Double-stranded DNA- and Nbs1-dependent ATM activation in vitro
Biochemical studies using purified proteins or Xenopus
extracts have shown that ATM can be activated by DNA fragments in vitro
(Dupre et al., 2006
; Lee and Paull, 2005
; Yoo et al., 2004
; You et al., 2007
). To reveal the DNA structural determinants for ATM activation, we devised an in vitro
ATM activation assay using human cell extracts and defined DNA structures. A 70-bp dsDNA fragment with blunt ends was generated using two complementary ssDNA oligomers. In HeLa cell nuclear extracts, dsDNA but not ssDNA induced the phosphorylation of ATM at Ser1981 in a concentration-dependent manner (). The phosphorylation of Chk2 at Thr68, a known ATM substrate site in cells, was also induced by dsDNA (). The dsDNA-induced phosphorylation of ATM and Chk2 was inhibited by KU-55933, a specific ATM inhibitor, suggesting that these phosphorylation events are ATM-dependent (). The dsDNA-induced phosphorylation of ATM and Chk2 was not detected in AT cell extracts, but was detected in the extracts of the AT cells complemented with ATM (), confirming that the phosphorylation of Chk2 is ATM-dependent.
Double-stranded DNA-induced ATM Activation in Human Cell Extracts
To further assess if the dsDNA-induced phosphorylation of ATM and Chk2 indeed reflects the activation of ATM in extracts, we asked if it is dependent on Nbs1 or Ku70. In cells, Nbs1 is critical for the activation of ATM at DSBs, whereas Ku70 is required for the activation of DNA-PKcs, another kinase responsive to DSBs. We generated extracts from the HeLa cells in which Nbs1 or Ku70 was depleted by siRNA. The induction of ATM and Chk2 phosphorylation by dsDNA was significantly diminished in the Nbs1-depleted extracts compared to the controls (). In marked contrast, in the extracts with reduced levels of Ku70, ATM and Chk2 were substantially phosphorylated even when no dsDNA was added (). This phosphorylation of ATM and Chk2 may be due to the genomic instability in Ku70-depleted cells, or the binding of MRN to the residual genomic DNA in extracts when Ku70 was removed. Despite this basal phosphorylation, ATM and Chk2 were further phosphorylated when dsDNA was added to the Ku70-depleted extracts. These results suggest that the DSB-induced phosphorylation of ATM and Chk2 in extracts, like that in cells, is dependent on Nbs1 but not DNA-PKcs.
To directly determine if ATM is activated by dsDNA in extracts, we measured the kinase activity of ATM. As revealed by in vitro
kinase assays with immunoprecipitated ATM, dsDNA stimulated the kinase activity of ATM by approximately 2-fold in extracts (Fig. S1
). Similar elevations of ATM kinase activity were observed in cells treated with ionizing radiation (IR) (Pandita et al., 2000
). Collectively, these results suggest that the activation of ATM by dsDNA in extracts closely resembles the activation of ATM by DSBs in cells.
dsDNA regulates ATM activation through length- and end-dependent mechanisms
Using the in vitro
assay above, we sought to systematically characterize the DNA structural determinants for ATM activation. Studies using purified proteins or Xenopus
extracts have shown that ATM is activated by dsDNA in a length-dependent manner (Lee and Paull, 2005
; You et al., 2007
). In these studies, only the DNA fragments longer than 200 bp efficiently activated ATM (Lee and Paull, 2005
; You et al., 2007
). In HeLa extracts, however, even 12.5 nM of 70-bp dsDNA (1.5×1010
DNA ends μl−1
) induced substantial ATM phosphorylation (). The high sensitivity of this assay allowed us to analyze short dsDNA fragments with defined structural features.
The Activation of ATM by dsDNA is Length- and End-Dependent
We first asked whether and how ATM is activated by dsDNA in a length-dependent manner in human cell extracts. When present at the same molar concentrations or the same DNA mass, dsDNA of 70 bp, 40 bp, or 20 bp induced ATM phosphorylation in a length-dependent manner (). Since these dsDNA fragments are much shorter than the DNA of a single nucleosome, a length-dependent mechanism for ATM activation may operate on the nucleosome-free dsDNA immediately flanking the breaks. To investigate how the length of dsDNA contributes to ATM activation, we generated a “bubble” DNA structure by converting an internal 30-bp region of the 70-bp dsDNA into a single-stranded region (). The ability of the bubble structure to induce ATM phosphorylation was between those of the 70-bp and the 40-bp dsDNA (), showing that the internal region of the 70-bp dsDNA contributes to the length-dependent activation of ATM. Using purified MRN complexes, we found that greater amounts of Nbs1 and Rad50 associated with 70-bp dsDNA than 40- and 20-bp dsDNA (). These results suggest that the MRN complex associates with nucleosome-free dsDNA in a length-dependent manner, providing a possible mechanism for ATM activation along dsDNA.
To assess if the ends of dsDNA are critical for ATM activation, we biotinylated all four DNA ends of the 20- and 70-bp fragments (5′ and 3′ ends of both strands). The biotinylated dsDNA efficiently induced ATM phosphorylation in the absence of streptavidin, but lost this activity when the ends were blocked by streptavidin (). When only the 5′ or 3′ ends of 70-bp dsDNA were blocked, the ability of the fragment to activate ATM was substantially reduced (Fig. S2
). Since the streptavidin on one DNA strand may block access to both strands, it was not possible to resolve how 5′ or 3′ ends contribute to ATM activation. Nonetheless, blockage of DNA ends inhibited ATM activation regardless of the length of dsDNA, suggesting that the length-dependent mechanism for ATM activation needs to be initiated from DNA ends, or act through the ends.
The ends of dsDNA could potentially be processed by helicases and/or nucleases in extracts. To assess how unwinding of dsDNA affects ATM activation in extracts, we generated a fork-like DNA structure that possesses both paired and unpaired DNA ends (). The ability of the fork structure to activate ATM was lost when the paired ends were blocked, but was unaffected when the unpaired ends were blocked (). Therefore, paired DNA ends are required for initiating ATM activation in extracts. These results suggest that ATM cannot be directly activated by unwound DNA ends or by the fork-like DNA structures associated with DNA replication or DNA repair.
Single-strand overhangs interfere with ATM activation by attenuating MRN binding
DSBs are not always blunt-ended in cells. The DSBs generated by the HO or I-Sce
I endonuclease initially have 4-nt 3′ single-strand overhangs (SSOs) (Colleaux et al., 1988
; Kostriken et al., 1983
). V(D)J recombination and meiosis produce DSBs with 3′ and 5′ SSOs, respectively (Schlissel, 1998
; Xu and Kleckner, 1995
). The DSBs resulting from collapsed replication forks or broken ssDNA gaps may possess either 3′ or 5′ SSOs (Lopes et al., 2006
). The “uncapped” telomeres resemble DSBs with 3′ SSOs (Celli and de Lange, 2005
). When exposed in cells, DSBs can be resected by exo- or endonucleases in the 5′-to-3′ direction (Lee et al., 1998
). ATM has been implicated in the response to the various types of DSBs above, indicating that it can be activated by DSBs with SSOs. In extracts, while the bulk of dsDNA appeared unaltered (Fig. S3
), a small fraction of it might be processed by nucleases. The in vivo
functions of ATM in the response to SSO-bearing DSBs prompted us to investigate the role of SSOs in ATM activation.
To directly assess the effects of SSOs on ATM activation, we analyzed the 20- and 70-bp dsDNA bearing either 5′ or 3′ SSOs of random sequences (). Both 5′ and 3′ SSOs of 5 nt slightly enhanced ATM phosphorylation. Interestingly, both 5′ and 3′ SSOs of 25 or 50 nt attenuated ATM and Chk2 phosphorylation (), suggesting that SSOs interfere with ATM activation in a length-dependent manner. SSOs of poly A also hindered ATM activation in a length-dependent manner (). SSOs not only attenuated the activation of ATM by 20- and 70-bp dsDNA, but also that by linear plasmids (see ). Together, these results suggest that SSOs may interfere with a DNA end-dependent event in ATM activation, which is independent of the length of dsDNA.
Regulation of ATM Activation by SSOs
Resection of DNA Ends Promotes an ATM-to-ATR Switch in vitro
The ssDNA generated by resection may interfere with ATM activation in cis or in trans. When blunt-ended 70-bp dsDNA was added to extracts with 25-nt ssDNA at 1:2 or 1:5 molar ratios, a modest reduction of ATM activation was observed (). When the 25-nt ssDNA was linked to 70-bp dsDNA as overhangs, it interfered with ATM activation more effectively. Thus, while ssDNA can interfere with ATM activation both in cis and in trans, SSOs are more potent than free ssDNA for this function.
To reveal the mechanism by which SSOs interfere with ATM activation, we asked if SSOs affect the binding of MRN to dsDNA. Indeed, 3′ SSOs of 25 nt substantially reduced the amounts of Nbs1 and Mre11 associated with 70-bp dsDNA in extracts (). However, purified MRN bound to dsDNA efficiently regardless of the presence or absence of SSOs (). Together, these results suggest that SSOs do not directly interfere with the binding of MRN to dsDNA, but they reduce MRN binding in the presence of other proteins.
ATM activation requires junctions of single- and double-stranded DNA
Although less potent than blunt-ended dsDNA, dsDNA bearing short SSOs retain some ability to associate with MRN and to active ATM in extracts. Our analysis of blunt-ended dsDNA suggest that ATM activation is dependent on DNA ends (). Two types of DNA ends are present in the DNA fragments with SSOs: the ends of the dsDNA region (the junctions of dsDNA/ssDNA) and the ends of SSOs (). To assess how these DNA ends contribute to ATM activation, we tested three sets of DNA structures (20-bp dsDNA with 5′ or 3′ 25-nt SSOs and 70-bp dsDNA with 3′ 25-nt SSOs) in which either the junctions or the SSO ends were biotinylated (). In the absence of streptavidin, all of the DNA structures with SSOs induced ATM phosphorylation at reduced levels compared to blunt-ended dsDNA (). When the ends of the 5′ or 3′ SSOs were blocked by streptavidin, the ability of the DNA fragments to activate ATM and Chk2 was not affected (). In striking contrast, when the 5′ or 3′ junctions of dsDNA/ssDNA were blocked by streptavidin, the DNA fragments failed to activate ATM and Chk2 (). These results suggest that the junctions of dsDNA/ssDNA, but not the ends of SSOs, are critical for ATM activation. Furthermore, the junctions of dsDNA/ssDNA are required for ATM activation regardless of the length of dsDNA (), suggesting that these ends are involved in an initiating event for ATM activation, possibly the DNA recognition by MRN.
ATM Activation Requires the Junctions of ssDNA and dsDNA
The junctions of dsDNA/ssDNA are present not only at DSBs, but also at single-strand DNA breaks, gaps, and DNA replication forks. To assess if dsDNA/ssDNA junctions are sufficient to activate ATM, we generated a plasmid carrying a single cleavage site of the nicking enzyme N. BbvCI (Fig. S4A
). Using the nicking enzyme or a restriction enzyme that cuts the plasmid in both DNA strands, we generated nicked plasmids and linear plasmids bearing blunt ends (Fig. S4B
). Like the short dsDNA fragments, linear plasmids induced ATM phosphorylation (). In contrast, nicked plasmids were unable to induce any ATM phosphorylation (). Moreover, when the DNA nicks were extended into ssDNA gaps by Exonuclease III (Fig. S4C
), the gap-carrying plasmids were still unable to activate ATM (). Thus, while the junctions of dsDNA/ssDNA are required for ATM activation at DSBs, they are not sufficient to elicit ATM response when present internally on DNA. These internal junctions may be recognized by proteins that inhibit ATM activation. Alternatively, additional structural features of DSBs, such as the topological state of DNA (Fig. S4B
), may be involved in ATM activation.
Resection of DNA ends promotes an ATM-to-ATR switch
The involvement of dsDNA/ssDNA junctions in ATM activation is surprising because these structures have been implicated in the activation of ATR (MacDougall et al., 2007
; Zou, 2007
). These results raise the question as to how the DNA-damage specificities of ATM and ATR are distinct from each other at DSBs, and how ATM and ATR are coordinated at the DSBs undergoing resection. In human cells, ATM is required for DSB resection and ATR activation (Jazayeri et al., 2006
). In this study, we show that SSOs interfere with ATM activation in a length-dependent manner (). In addition, we have previously shown that ssDNA coated by RPA binds to ATRIP in a length-dependent manner, allowing the ATR-ATRIP kinase complex to recognize DSBs (Zou and Elledge, 2003
). These findings led us to hypothesize that following the activation of ATM and the initiation of resection, SSOs might promote an ATM-to-ATR switch at DSBs.
To directly investigate if the process of SSO generation can restrict ATM activation and induce ATR activation, we used exonucleases to resect the DNA ends of a linear plasmid (Fig. S5A
). In a time-dependent manner, T7 exonuclease progressively resects DNA ends in the 5′-to-3′ direction, whereas Exonuclease III cleaves in the 3′-to-5′ direction (Fig. S5B
). When the same amounts of processed or unprocessed plasmids were added to extracts, the processed plasmids exhibited a reduced ability to activate ATM compared to the unprocessed plasmids (). These results confirm that the generation of SSOs progressively interferes with ATM activation ().
When generated at DSBs in cells, SSOs are recognized by RPA, leading to a DNA-protein structure recruiting ATR-ATRIP (Zou and Elledge, 2003
). In extracts, SSOs of 50 nt associated with both RPA and ATRIP (). To assess if SSOs induce ATR activation, we monitored the phosphorylation of RPA32 at Ser33 (Olson et al., 2006
). The phosphorylation of RPA32 was induced by linear plasmids even in the absence of exonuclease (). This phosphorylation of RPA32 was partially inhibited by KU-55933 alone, and virtually abolished by the combination of KU-55933 and NU7026, a specific inhibitor of DNA-PKcs, suggesting that the RPA32 phosphorylation induced by linear plasmids involves both ATM and DNA-PKcs. Interestingly, the phosphorylation of RPA32 was progressively enhanced by the exonuclease-mediated resection of DNA ends (). This is in marked contrast to the decline of ATM phosphorylation when SSOs were generated (). Furthermore, the phosphorylation of RPA32 induced by resected ends was not inhibited by the combination of KU-55933 and NU7026, but was inhibited by Wortmannin, a pan-inhibitor of ATR, ATM, and DNA-PKcs (). These results suggest that the SSO-induced RPA32 phosphorylation is independent of ATM and DNA-PKcs, but may be dependent on ATR.
To directly address if the increased phosphorylation of RPA32 reflects the activation of ATR, we used siRNA to knockdown ATR in HeLa cells and generated nuclear extracts from these cells. KU-55933 and NU7026 were added to the extracts to eliminate the potential contributions of DNA-PKcs and ATM to RPA32 phosphorylation. In the absence of exonucleases, linear plasmids did not induce RPA32 phosphorylation in the ATR-depleted or the control extracts (). Importantly, in the presence of exonucleases, the phosphorylation of RPA32 was efficiently induced in the control extracts, but not in the extracts lacking ATR (). Together, these results demonstrate that the generation of SSOs not only interferes with ATM activation, but also promotes ATR activation.
Consecutive activation of ATM and ATR in cells
The opposite effects of SSOs on ATM activation and ATR activation prompted us to investigate if ATM and ATR are reciprocally regulated in cells. We followed the IR-induced phosphorylation of Chk2 and Chk1, two specific substrates of ATM and ATR, respectively. Within 5 min after IR, Chk2 but not Chk1 was strongly phosphorylated, showing that Chk2 is phosphorylated more rapidly than Chk1 (). Furthermore, Chk2 phosphorylation started to decline 30 min after IR, whereas Chk1 phosphorylation remained at high levels until 2 hr (). These results show that Chk2 is transiently phosphorylated during a window that precedes the window of Chk1 phosphorylation. Nevertheless, there was a short period (10 to 30 min) in which both Chk2 and Chk1 were strongly phosphorylated. This may be due to asynchronous resection of DSBs in individual cells (Barlow et al., 2008
; Zierhut and Diffley, 2008
), or asynchronous resection in different cell sub-populations (Jazayeri et al., 2006
). It should be noted that IR not only induces DSBs but also other types of DNA damage (Ward, 2000
), some of which may interfere with DNA replication and lead to delayed ATR/ATM response. Overall, the decline of Chk2 phosphorylation coincided with strong Chk1 phosphorylation, which is consistent with an ATM-to-ATR switch in these cells.
Consecutive Activation of ATM and ATR in Cells
Chk2 is phosphorylated by ATM locally at DSBs and then moves throughout the nucleus (Lukas et al., 2003
), prompting us to assess more directly if ATM is transiently activated at DSBs. The accumulation of phosphorylated ATM at DSBs marks one of the events during the process of ATM activation (Berkovich et al., 2007
; You et al., 2005
). Phosphorylated ATM appeared on chromatin within 5 min after IR and started to decline after 30 min (). Since the overall levels of phosphorylated ATM in cells remain high for many hours after IR (; Bakkenist and Kastan, 2003
), our data indicates that Chk2 is primarily targeted by the phosphorylated ATM associated with DSBs. These results suggest that ATM is transiently activated at DSBs.
To monitor the resection of DSBs in cells, we analyzed the IR-induced RPA foci at different times after irradiation (). Few RPA foci were detected 5 min after IR. At 30 min, RPA foci appeared in a significant fraction of cells. At 120 min, approximately half of the cells exhibited intense RPA foci. Thus, RPA gradually accumulated at DSBs as ATM activation was attenuated. The numerous RPA foci at 120 min suggest that the attenuation of ATM activation was not due to the completion of DNA repair. These results provide further evidence that the activation of ATM is gradually attenuated by DSB resection in cells. Furthermore, because RPA-ssDNA is a key structure involved in ATR activation, these results also provide in vivo evidence linking the attenuation of ATM activation to ATR activation.
Regulation of the ATM-to-ATR switch by ATM and nucleases
The attenuation of ATM activation at resected DSBs could be attributed to the generation of SSO or the consequent ATR activation. To distinguish these possibilities, we analyzed the phosphorylation of Chk1 and Chk2 in cells treated with ATR siRNA. While Chk1 phosphorylation was abolished in cells lacking ATR, the kinetics of Chk2 phosphorylation was not altered in these cells (). These results strongly suggest that the process of DSB resection, rather than the activation of ATR, is responsible of the loss of ATM activation.
Regulation of the ATM-to-ATR switch by ATM, CtIP, and Exo1
If the ATM-to-ATR switch is indeed driven by DSB resection in cells, one would expect that this transition is controlled by the regulators of DSB resection. The resection of DSBs in cells is regulated by ATM and exonucleases including MRN-CtIP and Exo1. To test if ATM regulates the ATM-to-ATR switch, we treated cells with KU-55933 prior to IR irradiation. In the presence of 5 μM of KU-55933, the phosphorylation of both Chk2 and Chk1 was virtually abolished (). When ATM was partially inhibited by 1 μM of KU-55933, both Chk2 phosphorylation and Chk1 phosphorylation were reduced and delayed, and the two events became increasingly overlapped (). Thus, a timely and synchronous transition from ATM to ATR relies on ATM activity.
To more vigorously test if the ATM-to-ATR switch is driven by DSB resection in cells, we asked if it is possible to promote the switch by enhancing the functions of nucleases. CtIP, an activator of MRN (Sartori et al., 2007
), and Exo1 were co-expressed in cells (Fig. S6
). In the absence of IR, expression of CtIP and Exo1 did not induce significant ATM and Chk1 phosphorylation (). In response to IR, CtIP and Exo1 diminished Chk2 phosphorylation but enhanced Chk1 phosphorylation (), consistent with a more efficient ATM-to-ATR switch. Although ATM phosphorylation was not compromised in cells expressing CtIP and Exo1 (Fig. S6
), its retention to chromatin was reduced (). Interestingly, the IR-induced Chk1 phosphorylation was abolished by KU-55933 in cells expressing CtIP and Exo1 (), suggesting that ATM may act before CtIP and Exo1 during DSB resection. These results lend strong support to the conclusion that the ATM-to-ATR switch is driven by DSB resection in cells.