Homologous chromosomes make contact at the sites of double-strand breaks in genes in somatic G₀/G₁-phase human cells

Abstract

Double-strand DNA breaks (DSBs) are continuously induced in cells by endogenously generated free radicals and exogenous genotoxic agents such as ionizing radiation. DSBs activate the kinase activity in sensor proteins such as ATM and DNA-PK, initiating a complex DNA damage response that coordinates various DNA repair pathways to restore genomic integrity. In this study, we report the unexpected finding that homologous chromosomes contact each other at the sites of DSBs induced by either radiation or the endonuclease I-PpoI in human somatic cells. Contact involves short segments of homologous chromosomes and is centered on a DSB in active genes but does not occur at I-PpoI sites in intergenic DNA. I-PpoI-induced contact between homologous genes is abrogated by the transcriptional inhibitors actinomycin D and α-amanitin and requires the kinase activity of ATM but not DNA-PK. Our findings provide documentation of a common transcription-related and ATM kinase-dependent mechanism that induces contact between allelic regions of homologous chromosomes at sites of DSBs in human somatic cells.

Double-strand DNA breaks (DSBs) are continuously induced in eukaryotic cells by the free radicals generated by endogenous metabolic processes and by exposure to environmental genotoxic agents such as ionizing radiation (IR) and are considered to be the most dangerous DNA lesions (1–3). DSBs induce a complex DNA damage response that activates DNA repair pathways and cell cycle checkpoints and induces chromatin remodeling and apoptosis. ATM and DNA-PK, two phosphatidylinositol 3-kinase–like kinases, are primary mediators of the DNA damage response to DSBs. Although ATM phosphorylates hundreds of substrates that are believed to regulate multiple DNA repair pathways including DSB repair by homologous recombination (HR) (4, 5), DNA-PK primarily regulates a limited group of effectors that mediate DSB repair by nonhomologous end-joining (NHEJ) (3, 6). Misrepair of DSBs can result in the intrachromosomal and interchromosomal rearrangements that generate oncogenic gene fusions. Previously, we (7, 8) and others (9, 10) have shown that cancer-specific chromosomal rearrangements commonly arise as a result of exchange between chromosomal loci that are located in close spatial proximity at the time of DSB formation.

While exploring the role of nuclear architecture and gene topology in the generation of chromosomal rearrangements in human somatic cells, we observed an unexpected finding that homologous chromosomes frequently contact each other at the sites of DSBs induced in G₀/G₁ cells by either IR or the restriction enzyme I-PpoI. Further characterization demonstrated that this contact between homologous chromosomes is initiated by DSBs in genes but not by DSBs in intergenic DNA, requires active gene transcription, and depends on ATM kinase activity.

Results

Homologous Chromosomes Form Arm-Specific Contact Spontaneously and After Exposure to Ionizing Radiation.

We used four-color 3D-FISH and confocal microscopy (11, 12) to analyze the pattern and frequency of interaction between homologous chromosomes in untreated and irradiated primary cultured human epithelial thyroid cells and fibroblasts. Cells were highly enriched in G₀/G₁ by plating at high density in the absence of serum and other growth factors (84% of thyroid cells and 91% of fibroblasts were in G₀/G₁ phase by FACS). For each chromosome, arm-specific paints were used to visualize the p and q arm separately, and a centromeric probe was used to monitor the location and duplication of centromeres (Fig. 1A). Contact between homologous chromosomes was detected by using intensity-based image segmentation analysis (Fig. S1). S/G₂-phase cells, which were identified by the presence of centromere duplication that occurs in S phase (13), were excluded from further analysis. The accuracy of the exclusion criterion was confirmed by combining PCNA (marker of S phase) or cyclin A (marker of S and G₂ phase) (14) immunostaining and FISH analysis in selected experiments (Fig. 1 E and F).

Fig. 1.

Arm-specific contact between homologous chromosomes in G₀/G₁-phase human thyroid cells (HT) and fibroblasts (HF). (A) Color scheme of probes for FISH. (B) q arm contact. (C) p arm contact. (D) p:p and q:q contact. (E) G₀/G₁ nuclei on the right show either p:p and q:q contact (upper) or p arm contact (lower), whereas S/G₂-phase nucleus on the left is cyclin A positive (orange). (F) S-phase nucleus in Upper shows a duplicated centromere (arrows) and is positive for PCNA (orange), whereas G₀/G₁ nucleus in Lower shows q arm contact. (G and H) Frequency of contact between homologous chromosomes: Gray bars represent p:p or/and q:q contact in untreated cells; black bars represent p:p or/and q:q contact after IR; white bars represent contact between opposite arms (p:q) of homologous chromosomes. Data based on 300 cells analyzed from each of two donors shown as mean ± SEM. The differences between frequency of contact before and after IR were highly statistically significant (P < 0.0002 for all chromosomes, two sample t test). (Scale bars: 5 μm.)

Arm-specific contact between homologous chromosomes was readily identified for all six chromosomes studied in epithelial cells and all three chromosomes studied in fibroblasts. The common patterns of contact were between two homologous q arms (Fig. 1B), between two homologous p arms (Fig. 1C), and simultaneous contact between two p arms and two q arms of homologous chromosomes, i.e., p:p and q:q contact (Fig. 1D and Fig. S2).

Analysis performed on primary thyroid epithelial cells from two donors revealed that 15–26% (mean, 19%) of homologous chromosomes had arm-specific contact in untreated cells (Fig. 1G). Fifteen minutes after exposure to 5-Gy IR, the frequency of arm-specific contact between homologous chromosomes doubled to 28–45% (mean, 38%). Similar findings were observed in untreated and irradiated human fibroblasts (Fig. 1H). The frequency of contact between homologous chromosomes was ∼20-fold lower than the predicted frequency of DSBs induced by 5 Gy of IR (150–200 DSBs per cell; 6–8 DSBs per average-size chromosome), indicating that if contact is due to formation of DSBs, a fraction of induced DSBs are participating in contact at any given time. Contact between opposite arms of homologous chromosomes (i.e., p:q contact) was observed in only ∼1% of cells (range 0.6–1.6%) for all chromosomes, whether cells were irradiated or not irradiated, which likely represents a background level of random contact.

Homologous Chromosomes Form Contacts at the Sites of DSBs Induced by Homing Endonuclease I-PpoI in Gene Regions but Not Outside of Genes.

To determine whether contact between the corresponding arms of homologous chromosomes was induced by a DSB, and to find whether it involves the same allelic region on both homologs, we induced DSBs at specific chromosomal sites by expressing the homing endonuclease I-PpoI (15, 16) in the TPC-1 human thyroid cancer cell line (17). To introduce the I-PpoI gene, cells were transduced with the HA-ER-I-PpoI retrovirus (gift of M. Kastan Duke Cancer Institute, Durham, NC) by using either a single infection or three sequential infections. To induce DSBs, the infected cells were treated with 4-hydroxytamoxifen (4-OHT), which causes I-PpoI to enter the nucleus (16, 18). I-PpoI induction for 6 h produced multiple γH2AX foci in ∼55% and ∼75% of cells after single and triple infections, respectively (Fig. 2A).

Fig. 2.

Contact between homologous chromosomes at the sites of DSBs induced by endonuclease I-PpoI in genes. (A) Examples of multiple γH2AX foci (orange) formed 6 h after I-PpoI induction in human TPC-1 cells, indicating the efficiency of I-PpoI in generating DSBs in these cells. (B) Time course and percentage of DNA cleavage at four I-PpoI sites detected by real-time qPCR. (C) FISH visualization of the I-PpoI site in the DAB1 gene showing contact between DAB1 signals (orange) in one cell (arrow), which is in G₀/G₁ phase because it lacks cyclin A staining (green), compared with a cell on the left in S/G₂ phase. (D) Frequency of contact between homologous regions with an I-PpoI site located in gene regions and intergenic regions and between homologous regions lacking an I-PpoI site. Data based on 1,000 cells scored in each of two experiments shown as mean ± SEM. **P < 0.001, *P < 0.01, two sample t test. (Scale bars: 5 μm.)

Using site-specific DNA probes, we analyzed six chromosomal regions, all diploid in TPC-1 cells (Table S1). Two of the regions have an I-PpoI site located within genes (DAB1 on 1p and GRIP1 on 12q). Two other studied regions (2qIG and 5qIG) are intergenic loci containing an I-PpoI site. Two regions that lack an I-PpoI site (16pNC and 16qNC) were also analyzed. Quantitative PCR (qPCR) analysis revealed that between 9 and 16% of the analyzed target sites were cut at 6 h after I-PpoI induction (Fig. 2B), as expected based on previous observations showing that in human cells I-PpoI cleaves approximately 10% of the 200–300 target sites (16, 18).

FISH analysis with site-specific probes combined with cyclin A staining revealed that, in G₀/G₁-phase cells, I-PpoI cleavage induced contact between gene regions on homologous chromosomes, but did not induce contact between homologous intergenic regions or regions that do not contain an I-PpoI site (Fig. 2 C and D). For both genes (DAB1 and GRIP1), the proportion of cells showing contact at the I-PpoI site corresponded to the proportion of cells with expected cleavage at that site, estimated based on the known efficiency of cutting at each site and number of infected cells after single and triple infection. By contrast, the two intergenic regions containing an I-PpoI site exhibited no increase in allelic contact above the level observed in cells before I-PpoI entering the nucleus and in DNA regions lacking an I-PpoI site. Additionally, the frequency of contact between heterologous gene pairs, i.e., between DAB1 and GRIP1 genes, was low and unchanged when studied before (0.93 ± 0.24%) and after (0.90 ± 0.21%) DSB induction by I-PpoI.

To confirm that contact between homologous genes is a general phenomenon and not restricted to specific chromosomes, we expanded the analysis to include three additional genes that contain endogenous I-PpoI cleavage sites, all located on different chromosomes: ARID5B (10q), ERC2 (3q), and PDE1A (2q) (Table S2). Induction of I-PpoI for 6 h in TPC-1 cells stably transfected with the HA-ER-I-PpoI retrovirus resulted in the increased contact among all three pairs of homologous genes (Fig. S3).

Contact Between Homologous Chromosomes Involves a Relatively Small Region Flanking the Break.

Experiments illustrated in Fig. 1 showed that when contact involved single arms of homologous chromosomes, most of that arm and all of the other arm were apart. This finding suggested that contact between homologous chromosomes involved only a limited section of the chromosomes. To estimate the length of the contact between homologous chromosomes, we used an array of DNA probes extending up to 3 Mb centromeric and telomeric from the I-PpoI site in the DAB1 gene (Table S3). In a series of high-resolution four-color 3D-FISH analyses (Fig. 3 A and B), we observed nuclei in which probes spanning the I-PpoI site were colocalized, whereas probes located 300 kb and further from the I-PpoI site were at a distance (Fig. 3C). These data demonstrated that contact formed at the site of a DSB can involve less than 300 kb on either side of the break. In addition, in this experiment, we observed nuclei that had p arm territories of both chromosome homologs positioned near each other, and aligned in the same direction, but not in contact (Fig. 3D). Such mutual positioning of homologous arms of chromosome 1p was seen more often than contact at the site of I-PpoI cleavage (∼5:1) and may represent stages of approximation and alignment of homologous regions before the initiation of contact at a DSB.

Fig. 3.

Contact between homologous chromosomes is centered on a DSB and extends less than 300 Kb telomeric and centromeric from the break. (A) Scheme of the FISH probes used to map the extent of contact between chromosome 1 homologs at the I-PpoI site in the DAB1 gene region 6 h after I-PpoI induction in TPC-1 cells. (B) Representative nucleus with two 1p territories located at a distance. (C) Representative nucleus with full colocalization of the I-PpoI site probes (yellow), whereas flanking regions (red and aqua) remain at a distance. (D) Representative nucleus with 1p arm territories (green) near each other and aligned with signals in the same order. (Scale bars: 5 μm.)

Frequency of Contact Is Higher After I-PpoI Cleavage of Multiple Sites.

Experiments with I-PpoI using chromosome 1p paint, illustrated in Fig. 3, also allowed us to evaluate contact between any regions of homologous 1p arms after the induction of I-PpoI for 6 h. At least four I-PpoI cleavage sites are located on 1p, including three sites in gene regions and one in intergenic DNA (Table S4). The frequency of contact between homologous 1p arms was found to be 24.23 ± 1.21%, significantly higher than that seen at the DAB1 site alone (7.10 ± 0.57%). This finding demonstrates that induction of multiple I-PpoI cleavage sites on one chromosome arm results in the higher frequency of homologous arm contact, which becomes comparable to that seen after 5 Gy of IR.

Break-Induced Contact Between Homologous Chromosomes Requires Active Gene Transcription.

Because all five genes that exhibited contact between homologous gene pairs upon induction of I-PpoI cleavage are expressed in TPC-1 cells (Fig. 4A), we tested whether active transcription is required for contact. Cells were treated with actinomycin D, which blocks transcription elongation by all three RNA polymerases (19) and with α-amanitin, which inhibits RNA polymerases II and III (20). Treatment with either of the transcription inhibitors resulted in abrogation of the contact between each of five homologous gene pairs (Fig. 4B).

Fig. 4.

Contact between homologous genes upon DSB induction requires active transcription. (A) Expression of five studied genes in TPC-1 cells relative to expression of GAPDH. NC, a gene known to be not expressed in TPC-1 cells. (B) Frequency of contact between homologous gene regions after I-PpoI induction and influence of transcription inhibition by actinomycin D and α-amanitin for 6 h. Data based on 1,000 cells scored in each of two experiments depicted as mean ± SEM. The difference between contact before and after transcription inhibition with either agent was highly statistically significant (P < 0.0001 for all genes, two sample t test).

Break-Induced Contact Between Homologous Chromosomes Requires Function of ATM, but Not DNA-PK.

We proceeded to investigate whether the site-specific contact between homologous chromosomes depends on the kinase activities of either ATM or DNA-PK, which are required for DSB recognition and repair by various pathways including HR (ATM) and classical NHEJ (DNA-PK) (6, 21). We observed that inhibition of the kinase activity of ATM using the ATM inhibitor KU55933 abrogated the I-PpoI–induced contact between gene regions in TPC-1 cells, whereas the DNA-PK inhibitor Nu7441 had no effect (Fig. 5 A and B). To confirm the ATM kinase dependence of contact, we knocked down ATM expression by using a selective shRNA and showed that knockdown abrogated contact (Fig. 5 C and D). These results indicate that contact between homologous gene regions at the site of a DSB requires functional ATM kinase.

Fig. 5.

Contact between homologous genes upon DSB induction requires the kinase activity of ATM but not that of DNA-PK. (A) Western blot showing selective inhibition of ATM autophosphorylation by KU55933 (Upper) and of selective inhibition DNA-PK autophosphorylation by NU7441 (Lower) in TPC-1 cells. (B) Contact frequency between gene regions and intergenic regions after I-PpoI induction with and without KU55933 and NU7441. Data based on 1,000 cells scored in each of two experiments shown as mean ± SEM. *P < 0.0001, **P > 0.55, two sample t test. (C) Western blot showing sequence specific knockdown of ATM protein by an shRNA that targets the ATM mRNA but not a scrambled shRNA sequence. (D) Frequency of contact between homologous gene regions after I-PpoI induction and influence of ATM knockdown by using shRNA. Data for DAB1 and GRIP1 from 1,000 cells scored in each of two experiments are combined and shown as mean ± SEM. *P < 0.0001, two sample t test.

Discussion

Homologous chromosome pairing is known to occur during meiosis, where it is essential for high levels of recombination and the correct segregation of homologs into haploid germ cells (22). In this study, we used high-resolution confocal microscopy and chromosome arm paints or gene-specific probes to demonstrate that contact between allelic regions of homologous chromosomes frequently occurs in somatic human cells and can be induced by ionizing radiation or DNA cleavage by a site-specific endonuclease. The contact was observed between homologous chromosome arms and homologous genes on multiple different chromosomes and, therefore, appears to represent a general phenomenon occurring in human somatic cells.

However, in contrast to homologous chromosome pairing in meiosis, which occurs by multiple interstitial interactions along the entire length of homologs, contact in somatic cells observed in this study typically involved a limited section of the chromosome, centering on a site of DSB. A more extensive pairing was rarely seen and likely reflected juxtaposition of homologs in response to several breaks occurring along both arms of the chromosome.

Importantly, the I-PpoI–induced contact was restricted to DSBs in coding regions, and we were unable to observe contact at cleavage sites in intergenic DNA. Moreover, contact between homologous active genes was reversed after RNA polymerase inhibition and arrest of RNA synthesis by actinomycin D and α-amanitin. Interestingly, it has been shown that recruitment of some DNA repair factors such as 53BP1 to the sites of DNA damage requires direct interaction with RNA molecules, and treatment with RNase results in dissociation of 53BP1 from the IR-induced nuclear foci (23). Moreover, homologous chromosomes that pair in meiosis are transcriptionally active, and it has been proposed that RNA:RNA base pairing involving sense and antisense transcripts tethered through their polymerases may be a driver of homolog pairing in meiosis (24). In light of these reports, the results of our study raise at least a theoretical possibility that sequence-specific RNA molecules may be involved in homology search and/or establishing contact between allelic gene loci in somatic human cells.

Although our findings that contact centered on a break and required functional ATM kinase suggest that the contact may have a role in repair of DSBs, the DNA repair pathway involved remains unknown. ATM is known to regulate multiple DNA repair pathways including HR (25, 26). It is also well established that nucleotide excision repair proceeds more actively on the transcribed strands of expressed genes, removing the damage by excising a short DNA fragment containing DNA lesion and filling the gap by using the undamaged strand as a template (27, 28). The results of our study raise the possibility that a specialized DSB repair pathway operates on DSBs in expressed genes and, conceivably, uses a homologous chromosome as a repair template. However, it is also possible that contact between homologous chromosomes at the site of DSBs does not directly provide a template for repair of broken DNA molecule but contributes to other aspects of DSB recognition and repair. Another possibility is that contact is not associated with repair at all and is a product of other gene functions, such as of sharing transcription factories, which has been suggested to exist between allelic genes on homologous chromosomes (24) and between active heterologous genes (29). This possibility is, however, unlikely, because it would be expected to result in pairing along the full length of homologous chromosomes as seen in meiosis, would not depend on the induction of DSBs and ATM kinase, or would be expected to bring together active heterologous genes, such as DAB1 and GRIP1, which was not seen in this study.

In summary, our findings demonstrate that homologous gene regions in human somatic cells make contact at the sites of DSBs. As such, we also demonstrate that cells possess a mechanism that uses sequence homology to bring otherwise distant DNA molecules into contact. Our data suggest that this process may involve RNA molecules or active gene transcription, although the details of this mechanism remain unknown. Nevertheless, the data described here provide documentation of a common transcription-related and ATM kinase-dependent mechanism that induces contact between homologous chromosomes at sites of DSBs in human somatic cells.

Materials and Methods

Cell Culture and Irradiation.

Primary cultures of human thyroid cells were established from surgically removed thyroid glands as described (7) by following the University of Pittsburgh Institutional Review Board approval. Cells were cultured in RPMI medium 1640 with 15% FBS and 10 mU/mL of thyroid stimulation hormone (Sigma) for 1–2 wk. Primary fibroblast cultures were established from fragments of perithyroidal fibrous tissue. Cells were cultured for 3–4 wk in DMEM supplied with 10% FBS, nonessential amino acids, l-glutamine, and 4 ng/mL basic fibroblast growth factor (Sigma). To enrich cell populations for cells that are in the G₀/G₁ phase of the cell cycle, cells were plated at high density and cultured without FBS and growth factors for 72 h. When irradiation was used to induce DNA breaks, a single dose of γ-irradiation from a cesium-137 source was delivered at a dose rate of 301.66 cGy/min. Induction of DSBs was monitored by immunofluorescence with anti-γH2AX antibody (05–636; Upstate Biotechnology) (30).

Three-Dimensional Cell Fixation, FISH, and Immuno-FISH.

Cells plated in chamber slides were subjected to 3D fixation in 4% paraformaldehyde followed by repeated freeze-thaw cycles in liquid nitrogen (11). Centromeric enumeration probes prelabeled with either SpectrumAqua or SpectrumOrange (pseudocolored in blue for illustration) (Abbott Laboratories) and prelabeled arm-specific chromosome painting probes (p arm, FITC; q arm, Texas red) (Metasystems) were used. BAC clones for region-specific DNA probes (BAC/PAC Resources, Children's Hospital, Oakland, CA) were labeled by nick translation using d-UTP tagged Spectrum dyes (Abbott Laboratories). Pretreatment and hybridization were as described (8). For the simultaneous detection of specific DNA targets and nuclear proteins, FISH was combined with immunostaining (31). Mouse monoclonal antibodies for PCNA (Sigma-Aldrich; P-8825, 1:1,500) and cyclin A (Santa Cruz Biotechnology; sc 56299, 1:100) were used. Immunostaining was immediately followed by 3D-FISH.

Image Acquisition and Analysis of Chromosome Contact.

Confocal microscopy was performed by using a Leica SP5 TCS 4D confocal laser scanning fluorescence microscope using a 63×, 1.4 N.A. oil PlanApo objective. For most experiments, 300 nuclei from each of two different donors were scanned and analyzed. Nuclear boundaries were identified by DAPI staining, and image stacks were acquired with z steps of 0.13 μm. The digital image stacks were reconstructed by using Volocity software (PerkinElmer). Image stacks were subjected to uniform noise reduction by using the fine filter algorithm provided in the software. The analysis of spatial contact between chromosome arms was performed by using the intensity-based image segmentation technique (32).

Induction of DSBs by I-PpoI.

HA-ER-I-PpoI retrovirus (gift of M. Kastan, Duke Cancer Institute, Durham, NC) was propagated in Phoenix Ampho packaging cell line as described (33). Retroviral transductions of TPC-1 thyroid cancer cell line were carried out by using ViraDuctin Retrovirus Transduction Kit (Cell Biolabs). To increase transduction efficacy, three sequential transductions of cells were used. 4-hydroxytamoxifen (4-OHT) was added to a final concentration of 1 μM for 6–48 h to induce DNA digestion by I-PpoI (16, 18). To establish a cell line stably expressing HA-ER-I-PpoI, retrovirus was generated as described and used to infect TPC-1 cells, then selected with 1 μg/mL puromycin for 2 wk. Induction of DSBs was monitored by γH2AX staining (30). Cells with >5 γH2AX foci per nucleus were considered positive for I-PpoI–induced cleavage.

Identification of I-PpoI Sites in the Human Genome and qPCR Assay for Cleavage Efficiency.

The I-PpoI site located in the DAB1 gene has been characterized by Berkovich et al. (16), whereas other I-PpoI cleavage sites were mapped by us (Tables S1, S2, and S4). The efficiency of cleavage at these sites was assessed by quantitative real-time PCR with primers flanking each I-PpoI site (Table S5) using amplification of the GAPDH gene that does not contain an I-PpoI site for normalization as described by Berkovich et al. (16, 18).

Analysis of Contact Between Specific Regions of Homologous Chromosomes After I-PpoI Induction.

I-PpoI was induced for 6 h in TPC-1 cells after single and triple infection with HA-ER-I-PpoI retrovirus followed by 3D fixation and FISH with region-specific probes (Tables S1 and S2). For each experimental condition, 1,000 nuclei of TPC-1 cells were visually assessed by using a microscope with 63×, 1.4 N.A. oil PlanApo objective. Site-specific probes in each nucleus were identified by using the appropriate filter cube, and signals were scored as positive for contact when the space between them was smaller than size of one signal. The experiment was performed in duplicate.

Detection of Gene Expression and Transcription Inhibition.

Expression levels of studied genes were detected by using qRT-PCR with CyberGreen (ABI Biosystems) with primers shown in Table S6. For transcription inhibition, TPC-1 cells stably expressing HA-ER-I-PpoI retrovirus were plated in chamber slides, grown overnight, and treated with actinomycin D (5 μg/mL) or α-amanitin (50 μg/mL) (Sigma-Aldrich) for 6 h with or without presence of 4-OHT, washed with PBS, and fixed.

ATM and DNA-PK Inhibitors and ATM Knockdown Using shRNA.

ATM kinase inhibitor KU55933 and DNA-PK kinase inhibitor NU7441 were reconstituted in DMSO and used at 10 μM and 5 μM, respectively (34). I-PpoI was induced in TPC-1 cells infected with HA-ER-I-PpoI retrovirus by using 4-OHT for 6 h in the presence of vehicle, KU55933, or NU7441. In a separate experiment, ATM was disrupted in TPC-1 cells by using a short hairpin RNA (shRNA) that targeted the ATM sequence 5′-TGATGGTCTTAAGGAACATCT-3′. Stable cell lines expressing both the shRNA and the puromycin-resistance gene were selected in 1 μg/mL puromycin for five cell doublings after lentivirus transduction (University of Pittsburgh Cancer Institute lentiviral core). ATM knockdown and the efficacy of KU55933 and NU7441 in TPC-1 cells was determined by immunoblotting. Mouse monoclonal anti-ATM antibody (MAT3-4G10/8; Sigma-Aldrich), rabbit monoclonal anti-phosphoserine-1981 ATM (EP1890Y; Epitomics), rabbit polyclonal anti-DNA-PK (4062; Cell Signaling), and mouse monoclonal anti-phosphoserine-2056 (18192; Abcam) were used.