation promotes these changes.

Consistent with the mitophagy observation, we also observed that cells expressing the ATM CL allele are specifically deficient in arsenite-induced macroautophagy, a process that recycles proteins and organelles. ATM has previously been shown to regulate autophagy through a pathway that is induced by ROS and controls mTOR activity through LKB1, AMPK, and TSC2 (51). We found that cells expressing the 2RA ATM allele are deficient in CPT-induced macroautophagy, reminiscent of the role for ATM in Adriamycin induction of autophagy reported recently (54), and also exhibit a deficit in response to arsenite. It is not clear why the 2RA-expressing cells show this sensitivity, whereas most of the other readouts of ATM function in oxidative stress are normal in these cells, although it is clear that the 2RA cells also show a modest increase in protein aggregation in response to arsenite.

Our analysis of global phosphorylation patterns in cells expressing the ATM separation-of-function alleles in the absence of exogenous stress pointed toward effects of ATM on other kinases: CAMKII and CK2. Here, we focused on CK2, an enzyme that is often considered to be constitutively active, although cellular localization and protein-protein interactions appear to be critical for modulating its effects and substrates (76). Importantly for this study, CK2 is known to aggregate in vivo and in vitro, and this property has been shown to negatively affect its kinase activity (6064, 77, 78). Our results indicate aggregation of a subset of CK2 in cells expressing the CL allele of ATM that is specific to the CK2α′ subunit and CK2β. This aggregation is relieved by treatment with NAC; thus, it appears to be a result of increased ROS, although we cannot rule out the possibility that downstream effects of ATM deficiency may be alleviated by antioxidants indirectly. This collaborative relationship between ATM function and CK2 activity is consistent with previous observations that a significant fraction of DNA damage–induced phosphorylation events are predicted to be catalyzed by CK2 and that these are reduced upon inhibition of ATM (79). Because CK2 is an important enzyme in the DNA damage response, phosphorylating MDC1 and promoting Nbs1 association with DSBs, as well as many other critical interactions between signaling and repair proteins (80), the partial loss of CK2 function in the absence of ATM oxidative activation is likely to also have effects on the DNA damage response.

Our analyses of protein aggregates formed with expression of the CL allele of ATM during arsenite exposure identified about 500 polypeptides that form detergent-resistant precipitates more efficiently than in the cell line expressing wild-type ATM. These proteins only appear in the detergent-resistant aggregate fraction with low-level arsenite treatment and expression of the CL allele, suggesting that the combined effects of these treatments result in large-scale protein aggregation. It is interesting that the most overrepresented groups of polypeptides among these are those that function in the DNA metabolism and gene expression. ATM has been implicated in the control of RNA splicing (81), and many groups have noted the relationships between efficient RNA processing and genomic stability (82). It is not yet clear why these specific proteins show aggregation with the loss of ATM function, but the predicted aggregation propensity of the proteins we identified is significantly higher than the total proteome. The aggregation of proteins, particularly in neurons, is very likely relevant to the loss of Purkinje cell function that is the hallmark of the A-T disorder because protein aggregation is a common link between many, if not all, neurodegenerative diseases (83).

Loss of normal proteostasis in A-T patients, particularly if it is centered on proteins relevant to DNA metabolism, may play a critical role in the pathogenesis of A-T. Further investigation will be required to determine how this phenomenon relates to the A-T phenotype and also how this relates to ATLD patients who have deficiencies in Mre11 or Rad50 (8487); these patients are hypomorphic for MRN function yet also experience progressive neurodegeneration. Last, it is important to investigate the basis of the extreme dominant-negative effect of kinase-deficient ATM on cells. This is almost certainly related to the embryonic lethal phenotype of this allele in the mouse (33, 34), but it is currently unknown whether this is caused by stable binding of the mutant ATM to certain substrates and, if so, which substrates are the cause of the cellular toxicity during stress responses and during development.


Cell culture and ATM induction

U2OS Flp-In T-REx cells containing control vector, wild-type, 2RA, CL, or DA ATM alleles were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen) containing blasticidin (15 μg/ml; A1113903, Life Technologies), penicillin-streptomycin (100 U/ml; 15140-122, Life Technologies), and hygromycin (200 μg/ml; 400052-50ml, Life Technologies). Depletion of endogenous ATM was performed by incubating cells with lentivirus containing shRNA toward ATM (sc-29761-SH, Santa Cruz Biotechnology) overnight and selecting with media containing puromycin (1 μg/ml; Invitrogen) for 5 to 7 days. To induce wild-type ATM or mutant ATM, doxycycline (1 μg/ml or as indicated) was added to the medium as final concentration 3 days before treatment with DNA damage or ROS-inducing agent. For experiments with H2O2 (H325-100, Thermo Fisher Scientific) or arsenite (S7400-100G, Sigma), the medium was changed to serum-free medium when cells were treated and changed back to medium containing serum and appropriate antibiotics after treatment.

HEK-293T cells (CRL-11268, American Type Culture Collection) were grown in DMEM supplemented with 10% FBS. AT1ABR cells and AT1ABR cells expressing the various pMAT1-ATM constructs were a gift from M. Lavin. AT1ABR cells were grown in RPMI 1640 medium (R8758, Sigma) supplemented with 15% FBS. The medium for AT1ABR cells with pMAT1-ATM constructs was supplemented with hygromycin B (200 μg/ml; 400052, EMD Millipore) to select for pMAT1-ATM retention. ATM expression was induced from the metallothionein II promoter for 16 hours using 2 μM CdCl2, as described previously (42).

Virus production and transduction

HEK-293T cells were plated in 10-cm dishes and allowed to grow to near confluence. A solution of Opti-MEM and plasmids was made using 20 μg of pLX304 vector containing the gene of interest, 12 μg of pCMV-dR8.91 (Delta 8.9), and 8 μg of VSV-G and brought to 500 μl with Opti-MEM. In another tube, 60 μl of Lipofectamine 2000 (11668-027, Invitrogen) was mixed with 440 μl of Opti-MEM. The plasmid and Lipofectamine 2000 solutions were combined, incubated for 5 min at room temperature, and then added to the HEK-293T cells. The medium was changed the next day and left for 48 hours. The medium was harvested after 48 hours and replaced and then harvested again 24 hours later. The medium with virus was combined, filtered with 0.45-μm filters, aliquoted in 500-μl aliquots, and stored at −20°C for transduction. HEK-293T or U2OS cells were plated in 24-well plates and allowed to reach confluency. The medium was removed and replaced with 500 μl of the viral aliquots. Polybrene (TR-1003-G, EMD Millipore) was added at a final concentration of 10 μg/ml to increase viral transduction. The cells were grown overnight, and the medium was replaced the next day. The following day, selection agent was added—for pLX304 plasmids, cells were treated with blasticidin (10 μg/ml; ant-bl-1, InvivoGen)—and cells were allowed to grow for about 1 week until the control untransduced cells had died. The stable cell lines were checked for expression of V5-tagged protein and used in subsequent assays.

V5 immunoprecipitation and Western blotting

HEK-293T cells stably expressing V5-tagged CK2 subunits were lysed with 1× cell lysis buffer (9803, Cell Signaling Technology) according to the manufacturer’s instructions. Protein quantitation was performed by Bradford assay with Coomassie Plus Protein Assay Reagent (23236, Pierce). For each immunoprecipitation, 2.5 μg of the lysate was combined with 4 μl of mouse anti-V5 magnetic beads (M167-9, Medical and Biological Laboratories) and brought to a final volume of 50 μl with 1× cell lysis buffer. Tubes were rotated for 30 min at 4°C and then spun briefly. A magnetic stand was used to wash the beads three times with 500 μl of 1× cell lysis buffer. Beads were immediately used for the CK2 kinase reaction. The following antibodies were used for Western blotting: mouse anti-V5 (R960-25, Invitrogen), rabbit anti-CK2α (PA1-86381, Pierce), mouse anti-CK2β (218712, EMD Millipore), goat anti-mouse immunoglobulin G (IgG) IRDye 800 Conjugated (RL-610-132-121, Rockland Immunochemicals), and goat anti-rabbit IgG IRDye 700 Conjugated (RL605-430-003, Rockland Immunochemicals).

AT1ABR tissue culture for staining and qPCR

AT1ABR cells expressing wild-type ATM or CL ATM were seeded at a density of 500,000 cells/ml in 15 ml. The following day, four aliquots of 1.5 ml of each cell line were harvested for the uninduced controls. The aliquots were pelleted for 5 min at 100g. The supernatant was removed, and the cell pellets were washed with 1 ml of phosphate-buffered saline (PBS) and pelleted again. The pellets were immediately stained or frozen in liquid nitrogen for subsequent RNA or protein extraction. The remaining cells were induced for 16 hours with 2 μM CdCl2. After induction, the cells were gently spun down for 5 min at 100g and resuspended in fresh medium without CdCl2. The cells were allowed to recover for 48 hours, and four aliquots of 1.5 ml of each cell line were harvested and immediately stained or pelleted and frozen in liquid nitrogen for subsequent protein extraction.

Analysis of carnitine derivatives in AT1ABR cells

AT1ABR cells expressing wild-type ATM or CL ATM were grown and induced for ATM expression, as described above. Six biological replicates with 5 × 106 cells per sample were analyzed. Metabolic profiling was performed by Metabolon Inc., and identification of compounds was done by comparison to known purified standards.

Sucrose gradient sedimentation

Cell lysates were prepared in buffer containing 25 mM tris (pH 8), 100 mM NaCl, and 10% glycerol supplemented immediately before use with 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM sodium orthovanadate, 1 mM β-glycerophosphate, and 2.5 mM sodium pyrophosphate. Lysis was performed by douncing 50 times on ice. For both methods, the protein concentration was quantitated by Bradford assay with Coomassie Plus Protein Assay Reagent (23236, Pierce). The different concentrations of sucrose were made in 20 mM tris (pH 7.4) and 150 mM NaCl and layered in Thinwall Ultra-Clear Tubes (344059, Beckman Coulter) using gravity flow from a pipette bulb. The sucrose percentages from the bottom were 50, 45, 40, 35, 30, 25, 20, 15, 10, and 5% (w/v), and each layer was 1 ml. Between 10 μg and 2 mg of the lysate in 500 μl of the appropriate cell lysis buffer was added to the top of the sucrose gradient and then spun for 16 hours at 4°C and 180,000g in a swinging bucket SW 41 Ti rotor (331362, Beckman Coulter). Fractions of 500 μl were collected using a pipette. Western blots were performed with 16 or 120 μl of each fraction combined with 4 or 30 μl 5× SDS, and loading controls of 2.5 or 100 μg of lysates were run on each blot. The blots were probed as before, and the percentage of probed protein was quantitated using the LI-COR Odyssey system.

An identical sucrose gradient was performed with molecular weight markers to determine the fractions corresponding to different size complexes. The fractions were run on 4 to 12% NuPAGE bis-tris gel (NP0336BOX, Invitrogen) and the colloidal Coomassie-stained gels. Aldolase (40-kDa monomer) appeared mostly in fractions 6 and 7. Catalase (60-kDa monomer and 240-kDa tetramer) ran in fractions 8 to 10. Ferritin (21-kDa monomer and 500-kDa 24-mer) and thyroglobulin (330-kDa monomer and 660-kDa dimer) appeared in the sucrose gradient starting in fraction 12. However, some amount of ferritin and thyroglobulin are present through the last fraction, suggesting that there are larger complexes or aggregates in the sucrose gradient.

Expression constructs and protein expression

Wild-type MRN complex was expressed in Sf21 insect cells by coexpression with baculovirus for wild-type Rad50, wild-type Mre11, and wild-type Nbs1, as described previously (88). To make biotinylated MRN or MR, Nbs1 or Mre11 was modified at its C terminus with a biotinylation epitope for the BirA enzyme and coexpressed with other component(s) as described previously (89) and with baculovirus expressing BirA. Expression constructs for Flag- and HA-tagged wild-type ATM were gifts from M. Kastan and R. Abraham, respectively. 2RA ATM and CL ATM were generated using QuikChange XL site-directed mutagenesis (Stratagene) from wild-type ATM pcDNA3 expression plasmid (sequences available upon request). Cesium-purified plasmid DNA was used to transfect HEK-293T cells and express recombinant ATM as described previously (24). The Escherichia coli expression constructs for GST-p53 were described previously (88, 90). To generate U2OS Flp-In T-REx cells expressing wild-type, 2RA, CL, or DA ATM, ATM gene was digested from ATM pcDNA3 expression plasmid with Not I/Apa I and inserted into pcDNA5-FRT/TO-intron vector, which was a gift from B. Xhemalce, and cotransfected with pOG44 Flp recombinase expression vector into U2OS Flp-In T-REx cells (a gift from J. Parvin). Transfected U2OS cells containing inducible ATM genes were selected with hygromycin (200 μg/ml) in DMEM [10% Tet System Approved FBS (catalog no. 631106, Clontech), blasticidin (15 μg/ml), and penicillin-streptomycin (100 U/ml; 15140-122, Life Technologies)].

ATM expression and purification

Flag- and HA-tagged wild-type ATM were gifts from M. Kastan and R. Abraham, respectively. 2RA ATM and CL ATM were generated using QuikChange XL site-directed mutagenesis kit (Stratagene) from wild-type ATM pcDNA3 expression plasmid. Plasmid DNA prepared with QIAFilter plasmid Maxi kit (Qiagen) was used to transfect HEK-293T cells and express recombinant ATM. To generate ATM heterodimer, plasmid containing Flag-ATM (2RA) was cotransfected with plasmid containing HA–wild-type ATM or HA-ATM (CL) in HEK-293T cells and purified through Flag and HA columns, sequentially as described previously (24).

In vitro kinase assay

ATM kinase assays were performed in kinase buffer [50 mM Hepes (pH 7.5), 50 mM potassium chloride, 5 mM magnesium chloride, 10% glycerol, 1 mM adenosine 5′-triphosphate (ATP), and 1 mM DTT] for 90 min at 30°C in a volume of 40 μl as previously described (24). To test for MRN- and DNA-dependent ATM activation, 9.6 nM MRN and 10 ng of linear DNA (~140 nM) were used in kinase assays with 6.25 nM GST-p53. For H2O2-dependent ATM activation, 12.5 nM GST-p53 was incubated with various concentrations of H2O2 as indicated without additional DTT. Phosphorylated p53 (Ser15) was detected as described previously (90) using a phospho-specific antibody from Calbiochem (PC461).

In vitro binding assay

Biotinylated proteins (20 nM MRN or MR) were incubated with 50 nM wild-type or 2RA ATM protein in buffer A [100 mM NaCl, 25 mM tris (pH 8), 1 mM DTT, and 10% glycerol] for 30 min at 30°C in a final volume of 100 μl and then incubated with streptavidin-coated magnetic beads (Dynal) and 0.1% CHAPS (Sigma), rotating at 4°C for 15 min. Beads with associated proteins were washed three times with buffer A containing 0.1% CHAPS, and bound proteins were eluted by boiling the beads in SDS loading buffer and analyzed by SDS–polyacrylamide gel electrophoresis (PAGE) and Western blotting using antibodies directed against ATM (sc-135663, Santa Cruz Biotechnology), Rad50 (GTX70228, GeneTex), and Nbs1 (MSNBS10PX1, GeneTex).

ATM-dependent phosphorylation in cells

For the experiment with IR, cells were irradiated with 10 Gy and incubated for 1 hour before harvesting. For the experiments with CPT, H2O2, and arsenite, cells were incubated with medium containing 10 μM CPT, 100 μM H2O2, or arsenite for 1 hour before harvesting. Cells were lysed in 10× cell lysis buffer (9803, Cell Signaling Technology), and the lysate (20 μg) was separated by 8% SDS-PAGE gel and analyzed by Western blotting. Proteins were transferred to PVDF-FL membrane (Millipore) and probed with antibodies directed against ATM (sc-135663, Santa Cruz Biotechnology), phospho-ATM Ser1981 (AF-1655, R&D Systems), Kap1 (ab22553, Abcam), phospho-Kap1 Ser824 (A300-767A, Bethyl Laboratories), Chk2 (GTX70295, GeneTex), and phospho-Chk2 Thr68 (2661S, Cell Signaling Technology) followed by detection with IRDye 800 anti-mouse (RL-610-132-121, Rockland Immunochemicals) or Alexa Fluor 680 anti-rabbit (A21076, Invitrogen) secondary antibodies. Western blots were analyzed and quantitated using a LI-COR Odyssey system.

Clonogenic cell survival assay

U2OS cells were treated with various doses of IR, CPT, or arsenite as indicated. CPT and arsenite were added to the medium for 1 hour and removed by washing cells with PBS followed by changing to complete medium without drugs. Cells were incubated for 10 to 14 days as described (91). Colonies were stained with staining solution (0.5% crystal violet and 20% ethanol) for 30 min followed by washing with distilled water and were counted with ImageJ. The percentage of cell viability was calculated with untreated group, and error bars show SD from three independent experiments.

Quantitation of DNA end resection

U2OS Flp-In T-REx cells with wild-type or mutant ATM alleles were infected with ER–AsiSI retrovirus and selected with puromycin (1 μg/ml) for 2 weeks. Cells were further infected with lentivirus expressing control shRNA or shRNA against ATM, and expression of exogenous ATM alleles was induced with doxycycline treatment (1 μg/ml). DNA end resection in cells was measured using a method developed previously (36). Briefly, after 4-hour treatment with 600 nM 4-hydroxytamoxifen, cells were trypsinized and resuspended in 0.6% low–gelling temperature agarose at a final concentration of 2 × 106 cells/ml. The agar balls with cells were used for genomic DNA extraction, and DNA end resection at selected DSB sites was quantitated by qPCR as detailed previously (92).

Cell cycle checkpoint assay

U2OS Flp-In T-REx cells after knocking down endogenous ATM and induction of wild-type or mutant ATM were synchronized at G1 with aphidicolin (2 μg/ml; 14007, Cayman Chemical) for 17 hours and treated with 1 μM CPT in serum-containing medium or 100 μM arsenite in serum-free medium for 1 hour. Cells were washed with PBS and incubated in medium containing nocodazole (400 ng/ml) for 17 hours followed by harvesting with low spin (100g). For PI staining, cell pellets were washed two times with PBS and resuspended with cold PBS followed by mixing with 100% ethanol to generate a 70% final concentration and were stored at 4°C overnight. Cells were washed two times with PBS and incubated with PI staining solution [3.8 mM sodium citrate, PI (40 μg/ml), and ribonuclease A (0.5 μg/ml)] overnight followed by analysis of the percentage of G2-M cells by flow cytometry.

H2DCFDA, DHE, MitoTracker Green, and MitoTracker Red staining

For staining, all centrifugation occurred for 5 min at 100g unless otherwise stated. AT1ABR pellets were resuspended in PBS and stained with H2DCFDA (D399, Thermo Fisher Scientific), DHE (D1168, Thermo Fisher Scientific), MitoTracker Green FM (M7514, Thermo Fisher Scientific), or MitoTracker Red CMXRos (M7512, Thermo Fisher Scientific) as follows. For analysis of ROS levels, washed cells were incubated for 30 min with 1 μM H2DCFDA and then centrifuged, and the supernatant was removed. The pellets were washed three times with PBS and then transferred to 5 ml of polystyrene round bottom tubes (60818-496, VWR) in 0.3 to 1 ml of PBS for analysis by flow cytometry. For analysis of superoxide levels, washed cells were incubated for 30 min with 5 μM DHE and then centrifuged, and the supernatant was removed. The pellets were washed three times with PBS and then transferred to 5 ml of polystyrene round bottom tubes (60818-496, VWR) in 0.3 to 1 ml of PBS for analysis by flow cytometry. For mitochondrial mass staining, washed cells were incubated with 0.2 μM MitoTracker Green FM for 20 min and then centrifuged, and the supernatant was removed. The pellets were washed three times with PBS and then transferred to 5 ml of polystyrene round bottom tubes in 0.3 to 1 ml of PBS for analysis by flow cytometry. For mitochondrial staining based on membrane potential, washed cells were incubated with 0.2 μM MitoTracker Red CMXRos for 20 min and then centrifuged, and the supernatant was removed. The pellets were washed three times with PBS and then transferred to 5 ml of polystyrene round bottom tubes in 0.3 to 1 ml of PBS for analysis by flow cytometry.

Flow cytometry

Stained samples were analyzed on a BD LSRFortessa cell analyzer using the fluorescein isothiocyanate (FITC) setting (excitation, 488 nm; emission, 515 to 545 nm) for samples stained with H2DCFDA, the phycoerythrin (PE) setting (excitation, 561 nm; emission, 578 to 590 nm) for samples stained with DHE, and the PE–Texas Red setting (excitation, 488 nm; emission, 600 to 620 nm) for samples stained with MitoTracker Red CMXRos. Live cells were gated according to the forward scatter and side scatter before analysis of fluorescence. For each sample, 10,000 cells were analyzed. The median values from four to six biological replicates were used to calculate the means and SDs.

mKeima mitophagy assay

pCHAC-mt-mKeima was a gift from R. Youle (Addgene plasmid #72342) (47). The plasmid was transfected with retroviral helper plasmids in HEK-293T cells to make recombinant virus. U2OS Flp-In T-REx cells inducibly expressing wild-type, CL, or 2RA ATM were seeded into six-well plate with DMEM containing doxycycline (1 μg/ml) 1 day before infection. After infection with mKeima retrovirus overnight, the medium was changed and cells were incubated two more days. Cells were harvested with trypsin, and cell pellets were washed once with PBS. For analysis by flow cytometry, cells were resuspended with 1 ml of PBS. To measure mKeima, 488- and 561-nm lasers with 610/20-nm emission filters were used. For each sample, 20,000 events were collected for each of four biological replicates and analyzed with BD FACSDiva software.

Acridine orange staining of acidic lysosomal vesicles

U2OS cell lines expressing inducible ATM alleles were stained for acidic vesicular organelles using acridine orange as previously described (54). Cells incubated with doxycycline (1 μg/ml) were added to the medium as final concentration 3 days before treatment with 5 μM CPT or 100 μM arsenite for 2 hours, and cells were washed with PBS and incubated with fresh medium without CPT or arsenite for 48 hours for fluorescence-activated cell sorting (FACS) or 72 hours for confocal microscopy. For FACS analysis, cells were trypsinized, washed two times with PBS, and stained with acridine orange (1 μg/ml) in PBS for 15 min, followed by washing two times with PBS. Cells were fixed with 4% paraformaldehyde for 5 min and washed once with PBS. For confocal microscopy, cells, which were seeded on a glass bottom dish (HBST-3522, WillCo Wells) before adding doxycycline, were washed two times with PBS and stained with acridine orange (1 μg/ml) in the medium for 20 min, followed by replacing with PBS for microscopy. FACS analysis of the acridine orange signal was performed as previously described (54), and the ratio of cells with high 585-nm signal in treated versus untreated cells was calculated for each of three biological replicates.

In vitro CK2 kinase reactions

CK2 kinase assays were performed with either 1 μl of CK2 (P6010L, NEB) or immunoprecipitated CK2 on beads. The reactions were performed in 40 μl total with 1× CK2 Reaction Buffer (NEB) [20 mM tris-HCl (pH 7.5), 50 mM KCl, 10 mM MgCl2] supplemented with 1 mM ATP, 1 μl of [γ-32P]ATP (BLU002H, NEN Radiochemicals), and 1.15 μg of GST-CK2 substrate. For the time course, a master mix of immunoprecipitated CK2β-V5 in the kinase reaction was made and split into five tubes. The tubes were incubated at 30°C for various times and stopped by the addition of 6 μl of 5× SDS loading buffer. Quantitation was performed using GelQuant and normalized to protein levels quantified through Western blotting. Proteins were run on SDS-PAGE gels and transferred to polyvinylidene difluoride (PVDF) membrane following standard protocols. V5-tagged CK2 levels were probed using rabbit anti-V5 (NB600-381, Novus Biologicals) and Alexa Fluor 680 goat anti-rabbit (A21076, Invitrogen) and scanned and quantitated on the LI-COR Odyssey system. Then, the membrane was exposed to a PhosphorImager screen for at least 16 hours. The screen was scanned using Typhoon Imager to analyze 32P incorporation.

Protein aggregate isolation and analysis

Cell pellets from 150-mm dishes were frozen in liquid N2. To prepare cell lysates, the pellets were resuspended in lysis buffer [20 mM Na-phosphate (pH 6.8), 10 mM DTT, 1 mM EDTA, 0.1% Tween 20, 1 mM PMSF, and EDTA-free protease inhibitor mini tablets (Pierce)] and rotated at 4°C for 30 min. Cells were lysed in a 4°C water bath–based sonicator (Bioruptor; eight times at level 4.5 and 50% duty cycle) and centrifuged for 20 min at 200g at 4°C. Supernatants were adjusted to a similar concentration, and protein aggregates were pelleted at 16,000g for 20 min at 4°C. After removing supernatants, protein aggregates were washed twice with 2% NP-40 [20 mM Na-phosphate (pH 6.8), 1 mM PMSF, and EDTA-free protease inhibitor mini tablets (Pierce)], sonicated (six times at level 4.5 and 50% duty cycle), and centrifuged at 16,000g for 20 min at 4°C. Aggregated proteins were washed in buffer without NP-40 (sonication; four times at level 3 and 50% duty cycle), boiled in 2.5× SDS sample buffer, ran in 4 to 12% bis-tris protein gels (NuPAGE Novex), and analyzed by Western blotting.

Aggregate samples were prepared in triplicate, and each sample was resuspended in 15 μl of 10% SDS with 50 mM β-mercaptoethanol and boiled at 100°C for 5 min. The samples were diluted with 200 μl of UA buffer [8 M urea and 0.1 M tris (pH 8.8)]. Microcon-30 centrifugal filter units (MRCF0R030, Millipore) were equilibrated with 20% acetonitrile (ACN) and 2% formic acid solution (14,000g for 10 min) before use. Diluted samples were loaded on the filters and then washed three times with 400 μl of UA buffer [8 M urea and 0.1 M tris (pH 8.8)]. After washing, samples were reduced with 400 μl of 50 mM DTT in UA buffer by addition of buffer to filter, incubation for 5 min at room temperature, and centrifugation. Samples were alkylated with 400 μl of 50 mM iodoacetamide in UA buffer with incubation for 5 min at room temperature and centrifugation. Samples were desalted three times with 400 μl of 40 mM ammonium bicarbonate (ABC). One hundred microliters of 40 mM ABC with 0.5 μl of trypsin gold (V528A, Promega) in PBS was added to the sample and incubated overnight (37°C). The next morning, peptides were eluted by centrifugation; filters were washed with 20% ACN and 2% formic acid solution, and filtrate was combined with eluted peptides in ABC buffer. Collected samples were lyophilized at room temperature. Dried samples were resuspended in 10 μl of 0.1% formic acid with 0.1% trifluoroacetic acid (TFA) and then desalted with C18 tips (QK224796, Pierce) according to the manufacturer’s protocol. Finally, samples were resuspended in 80% ACN and 2% formic acid for liquid chromatography (LC)–MS analysis. All centrifugations were done at 14,000g for 20 min at room temperature unless otherwise noted. Protein identification by LC-MS/MS was provided by the University of Texas at Austin Proteomics Facility on an Orbitrap Fusion following previously published procedures (93). Raw files were analyzed using label-free quantification with Proteome Discoverer 2.15. Proteins with at least two peptides identified by MS/MS were used for quantification, and results show only proteins with P ≤ 0.05 using Student’s t test with a two-tailed distribution.

Cell lysis for MS and phosphopeptide analysis

For cell lysis and protein quantitation, AT1ABR cell pellets were lysed in four volumes of cell lysis buffer [8 M urea, 50 mM tris (pH 8), 5 mM CaCl2, 30 mM NaCl, 50 mM NaF, 1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 1× mini EDTA-free protease inhibitor (Roche Diagnostics), and 1× PhosSTOP phosphatase inhibitor (Roche Diagnostics)]. The lysate was sonicated until clear and centrifuged for 10 min at 20,000g. The supernatant was collected, and the protein concentration was quantitated by bicinchoninic acid (BCA) assay (23227, Pierce).

Protein digestion

Equal amounts of lysate from each cell line (1 mg) were reduced with 5 mM DTT for 30 min at 37°C. After reduction, the cysteine residues were alkylated by the addition of 15 mM iodoacetic acid for 30 min at room temperature protected from light. The lysates were then reduced again with 5 mM DTT for 15 min at room temperature. The protein sample was diluted with 50 mM tris (pH 8) and 5 mM CaCl2 to a final concentration of 1.5 M urea. To digest the proteins, LysC (10 μg; 129-02541, Wako Chemicals) was added, and the samples were digested for 2 hours at 37°C. Trypsin (20 μg; V5113, Promega) was added, and the samples were digested overnight at room temperature.

Peptide desalting

Sep-Pak Accell Plus CM Vac cartridges (WAT023625, Waters) were prepared under light vacuum using the following washes: 3 ml of 100% (v/v) ACN; 1 ml of 70% (v/v) ACN, 0.25% (v/v) acetic acid (AA); 1 ml of 40% (v/v) ACN, 0.5% (v/v) AA; 1 ml of 20% (v/v) ACN, 0.5% (v/v) AA; and 3 ml of 0.1% (v/v) TFA. The peptide samples were brought to pH 2 with TFA and loaded onto the prewashed cartridge. The bound peptides were washed with 3 ml of 0.1% TFA followed by 300 μl of 0.5% AA. The peptides were eluted in sequential applications of 1 ml of 40% ACN and 0.25% AA and 750 μl of 70% ACN and 0.25% AA, and the samples were dried in a speed vacuum concentrator.

Peptide TMT labeling

The TMTsixplex Isobaric Label Reagent Set (90061, Pierce) was used to individually tag the samples with six unique mass tags. The dried peptide samples were resuspended in 100 μl of 200 mM tetraethylammonium bromide. TMT Label Reagents (0.8 mg per sample) were brought to room temperature, spun down, and resuspended in 50 μl of ACN. The peptide samples and TMT Label Reagents were combined and incubated at room temperature for 2 hours. The labeling reactions were quenched with 0.8 μl of 50% (v/v) hydroxylamine and then shaken for 15 min. Aliquots (5 μl) of the samples were combined with 50 μl of water and dried down, and the remaining samples were immediately frozen. The aliquots of labeled peptides were resuspended in 0.2% formic acid and then run on the mass spectrometer to determine the ratios of the samples to load for equal protein abundance. The frozen samples were thawed, mixed according to the determined ratios, dried down, and desalted as before.

Strong cation exchange fractionation

Strong cation exchange fractionation was performed at a flow rate of 3.0 ml/min using a polySULFOETHYL A column (9.4 × 200 mM; 209-SE05, PolyLC) on a Surveyor LC quaternary pump. Tagged samples were resuspended in buffer A [30% ACN and 5 mM KH2PO4 (pH 2.6)] and separated using the following gradients: 0 to 2 min, 100% buffer A; 2 to 5 min, 0 to 15% buffer B [30% ACN, 350 mM KCl, and 5 mM KH2PO4 (pH 2.6)]; 5 to 35 min, 15 to 100% buffer B. Buffer B was then held at 100% for 10 min. The column was washed with buffer C [500 mM KCl and 50 mM KH2PO4 (pH 7.5)] and water and reequilibrated with buffer A. All fractions were collected by hand, frozen, lyophilized, and desalted over a Sep-Pak Accell Plus CM Vac cartridges (WAT023625, Waters).

Phosphopeptide enrichment

Phosphopeptide enrichment was performed using immobilized metal affinity chromatography with metal beads (Qiagen). Before loading the peptides, the beads were equilibrated with water and then incubated with 40 mM EDTA (pH 8) for 30 min with shaking. EDTA was removed by three successive water washes, and then the beads were incubated with 100 mM FeCl3 for 30 min with shaking. For a final wash, the beads were washed with four rounds of 80% ACN and 0.1% TFA. Peptides were resuspended in 80% ACN and 0.1% TFA and incubated with the beads for 1 hour with shaking. Nonphosphorylated peptides were removed by washing the beads with 80% ACN and 0.1% TFA, and these were used for the proteomic analysis. Phosphorylated peptides were eluted from the beads with 50% ACN and 0.7% NH4OH and immediately acidified with 4% formic acid.

Liquid chromatography–tandem mass spectrometry

LC-MS/MS was performed using a nanoACQUITY UPLC system (Waters) coupled to an Orbitrap Elite (Thermo Fisher Scientific). Peptides were loaded onto a 75–μm–inner diameter and 360–μm–outer diameter bare fused silica capillary packed with 10 cm of Magic C18 particles (Bruker-Michrom) for 12 min at a flow rate of 1 μl/min. Peptides were then eluted onto a 50–μm–inner diameter and 360–μm–outer diameter analytical column packed with 17 cm of Magic C18 particles. Peptides were separated over a 90- or 120-min gradient by ramping from 2 to 35% ACN and 0.2% formic acid at a flow rate of 0.3 μl/min. Peptides were fragmented by high-energy collision dissociation (HCD) and analyzed using the Coon OMSSA Proteomics Software Suite (COMPASS) (94). The mass tolerance was set to 20 parts per million (ppm) for precursors and 0.01 Th for fragment ions. The carbamidomethylation of cysteines and the six unique mass tags of TMTsixplex reagents were searched as fixed modification. Oxidation of methionine, TMTsixplex modification of tyrosine, phosphorylation of tyrosine, and phosphorylation with neutral loss on serine and threonine were searched as variable modifications. COMPASS software was used to filter peptides to a 1% FDR, which were then combined and used to filter proteins at a 1% FDR. Comparisons of quantitative values were performed by converting all values to log2 and subtracting the value for uncomplemented AT1ABR (see table S1).

Hierarchical clustering

Protein and phosphopeptide levels were clustered in Cluster 3.0 using hierarchical clustering with centered correlation and average linkage (95). Results were visualized, and gene lists were extracted using Java Treeview (96).

Pairwise comparisons

Pairwise comparison graphs were made in Microsoft Excel 2010. The proteins or phosphopeptides were ranked from lowest intensity to highest intensity separately in each cell line, and the Spearman’s coefficients were calculated as ρ=16 Σdi2n(n21) where di = xi − yi is the difference in ranks for one protein between two cell lines. The t statistic for each Spearman’s coefficient was calculated as t=ρn21ρ2 and the significance at the 5 × 10−16 level was calculated using the qt() function in R version 3.0.3.

Sequence analysis

Six–amino acid peptide sequences N-terminal and C-terminal to each phosphorylated residue were extracted and formatted for submission to motif-x v1.2 10.05.06 (57, 58). The following settings were used for analysis of the phosphoproteome: foreground format, MS/MS; extend from International Protein Index (IPI) human proteome; central character, s, t, or y in three separate runs; width, 13; occurrences, 20; significance, 0.000001; background, IPI human proteome. For the C2991 Dependent Cluster, the occurrences were reduced to 10 to reflect the smaller number of phosphopeptides. The analysis was performed using the default significance value of 0.000001, which corresponds to a maximal P value of <0.0002.

Kinase motif prediction

Six–amino acid peptide sequences N-terminal and C-terminal to each phosphorylated residue were extracted and formatted for submission to GPS version 3.0 (97). The program was run with the threshold set to high, and the results for multiple kinases were combined in one table for further analysis.

K-S tests

The empirical distribution functions (EDFs) were generated, and K-S tests were performed in R version 3.0.3. Briefly, the ratios of intensities from AT1ABR cells expressing wild-type ATM and CL ATM for the phosphopeptides from the entire phosphoproteome or those predicted to be phosphorylated by a specific kinase using GPS were calculated. These ratios were entered into R as a table and converted to numeric vectors. The ECDF for the phosphoproteome and the phosphopeptides predicted to be substrates of a specific kinase were plotted with the plot() and ecdf() functions, and the K-S test was performed using the ks.test() function.


Fig. S1. Recombinant ATM protein expression.

Fig. S2. Arsenite and peroxide treatments do not induce DSBs in ATM-inducible U2OS cell lines.

Fig. S3. The 2RA ATM mutant is deficient in CPT-induced KAP1 phosphorylation.

Fig. S4. Expression of kinase-deficient ATM sensitizes cells to exogenous stress.

Fig. S5. U2OS cells expressing wild-type and mutant alleles of ATM have equivalent ratios of cells in S phase.

Fig. S6. Mitochondrial dysfunction in AT1ABR cells expressing CL ATM.

Fig. S7. The mKeima mitochondria-targeted pH probe responds to induction or repression of mitophagy.

Fig. S8. ATM depletion reduces macroautophagy responses of human cells in response to DNA damage or oxidative stress.

Fig. S9. ATM mutant expression does not alter the predicted phosphorylation targets of CK1 or CDK.

Fig. S10. CK2 subunit levels are not reduced in A-T cells expressing the CL ATM allele.

Fig. S11. Expression of CL ATM causes Rad50 accumulation in protein aggregates.

Table S1. Phosphorylated peptides observed in AT1ABR cells with inducibly expressed wild-type, CL, 2RA, or 2RA/CL ATM alleles.

Table S2. Proteins detected in aggregate fractions.


Acknowledgments: We thank B. Xhelmace for gifts of cell lines and expression constructs, E. Marcotte for helpful advice with statistical analysis, A. Harvey for assistance with analysis of MS data, J. Schymkowitz for assistance with TANGO/WALTZ analysis, and Paull laboratory members for helpful suggestions. Funding: C.-H.K. was supported by The Ministry of Education Technologies Incubation Scholarship grant of Taiwan. Studies in the Coon laboratory were supported by the NIH (P41 GM108538). Studies in the Paull laboratory were supported in part by the Cancer Prevention Research Institute of Texas (grant RP100670). Author contributions: J.-H.L. and M.R.M. conducted and analyzed experiments and contributed to the writing of the manuscript. C.-H.K., S.W.R., Y.Z., and A.L.R. conducted and analyzed experiments. J.J.C. analyzed experiments and helped edit the manuscript. T.T.P. analyzed experiments and wrote and edited the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The MS and phosphoproteomic data are deposited to CHORUS.

Stay Connected to Science Signaling

Navigate This Article