A Proteomic Analysis of Ataxia Telangiectasia-mutated (ATM)/ATM-Rad3-related (ATR) Substrates Identifies the Ubiquitin-Proteasome System as a Regulator for DNA Damage Checkpoints*

ATM (ataxia telangiectasia-mutated) and ATR (ATM-Rad3-related) are proximal checkpoint kinases that regulate DNA damage response (DDR). Identification and characterization of ATM/ATR substrates hold the keys for the understanding of DDR. Few techniques are available to identify protein kinase substrates. Here, we screened for potential ATM/ATR substrates using phospho-specific antibodies against known ATM/ATR substrates. We identified proteins cross-reacting to phospho-specific antibodies in response to DNA damage by mass spectrometry. We validated a subset of the candidate substrates to be phosphorylated in an ATM/ATR-dependent manner in vivo. Combining with a functional checkpoint screen, we identified proteins that belong to the ubiquitin-proteasome system (UPS) to be required in mammalian DNA damage checkpoint control, particularly the G1 cell cycle checkpoint, thus revealing protein ubiquitylation as an important regulatory mechanism downstream of ATM/ATR activation for checkpoint control.

response to ionizing radiation (IR): an initial rapid response, typically within 4 h after IR (4) and a sustained response that operates from 4 to 24 h after IR (2). While the sustained G 1 arrest is mediated by p53-dependent transcription, the rapid G 1 arrest is p53-independent and is poorly understood. While many proteins that regulate intra-S and G 2 /M checkpoints have been identified, the number of proteins known to participate in G 1 checkpoint is very limited.
Central to DDR are the two checkpoint kinases, ATM and ATR, that phosphorylate the effector kinases Chk1 and Chk2 and other downstream effectors (5). ATM/ATR substrates execute DDR directly or indirectly and are the keys for the understanding of this response. The identification of p53, NBS1, BRCA1, Chk1/Chk2, and SMC1 as substrates of ATM/ATR has significantly advanced our understanding of DDR (6 -12). A thorough understanding of DDR in mammalian cells awaits the identification of new ATM/ATR substrates. Unfortunately, few techniques are available today to systematically identify in vivo substrates of protein kinase.
ATM and ATR kinases preferentially phosphorylates the S/TQ motif as a consensus sequence (13). It has been observed that a phospho-specific antibody raised against a phosphopeptide is not strictly site or sequence specific; it also recognizes other proteins (cross-reacting proteins) with similar sequences flanking the phosphorylation site. It has been recently realized that these cross-reacting proteins could also be substrates of the same kinase (14 -17). Thus, the phospho-antibodies can be used to find potential kinase substrates.

RESULTS AND DISCUSSION
To screen for putative ATM/ATR substrates, we used phospho-antibodies raised against known ATM/ATR substrates that recognize pSQ motifs to immunoprecipitate potential new substrates. HeLa nuclear extracts were prepared from cycling cells and from those treated with IR and hydroxyurea (HU), conditions that activate ATM and ATR, respectively. Each lane of Coomassie Blue-stained gels was divided into 12 regions and proteins in each region were identified with mass spectrometry (MS). As shown as an example in Fig. 1a, structure maintenance of chromosomes 1 and 3 (SMC1/3) migrating at 170 and 150 kDa, respectively, are the most abundant proteins recognized by the ATM-pS1981 antibody (21). Many proteins were identified from cycling cells and IR-or HU-treated cells (supplemental Table I). Similar experiments were carried out with antibodies against phospho-BRCA1 (pS1387, pS1423, and pS1457), and the results are summarized in supplemental Table I. It is possible that some proteins identified using this approach were co-precipitated by association with phosphorylated proteins. To minimize precipitation of associated proteins, we used radioimmunoprecipitation buffer containing high detergent concentration to wash the immunoprecipitates * This work was supported in part by National Institutes of Health Grants CA98500 and CA84199 (to J. Q.) and DAMD W81XWH 04-1-0437 (to Y. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Material I and extensively to disrupt the weak interactions. Thus, proteins identified by this method represent both potential ATM/ATR substrates and core components of the complex that associate with the substrates and have survived the stringent wash. For example, SMC1 was co-precipitated by its strong interaction with SMC3, the substrate recognized by ATM-pS1981 antibody (Fig. 1, A and B).
Since interaction between two protein complexes is usually transient, the identification of protein complexes from different functional categories suggests that one or more proteins in each category are recognized by the phospho-antibodies as potential ATM/ATR substrates, and other core components in each protein complex come along with the substrate. We used the following criteria to validate the identified proteins as ATM/ATR substrates. First, they must be phospho-proteins; second, they are phosphorylated in an IR/HU-inducible and ATM/ATRdependent manner; and third, they must contain sequences similar to the SQ peptides used to raise the phospho-antibodies. As shown in Fig. 1B, when the lysate was treated with calf intestine phosphatase (CIP), no band was recognized by ATM-pS1981 antibody by Western blotting, demonstrating that all bands correspond to phospho-proteins. Furthermore, most of the proteins were phosphorylated in an IRand HU-inducible manner ( Fig. 1, B and C). To determine the identity of the protein band and to test whether the phospho-Western blot signal is ATM-dependent, we transfected siRNAs against ATM and the candidate protein that has a similar molecular weight to the protein band identified by immunoprecipitation/MS (supplemental Table 1). The cell lysate was analyzed by Western blotting of duplicate blots with either the phospho-antibody or a specific antibody against the candidate protein. As shown in Fig.  2A, the knockdown of either ATM or ARFBP1 reduced the p480 phospho-signal, suggesting that p480 is ARFBP1 and that this protein is phosphorylated in an ATMdependent manner; this is confirmed by the reduction of the Western blotting signal of ARFBP1 on the duplicate membrane. Using this approach, we found that ARFBP1, USP34, Each lane was divided into 12 regions and proteins were in-gel-digested by trypsin, and the resulting peptides were analyzed with mass spectrometry. B and C, Western blotting of HeLa nuclear extract prepared as described for A. An aliquot of each sample was also treated with 1 unit of calf intestine phosphatase (CIP) to de-phosphorylate the proteins. Fifty g of protein extracts per lane were resolved on an 8% SDS-PAGE gel. Western blotting (WB) of tubulin was used as loading controls. FIGURE 2. Validation of a subset of candidate ATM/ATR substrates identified from phospho-specific antibody cross-reacting proteins. A, Western blotting (WB) of HeLa whole cell lysates depleted of candidate proteins. HeLa cells transfected with siRNA to ATM or putative ATM substrates X (X ϭ ARFBP1, USP34, KIAA1794, SMC3, TDP1, TAOK3, DRIP205) were treated with 10-Gy IR and harvested 2 h later. Duplicate membranes were blotted for anti-ATM-pS1981 or antibody to protein X. The arrows indicate two isoforms of KIAA1794 that are resolved on the 8% gel. B and C, immunoprecipitation (IP)/Western blotting of HeLa whole cell lysates depleted of ATR (B) or ATM (C). HeLa cells transfected with mock siRNA or siRNA to ATR (B) or ATM (C) were treated with HU (1 mM for 16 h) or IR (10 Gy, 2 h recovery). The immunoprecipitations were carried out with the indicated phospho-specific antibodies, and Western blotting was carried out with antibodies specific to RFWD3 (Fig. 2B), RIF1 and VCP (Fig. 2C).
KIAA1794, SMC3, TAOK3, and TDP1 are phosphorylated in an IR-and HU-inducible and ATM/ATR-dependent manner. Analyses of amino acid sequences of these proteins reveal that they often contain a stretch of at least 4 amino acid residues or multiple stretches of 3 amino acid resides surrounding the SQ site that are identical to the phosphopeptide antigen, suggesting that these sequences are recognized by the phospho-antibodies employed in this study. Thus, these proteins are most likely to be ATM/ATR substrates ( Table 1). The protein band p250 is identified as TRAP220, but its phosphorylation is neither dependent on ATM/ATR nor IR/HU-inducible. Thus, some proteins cross-reacting to phospho-antibodies are not ATM/ATR substrates.
Some of the candidate proteins identified from the immunoprecipitation/MS analysis are not recognized by the same phospho-antibody in Western blots. To demonstrate that they are potential ATM/ATR substrates, we immunoprecipitated these proteins with the phospho-antibody and Western blotted with antibodies specific to these proteins. Using this method, we found that RFWD3 is phosphorylated in an ATR-dependent manner in response to HU (Fig. 2B), and RIF1 and VCP are phosphorylated in an ATM-dependant manner in response to IR (Fig. 2C). Substrates validated by this method could be associated proteins, but since they have survived the stringent wash and all contain sequences similar to the peptide antigen, thus it is equally possible that they are kinase substrates. Using both approaches demonstrated in Fig. 2, we validated nine ATM/ ATR substrates among our collection of candidates. We conclude that the above approach is capable of finding ATM/ATR substrates.
The large number of potential ATM/ATR substrates identified in the proteomic screen poses a significant challenge for functional characterization. When we categorized the proteins into functional modules based on the known function or functional motifs in these proteins, we found many proteins involved in DDR, DNA repair, chromosomal cohesion, and transcriptional regulation, which are expected from ATM/ATR substrates (supplemental Table I). We also found that proteins in the ubiquitin-proteasome system (UPS) are overrepresented (Table 1). This suggests the involvement of protein ubiquitylation in DNA damage response, a notion that is well known in yeast and begins to be appreciated in mammalian cells.
If these candidates are bode fide ATM/ATR substrates, they may participate in regulation of DNA damage checkpoint, DNA repair, or apoptosis. Ideally, we would like to map the precise phosphorylation sites of these candidates and test the loss-of-function mutants for their functions. Unfortunately, it is too labor intensive for a single lab to carry out such analysis on 10 proteins. We decided to examine whether the UPS proteins are required to regulate DNA damage checkpoints.
We carried out intra-S and G 2 /M checkpoint assays in HeLa cells depleted of these genes using siRNA. In response to IR, the intra-S checkpoint is activated to mainly suppress late origin firing and is measured by a temporary decrease of nucleotide incorporation following IR. Three days after siRNA knockdown, we irradiated the cells with 10-Gy IR and measured nucleotide incorporation 1 h post-irradiation. Of the 10 genes

Comparison of phospho-specific antibody epitopes with sequences within the identified cross-reacting proteins (S/T, D/E, and I/L are treated as the same residue in the analysis)
The identical residues to antibody epitopes are underlined. tested, six exhibited radio-resistant DNA synthesis significantly above a control siRNA (siVim) transfected cells (Fig. 3A), indicating that these UPS proteins are required to regulate the intra-S phase checkpoint. The G 2 /M checkpoint is activated to prevent cells with damaged DNA from entering mitosis. We irradiated siRNA-transfected cells with 10-Gy IR and measured the mitotic index 1 h post irradiation. The population of histone H3-pS10 positive cells, which correspond to mitotic cells, was visualized using a microscope. Knocking down 2 out of the 10 genes leads to a profound G 2 /M checkpoint failure that cells accumulate more than 4 times as many mitotic cells as in control transfected cells (compare Ͼ30% with 7.5% relative mitotic index for control transfected cells). Interestingly, siRNA transfection of ZBTB2 causes a marked increase in mitotic index before IR and virtually no further decrease after IR, indicating that ZBTB2 may play an essential role in mitotic progression, which complicates the interpretation of its role in G 2 /M checkpoint control. The G 1 checkpoint arrests cells with damaged DNA at the G 1 /S boundary. G 1 checkpoint is the least understood checkpoint among the DNA damage checkpoints, at least in part due to the lack of an accurate and sensitive assay. Current G 1 checkpoint assay, which measures G 1 accumulation by flow cytometry, is unable to accurately measure cell cycle progression immediately after DNA damage induction. We developed an imaging-based G 1 checkpoint assay by pulse-labeling cells with two bromodeoxyuridine analogs at various times before and after DNA damage induction. By distinguishing cells that have entered S phase after DNA damage induction from the replicating cells at the time of damage induction, this assay allows for accurate measurement of G 1 -S transition with a better time resolution. We have validated this assay by measuring the G 1 arrest of known checkpoint proteins and confirmed the existence of p53-independent rapid response (within 4 h post-IR) and p53dependent sustained arrest (after 4 h). 4 We used this assay to examine in details whether the UPS genes are involved in both rapid and sustained G 1 arrest in U2OS cells. We first pulse-labeled the cells with CldU to mark the starting point and then labeled 4 Y. Wang and J. Qin, submitted for publication. them with IdU 4 h later to mark the end point. Using CldU-and IdU-specific antibodies, we scored cells that were CldU-negative and IdU-positive as having entered S phase during the 4 h period after IR. We calculated the S phase entry as the percentage of CldU-negative and IdU-positive cells among DAPIstained cells. By comparing relative S phase entry (ratio of S phase entry after and before IR in the 4 h period) between 0 -4 h, 4 -8 h, and 8 -12 h following 10-Gy IR, we evaluated the G 1 checkpoint in U2OS cells depleted of USP genes by siRNA transfection. In control siRNA-transfected cells, relative S phase entry is decreased to 60% during the first 4 h after IR as a result of rapid G 1 checkpoint activation and is further decreased to and maintained below 20% starting from 4 h after IR. Of the 10 genes tested, 7 genes are required to regulate the rapid G 1 checkpoint 4 h after IR, and 6 genes are required to regulate the sustained G 1 arrest (Fig. 3C). Up to now, only Cyclin D1 degradation has been shown to regulate the rapid G 1 response (4), and Cdc25A is implicated in this process by biochemical evidence (31,32). Our screen therefore has significantly increased the number of potential G 1 checkpoint regulators and reinforced the notion that protein degradation is an important mechanism for the rapid G 1 arrest. Moreover, the finding that UPS proteins are also required for sustained G 1 arrest thus expands its regulators beyond p53-p21 CIP . It is conceivable that the UPS may regulate the sustained G 1 arrest through p53-dependent transcriptional program as transcription co-regulators or through other novel pathways yet to be discovered. Detailed mechanistic insights await the identification of substrates of the UPS proteins.
As summarized in Fig. 3D, 9 of 10 UPS genes tested are required for at least one cell cycle checkpoint, and 1 gene, ARFBP1, regulates three checkpoints. Our checkpoint assays utilize siRNA pools that are composed of four sequences to maximize knockdown, the possibility of off-target effect cannot be excluded. Nevertheless, the high success rate for identifying checkpoint genes demonstrates that a proteomic screen for ATM/ATR substrates combined with a functional screen is highly effective in selecting protein candidates for further elucidation of the signaling pathways mediated by ATM/ATR. The use of large number of phospho-SQ antibodies to screen for ATM/ATR substrates has revealed an extensive network of proteins in DNA damage response. 5 When combined with functional categorization and application of functional screens for checkpoint, DNA repair or apoptosis, potential substrates as listed in supplemental Table I can be similarly screened and placed into different pathways in the ATM/ATR-regulated DNA damage response network, expending the conceptual framework of DNA damage response. Our report is an example of applying this strategy to reveal the involvement of UPS among the ATM/ATR substrates in mammalian checkpoint control.