FATC Domain Deletion Compromises ATM Protein Stability, Blocks Lymphocyte Development, and Promotes Lymphomagenesis

Ataxia-telangiectasia mutated (ATM) kinase is a protein serine/threonine kinase and a master regulator of the DNA damage response (1). Germline inactivation of ATM causes ataxia-telangiectasia (A-T) syndrome, characterized by extreme sensitivity to ionizing radiation, cerebellar degeneration, and loss of fertility (2). Nearly 70% of A-T patients show immunodeficiency (3, 4).

ATM-null mice do not develop spontaneous ataxia but recapitulate the immunodeficiency of A-T patients (5–10).

At the cellular level, ATM kinase is recruited to and activated by the MRE11-RAD50-NBS1 (MRN) complex at the site of DNA damage (11). Activated ATM promotes efficient and accurate DNA double-strand break (DSB) repair. In developing lymphocytes, loss of ATM compromises physiological DNA DSBs repair during V(D)J recombination (12, 13), underlying the T cell development block in A-T patients and ATM-deficient mice (8). Loss of ATM abolished the residual V(D)J recombination in cells lacking XLF, a nonessential, nonhomologous end-joining (NHEJ) factor (13–16). ATM also plays an important role in DSB repair during Ig class-switch recombination (CSR) (17–19), upon which naive B cells express Abs with different effector function and, thus, different isotypes (e.g., IgG1 or IgE, instead of the initially expressed IgM) (20). Loss of ATM reduces CSR efficiency by ∼50% (17, 18, 21) and leads to the accumulation of DSBs at IgH locus in ∼30% of activating B cells (22), underlying the humoral/Ab defects in A-T patients (3, 4). High-throughput sequencing of CSR junctions recovered from ATM-deficient B cells reveals a significant more and longer microhomology (MH) at the junctions (19), suggesting a role of ATM in NHEJ-mediated repair of CSR breaks. In addition to lymphocyte development, ATM is also required for meiotic recombination in germ cells (23–25). A-T patients and ATM-deficient mice are sterile (26). Activated ATM phosphorylates p53, Chk2, and others to activate DNA damage-induced cell cycle checkpoints (27). The checkpoint function of ATM is critical for its tumor suppressor role (1). About 25% of A-T patients develop lethal lymphoid malignancies (28). Almost all ATM-null mice (5–10) succumb to lethal thymic lymphomas with recurrent t(12:14) translocations involving the TCRα/σ locus (10, 29, 30). In addition to DNA damage, purified ATM can also be activated by reactive oxygen species (ROS) in an MRN-independent process that requires the formation of disulfide bond between the C2991 residues of two ATM monomers (31). Whereas brain tissues from A-T patients show signs of protein and lipid oxidation (32–36), the physiological role of ROS-induced ATM activation remains elusive, in part because of the lack of mouse models that can distinguish ROS versus DNA damage-induced ATM activation (31, 37).

ATM belongs to the PI3K-related kinase (PIKK) family that also includes ATR, DNA-PK, and mTOR kinases (38). They all contain a conserved kinase domain followed by a PIKK regulatory domain (PRD) and an ∼33-aa FATC domain (39). Structure studies suggest an allosteric activation model, in which the conserved Kα9 and Kα10 helices in the PRD physically block the substrate and ATP from entering the catalytic center (40–42). Indeed, PRD contains several residues implicated in ATM activation, including the C2991 (31) involved in disulfide-bond–mediated ATM activation by ROS, K3016 (43) implicated in DNA damage-induced acetylation of ATM, and S2996 (44), an autophosphorylation site (Fig. 1A, 1B). The FATC domain (ATM aa 3023–3056) includes a conserved α helix and flexible tail. The FATC domain of DNA-PK and mTOR are essential for their respective kinase activity (45, 46). Truncating mutation with the ATM FATC domain (R3047X) selectively ablates ROS-induced, but not MRN- and DNA-induced, activation of purified ATM protein (31). To understand the physiological role of the FATC domain, we generated a mouse model carrying the knock-in R3057X (corresponding to R3047 in human ATM, Atm⁠^RX⁠) mutation. Our analyses reveal a critical role of the FATC tail in ATM protein stability and indicate that growth retardation and lymphomagenesis are more sensitive to residual ATM kinase activity, whereas lymphocyte development, CSR, and germ cell development require substantial ATM kinase activity.

The Atm^−/− mouse model was described previously (9). To generate the Atm⁠^RX/RX⁠ mouse carrying the R3057X truncating mutation (NP_031525.3, corresponding to R3047X of human ATM) (Fig. 1A, 1B), the 3.5-kb 5′ arm and the 3.9-kb 3′ arm were amplified from genomic DNA extracted from CSL5 embryonic stem cells (129/sv background). The mutation was introduced in the 3′ arm and cloned into the targeting vector backbone carrying a neomycin-resistant (NeoR) cassette flanked by flippase recognition sites, and an XbaI site for identification (Supplemental Fig. 1A). The correct targeting clones have the NeoR upstream of the R3057X mutation site (CGA [Arg] → TGA [stop]) (the mutated base is underlined) in exon 62 (Supplemental Fig. 1A, 1C) and were identified by Southern blotting upon XbaI digestion using a 3′ probe generated by PCR (primers, 5′-TCT CCT GGC TAC ATG CTA-3′ and 5′-AAC ACT CAG CCG TCG TC-3′) (Supplemental Fig. 1B). The expected germline band is ∼7.7 kb, and the correctly targeted band is ∼4.3 kb (with the insertion of NeoR) (Supplemental Fig. 1B). The mutation was confirmed by Sanger sequencing (Supplemental Fig. 1C). Two clones were independently injected for germline transmission and yielded identical phenotypes. They were discussed together thereafter. The resulting chimeras were bred with Rosa26a^FLP/FLP mice (stock no. 003946; The Jackson Laboratory) to remove the NeoR cassette. Tail DNA from Atm⁠^RX/RX⁠ mice were PCR amplified and sequenced to confirm the desired mutations. Genotyping was performed with primers (forward 5′-CGC ACA GTG TCG TCT G-3′ and reverse 5′-CGT GCC TTT TAA TTA TGT AG-3′), and the Atm⁠^RX⁠ allele was identified by a 592-bp PCR product versus a 474-bp product for the germline allele. The Atm⁠^RX/RX⁠ mice used in this study were all in a 129/Sv background. All animal work was conducted in a specific-pathogen-free facility, and all the procedures were approved by the Institutional Animal Care and Use Committee at Columbia University Medical Center.

Lymphocyte development and CSR were performed as described before (47–49). Briefly, ∼1 × 10⁠⁵⁠⁠⁠ nucleated cells from lymphoid organs of 5- to 8-wk-old mice were stained with fluorescence-conjugated Abs and analyzed on a FACSCalibur flow cytometer (BD Biosciences). The specific Abs used were as follows: PE-CD4 (clone GK1.5, 553730; BD Pharmingen), FITC-CD8α (clone 53-6.7, 100705; BioLegend), APC-TCRβ (clone H57-597, 553174; BD Pharmingen), PE/Cy7 TER-119 (116222; BioLegend), FITC-CD43 (clone S7, 553270; BD Pharmingen), PE-Cy5-B220 (clone RA3-6B2, 553091; BD Pharmingen), and PE-IgM (1020-09; Southern Biotech). For CSR, CD43⁠⁻ B cells were isolated from total splenocytes after depletion with anti-CD43 magnetic beads (MACS, Miltenyi Biotec, 130-049-801) and cultured at ∼1 × 10⁠⁶⁠⁠⁠ cells ml⁠⁻¹ ⁠in RPMI 1640 (11875-093; Life Technologies) supplemented with 15% FBS (SH30071.03; Hyclone), 1× MEM nonessential amino acids (11140-050; Life Technologies), 20 mM HEPES (15630080; Life Technologies), 2 mM l-glutamine (25030-081; Life Technologies), 1× penicillin/streptomycin (15140122; Life Technologies), 120 μM 2-ME (034461-100; Fisher), and IL-4 (20 ng/ml, 404-ML-050; R&D) and anti-CD40 (1 μg/ml, 553721; BD Bioscience). Cultured cells were maintained daily at ∼1 × 10⁠⁶⁠⁠⁠ cells ml⁠⁻¹⁠ and collected daily for flow cytometry analyses with FITC-conjugated IgG1 (clone A85-1, 553443; BD Pharmingen) and PECy5-conjugated B220 (clone RA3-6B2, 553091; BD Pharmingen). All flow cytometry data were analyzed using the FlowJo software package.

Murine Abelson virus-transformed B cells were generated by isolating total bone marrow from <5-wk-old Atm⁠^+/+⁠, Atm⁠^−/−, ⁠and Atm⁠^RX/RX⁠ mice and infecting them with a retrovirus encoding the v-abl kinase (12). Cells were cultured in DMEM (12430-054; Life Technologies) supplemented with 15% FBS (SH30071.03; Hyclone), 1× MEM nonessential amino acids (11140-050; Life Technologies), 2 mM l-glutamine (25030-081; Life Technologies), 1 mM sodium pyruvate (11360-070; Life Technologies), 1× penicillin/streptomycin (15140122; Life Technologies), and 120 μM 2-ME (034461-100; Fisher) for the next 6–8 wk to allow for the generation of the stable clones. For treatment with tert-butyl hydroperoxide (TBH) solution, cells were incubated in media without 2-ME. The proliferation of primary B cells was analyzed using the CellTrace Violet (CTV) kit (C34557; Thermo Fisher Scientific) according to the manufacturer’s protocol. Cells were analyzed by flow cytometry on an Attune NxT flow cytometer (Thermo Fisher Scientific). The following small molecules were used at the concentrations indicated in the figure legends: TBH solution (Luperox TBH70X, 458139; Sigma-Aldrich), Colcemid (KryoMAX Colcemid, 15212-012; Life Technologies), and neocarzinostatin (NCS) (N9162; Sigma-Aldrich).

PCR analyses of hybrid joints (HJs) were performed as previously described (12). PCR (50 μl) was carried out with 75 μm dNTPs and 15 pmol of each primer. Vκ6–23 HJs and coding joints (CJs) were amplified from the splenocyte DNA (0.5 μg) using nested PCR. The PCR condition for the first PCR was 95°C for 5 min (94°C 30 s, 64°C 30 s, 72°C 30 s) × 17 with the primers pκJa and pκ6a for HJ and pκJa and pκ6d for CJ. Secondary PCR was carried out under the same conditions (×25 cycles) with 1:4 serial dilutions of the first PCR products with primers pκJa and pκ6b for HJ and pκJa and pκ6c for CJ (Supplemental Fig. 2G). Vβ14 HJs and CJs were amplified from thymocyte genomic DNA (0.5 μg). For Vβ14 HJs, pßa and pßb were used for the first⁠ PCR, and pßc and pßd were used for the secondary PCR (same condition as above with annealing temperature at 55°C and extension time of the 60 s). Vβ14 CJs and IL-2 were directly amplified from 4-fold serially diluted genomic DNA with the primers pße and pßf for Vβ14 CJs and IMR-042 and IMR-043 for IL-2. The PCR condition was 94°C 5 min (94°C 30 s, 60°C 30 s, 72°C 30 s) × 30. PCR products were run on 1% TAE Gel followed by Southern blotting analysis with p⁠³²⁠-labeled oligonucleotide probes pßg for Vβ14 HJs and CJs, pκg1 for Vκ6–23 HJs and CJs, and IMR042-2 for IL-2. Vk6-23 CJ band is 0.68 kb, and Vk6-23 HJ band is 0.26 kb. Vβ14 CJ band is 0.3 kb, and Vβ14 HJ bands are 0.9 kb. The primers used are listed in Supplemental Fig. 5.

For G2/M checkpoint analysis combining the antiphosphorylated histone H3 (pH3) and BrdU staining (Fig. 6D–F), B cells undergoing CSR were collected at 2.5 d poststimulation and left untreated or treated with NCS (100 ng/ml) for 1 h. Cells were washed with PBS and then incubated in the presence of Colcemid (100 ng/ml final) for the next 3 h. Half an hour before collection, cells were pulse-labeled with 10 μM BrdU, then collected and fixed with 70% ethanol for 24 h. The cells were permeabilized (0.3% Triton-X in PBS) and stained with an anti-pH3 (S10) Ab (06–570; EDM Millipore), followed by incubation with the secondary Ab Alexa594 (A-11012; Invitrogen). Cells were then neutralized (sodium phosphate citrate buffer, pH 7.4) before being stained with an anti-BrdU Ab (556028; BD Pharmingen) according to the manufacturer’s protocol. The stained cells were costained with propidium iodide (P4170; Sigma-Aldrich) for 30 min in the presence of RNase A (10109169001; Sigma-Aldrich). Cells were analyzed on an Attune NxT flow cytometer (Thermo Fisher Scientific), and the data were analyzed with FlowJo software package.

For Western blotting, whole-cell extracts were prepared using modified RIPA buffer (150 mM sodium chloride, 10 mM Tris–hydrogen chloride pH 7.4, 0.1% NaDodSO₄, 0.1% Triton X-100, 1% sodium deoxycholate, 5 mM EDTA) supplemented with 2 mM PMSF, 10 mM sodium fluoride, 10 mM β-glycerophosphate, and protease inhibitor mixture (11697498001; Roche). SDS-PAGE and immunoblots were performed following standard protocols. Primary Abs used in the study are anti-ATM (A1106; Sigma-Aldrich), anti-pKAP1 S824 (ab70369; Abcam), anti-KAP1 (4124; Cell Signaling), anti-CHK2 (611570; BD Biosciences), anti-pH2AX Ser139 (9718S; Cell Signaling), anti-H2AX (07-627; Millipore), anti-Vinculin (05-386; Millipore), and anti–β-actin (A1978; Sigma-Aldrich). Image quantification was carried out using ImageJ. Briefly, pKAP1 and KAP1 bands were selected and measured for the area under the curve (arbitrary units). The data were presented as the ratio of pKAP1/KAP1 (both area under the curve).

Metaphases were collected from activated B cells at 4.5 d after stimulation as previously described with 2 h incubation Insertions are regions that map to neither bait nor prey. The data with Colcemid (100 ng/ml, KaryoMax Colcemid Solution; Life Technologies) (22). Telomere fluorescence in situ hybridization (FISH) staining was performed as detailed before (22), counterstained with Vectashield + DAPI (H-1200-10; Vector Laboratories), 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).

High-throughput genome-wide translocation sequencing (HTGTS) was carried out as previously described (19, 50–53). Briefly, DNA from activated B cells (4 d with IL-4, anti-CD40) was sonicated to ∼1000 bp (Diagenode Bioruptor) before amplification via an Sμ-specific biotinylated primer (5′/5-BiosG/CAGACCTGGGAATGTATGGT-3′). The biotinylated products were isolated with magnetic beads, ligated to an adaptor, and amplified with a nesting primer 5′-CACACAAAGACTCTGGACCTC-3′. Endonuclease AflII was used before the nesting PCR to remove the germline sequence. The libraries were sequenced on Illumina Miseq (150 × 150 pair-ended platform). Because all mice used in this study are of 129 background, we replaced the IgH switch region (from JH4 to the last Cα exon, chr12 114, 494, 415–114, 666, 816) of the C57/BL6-based mm9 with the corresponding region from 129sv (GenBank accession no. ⁠AJ851868.3⁠) (1415966–1592715) to generate the mm9sr (switch region replacement) genome. The published HTGTS pipeline was used for mapping and filtration (19, 52). The best-path searching algorithm [related to YAHA (54)] has been deployed to identify optimal sequence alignments from Bowtie2-reported top alignments (alignment score >50). Mispriming events, germline (unmodified) sequence, sequential joints, and duplicated reads were removed. Duplication was defined by bait and prey alignment coordinates both within 2 nt of another read’s bait and prey alignments. Reads unequivocally mapped to an individual S-region were recovered based on mappability filter (please see Ref. 52 for simulation and details on plotting). MHs are defined as sequences that can be assigned to both bait and prey. Insertions are regions that map to neither bait nor prey. The data were then plotted in Excel using a Visual Basic tool (51, 53, 55). Sequence data have been deposited to the Gene Expression Omnibus under accession number GSE162568.

The R3057X (CGT → CAT) mutation (corresponding to the R3047X at the FATC domain of human ATM; Fig. 1A, 1B) was introduced into the endogenous Atm locus via homologous targeting (Supplemental Fig. 1A). The successfully targeted embryonic stem cell clones were identified by Southern blotting (Supplemental Fig. 1B). The mutation was confirmed by Sanger sequencing (Supplemental Fig. 1C). Two independently targeted embryonic stem cell clones were injected for germline transmission and analyzed together. Atm⁠^+/RX⁠ and Atm⁠^RX/RX⁠ mice were born at the expected Mendelian ratio (Fig. 1C). In contrast to the smaller size of the Atm⁠^−/−⁠ mice, Atm⁠^RX/RX⁠ mice were of normal weight and had a similar rate of weight gain as Atm⁠^+/+⁠ and Atm⁠^+/RX⁠ littermates (Supplemental Fig. 1D, 1E). Unexpectedly, Western blotting analyses of the splenic B cells from Atm⁠^RX/RX⁠ mice using a mAb against aa 1967–1988 of human ATM (56) showed that ATM-RX protein level reduced ∼10-fold from the ATM–wild-type (WT) control (Fig. 1D). The severe reduction of the ATM-RX protein is mainly due to the loss of protein stability because the Atm-RX mRNA level is comparable to that of ATM-WT in Atm⁠^+/RX⁠ B cells (Fig. 1E). R3057 is encoded in the last exon of the Atm gene (Supplemental Fig. 1A), and R3057X should not be a candidate for nonsense-mediated decay (57). In agreement with the much-reduced ATM-RX protein level and the essential role of ATM in meiotic recombination (23–25), male Atm⁠^RX/RX ⁠mice have empty epididymis and lack mature spermatids in testes (Fig. 1F). Taken together, these results suggest that the residual ATM-RX protein and activity can support normal weight but not meiosis (26).

ATM kinase activity plays an important role in lymphocyte development (58). So next, we analyzed the lymphocyte development in young (5–7 wk) Atm⁠^RX/RX⁠ mice. The total thymocyte count from the Atm⁠^RX/RX⁠ mice was not significantly lower than that of Atm⁠^+/+⁠ controls, although significantly reduced in Atm⁠^−/− ⁠mice (Supplemental Fig. 2A). Nevertheless, Atm⁠^RX/RX⁠ thymocytes showed a partial blockade at the CD4⁺CD8⁺ double-positive to CD4⁺ or CD8⁺ single-positive T cell transition (Fig. 2A, 2B, Supplemental Fig. 2B, 2C) and have correspondingly lower levels of surface TCRβ (Fig. 2A, 2C). The extent of the stage-specific T cell development defects in Atm⁠^RX/RX⁠ mice is similar to that of Atm⁠^−/−⁠ mice, suggesting that the residual ATM kinase activity in Atm⁠^RX/RX⁠ mice, if present, is not sufficient to support efficient TCRα rearrangement (10). The normal total thymocytes number coupled with the stage-specific developmental defects seen in Atm⁠^RX/RX⁠ mice is unique and suggests that there might be a TCR-independent component that underlies the reduced thymocyte number in Atm⁠^−/−⁠ mice, potentially rooted in early progenitor cells (10, 29). There are no measurable defects in the cellularity or the composition of B cells (Fig. 2D, 2E, Supplemental Fig. 2D) or the frequency of myeloid cells (CD11b⁠⁺⁠Gr1⁠⁺⁠) in either Atm⁠^RX/RX⁠ or Atm⁠^−/− ⁠mice (Supplemental Fig. 2E, 2F). Mechanistically, Atm⁠^RX/RX⁠ lymphocytes accumulate abnormally high levels of HJs at TCRβ locus in T cells (Fig. 2F) and Igκ in B cells (Fig. 2G, Supplemental Fig. 2G), like documented in Atm⁠^−/−⁠ mice (12).

Purified Atm⁠^RX/RX⁠ as well as Atm⁠^−/− ⁠splenic B cells underwent IgG1 switching in the presence of anti-CD40 and IL-4 at ∼50% reduced frequency as the WT control (Fig. 3A, 3B). CTV is a cell surface dye, which can mark cell division by reduced surface levels. Analyses of IgG1 switching levels by CTV in Atm⁠^RX/RX⁠ and Atm⁠^−/− ⁠B lymphocytes confirmed that reduced CSR cannot be explained by proliferation defects (Fig. 3C, 3D). Using a telomere-specific FISH probe that enhances the sensitivity to detect IgH locus breaks at the telomeric end of Chr12 (22), we found that the genomic instability of activating Atm⁠^RX/RX⁠ B cells is significantly lower than that of Atm⁠^−/− ⁠B cells (Fig. 3E). Telomeric FISH can identify two types of DNA breaks: chromosome breaks involving both sister chromatids and chromatid breaks involving one of the two sister chromatids (Fig. 3F). Whereas chromatid breaks occur during or after DNA replication, chromosome breaks might occur in G1 cells (22). Because CSR is initiated in the G1 phase of the cell cycle, CSR-associated IgH breaks make up the majority of chromosome breaks in activating Atm⁠^−/−⁠ B cells (22). Most instabilities in Atm⁠^RX/RX⁠ B cells were chromosome breaks (Fig. 3G). Proportional reduction of both chromosome and chromatid breaks in the metaphases from Atm⁠^RX/RX⁠ B cells together with the severe CSR defects suggests that the residual ATM kinase activity in Atm⁠^RX/RX⁠ cells might activate a cell cycle checkpoint to prevent the damaged cells from entering metaphase without restoring CSR.

To determine how ATM-RX protein affects the repair phase of CSR, we analyzed thousands of CSR and internal deletion junctions from activated Atm⁠^RX/RX⁠, Atm⁠^+/+⁠, and Atm⁠^−/− ⁠control B cells using the HTGTS technology developed by the laboratory of Dr. Frederick Alt (51, 52, 59, 60). Specifically, we placed the linear amplification primer near the 5′ Sμ region (bait breaks) and isolated Sμ-Sμ junctions corresponding to internal deletion as well as Sμ-Sγ1 and Sμ-Sε junctions corresponding to IgG1 and IgE CSR, respectively (Fig. 4A). The translocation partners of the initial Sμ breaks are referred to as the preys, as they were called in the original HTGTS publications (50, 52). In addition to the location of the preys, HTGTS identified the orientation of junctions (Watson or Crick strands) based on the orientation of the junctional sequences. We defined the prey orientation as plus if it aligned from centromere to telomere starting from the junction site (as blue arrowheads in Fig. 4A) and as minus if it aligned from telomere to centromere starting from the junction site (as red arrowheads in Fig. 4A). Because the IgH locus resides at the minus strand of murine chromosome 12, the productive Sμ (telomeric) to Sγ1 or Sε (centromeric) joints are minus. The plus orientation joints can be generated by true inversion or could be intersister or interhomologue translocation between the Sμ breaks from one chromosome to the Sγ1 breaks on another chromosome 12 (sister or homologue). This is due to the limitation that the digital genome is haploid whereas the physical genome is diploid. We are not able to distinguish the two homologous chromosomes in this pure background strain. As previously described for Atm⁠^−/− ⁠cells (19), the IgH junctions recovered from Atm⁠^RX/RX ⁠B cells show an increased proportion of plus-oriented joints, suggesting increased interchromosomal translocations (Fig. 4B). This trend of increased IgH⁺-oriented junctions holds for each switch region (Sμ, Sγ1, and Sε) analyzed, as well (Supplemental Fig. 3A). In Atm⁠^+/+⁠ B cells, ∼30% of the IgH preys (both orientations) reside in each of the three switch regions analyzed: Sμ, Sγ1, and Sε (Fig. 4C). Loss of ATM and the R3057X mutation preferentially reduced Sμ-Sε junctions, while correspondingly increasing Sμ-Sμ internal deletions (Fig. 4C). Similar trends were also noted in other classical NHEJ–deficient B cells (19, 52, 51, 61). In adult B cells, the majority of the IgE switching is achieved through sequential switching to Sγ1 first, then to Sε (62). This need for two end-ligation events might explain the hypersensitivity of IgE switching to DNA repair defects. Finally, CSR junctions recovered from ATM-deficient cells have increased usage of MH at the junctions. Junctions recovered from Atm⁠^−/−⁠ and Atm⁠^RX/RX ⁠B cells show similar skews toward short MHs in IgH together or in each switch region analyzed (Fig. 4D, Supplemental Fig. 3B, 3C). The number of Sμ-Sε junctions was too low for MH analyses in the ATM-deficient B cells. Together, the increased translocation (plus-oriented junction), preferential loss of Sμ-Sε junctions, and the increased MH usage in the HTGTS analyses of Atm⁠^RX/RX ⁠B cells indicate that the R3057X mutation impairs the DNA repair function of ATM during CSR, similar to the loss of ATM.

Atm⁠^−/−⁠ mice routinely succumb to thymic lymphomas by 4 mo of age (10). Despite similar T cell development and CSR defects, Atm⁠^RX/RX⁠ mice developed thymic lymphomas nearly 140 d later than Atm⁠^−/−⁠ mice (t⁠_1/2⁠ = 115 d for Atm⁠^−/−⁠ versus t⁠_1/2⁠ = 255 d for Atm⁠^RX/RX⁠, p < 0.002, Mantel-Cox test) (Fig. 5A). Because Atm⁠^+/+ ⁠mice in the pure 129sv background rarely develop spontaneous tumors by 365 d of age (1 y) (63, 64), these findings suggest that ATM-RX protein has a reduced tumor suppression function, but some residual activity of ATM-RX can effectively delay lymphomagenesis. All Atm⁠^RX/RX⁠ mice analyzed succumbed to thymic lymphomas formed by immature (surface TCRβ low) T cells (Fig. 5B), like previously characterized Atm⁠^−/−⁠ mice (10, 29). Together with the lack of major genomic instability in the metaphases from activated Atm⁠^RX/RX⁠ B cells, these results suggest that the residual ATM-RX protein might have checkpoint functions that delay lymphomas and prevent cells with broken chromosomes from entering mitosis.

To understand whether Atm⁠^RX/RX⁠ cells have residual ATM kinase activity, we measured irradiation (IR)-induced phosphorylation of ATM substrates in Atm⁠^RX/RX⁠ B cells together with Atm⁠^−/−⁠ and Atm⁠^+/+⁠ controls. IR-induced phosphorylation of Kap1, a relatively specific substrate of ATM (65), is much reduced in Atm⁠^RX/RX⁠ cells, comparable to Atm⁠^−/−⁠ control (Fig. 6A–D). The residual radiation-induced Kap1 phosphorylation in Atm⁠^RX/RX⁠ and Atm⁠^−/−⁠ cells can be completely eliminated by DNA-PKcs inhibitor (NU7441, 15 μM) (Fig. 6C, Supplemental Fig. 4A, 4C). ATM inhibitor (KU55933 15 μM) also significantly reduced the residual KAP1 phosphorylation in Atm⁠^RX/RX⁠ (Fig. 6C, Supplemental Fig. 4A). But because KU55933 (15 μM) also consistently reduced Kap1 phosphorylation in Atm⁠^−/−⁠ cells (Fig. 6C, Supplemental Fig. 4A, 4C), we propose that this might be due to cross inhibition of DNA-PKcs (IC₅₀ 2.5 μM in the cell-free assay). Long exposure of the Western blot saturates the signal and made it less obvious (Supplemental Fig. 4C). The residual phosphorylation of two other ATM substrates (Chk2 and H2AX) is also comparable between Atm⁠^RX/RX⁠ and Atm⁠^−/−⁠ cells (Fig. 6A–C, Supplemental Fig. 4A–C). We reason that Western blotting might not be sensitive enough to measure residual ATM kinase activity. To directly measure DNA damage-induced G2/M checkpoint activation, we treated Atm⁠^RX/RX⁠ primary B cells with the radiation mimetic agent NCS (100 ng/ml) for 1 h, performed BrdU labeling to identify G2/M cells (BrdU-4N cells), and stained for Histone H3 Ser10 phosphorylation (pH3⁠⁺⁠), a mitotic marker (Fig. 6E). The frequency of G2/M cells and the percentage of mitotic pH3⁠⁺⁠ cells among the G2/M cell population are comparable in untreated Atm⁠^RX/RX⁠, Atm⁠^−/−⁠, and Atm⁠^+/+⁠ cells (Fig. 6E, 6F). As expected, based on literature, loss of ATM severely compromised the radiation-induced G2/M checkpoint (Fig. 6F, 6G). Atm⁠^RX/RX⁠ cells show significantly more robust G2 arrest than Atm⁠^−/−⁠ cells (Fig. 6F, 6G), suggesting residual checkpoint activities, which might contribute to the delayed lymphomas and lack of measurable instability in metaphase preps. Next, we measured ROS-induced activation of ATM by treating cells with TBH, a relatively stable ROS inducer. We observed no ATM-dependent residual phosphorylation of Kap1, Chk2, or H2AX in Atm⁠^RX/RX⁠ cells, consistent with the lack of ROS-induced activation of purified human ATM-R3047X protein (31) (Fig. 6H). Together, these results suggest that the residual Atm-RX protein in the Atm⁠^RX/RX⁠ cells supports some DNA damage-induced G2/M checkpoints but cannot be activated by ROS.

Using knock-in mouse models, we found that a truncation mutation within the FATC domain (R3057X in mouse, corresponding to R3047X in human) destabilizes ATM protein, indicating a critical role of the FATC domain in ATM protein folding and stability. As such, Atm⁠^RX/RX⁠ mice display infertility, stage-specific T cell development defects, and CSR defects that are characteristic of ATM-null mice. However, in comparison with Atm⁠^−/−⁠ mice, Atm⁠^RX/RX⁠ mice have no growth retardation, nearly no thymus atrophy, substantially lower cytogenetic instability in metaphase, and much-delayed lymphomagenesis. These differences in phenotypes can be in part attributed to the residual DNA damage-induced checkpoint function of the ATM-RX protein. In this context, Atm⁠^RX/RX⁠ cells show no sign of ROS-induced activation of ATM, like previously described for A-T cells with compound ATM mutations including R3047X (31, 66). Unfortunately, the severely reduced ATM-RX protein levels abrogate meiosis and lymphocyte development associated with DSB repairs, preventing us from analyzing the physiological role of ROS-dependent ATM activation. In this context, the patients carrying R3047X mutation also show 5- to 10-fold decrease of ATM protein levels, a mild form of A-T, primary immunodeficiency, and no lymphomas at age 20 y (66). In the future, other separation of function mutations, if possible, would be needed to address the role of ROS-induced ATM activation in vivo.

A-T is a heterogeneous disease. Whereas all A-T patients develop different degrees of ataxia, ∼70% of A-T patients show immunodeficiency (3, 4), and 25% of patients develop lymphoid malignancies (28). The heterogeneity in A-T patients could be caused by the pleiotropic function of ATM kinase as a stress response kinase and can also be influenced by the diverse genetic backgrounds and life experiences of individual patients. Inbred mouse models with knock-in mutations of ATM provide a valuable tool to address the heterogeneities associated with A-T. In the past three decades, nine germline mouse models of A-T have been made, including five null alleles (5, 7, 8, 67, 68), two kinase-dead alleles [one knock-in (50) and one with BAC-transgene (69)], two autophosphorylation site alanine substitutions (both BAC transgenic) (70, 71), one in-frame deletion (deltaSRI) (72), the R3047X described in this article, and the R3008H generated by us in another study (73). Among them, the autophosphorylation site mutation does not have any measurable impact on ATM kinase activity (70, 71). The phosphorylation site mutant mice are virtually identical to normal mice. The two mouse models expressing kinase-dead ATM result in embryonic lethality (49, 69), explaining why nearly all live-born A-T patients have truncating mutations with nearly no ATM protein expression. The deltaSRI, R3047X, and R3008H all have substantially reduced kinase activities, manifested by infertility and T cell development and CSR defects. Meanwhile, all three models have delayed lymphomagenesis, suggesting that the amount of ATM needed for its tumor suppressor function is much lower than that needed to support V(D)J or CSR recombination. Given the residual checkpoint function of ATM-R3057X detected in this study, the data suggest that ATM-mediated phosphorylation of checkpoint targets, in particular G2/M targets CHK2 and CHK1, might be less vulnerable to ATM activity loss. It remains unclear whether this substrate selectively reflects a direct role of ATM on its substrates (e.g., structure or activation difference) or reflects the signal amplification in the checkpoint pathway through mediator kinases (e.g., Chk1, Chk2, Wee1, etc.). Nevertheless, this finding highlights the important role of ATM-mediated checkpoint function in tumor suppression. In addition to lymphomagenesis, growth retardation and pan-lymphocytopenia can also be rescued with a minimal amount of ATM. This might be a feature to all hypomorphic ATM mutations, including but not limited to R3047X. Consistent with these observations, patients with residual ATM activity show mild phenotype, underlying the 70% immunodeficiency rate versus 25% lymphoma rate in A-T patients. Systematic characterization of ATM missense mutations might be necessary to understand the full spectrum of A-T and how to prevent lethal lymphomas by restoring minimal ATM kinase activity that might not be visible by Western blotting but detectable in cell cycle assays.

Finally, we note that ATM-specific inhibitor KU55933 used at 15 μM has unexpected cross activity toward DNA-PKcs in murine cells. This effect is more obvious with low exposure, when the signals are not saturated. Lower dose of ATM kinase inhibitor, such as 7.5 μM, should be considered for future experiments. We note the current experiments are conducted in murine B cells; whether a similar dose also has off-target effect in other cell types and in human cells remains elusive.

We thank all current and past members of the Zha laboratory for their helpful discussion. We thank Dr. Tanya Paull for discussing the R3047X allele and ATM biochemical analyses. The authors apologize to colleagues whose original work could not be cited because of space limitations and were covered by reviews instead.

The authors have no financial conflicts of interest.