The Cancer-Associated ATM R3008H Mutation Reveals the Link between ATM Activation and Its Exchange

Ataxia-Telangiectasia (A-T) mutated (ATM) is a master regulator of the DNA damage response (1) and a tumor-suppressor gene. Germline ATM inactivation causes A-T syndrome, characterized by cerebellar degeneration, immunodeficiency, and early-onset lymphomas. Over 90% of patients with A-T have truncating or frameshift mutations with little or no ATM protein expression (2, 3). Missense mutations of ATM are rare in A-T. Mouse models with complete loss of Atm (null; refs. 4–8) recapitulate the immunodeficiency and lymphoma predisposition of A-T but do not develop overt neurological defects (9).

To ask whether ATM protein has a role in tumor suppression beyond its kinase activity, we and others generated mice expressing kinase-dead (KD) ATM. In contrast with the normal development of Atm^−/− mice, Atm^KD/KD and Atm^KD/− mice died at embryonic day 9.5 with severe genomic instabilities (10, 11), explaining the lack of missense mutations in live-born A-T patients. ATM ranks as the third most common pathogenic germline variants in 10,389 adult cancers (12). Hematopoietic specific inactivation of the conditional allele in Atm^C/KD mice causes earlier and more aggressive lymphomas than Atm deletion (Atm^C/−; refs. 3, 10). Notably, >70% of cancer-associated ATM alterations in The Cancer Genome Atlas are missense mutations that are highly enriched in the C-terminal kinase domain and the adjacent PI3-Kinase regulatory domain (PRD; ref. 3). Notably, neuron-specific expression of Atm-KD does not elicit Ataix in mice (13).

ATM suppresses lymphomagenesis by promoting DNA repair and activating cell-cycle checkpoints. ATM-deficient lymphocytes accumulate DNA double-strand breaks (DSB) during V(D)J recombination (14, 15) and class switch recombination (CSR; refs. 16–18). In the absence of checkpoints, the breaks participate in chromosomal translocations, leading to lymphomas in approximately 25% of patients with A-T and nearly all Atm-null mice (1). Cells expressing Atm-KD display similar V(D)J recombination and CSR defects as of the Atm^−/− cells (3, 10). Both Atm^KD/− and Atm^−/− cells cannot phosphorylate Chk2 and p53, two checkpoint effectors (1).

How does Atm-KD cause embryonic lethality and accelerate tumors? Atm^KD/− lymphocytes have more chromatid breaks than Atm^−/− control and are hypersensitive to Topoisomerase I inhibitors, suggesting that Atm-KD impairs replication-associated DNA repair beyond the loss of ATM (3, 10, 11). Atm^KD/−, but not Atm^−/− cells, show defects in the DR-GFP assay that measures homologous recombination (3, 10, 11). ATM is activated at the DNA damage sites by the MRE11–RAD50–NBS1 (MRN) complex. ATM Activation is concurrent with a dimer to monomer transition and inter-molecular auto-phosphorylation (1). ATM auto-phosphorylation at S1981 (corresponding to mouse S1987) are widely used as a marker for ATM activation (19, 20). Transgenic mice carrying alanine substitution at one (Atm-S1987A) or two additional auto-phosphorylation sites were born alive with no detectable defects in ATM activation (21, 22), suggesting auto-phosphorylation alone cannot explain the severe genomic instabilities in Atm^KD/− mice. In addition to DNA damage, reactive oxygen species (ROS) also activate ATM (23). But the physiological role of ROS induced ATM activation remains elusive.

The allosteric changes of PRD-I loop between Kα9 and Kα10 helices are implicated in the ATM and Tel1 activation (24–27; Fig. 1A–D). Within PRD, C2991 is essential for ROS-induced ATM activation in vitro (23), and K3016 and S2996 are important for DNA damage-induced ATM acetylation and phosphorylation (Fig. 1B). The R3008 (R3016 in mice) in the Kα10 helix is the most common cancer-associated ATM missense mutation (Supplementary Fig. S1A and S1B; ref. 3). To understand the impact of the PRD in ATM and its tumor-suppressor functions, we generated the Atm^R/R mice carrying R3016H mutation. Atm-R3016H protein is stable but cannot be efficiently activated by DNA damage or ROS. Yet, Atm^R/R mice are viable and display spontaneous craniofacial defects and lymphoma spectrum that are distinct from both Atm^−/− and Atm^KD/KD mice (see Supplementary Table S1). In cells, the R3008H mutation rescues the delayed exchange of ATM-KD at the DNA damage, suggesting that R3008 allosterically coordinates ATM activation and exchanges, with implications for patients with A-T and ATM-mutated cancers.

The Atm^−/− mouse was described previously (28). The R3016H mutation (NP_031525.3; Fig. 1A and B) and a silent Bgl II digestion site (no protein change) were embedded in the 3′ arm and cloned into the pEMC vector with a neomycin-resistant (NeoR) cassette flanked by FRT sites and additional KpnI and EcoRV digestion sites for Southern blotting (Supplementary Fig. S1C). The correctly targeted clones have the NeoR upstream of the R3016H mutation site [CGT(Arg)→ CAT(His)] in exon 62 (Supplementary Fig. S1C and S1E) and were identified by Southern Blotting (EcoRV+KpnI digestion, a 3′ probe (generated 5′-TCT CCT GGC TAC ATG CTA-3′ and 5′-AAC ACT CAG CCG TCG TC-3′; Supplementary Fig. S1C and S1D). The germline and targeted bands are approximately 13.1 kb and 4.1 kb, respectively (Supplementary Fig S1D). The mutation was confirmed by Sanger sequencing (Supplementary Fig. S1E). The resulting chimeras were bred with Rosa26a^FLP/FLP mice (the Jackson Laboratory; stock no. 003946) to remove the NeoR cassette. Two independently targeted clones were injected for germline transmission and discussed together thereafter. Genotyping primers are 5′- CGC ACA GTG TCG TCT G-3′ and 5′-CGT GCC TTT TAA TTA TGT AG-3′ and Atm^R allele = 592 bp versus Germline = 474 bp. Only 129/sv Atm^R/R mice were used in this study. All animal procedures were approved by the Institutional Animal Care and Use Committee at Columbia University Irving Medical Center.

The wild-type Flag-tagged ATM was a gift from Dr. Michael Kastan and the R3008H and D2889A versions were made using Quikchange mutagenesis (Stratagene). Recombinant human ATM was made by transient transfection of expression constructs into 293T cells using calcium phosphate and purified as described previously (29). ATM kinase assays with MRN and DNA were performed with 1.35 nmol/L ATM, 6.25 nmol/L GST-p53 substrate, 75 nmol/L MR, 100 nmol/L Nbs1, and 140 nmol/L linear double-stranded DNA as indicated in the figure legends. Kinase assays were performed in kinase buffer (50 mmol/L HEPES, pH 7.5, 50 mmol/L potassium chloride, 5 mmol/L magnesium chloride, 10% glycerol, 1 mmol/L ATP, and 1 mmol/L DTT) for 90 minutes at 30°C in a volume of 40 μL as described previously (2). Kinase assays with oxidation were performed with 817 μmol/L hydrogen peroxide (H₂O₂; Thermo Fisher Scientific, H325-100), 2.7 nmol/L ATM, and approximately 50 μmol/L dithiothreitol (DTT). Phosphorylated p53 (pS15) was detected as described previously (30) using a phospho-specific antibody (Calbiochem, PC461).

Mice were transcardially perfused with 4% paraformaldehyde (PFA), and tissue was cryoprotected in 25% PBS-buffered sucrose solution, embedded in O.C.T. and sectioned sagittally at 10 μm using an HM500M cryostat (Microm). IHC was performed after antigen retrieval (HistoVT One, Nacalai USA Inc.). The following primary antibodies were used: phospho-H2ax-Ser-139 (1:500, Cell Signaling Technology, #2577); NeuN (1:500, Chemicon, MAB377), and PCNA (1:500, Santa Cruz Biotechnologies, SC-56), and Cy3-conjugated secondary antibodies (Jackson Immunologicals) were used for fluorescent visualization and counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories). Three independent tissue samples from each genotype were stained and analyzed and one representative picture is shown for each genotype.

Lymphocyte development and in vitro CSR were performed as described before (10, 31, 32). Hematopoietic cells from 5 to 8 week mice were stained and analyzed on a FACS Calibur flow cytometer (BD Biosciences). The antibodies are: PE-CD4 (clone GK1.5, BD Pharmingen, 553730), FITC-CD8α (clone 53-6.7, BioLegend, 100705), APC-TCRβ (clone H57-597, BD Pharmingen, 553174), PE/Cy7 TER-119 (clone TER-119, BioLegend, 116222), FITC-CD43 (clone S7, BD Pharmingen, 553270), PE-Cy5-B220 (clone RA3-6B2, BD Pharmingen, 553091), and PE-IgM (Southern Biotech, 1020-09). For in vitro CSR, CD43⁻ splenic cells (anti-CD43 magnetic beads, Miltenyi, 130-049-801) were cultured (∼1  ×  10⁶ cells ml⁻¹) in RPMI (Gibco, 11875-093), serum supplements (10, 31, 32), IL4 (20 ng/mL; R&D Systems, 404-ML-050), and anti-CD40 (1 μg/mL; BD Biosciences, 553721) and analyses with FITC-conjugated IgG1 (clone A85-1, BD Pharmingen, 553443) and PECy5-conjugated B220 (clone RA3-6B2, BD Pharmingen, 553091). FlowJo software package was used for data analyses

Murine embryonic fibroblasts (MEF) were harvested at embryonic day 14.5 and cultured in DMEM (Gibco, 12430-054) with 15% FBS. Tert-Butyl hydroperoxide (TBH)–treated cells were incubated in 2-mercaptoethanol–free medium. SV40 antigens immortalized the MEFs (33). MEF proliferation was measured by CellTiter Glo (Promega, G7572, 2 × 10³ cells/well on 96 wells) on a GloMax microplate reader (Promega). The relative growth was plotted as the fold increase over the first day (noted as day 0). B-cell proliferation was analyzed using the Cell Trace Violet (CTV; ThermoFisher Scientific, C34557) and collected on an LSRII flow cytometer (BD Biosciences). The following chemicals were used: camptothecin (CPT; Calbiochem, 208925), etoposide (EtOP; Sigma, E1383), TBH solution (Luperox TBH70X, Sigma, 458139), colcemid (KryoMAX Colcemid, Gibco, 15212-012), and neocarcinostatin (NCS; Sigma, N9162).

For G₁–S checkpoint, MEFs were irradiated at 0 or 5 Gy and recovered for 12 hours before pulse-labeled with 10 μmol/L 5-bromo-2′-deoxyuridine (BrdUrd, Sigma, B9285) for 30 minutes, stained with anti-BrdUrd kit (BD Pharmingen, 556028) and propidium iodide (PI; Sigma, P4170) + Rnase A (Sigma, 10109169001). For G₂–M checkpoint, MEFs were irradiated with 0 or 10 Gy, incubated with 100 ng/mL of colcemid for 3 hours, fixed with 70% ethanol, and stained with anti-phospho Histone H3 (S10; pH3) antibody (EDM Millipore, 06-570). To combine the anti-pH3 and BrdUrd staining (Fig. 5D and E), the 2.5 days activated B cells were pulse treated with NCS (100 ng/mL) for 1 hour and collected after 3 hours with colcemid (100 ng/mL final) and 30 minutes with BrdUrd (10 μmol/L). Cells were analyzed on an Attune NxT flow cytometer (Thermo Fisher Scientific).

Whole-cell extracts were prepared using modified RIPA buffer (150 mmol/L sodium chloride, 10 mmol/L Tris–hydrogen chloride pH 7.4, 0.1% sodium dodecyl sulfate, 0.1% Triton X-100, 1% sodium deoxycholate, 5 mmol/L ethylenediaminetetraacetic acid) with protease inhibitor cocktail (Roche, 11697498001). SDS-PAGE and immunoblots were performed following standard protocols. The primary antibodies are: anti-ATM (Sigma, A1106), anti-pATM S1981 (Cell Signaling Technology, 4526), anti-pKAP1 S824 (Abcam, ab70369), anti-KAP1 (Cell Signaling, 4124), anti-CHK2 (BD Biosciences, 611570), anti-pH2AX Ser139 (Cell Signaling Technology, 9718S), anti-H2AX (Millipore, 07-627), anti-Vinculin (Millipore, 05-386), anti–β-actin (Sigma, A1978), and anti-α-tubulin (Calbiochem, CP06).

Metaphases were collected from activated B cells at 4.5 days after stimulation after 2 hours with colcemid (KaryoMax Colcemid Solution, GIBCO, 100 ng/mL final) and stated with telomere-PNA probe (34). The slides were counterstained with Vectashield mounting media containing DAPI (Vector Laboratories, H-1200-10) and analyzed on a Carl Zeiss AxioImager Z2 microscope equipped with a CoolCube 1 camera and a 63 × /1.30 oil objective lens, driven by Metafer4 and the ISIS fluorescence image software (MetaSystems).

U2OS cells with stable shRNA ATM knockdown (U2OS shATM) and plasmids encoding shRNA-resistant human ATM-WT and ATM-KD (pcDNA3; NotI-XhoI containing the entire ORF) were provided by Dr. Christopher Bakkenist (University of Pittsburgh, Pittsburgh, PA; ref. 20). We inserted a GFP tag and the SV-40 nuclear localization sequence at the N-terminus of ATM (GFP-ATM and GFP-ATM KD). The R3008H mutation was introduced using: FWD 5′-CAA CAA AGT AGC TGA ACA TGT CTT AAT GAG ACT AC-3′ and REV 5′-GTA GTC TCA TTA AGA CAT GTT CAG CTA CTT TGT TG-3′ (GFP-ATM R3008H and GFP-ATM KD/R3008H). For live-cell imaging, U2OS shATM were transiently transfected GFP-ATM plastic with Lipofectamine 2000 (Invitrogen, 11668019). Two days (48 hours) later, the cells were incubated with BrdUrd (10 μmol/L) for 16-18 hours before imaging. Laser microirradiation and time-lapse imaging were conducted via the NIS Element High Content Analysis software (Nikon Inc.) using a 405-nm laser (energy level ≅500 μW for a ∼0.8 μm diameter region; refs. 32, 35, 36). Relative fluorescence intensity = ratio of the mean intensity at the microirradiation sites/the mean nuclear intensity. Fiji: ImageJ software was used for data analyses. For fluorescence recovery after photobleaching (FRAP), photobleaching with a GFP-specific 488-nm laser (for a ∼1.2 μmol/L diameter region) was performed at 5 minutes after initial microirradiation when the relative intensity of the ATM foci was at its peak, and the fluorescence recovery was documented every 2 seconds for up to 60 seconds as previously described (35, 36). Normalized fluorescence intensity for each time point was determined by setting the intensity immediately before and after photobleaching as 100% and 0%, respectively. The maximal recovery (or mobile fractions) and t_1/2 were calculated on the basis of earlier publications with minor modifications (36). The dissociation constants (K_d) were calculated using the Prism software (GraphPad Software) assuming the one-binding site model. More than 10 cells were acquired and analyzed for each technical repeat. At least two and often three independent technical repeats were performed.

Homologous targeting was used to introduce the R3016H (CGT→ CAT) mutation (Supplementary Fig. S1C) to Atm locus in ES cells. The targeted clones were identified by Southern blotting (Supplementary Fig. S1D), and confirmed by Sanger sequencing (Supplementary Fig. S1E) before injection. Atm^R/R mice express Atm protein at comparable levels with Atm^+/+ control but lack ionizing radiation (IR) induced phosphorylation of Kap1, H2ax, or Chk2 (Fig. 1E). ROS-inducing agent TBH also cannot induce Kap1 or Chk2 phosphorylation in Atm^R/R cells (Fig. 1F). A much-attenuated activation of Atm-R3018H protein is evident after prolonged treatment with TBH (2-4 hours; Fig. 1F). Purified human ATM-R3008H also cannot be efficiently activated by MRN+DNA (DNA damage; Supplementary Fig. S1F) or H₂O₂ (ROS-generating agents; Supplementary Fig. S1G). When reducing agent DTT was omitted from the reaction buffer, purified ATM-R3008H protein showed reduced yet significant basal activity toward a p53 peptide (Supplementary Fig. S1H), suggesting that the R3008H mutation compromised ATM activation without abrogating its intrinsic kinase activity. Atm^+/R mice were born at the expected rate, normal size, and fertile (Fig. 1G and I), suggesting that it is not a classical dominant-negative allele. In contrast with the embryonic lethality of Atm^KD/KD mice, Atm^R/R mice were born alive, but under-represented (P < 0.0001; Fig. 1G), and smaller (Fig. 1H and I) than Atm^−/− controls. Adolescent Atm^R/R mice (1–6 weeks) gain weight normally after birth (Supplementary Fig. S1I).

The 2- to 4-weeks-old Atm^R/R mice have notably short noses (Fig. 2A). The eye-to-nose/eye-to-eye distance is significantly smaller in Atm^R/R mice than Atm^+/+ and Atm^−/− mice (Fig. 2B). Yet, the rotarod test on 3-week-old and 6-week-old Atm^R/R mice revealed similar defects in Atm^R/R and Atm^−/− mice with no craniofacial defect (Fig. 2C). Histological analyses of 2-week- (Fig. 2D) and 8-week-old (Fig. 2E) Atm^R/R mice also show no major changes in γH2ax staining or PCNA (S phase marker). Lig4-deficient neurons accumulate more γH2ax, whereas Atm-deleted neurons also show no measurable accumulation of γH2ax (Fig. 2D; refs. 37, 38). Thus, Atm^R/R mice were born alive with idiopathic craniofacial defects that do not correlate with additional behavior or histological changes beyond those on the Atm^−/− mice.

Next, we analyzed lymphocyte development. B-cell development in both Atm^R/R and Atm^−/− mice are largely normal (Supplementary Fig. S2A and S2B). T cells from Atm^R/R and Atm^−/− mice show a prominent blockade at the CD4⁺CD8⁺ double-positive (DP) to CD4⁺ or CD8⁺ single-positive (SP) transition (Fig. 3A and B; Supplementary Fig. S2C) and decreased surface TCRβ levels (Fig. 3A), consistent with defects in Vα-Jα recombination described for Atm-deficiency (8). Consequently, total thymic cellularity is also decreased in Atm^R/R mice (Supplementary Fig. S2C). The extent of all T-cell development defects in Atm^R/R mice are largely comparable with those of Atm^−/− mice, despite a statistically significant but at most moderate improvement of SP/DP ratio (Fig. 3A and B; Supplementary Fig. S2C). To examine CSR, purified splenic B cells were activated in the presence of anti-CD40 and IL4, which stimulate CSR to IgG1 and IgE. Both Atm^R/R and Atm^−/− B cells showed a >50% decrease in IgG1 switching (Fig. 3C and D). At day 4.5, Atm^R/R B cells show a statistically significant, but very moderate improvement of IgG1 switching than the Atm^−/− B cells Fig. 3C and D). No statistically significant difference in IgG1 CSR was noted at day 3.5 after stimulation (Fig. 3C and D). To understand this “late” switching in Atm^R/R B cells, we performed CTV labeling and plotted the frequency of IgG1 CSR in Atm^R/R and Atm^−/− B lymphocytes by cell division. However, no significant proliferation defects or preference for IgG1 CSR at later days were noted between in Atm^R/R and Atm^−/− B lymphocytes (Supplementary Fig. S2D and S2E).

Because the IgH gene resides near the telomere end of murine chromosome 12, telomere-FISH (T-FISH) analyses have been used to visualize and quantify CSR-associated DNA repair defects in activating B cells (10, 32, 34). T-FISH can recognize two types of breaks—chromosome breaks involving both sister chromatids and chromatid breaks involving one of the two sister chromatids (Fig. 3G; refs. 10, 32, 34). Chromatid breaks occur during or after DNA replication, reflecting the different fates of the two sister chromatids, whereas chromosome breaks are likely initiated in G₁ cells (34). In Atm^−/− B cells, the CSR-associated IgH breaks were almost all chromosome breaks (34). T-FISH analyses show that both Atm^R/R and Atm^−/− B cells have increased metaphases with breaks (either kind; Fig. 3E and G; Supplementary Fig. S3), with a very moderate yet significant lower frequency of chromosome breaks in the Atm^R/R cells (Fig. 3F and G; Supplementary Fig. S3). In addition to breaks, T-FISH can also reveal telomere instability, characterized by fragmented telomere signals (Fig. 3G). Telomere fragility has been noted in Pot1a or ATR-deficient cells and linked to replication stress at the telomeres (36, 39). Atm^R/R B cells have at most a moderate increased frequency of telomere fragility (Fig. 3G and H; Supplementary Fig. S3). Taken together, lymphocyte analyses of Atm^R/R mice confirmed that lymphocyte-specific gene rearrangement is significantly compromised in Atm^R/R mice. The extent of repair defects is quite comparable with those seen in Atm^−/− mice.

Atm^−/− mice routinely succumbed to thymic lymphomas by 4 months of age (t_1/2 = 115 days; Fig. 4A and C). Despite the ongoing genomic instabilities noted above, the expression of Atm-R3016H significantly lengthened overall life expectance (t_1/2 = 211 days; Fig. 4A) and the occurrence of thymic lymphomas specifically in both Atm^−/R mice (t_1/2 = 194 days overall, and 190 days for thymic lymphomas only, P < 0.01, Mantel–Cox test) and Atm^R/R mice (t_1/2 = 211 days overall, and 183 days for thymic lymphomas, P < 0.0001, Mantel–Cox test; Fig. 4C). Atm^+/+ mice in pure 129sv background rarely develop spontaneous tumors before 1 year (365 days) of age, indicating that R3008H mutation confers the partial loss of ATM tumor-suppressor function. Moreover, Atm^R/R and Atm^−/R mice displayed a shift in overall tumor types (Fig. 4B). In addition to the (CD3 low) thymic lymphomas (Fig. 4D), the Atm^−/R and Atm^R/R mice developed other hematologic malignancies (e.g., pro–B-cell lymphomas with the expansion of B220⁺IgM⁻CD43⁺cells, Fig. 4E). Thymus development peaks at 1 month and attenuates afterward. The older Atm^R/R and Atm^−/R mice that escaped lethal thymic lymphomas, eventually developed other hematological malignancy and sarcomas later in life (Fig. 4A). Those findings suggest that a mechanism beyond the programmed DSBs repair defects might contribute to the delayed lymphomagenesis and craniofacial abnormalities in Atm^R/R mice.

So next, we analyzed the proliferation and the cell-cycle checkpoints in Atm^R/R and Atm^−/− MEFs and B cells. Both Atm^R/R and Atm^−/− MEFs grew significantly slower without a major difference between the two alleles (Fig. 5A), had a reduced S phase (BrdUrd⁺) fraction (Fig. 5B; Supplementary Fig. S4A), and a decreased mitotic fraction (pH3⁺) among all G₂–M cells (Fig. 5C and D), suggesting a delayed mitotic entry. ATM regulates DNA damage-induced G₁–S checkpoint upstream of p53 and the G₂–M checkpoint upstream of CHK2 (40). Both functions require DNA damage-induced kinase activity of ATM. Accordingly, both Atm^R/R and Atm^−/− B cells lost the genotoxic-induced G₂–M checkpoint (Fig. 5E).

Atm^R/R and Atm^−/− primary MEFs showed defects in ionizing-radiation induced G₁–S (Supplementary Fig. S4A) and G₂–M checkpoints (Fig. 5F; Supplementary Fig. S4B), consistent with the lack of IR-induced acute activation of Atm-R3016H (Fig. 5H and I, at 0.5 and 1 hour). Thus, those analyses in MEFs and B cells suggest that DNA damage-induced acute checkpoint activation is similarly affected in Atm^R/R and Atm^−/− mice in general.

Although IR-induced “transient” DNA damage and ATM activation, we note that the prolonged TBH treatment (2–4 hours) evoked significant activation of Atm-R3016H (Fig. 1F). To determine whether Atm-R3016H can respond to ROS-induced checkpoint activation, we treated Atm^R/R and Atm^−/− primary MEF cells with TBH, a ROS inducer. As shown in Fig. 1F, acute treatment with TBH, even at a high dose (700 μmol/L), did not induce significant Chk2 phosphorylation, but the extended treatment of 2 and 4 hours caused noticeable Atm-R3016H activation beyond the loss of Atm. Thus, we treated primary MEFs for 4 hours with 200 μmol/L TBH, which induced approximately 40% reduction of the mitotic entry in Atm^+/+ MEFs (Fig. 5F). Although Atm^R/R and Atm^−/− MEFs fail to activate the G₂–M checkpoint upon radiation, Atm^R/R MEFs, and, to a lesser extent, Atm^−/− MEFs show a stronger G₂–M arrest in response to TBH than Atm^+/+ MEFs (Fig. 5F; Supplementary Fig. S4B).

To determine whether this difference reflects the ability of Atm-R3016H to respond to chronic stress in general or ROS specifically, we treated Atm^R/R cells with DNA damage agents for 4 hours. Although very much diminished, Atm^R/R cells could phosphorylate Chk2 beyond the levels observed in Atm^−/− cells (Fig. 5G). Those findings are consistent with the retention of basal kinase activity, but the lack of efficient DNA damage- and ROS-induced activation of ATM-R3008 in vitro (Supplementary Fig. S1F–S1H). We proposed that chronic activation of Atm-R3016H might contribute to the delayed lymphomagenesis and craniofacial defects in Atm^R/R mice.

To determine whether the R3008H mutation affects ATM–MRN interaction to attenuate ATM activation, we expressed shRNA-resistant GFP-tagged ATM-R3008H or ATM-WT in U2OS shATM cells (Supplementary Fig. S4C). Both ATM-R3008H and the previously characterized ATM-KD could be efficiently recruited to the site of DNA damage (Fig. 6A–C). Live-cell imaging further showed that the recruitment kinetics (Fig. 6B) and relative fluorescence intensity (Fig. 6C) of damage-induced foci formed by ATM-R3008H and ATM-KD were indistinguishable from ATM-WT. Correspondingly, purified ATM-R3008H can efficiently bind to both MRN and MR (Supplementary Fig. S4D and S4E). In an earlier study, we found that the ATR kinase inhibitors affect ATR exchange without affecting the initial recruitment of ATR at the DNA damage site (36, 41). To test whether this is also true for ATM, we measured FRAP at 5 minutes after initial microirradiation, when the ATM foci intensity reaches a plateau (Fig. 6B). ATM-KD showed a moderate, yet consistent, delay of exchange evidenced by a significantly delayed t_1/2 (9.46 ± 1.91 vs. 7.56 ± 1.79 for ATM-WT, P = 0.02, unpaired two-tailed t test) and moderate reduction of maximal recovery (0.67 ± 0.08 vs. 0.75 ± 0.1 for ATM-WT, P = 0.05, unpaired two-tailed Student t test; Fig. 6D–F), suggesting that ATM kinase activity might also be linked to ATM exchange. Surprisingly, ATM-R3008H displays no exchange defects (t_1/2 = 7.98 ± 1.83 vs. 7.56 ± 1.79 for ATM-WT, P = 0.59, unpaired two-tailed Student t test) and even higher maximal recovery (0.91 ± 0.05 vs. 0.75 ± 0.1 for ATM-WT, P < 0.001 unpaired two-tailed Student t test; Fig. 6D–F). On the basis of these findings, we propose a model in which the PRD is critical in connecting ATM activation with the dynamic exchange of ATM at the DNA damage sites. The R3008H mutation at the Kα10 helix (Fig. 1B and C) abrogates this link, impairing ATM activation and high-affinity engagement of ATM with the MRN–DNA complex. Consistent with this hypothesis, ATM carrying both R3008H and KD mutations at the same cDNA can exchange effectively, suggesting that R3008H mutation rescued the exchange defects in ATM-KD in cis (Fig. 6D–F). Together, our data support a model in which the R3008H mutation interferes with ATM activation by impairing the allosteric changes that link ATM kinase activation and ATM exchange.

The embryonic lethality of Atm^KD/− mice and the enrichment of missense ATM mutation in cancers suggest that the frequency of pathogenic ATM mutations in cancer might be greatly underestimated. The much more severe genomic instability of Atm^KD/− cells suggests that the presence of ATM protein without kinase activity might physically block DNA processing and/or repair (3, 10, 11). Because of the large size, functional validation of cancer-associated ATM mutations has been technically challenging. Most genetic screens only consider mutations identified in patients with A-T as bona fide ATM-inactivation mutations. Missense mutations in the PRD domain are frequently found in human cancer. Here, we showed that Atm^R/R mice carrying a PRD domain missense mutation are viable and express a stable ATM protein that cannot be activated by either DNA damage or ROS. This is unexpected, given the embryonic lethality of Atm^KD/KD mice, and suggest that the kinase-dependent structural function of ATM can be uncoupled from the phosphorylation of its substrates, including itself. Instead, we propose that ATM activation and the completion of the phosphoryl group transfer might be coupled with ATM exchange. Specifically, FRAP analyses provide direct evidence for the rapid exchange of ATM at the DNA damage sites (42). Mechanistically, structural analyses of Tel1, the yeast homolog of ATM, suggest that the PRD regulates the cap that limits access of the substrate and ATP to the well-positioned catalytic center (24–26). On the basis of the successful development of Atm^R/R mice versus the embryonic lethality of Atm^KD/KD mice, and the ability of R3008H mutation to rescue the exchange defects of ATM-KD, we propose a model in which the PRD supports ATM activation by allowing the substrate to engage with the catalytic center, where this engagement is necessary to lock Atm-KD, where it prevents replication-associated DNA repair. At the molecular level, the arginine 3008 residue is one of the several conserved basic amino acids spaced 3-4 aa apart within the Kα10 helix implicated in stabilizing the helix structure by providing stacking interaction (Fig. 1B and C; ref. 43). Thus, R3008H mutation might cause partial denaturation of the Kα10 helix and impairs the “cap (PRD-I)” opening necessary for ATM activation, thereby diminishing the substrates' and ATP's access to the catalytic loop. In this role, the R3008H mutation reduces the unproductive engagement of the ATM-KD protein with its substrate and preventing stalling. This mechanism would also explain why the R3008H mutation (or R3016H of mouse Atm) abolished acute activation of ATM necessary for DNA repair but retained some basal activity to phosphorylate CHK2 during chronic stress, which might contribute to the delayed tumor onset. Alternatively, R3008H might alter substrates specificity, as IR-induced Atm-S1987 auto-phosphorylation seems to be less affected than Kap1 phosphorylation, and in chronically stressed Atm^R/R cells, Chk2 phosphorylation is less affected than Kap1 phosphorylation. In this case, we caution that in addition to ATM activity, the phosphorylation levels of ATM substrates at any given time are also influenced by the availability of other kinases (e.g., DNA-PKcs), the half-life of the phosphorylated proteins, and the accessibility to phosphatase.

In patients with A-T, Ataxia is largely uniform, but immunodeficiency and cancer risk vary extensively. Using inbred mouse models, our data, and previous analyses of Atm-KD mice (summarized in Supplementary Table S1) support a specific contribution of ATM mutation in the diverse immunological and cancer predisposition phenotypes beyond the genetic background of the patients. Specifically, ATM kinase activity affects clean DSB repair, which is similarly affected by null, KD, and R3008H mutations and correlates with lymphocyte-specific recombination defects. We note that lymphocyte development can also be affected by proliferation (as those caused by Atm-KD) in addition to rearrangement defects. In the absence of ATM kinase activation, the expression level of the inactive ATM protein and the type of mutation (KD or inactivation mutation—R3008H) work together to influence the stalling of ATM at DNA damage sites and the genotoxicity associated with the expression of KD ATM. In this regard, we speculate that the craniofacial defects of Atm^R/R mice without an additional decline of motor neural function might reflect proliferation defects in olfactory neurons that undergo extensive post-natal neurogenesis.

Finally, the difference between KD versus null is not limited to ATM kinase but also observed for DNA-PK (44, 45) and ATR (36, 41). In each case, the cytotoxicity is specific to the DNA substrates that recruit and activate the respective kinases—DSBs for DNA-PKcs and single-stranded DNA for ATR. In the case of DNA-PKcs, the deletion of KU, which recruits DNA-PKcs to the DNA ends, rescues the embryonic lethality of DNA-PKcs^KD/KD mice (44–46). The dynamic exchange of ATR at RPA-coated single-strand DNA is markedly reduced by ATR-specific inhibitors (36, 41), suggesting that the coupling between catalysis and the release of the kinase might be a common feature for all PIKK. The sequence corresponding to the Kα10 also exists in ATR, DNA-PKcs, and the related mTOR (47). Therefore, the phenotype and the mechanism of exchange regulated by R3008H identified here can potentially be applied to understand the kinase-dependent regulation of other PIKKs.

D. Menolfi reports receiving special fellowship of Lymphoma and Leukemia (LLS) society. J.-H. Lee reports grants from National Institutes of Health (NIH) during the conduct of the study. S. Zha reports grants from NIH NCI R01CA158073, NIH NCI R01CA215067, NIH NCI R01CA226852, NIH P01 CA174653, Leukemia Lymphoma Society, NIH/NCI T32 CA09503-29, Howard Hughes Medical Institute, NIH NS-37956, NIH CA-21765, American Lebanese and Syrian Associated Charities of St. Jude Children's Research Hospital, NIH GM102362, and NIH P30CA013696 during the conduct of the study. No disclosures were reported by the other authors.

M. Milanovic: Data curation, writing-original draft. L. Sprinzen: Data curation, formal analysis. D. Menolfi: Data curation, writing-review and editing. J.-H. Lee: Data curation. K. Yamamoto: Data curation, formal analysis, investigation, methodology. Y. Li: Resources, data curation. B.J. Lee: Resources, data curation, project administration. J. Xu: Resources, data curation, formal analysis. V.M. Estes: Data curation, formal analysis. D. Wang: Resources, data curation, software, visualization. P.J. McKinnon: Resources, data curation, supervision, funding acquisition, validation. T.T. Paull: Conceptualization, resources, data curation, funding acquisition, investigation, writing-review and editing. S. Zha: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration.

We thank members of the Zha laboratory for discussion, Dr. Bakkenist for providing ATM plasmids and ATM knockdown U2OS cells, Drs. Lasorella and Tala for helping with perfusion, Dr. Claudio Scuoppo for comments on the article. This work was supported by the NIH/NCI grants R01CA158073, R01CA215067, R01CA226852, and P01 CA174653 (to S. Zha). S. Zha is a Leukemia Lymphoma Society scholar. L. Sprinzen was supported by NIH/NCI T32 CA09503-29. J.-H. Lee and T.T. Paull were supported by the Howard Hughes Medical Institute. P.J. McKinnon is supported by the NIH (NS-37956, CA-21765), the CCSG (P30 CA21765), and the American Lebanese and Syrian Associated Charities of St. Jude Children's Research Hospital. D. Wang and J. Xu were supported by NIH GM102362. This research was funded in part through the NIH/NCI Cancer Center Support Grant P30CA013696 to the Herbert Irving Comprehensive Cancer Center of Columbia University.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.