|Home | About | Journals | Submit | Contact Us | Français|
The ATR (ATM and Rad3-related) kinase is essential to maintain genomic integrity. ATR is recruited to DNA lesions in part through its association with ATR-interacting protein (ATRIP), which in turn interacts with the single-stranded DNA binding protein RPA (replication protein A). In this study, a conserved checkpoint protein recruitment domain (CRD) in ATRIP orthologs was identified by biochemical mapping of the RPA binding site in combination with nuclear magnetic resonance, mutagenesis, and computational modeling. Mutations in the CRD of the Saccharomyces cerevisiae ATRIP ortholog Ddc2 disrupt the Ddc2-RPA interaction, prevent proper localization of Ddc2 to DNA breaks, sensitize yeast to DNA-damaging agents, and partially compromise checkpoint signaling. These data demonstrate that the CRD is critical for localization and optimal DNA damage responses. However, the stimulation of ATR kinase activity by binding of topoisomerase binding protein 1 (TopBP1) to ATRIP-ATR can occur independently of the interaction of ATRIP with RPA. Our results support the idea of a multistep model for ATR activation that requires separable localization and activation functions of ATRIP.
ATR (ATM and Rad3-related) kinase is a protein kinase that coordinates cellular responses to genotoxic stress. ATR activation occurs primarily in S phase due to replication stress induced by DNA-damaging agents or replication inhibitors. More specifically, ATR activation is stimulated when the replication machinery encounters a DNA lesion and becomes uncoupled (the helicase continues to unwind DNA while the polymerase becomes stalled at the site of DNA damage) (9).
The critical factor that promotes ATR activation is believed to be the accumulation of RPA (replication protein A)-coated single-stranded DNA (ssDNA) (11, 33, 43). At least two separate checkpoint complexes accumulate in distinct foci that colocalize with RPA. Rad17, a PCNA-like clamp loader protein, is recruited to RPA-ssDNA and loads the Rad9-Rad1-Hus1 checkpoint clamp at the junction of double-stranded and single-stranded DNA (4, 14, 53). Independently, ATR is recruited by ATR-interacting protein (ATRIP), which binds the RPA-ssDNA that accumulates at DNA lesions (3, 15, 37, 52).
ATRIP is required for ATR function, and mutation of either ATR or ATRIP causes the same phenotypes (3, 12). The strict requirement for ATRIP is conserved in Schizosaccharomyces pombe (Rad3 and Rad26), Saccharomyces cerevisiae (Mec1 and Ddc2/Lcd1/Pie1), and Xenopus laevis (xATR and xATRIP) (13, 38, 41, 51). An N-terminal domain of ATRIP binds RPA-ssDNA and is necessary for stable ATR-ATRIP localization to damage-induced nuclear foci (3, 25).
The ATR signaling pathway is currently viewed as an important target for the development of cancer therapies (10, 22, 24, 32, 34). However, the mechanism by which ATR is activated remains unclear. Localization to sites of DNA damage or replication stress has been suggested to be essential and perhaps sufficient to promote ATR signaling. However, mutations in ATRIP that disrupt the stable RPA-ATRIP interaction and impair the accumulation of ATR-ATRIP complexes in DNA-damage-induced foci have minimal effects on ATR activation and signaling (3, 25). Furthermore, topoisomerase binding protein 1 (TopBP1) was recently discovered to stimulate ATR kinase activity, suggesting regulation by a means other than localization (28). To clarify the functions of ATRIP, RPA, and TopBP1 in mediating ATR-dependent checkpoint response we have performed a series of biochemical and genetic experiments in human and yeast systems. We report structural and functional data that support a model for ATR activation in which two separable ATRIP activities—localization and activation—cooperate to promote ATR signaling.
All strains used in this study are described in Table Table1.1. Myc-Ddc2 and Myc-Ddc2 mutant strains were generated by expressing mutants from a centromeric plasmid under the control of the endogenous DDC2 promoter in strain DMP2995/1B (MATa sml1D::KanMX4 ddc2D::KanMX4) (38). GFP-Ddc2ΔN was generated in JK8-1 (36) by use of the delitto perfetto system (47). Strain yHB244 was generated by expressing RNR3 by use of pBAD79 and deleting DDC2 by use of pGEM499 in the JKM179 strain (30). Myc-DDC2 and Myc-Ddc2ΔN were expressed in strain yHB244 from the pNML1 centromeric plasmid (42).
The 14 kDa and 70 kDa RPA subunits were tagged with an His6 epitope tag (45). RPA was purified from Escherichia coli by use of nickel affinity chromatography followed by Superdex fractionation. A 20-pmol volume of biotin-labeled 69-bp single-stranded oligonucleotide was bound to streptavidin beads and incubated with binding buffer (10 mM Tris [pH 7.5], 100 mM NaCl, 10% glycerol, 0.02% Igepal CA-630, 10 μg/ml bovine serum albumin) alone or with a 4 M excess of RPA in binding buffer. The RPA-ssDNA-streptavidin beads were washed three times with binding buffer prior to use. Hemagglutinin-ATRIP (HA-ATRIP) fragments were generated using in vitro transcription/translation (Promega) and added to recombinant His-RPA or His-RPA-ssDNA beads in binding buffer, and RPA was isolated using His-Select (Sigma Aldrich) or ssDNA-Sepharose beads. Proteins bound to beads were washed with binding buffer three times, eluted, and separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) prior to blotting.
ATR kinase assays were performed essentially as described previously (28) with the following alterations. HA-ATRIP and Flag-ATR expression vectors were transfected into 293T cells and the ATR-ATRIP complexes purified by immunoprecipitation. Kinase reactions were performed with the antibody-linked ATR-ATRIP complex. Recombinant TopBP1 and RPA heterotrimer was purified from E. coli. The Phas1 substrate was purchased from A.G. Scientific. The in situ Rad53 autophosphorylation assay following denaturation/renaturation was performed as previously described (39).
All cells were grown in Dulbecco modified Eagle medium (DMEM)-7.5% fetal bovine serum. Plasmid transfections were performed with Lipofectamine 2000 (Invitrogen). The ATRIP wild-type (WT) and ATRIPΔN 293T cells were generated by retroviral infection and selection (3). ATRIP small interfering RNA (siRNA) transfections and immunofluorescence methods were also performed as described previously (3).
Tandem affinity purification (TAP)-Rfa1 was purified from soluble yeast extracts as described previously (1, 17). RPA70N was kindly provided by Cheryl Arrowsmith in a pET15b vector (Novagen) and was grown on defined M9 medium supplemented with 15N-NH4Cl and d-glucose and purified over nickel-nitrilotriacetic acid. γH2AX (Upstate), Myc9E10 (Covance), and HA.11 (Covance) were purchased from the indicated companies. Rad53 antibody was a gift from Stephen Elledge. Rfa1 antibody was a gift from Steven Brill.
Nuclear magnetic resonance (NMR) spectra were collected on ~100 μM 15N-RPA70N in a buffer containing 2 mM β-mercaptoethanol, 50 mM NaCl, and 20 mM Tris-d11 at pH 7.4. ATRIP1-107 and ATRIP54-70 were added at a four- to sixfold molar excess to maximize the bound-state population of the observed component, RPA70N. NMR experiments were performed at 25°C using Bruker AVANCE 500 MHz or 600 MHz NMR spectrometers equipped with a 5 mm single axis z-gradient Cryoprobe. Two-dimensional, gradient-enhanced 15N-1H heteronuclear single-quantum correlation (HSQC) spectra were recorded with 1,024 complex datum points in the 1H dimension and 96 complex points in 15N dimension. 1H and 15N backbone NMR assignments for RPA70N were kindly provided by Cheryl Arrowsmith.
HO expression in strains yHB245 (vector), yHB246 (Myc-DDC2), and yHB247 (Myc-Ddc2ΔN) containing galactose-inducible expression of HO endonuclease was performed as described previously (30). Cells were cross-linked with 1% formaldehyde, lysed, and sonicated to generate DNAs with an average size of 500 bp. Myc-Ddc2 protein-DNA complexes were isolated using Myc9E10 antibody and protein G-Sepharose beads, washed extensively, and eluted from beads. Cross-links were reversed by overnight incubation at 65°C. DNA was precipitated and amplified using the following primers specific to a region adjacent to the HO break site (HO-A or HO-B) or to a region of SMC2: HO-A1 (5′-CTCATCTGTGATTTGTGG-3′), HO-A2 (5′-AGAGGGTCACAGCACTAATACA-3′), HO-B1 (5′-CCAGATTTGTATTAGACGAGGGACGGAGTGA-3′), HO-B2 (5′-AGAGGGTCACAGCACTAAATACAGCTCGAAT-3′), SMC2-1 (5′-AAGAGAAACTTTAGTCAAAACATGGG-3′), and SMC2-2 (5′-CCATCACATTATACTAACTACGG-3′).
Our analysis started with an examination of the binding of ATRIP fragments to RPA in the absence and presence of ssDNA. Previous analysis of ATRIP identified at least three domains: an N-terminal RPA-ssDNA binding domain, a dimerization domain predicted to fold into a coiled-coil structure, and a C-terminal ATR-interaction domain (2, 3). HA-tagged, intact ATRIP and ATRIP fragments spanning the various domains were generated using a coupled transcription/translation system. These ATRIP proteins were added to purified His-tagged RPA heterotrimer bound to ssDNA displayed on Sepharose beads or His-tagged RPA heterotrimer bound to nickel beads. After incubation and washing, the bound ATRIP proteins were detected by Western blot analysis. We found that all ATRIP fragments containing the N-terminal 107 amino acids bound well to RPA-ssDNA and His-RPA in the absence of DNA (Fig. 1A and B). Thus, the first 107 amino acids of ATRIP contain a protein-protein interaction domain that mediates binding to the RPA heterotrimer.
HA-ATRIP fragments lacking this N-terminal RPA binding domain of ATRIP (N-RBD) were deficient in binding RPA-ssDNA and His-RPA (Fig. 1A and B). Long exposures of Western blots did show a small degree of association of these ATRIP fragments with RPA. In fact, all of the protein fragments that we tested, including a fragment of Brca1, bound weakly to RPA-ssDNA (Fig. (Fig.1A).1A). These interactions may reflect additional RPA-ssDNA binding domains on ATRIP, as has been previously reported (37).
To determine which subunit of the RPA heterotrimer interacts with ATRIP, we purified recombinant RPA domains individually or in combination as His-tagged proteins (Fig. (Fig.1C).1C). Using pull-down assays with in vitro-translated ATRIP proteins, we found that full-length HA-ATRIP or the isolated N-RBD bound only to RPA fragments containing the N-terminal RPA70 oligonucleotide/oligosaccharide (OB) fold domain (RPA70N) (Fig. (Fig.1D).1D). No significant binding to other RPA domains was detectable, and no binding of ATRIPΔN (ATRIP108-791) protein lacking the N-RBD to any RPA fragment (Fig. (Fig.1D)1D) was detectable in this assay. Taken together, these data suggest that the ATRIP N-RBD interacts directly with the 70N domain of RPA.
The specific residues involved in the interaction of RPA and ATRIP were identified using an NMR chemical-shift-mapping approach. This strategy involves monitoring NMR chemical shifts of one protein over the course of a titration with a binding partner. Measurement of the RPA 15N-1H-HSQC NMR spectrum of 15N-enriched RPA70N as ATRIP N-RBD after titration in solution showed that only a subset of the RPA70N signals was affected (Fig. (Fig.2A).2A). The observation of effects in the fast-to-intermediate-exchange regimen on the NMR timescale suggests that binding was occurring with a dissociation constant in the low micromolar range. When the chemical shifts are mapped onto the crystal structure of RPA70N (7), it is apparent that ATRIP N-RBD interacts within the basic cleft of RPA70N (Fig. (Fig.2B2B).
Initial insight into characteristics of the RPA70N binding site of the ATRIP N-RBD was obtained from sequence analysis. When the N termini of five ATRIP orthologs were aligned, minimal sequence similarity was observed, with the notable exception of a small, acidic region spanning approximately 15 amino acids (Fig. (Fig.2C).2C). On the basis of putative electrostatic complementarity, we hypothesized that this small acidic region made contact with the basic surface in the cleft of the OB fold of RPA70N. To test this hypothesis, an ATRIP peptide spanning the conserved acidic region (ATRIP54-70) was synthesized, and the RPA70N titration was repeated. The titration with the peptide perturbed most of the same residues as the titration with ATRIP N-RBD, indicating that the ATRIP peptide binds in the same manner within the basic cleft of RPA70N (Fig. (Fig.2D).2D). In addition, we analyzed the binding of ATRIP1-107 containing charge-reversal mutations at positions D58 and D59 to RPA70N by use of NMR. This mutant binds much more weakly than wild-type ATRIP.
The basic cleft of RPA70N has been shown to bind peptides that can mimic DNA in a manner similar to the binding of ssDNA to the A and B domains of RPA70 (5, 7). RPA70N binds an acidic helical peptide of p53, and the crystal structure of the p53 peptide bound in the cleft was determined by a method previously reported (7). Alignment of this p53 peptide with ATRIP54-70 indicates significant homology between the two peptides (Fig. (Fig.2E).2E). Therefore, the crystal structure of RPA70N bound to the p53 peptide was used to generate a homology model for the ATRIP peptide-RPA70N interaction. The strategy involved using the backbone coordinates of the RPA70N and the p53 peptide along with the side chains of RPA70N from the crystal structure. The p53 amino acid side chains were replaced with the ATRIP amino acid 55 to 66 side chains, and the best fit of the ATRIP peptide into the constrained RPA70N was determined using ROSETTA (Fig. (Fig.2F)2F) (40).
The model predicts that there are several specific electrostatic interactions between the acidic residues on ATRIP and the basic residues on RPA. In particular, the absolutely conserved aspartic acid residues D58 and D59 of ATRIP are likely to make contact with R41 and K88 of RPA70N (Fig. (Fig.2G).2G). Basic residues at these positions in RPA are highly conserved. The NMR data and molecular modeling are fully consistent with the previously described pull-down experiment results, indicating that the N terminus of ATRIP binds directly to RPA70N. Importantly, these data create a structural framework within which specific ATRIP-RPA binding mutants can be designed and used for functional analysis.
The functional consequences of disrupting the ATRIP-RPA interaction in human cells were previously characterized using an ATRIP mutant lacking the entire N-RBD (ATRIPΔN). Unlike wild-type ATRIP, ATRIPΔN has a severe defect in localizing to damage- or replication stress-induced nuclear foci (3). Despite this localization defect, cells depleted of endogenous ATRIP and complemented with ATRIPΔN exhibit normal ATR-dependent signaling following DNA damage (3). The only checkpoint defect that we have uncovered in the ATRIPΔN-expressing cells is a slight delay in recovery from hydroxyurea (HU)-induced stalling of replication (H. L. Ball, unpublished data). The use of RNA interference in endogenous ATRIP synthesis is not 100% effective, and the results are variable from cell to cell. In addition, the level of retrovirally expressed ATRIP or ATRIPΔN after integration of the retroviral vector is variable and not equivalent to the endogenous protein levels (3). These technical limitations to performing genetic analysis of human cell cultures may confound our ability to detect the phenotypic consequences of abrogating the ATRIP-RPA interaction. Therefore, we sought to examine the physiological role of the ATRIP-RPA interaction in the genetic system of another organism—S. cerevisiae.
The RPA binding domain of ATRIP is N terminal to the predicted coiled-coil domain. To determine whether the equivalent region of yeast ATRIP (Ddc2) mediates binding to yeast RPA70 (Rfa1), we deleted 42 amino acids N terminal to the predicted coiled-coil domain of Ddc2 (Ddc2ΔN) (Fig. (Fig.3A).3A). Binding of Ddc2 and Ddc2ΔN to Rfa1 was assayed using coimmunoprecipitation. Myc-tagged Ddc2 (WT) or Ddc2ΔN (ΔN) was expressed from plasmids under the control of the endogenous DDC2 promoter in Δddc2 cells containing HA-tagged Mec1. Myc-Ddc2 and Myc-Ddc2ΔN were immunoprecipitated using a Myc antibody, and the coassociated Rfa1 and Mec1 were assayed by Western blotting. As expected, both Mec1 and Rfa1 were coimmunoprecipitated with Ddc2 (Fig. (Fig.3B).3B). In comparison, Rfa1 association with Ddc2ΔN was greatly reduced, although Ddc2ΔN continued to bind Mec1 (Fig. (Fig.3B).3B). These data suggest that the N terminus of Ddc2 is required for a stable Ddc2-Rfa1 interaction. The amount of Rfa1-associated Ddc2 was not altered by exposing cells to UV damage, suggesting that the Ddc2-Rfa1 interaction may not be regulated by DNA damage (Fig. (Fig.3B).3B). However, these experiments utilized soluble extracts, so it is possible that the interaction with DNA-bound RPA is regulated.
Sequence alignment of human and yeast ATRIP indicates that the small acidic region in the ATRIP N-RBD is conserved (Fig. (Fig.2C).2C). The homology model generated from NMR data predicts that the absolutely conserved aspartic acid residues in this region (D12 and D13 in Ddc2) could make contacts with conserved basic amino acids on yeast Rfa1. Therefore, we hypothesized that mutating these residues would disrupt Ddc2-Rfa1 binding. To test this hypothesis a Ddc2 mutant was generated with aspartic acid-to-lysine charge-reversal mutations in these two aspartic acids (Ddc2DK) (Fig. (Fig.3C).3C). In addition, a mutant (Ddc2N14) was generated replacing Ddc2 residues 14 to 19 with a peptide (NAAIRS) that is known to adopt a helical conformation (Fig. (Fig.3C).3C). Myc-Ddc2, Myc-Ddc2ΔN, Myc-Ddc2DK, or Myc-Ddc2N14 was expressed in a yeast strain containing TAP-tagged Rfa1. TAP purification of Rfa1 protein complexes indicated they contained Myc-Ddc2 whether or not cells were pretreated with methyl methanesulfonate (MMS) (Fig. (Fig.3D).3D). In contrast, TAP-Rfa1 purifications contained minimal Myc-Ddc2ΔN, Myc-Ddc2DK, or Myc-Ddc2N14 protein (Fig. (Fig.3C).3C). These results confirm that the conserved acidic region in the N terminus of Ddc2 is required for a stable Ddc2-Rfa1 association.
Human ATRIP lacking the N-terminal RPA binding domain (ATRIPΔN) is defective in DNA-damage-induced focus formation (3). To determine whether the interaction between the N terminus of Ddc2 and Rfa1 is also required for the localization of Ddc2 to sites of DNA damage we assayed Ddc2 localization by use of ChIP analysis and focus formation. To do this we used the inducible HO nuclease system, which introduces a single double-strand break in the yeast genome (29). The induction of a double-strand break in Δddc2, DDC2, or ddc2ΔN yeast strains harboring a galactose-inducible HO endonuclease was diagnosed by comparing PCR products generated using primer sets adjacent to or spanning the HO cleavage site, and the results were equal in all strains. One hour after HO induction, cells were treated with cross-linking agent, lysed, and sonicated, and Myc-Ddc2 protein-DNA complexes were isolated by immunoprecipitation. Myc-Ddc2-bound DNA fragments were recovered and amplified by PCR using two different HO primer sets (HO-A and HO-B) adjacent to the HO cleavage site. As a control we amplified a region of the SMC2 gene that is on a chromosome different than that with the HO cleavage site. WT Ddc2 specifically accumulated at the HO cleavage site but not at the SMC2 site after induction of the HO endonuclease (Fig. (Fig.4A).4A). Compared to wild-type Ddc2 results, the accumulation of Ddc2ΔN at the HO break site was severely reduced although not completely abrogated (Fig. (Fig.4A).4A). Quantitation of the results of ChIP experiments indicated that Ddc2 binding to the HO cleavage site is fivefold greater than Ddc2ΔN binding. Ddc2 and Ddc2ΔN were expressed at equal levels, and the levels of efficiency of immunoprecipitation were equal in all samples (Fig. (Fig.4B4B).
To determine whether the defect in Ddc2ΔN accumulation at sites of DNA double-strand breaks as detected using ChIP correlated with a defect in accumulation of Ddc2ΔN into DNA-damage-induced nuclear foci we fused a C-terminal green fluorescent protein (GFP) tag onto Ddc2 and Ddc2ΔN. Equal expression of GFP-Ddc2 and GFP-Ddc2ΔN was assessed by Western blotting with an antibody specific to the GFP tag (Fig. (Fig.4C).4C). HO-endonuclease expression in GFP-DDC2 and GFP-ddc2ΔN strains was induced and Ddc2 localization monitored by fluorescence microscopy. Induction of a DNA break caused GFP-Ddc2 to accumulate into one distinct focus per cell in 40% and 42% of the cells at 4 h and 6 h after HO induction, respectively (Fig. 4D and E). Unlike GFP-Ddc2, GFP-Ddc2ΔN formed a focus in only 13% and 19% of cells after 4 h and 6 h of HO induction, respectively (Fig. 4D and E). Additionally, in cells that did demonstrate HO-induced GFP-Ddc2ΔN foci, the foci were noticeably smaller than GFP-Ddc2 foci (Fig. (Fig.4D).4D). Taken together, these data demonstrate that, consistent with the role of ATRIP-RPA interaction in human cells, Ddc2-Rfa1 interaction is required for efficient localization of Ddc2 to sites of DNA damage. Since the N-RBD of both ATRIP and Ddc2 is required for recruitment of the ATR-ATRIP/Mec1-Ddc2 checkpoint complexes to DNA lesions, we have named this domain the checkpoint protein recruitment domain (CRD).
To examine the function of the Ddc2-Rfa1 interaction in Mec1-dependent checkpoint signaling, we first determined whether disrupting binding sensitized cells to replication stress or DNA damage. Δddc2 yeasts expressing Ddc2, Ddc2ΔN, Ddc2DK, or Ddc2N14 were grown to log phase in liquid culture and plated onto media containing increasing amounts of HU or MMS. Yeasts lacking Ddc2 altogether are extremely sensitive to even low doses of HU or MMS (Fig. 5A and B). In contrast, none of the mutant ddc2 strains were sensitive to low doses of HU, and only a very small difference was visible compared to the DDC2 strain results at the highest HU concentration (Fig. (Fig.5A).5A). The difference in sensitivity to genotoxic agents between WT and mutant strains was more apparent in response to the MMS. ddc2ΔN, ddc2DK, and ddc2N14 strains were more sensitive to high doses of MMS than the DDC2 strain but much less sensitive than Δddc2 yeast. For example, at a dose of 0.01% MMS, ddc2ΔN cell viability was reduced to 3% compared to 23% for DDC2 and less than 0.01% for Δddc2 yeast (Fig. (Fig.5B).5B). At 0.15% MMS there was a difference of an order of magnitude in the viability of ddc2ΔN ddc2DK and ddc2N14 strains compared to WT DDC2 results (Fig. (Fig.5B).5B). These results suggest that the Ddc2 CRD is important for survival of cells following exposure to the DNA-alkylating agent MMS.
To directly examine the role of the Ddc2 CRD in checkpoint signaling, we tested the ability of wild-type Ddc2 or Ddc2 mutants to support Mec1-dependent Rad53 phosphorylation. Yeast were grown to log phase, arrested in G1 with alpha factor, released in the presence or absence of 200 mM HU, and harvested at various time points after release. Cell lysates were generated, and proteins were separated by SDS-PAGE and blotted using antibodies to Rad53. Rad53 phosphorylation is detectable by an electrophoretic mobility shift and is defective in Δddc2 yeast, as seen by the absence of a slower-migrating form of Rad53 (Fig. (Fig.5C).5C). Consistent with the lack of HU sensitivity, Ddc2-Rfa binding mutants Ddc2ΔN, Ddc2DK, and Ddc2N14 all support Rad53 phosphorylation after exposure to HU as efficiently as Ddc2 (Fig. (Fig.5C).5C). Detailed time course and dose-response experiments also failed to detect a significant Rad53 activation defect in the Rfa1-binding mutant strains in response to HU (Fig. 5D and E). These results are consistent with the effects of equivalent mutations in human ATRIP which fail to disrupt ATR signaling in response to HU (3).
In contrast, we did observe an attenuation of Mec1 signaling in these yeast strains in response to the presence of MMS. Strains were grown to log phase, arrested in G1 with alpha factor, and released into media containing various doses of MMS. Phosphorylated Rad53 is visible in DDC2 cells after the addition of 0.01% MMS (Fig. (Fig.5F).5F). However, Rad53 phosphorylation is attenuated in ddc2ΔN cells, indicating that optimal Rad53 phosphorylation after exposure to MMS depends upon Ddc2-Rfa1 binding (Fig. (Fig.5F).5F). The defect in Mec1 signaling after MMS treatment was most apparent at early time points after release into S phase (Fig. (Fig.5G).5G). For example, at the 60-min time point in the presence of either 0.005% or 0.01% MMS both the phosphorylation-dependent shift of Rad53 and Rad53 kinase activity are significantly reduced in the ddc2ΔN strain compared to DDC2 results (Fig. (Fig.5G).5G). However, at later time points (90 min), cells expressing Ddc2ΔN showed considerable Rad53 activation whereas Δddc2 cells did not (Fig. (Fig.5G).5G). These defects at early time points were not due to a difference in the results of release of yeast from alpha factor arrest, since all strains released equivalently. Taken together, these results suggest that Ddc2-Rfa1 binding and localization to damage sites is required for optimal checkpoint activation after exposure to MMS.
TopBP1 was recently shown to bind and activate ATR (28). This activation activity was localized to a small fragment of TopBP1 between two BRCT repeat domains. These authors also found that TopBP1 binding and activation of xATR requires xATRIP. We confirmed that TopBP1 activates ATR-ATRIP complexes in an ATRIP-dependent manner (Fig. (Fig.66 and data not shown). To determine whether the ATRIP CRD influences TopBP1 activation of ATR, we purified either wild-type ATR-ATRIP complexes or ATR-ATRIPΔN complexes. Addition of the TopBP1 fragment but not of an equivalent fragment containing an inactivating mutation (W1145R) to ATR-ATRIP complexes stimulated ATR activity toward substrates in an immune complex kinase reaction (Fig. (Fig.6A).6A). Activation of the ATR-ATRIPΔN complex upon the addition of TopBP1 was equal to the activation of ATR-ATRIP (Fig. (Fig.6A).6A). These findings are consistent with those of Kumagai et al., who found that xTopBP1 stimulates activation of xATR-xATRIP complexes containing a xATRIP protein lacking the N terminus (28). Therefore, TopBP1-dependent ATR activation does not require the ATRIP CRD.
We next assayed whether RPA or RPA-ssDNA influences ATR activity or TopBP1-dependent ATR activation. Addition of TopBP1 to ATR-ATRIP stimulated ATR kinase activity (Fig. (Fig.6B).6B). In contrast, addition of RPA (data not shown) or RPA-ssDNA to ATR kinase assays failed to stimulate ATR activity (Fig. (Fig.6B).6B). RPA-ssDNA also had no influence on TopBP1 activation of ATR (Fig. (Fig.6B).6B). RPA32 phosphorylation by ATR is stimulated by TopBP1. In addition, we also observed significant phosphorylation of the TopBP1 fragment, ATRIP, and ATR in these experiments. However, in contrast to the results seen with other proteins added to the kinase assay, the amount of autophosphorylation on the ATR-ATRIP complex was not altered significantly by the addition of the TopBP1 fragment. These results suggest that RPA-ssDNA binding to ATR-ATRIP does not influence the kinase activity of ATR. Furthermore, the function of ATRIP required to promote TopBP1-dependent activation of ATR can be separated from its RPA binding activity. However, the results do not exclude the possibility that specific RPA-DNA structures found in cells might regulate kinase activity.
To confirm these results in cells, a GFP-TopBP1 fragment containing the region that activates ATR was transfected into human cells. The cells were engineered to stably express siRNA-resistant wild-type ATRIP, ATRIPΔN, or an empty vector and were transfected with the ATRIP siRNA prior to GFP-TopBP1 transfection. Depletion of endogenous ATRIP by siRNA transfection in these cells is approximately 80% (3). Twenty-four hours after transfection of GFP-TopBP1, cells were fixed and stained for a marker of ATR activation (γH2AX). Overexpression of GFP-TopBP1 in cells containing wild-type ATRIP or ATRIPΔN caused phosphorylation of H2AX throughout the chromatin (not in distinct foci, as would be observable in response to a DNA-damaging agent) (Fig. (Fig.6C).6C). However, both the intensity of phosphorylation and the number of cells containing phosphorylated H2AX were greatly reduced in cells depleted of ATRIP, indicating that this result was due to ATR-ATRIP signaling (Fig. 6C and D). These results confirm that TopBP1 can activate ATR in cells when highly overexpressed even when ATR-ATRIP complexes lack the RPA binding domain and fail to localize to specific sites of DNA damage or replication stress. The overexpression of the TopBP1 fragment likely bypasses the regulation of TopBP1-dependent ATR activation that exists under physiological conditions.
A checkpoint protein recruitment domain (CRD) has been identified in the N terminus of ATRIP and Ddc2. This domain binds directly to RPA70N, recruits ATR-ATRIP/Mec1-Ddc2 complexes to sites of DNA damage, and promotes ATR-dependent checkpoint signaling in response to MMS. These findings are consistent with those of Kim et al., who reported that an N-terminal domain of Xenopus ATRIP is required for binding to RPA (25). RPA is a modular protein, and it often makes more than one contact with its interacting partners. Indeed, Namiki and Zou identified three large regions of ATRIP that may interact with RPA-ssDNA (37). Since no functional data were reported, additional experiments will be required to define and study the function of any other ATRIP surfaces that make direct contacts with RPA subunits. However, our data indicate that the N-terminal CRD domains of ATRIP and Ddc2 are required for the stable binding of ATRIP/Ddc2 to RPA and are necessary for retention of ATR-ATRIP/Mec1-Ddc2 at sites of DNA damage in cells.
A model of the interaction of RPA70N with a conserved ATRIP peptide within the CRD was generated using NMR data and molecular modeling from the crystal structure of a p53 peptide bound to RPA70N. The model predicts that acidic ATRIP residues (D58 and D59) make direct contacts with basic RPA70N residues (R41 and K88) in the basic cleft of the RPA70N OB fold domain. All of these amino acids are highly conserved. As predicted by this model, mutations reversing the charges on the equivalent aspartic acid residues in Ddc2 (D12K and D13K) abrogate binding to Rfa1. Interestingly, the well-characterized rfa1-t11 mutant, which is known to be replication competent but DNA-damage-response deficient, contains a single charge-reversal mutation at K45, the residue equivalent to R41 in human RPA (49). Indeed, as our model would predict, rfa-t11 is deficient in recruiting Ddc2 to double-strand breaks (23, 52) and in binding Ddc2 (H. L. Ball, unpublished data). The rfa-t11 mutant is also recombination deficient, suggesting that this basic cleft in RPA70N may be a key ligand in DNA damage responses (44, 49). It will be interesting to determine whether other DNA damage response proteins also contain acidic helices that bind within this cleft of RPA70N. It is also noteworthy that an ATR phosphorylation site (S68) is located within the ATRIP CRD just downstream of the acidic peptide that binds to the RPA basic cleft (21). Moreover, RPA70N appears to interact with the RPA32 N terminus when it is phosphorylated by checkpoint kinases (6). Therefore, phosphorylation of either ATRIP or RPA may be a means to regulate the ATR-RPA interaction.
The phenotypic consequences of disrupting the ATRIP CRD-RPA70 interaction are similar in human and yeast cells. In contrast to ATRIP or Ddc2 loss of function, cells containing mutations that disrupt the CRD are only mildly sensitive to DNA-damaging agents and partially compromised in checkpoint signaling. In fact, the response to HU is nearly indistinguishable from wild-type results despite severe defects in ATR-ATRIP/Mec1-Ddc2 localization. Functions of ATRIP in addition to RPA binding are also critical for ATR signaling. These functions include oligomerization (2, 20), ATR stabilization (12), and an undefined activity important for TopBP1-dependent activation of ATR.
The reason for the increased sensitivity of Ddc2 lacking the CRD to damage that generates DNA adducts (MMS) compared to depletion of nucleotides (HU) is unknown. Both types of genotoxic stress activate Mec1 during replication and stall replication forks (48). One potential explanation for this difference may be the amount of RPA-ssDNA present at various types of DNA lesions. Mec1-Ddc2ΔN complexes may have some residual association with Rfa1 and can still partially localize to double-strand breaks. Perhaps there is more RPA-ssDNA at an HU-stalled fork than at an MMS-induced lesion, increasing the requirement for the Ddc2 CRD at the MMS lesion. Alternatively, the recruitment and activation mechanisms of ATR-ATRIP and Mec1-Ddc2 at MMS or HU lesions may be different. Accumulating evidence suggests that additional protein-protein and protein-DNA interactions other than the ATRIP-RPA interaction may help recruit ATR-ATRIP to DNA lesions (8, 50). Also, Ddc2 contains a DNA end-binding activity localized to a region C terminal to the predicted coiled-coil domain (42). Perhaps these alternative modes of ATR-ATRIP/Mec1-Ddc2 recruitment function differently at HU and MMS lesions.
Consistent with the report by Kumagai and coworkers, we have found that TopBP1 activates ATR and that TopBP1-dependent activation of ATR is ATRIP dependent and does not require the ATRIP CRD (28). RPA-ssDNA, in contrast, does not stimulate ATR kinase activity in immune complex in vitro kinase reactions, and TopBP1-dependent activation of ATR is not altered by adding RPA or RPA-ssDNA to the kinase reaction. These results suggest that TopBP1-dependent ATR activation can be separated from ATRIP-RPA binding. The affinity of TopBP1 for ATR-ATRIP is weak and difficult to detect by coimmunoprecipitations (28). The accumulation of ATR-ATRIP and TopBP1 at sites of damage may facilitate this low-affinity interaction by increasing the local concentration of these proteins.
Taken together, these data support a multistep model proposed by Dunphy and colleagues (27) for the activation of ATR checkpoint signaling. ATR recruitment to sites of DNA damage and replication stress occurs in part through a direct interaction between the ATRIP CRD and RPA70N. TopBP1 is recruited independently through an interaction with Rad9 (16, 18, 35, 46). The assembly of ATR-ATRIP and TopBP1 at the lesion facilitates TopBP1-dependent ATR activation and, in turn, phosphorylation of ATR substrates. Accessory proteins such as claspin are also required for phosphorylation of specific substrates (26, 31). Localization may also serve to bring ATR to the vicinity of key substrates involved in fork stabilization or other aspects of checkpoint regulation. Within this model, ATRIP is a key ATR regulator since it promotes both the localization and activation of ATR. The model suggests that ATR localization to a damage site precedes its activation. However, it remains possible that ATR can be activated without localization. Indeed, when TopBP1 is highly overexpressed it activates ATR throughout the nucleus in the absence of a DNA lesion (Fig. (Fig.6C).6C). Furthermore, stable retention of ATR at a damage site is not required for ATR activation, at least, not in response to relatively high doses of UV (3). Thus, further experiments are required to definitively determine whether localization must precede activation.
ATR, ATM, and other checkpoint signaling pathways are activated by many cancer therapies and regulate the cellular outcomes of these treatments. Disruption of ATR, ATM, and DNA-PK kinases sensitizes cells to radiation and chemotherapy. Since mutations in DNA repair and DNA-damage response pathways are common in cancer cells, these cells are particularly sensitive to disruption of additional pathways. This rationale has driven the development of small molecule inhibitors of DNA damage-responsive kinases for use as chemo- or radio-sensitizing agents (22, 32). Thus far, specific inhibitors of ATM and DNA-PK kinases have been developed, built around competitive inhibition by binding of ATP analogues (19). However, specific inhibitors of the ATR kinase have not been found. Unique properties of ATR, such as its requirement for ATRIP, may provide an alternative means of disrupting ATR signaling. Hence, the molecular characterization of ATRIP structure and function as described here may provide a useful starting point for the development of an ATR-targeted therapy.
We thank William Dunphy, Steve Jackson, Mike Resnick, Stephen Elledge, Steven Brill, Tony Weil, and Jim Haber for reagents. We thank Cheryl Arrowsmith for reagents and RPA70N NMR assignments and Susan M. Meyn, Marie-Eve Chagot, and Kristian Kaufmann for assistance in preparation of peptide and protein samples and homology modeling.
This work was supported by grants from the National Cancer Institute (R01CA102729 to D.C.) and the National Institute of General Medical Sciences (W.J.C.). D.C. is also supported by the Pew Scholars Program in the Biological Sciences, sponsored by the Pew Charitable Trusts. H.L.B. is supported by a Department of Defense predoctoral fellowship and M.R.E. by an institutional training grant from the National Institute of Environmental Health Sciences. Support for facilities was provided by grants to the Vanderbilt-Ingram Cancer Center (National Cancer Institute) and the Vanderbilt Center in Molecular Toxicology (National Institute of Environmental Health Sciences grant P30ES000267).
Published ahead of print on 5 March 2007.