Profiling of UV-induced ATM/ATR signaling pathways

Immunoaffinity Isolation and Mass Spectrometry Identification of Phosphopeptides.

To identify novel ATM/R substrates, we developed two antibodies that broadly react with the ATM/R substrate consensus motif. One ATM/R substrate motif antibody was generated and characterized as described (16). The second antibody was intended to be specific for phospho-Chk2 (T26/S28) but was found to detect both phosphorylated Chk2 and other ATM/R-dependent phosphorylation sites (17). Further characterization of this antibody as a bona fide ATM/R substrate motif antibody is shown in supporting information (SI) Fig. 4.

We used the ATM/R substrate antibodies combined with liquid chromatography–tandem mass spectrometry (LC-MS/MS) to isolate and identify phosphopeptides from UV-treated M059K glioblastoma cells. Sequences were assigned to MS/MS spectra with Sequest (18), allowing for phosphorylation at serine and threonine residues. Phosphorylation sites found at SQ/TQ motifs were considered highly likely to be correct, because the phosphorylated motif was not considered in the assignment process yet was expected based on antibody recognition specificity. A complete table of identified phosphorylation sites is provided in SI Table 3 and in the PhosphoSite bioinformatics resource (Cell Signaling Technology). SI Table 3 lists only phosphopeptides found to contain at least one SQ/TQ motif. These sites are likely to be direct ATM/R substrates or substrates of other SQ/TQ-directed kinases. In some cases, nonmotif phosphorylation sites are present on the same peptide as a motif site. These sites have also been reported, although because of lack of motif, their assignment is more ambiguous. Frequency maps derived by aligning all phosphopeptides found in the analysis (not just SQ/TQ phosphopeptides) for the two motif antibodies are shown in SI Fig. 5. The analysis yielded 570 SQ/TQ phosphorylation sites from 464 proteins, and 368 were localized with at least 95% certainty (SI Table 4) based on probability-based scoring of phosphorylation site localization (19). Nearly all (498, or 87%) of these sites were previously undescribed human phosphorylation sites, based on information contained in the PhosphoSite bioinformatics resource. One hundred seventy-eight of these 498 sites were also found in another recent study investigating IR damage-induced SQ/TQ phosphorylation (14). Before that study, only 304 human SQ/TQ sites were known, and only 90 of these sites were known substrates of ATM/ATR/DNA-PK (from the PhosphoSite bioinformatics resource). Thus, this study has increased the number of potential substrates of ATM/R phosphorylated in response to UV damage of DNA >2-fold from all previously known SQ/TQ sites and >5-fold from all previously known DNA damage-inducible SQ/TQ sites.

Phosphorylation of Diverse Protein Classes Induced upon UV Damage of DNA.

We next measured the changes in the abundance of these SQ/TQ phosphorylation sites in response to UV damage as a way of pinpointing sites associated with ATM/R signaling. For each phosphorylation site identified in M059K glioblastoma cells with or without UV damage, the intensity of the precursor peptide ion was measured at the apex of its chromatographic peak, using a semiquantitative analytical approach (see Table 1 for a partial list and SI Table 5 for the complete list). Peptide ions whose intensities were increased at least 2-fold by UV damage were included in the tables. This analysis yielded 216 UV-inducible SQ/TQ sites on 180 proteins. Of the 216 sites, 192 (88%) were previously undescribed. Approximately half (99 of 192) of these sites were also found by Matsuoka et al. (14). The 216 UV-inducible SQ/TQ sites represent a significant increase in the number of potential UV damage-dependent ATM/R substrates from the 90 damage-inducible sites previously known.

Semiquantitative analysis showed peptides differentially phosphorylated from a wide range of protein classes (Fig. 1). Interestingly, some protein classes expected to be found in the analysis, such as cell cycle and apoptosis proteins, have relatively low numbers of peptides, whereas other classes, such as adaptor/scaffold or ubiquitin system proteins, were found in higher numbers than expected. Overall, 21 different protein classes were found in the analysis, with an additional 16% of peptides having unknown functions. The diversity of protein classes found to be differentially phosphorylated in response to UV damage suggests that ATM/R-dependent phosphorylation affects cellular processes not previously thought to be involved in the DNA damage response. Our results clearly show that UV-inducible phosphorylation at SQ/TQ motifs is widespread among protein classes, as was also seen in response to IR damage of DNA (14).

Confirmation of Differential Phosphorylation.

Confirmation of differential phosphorylation between undamaged and UV-damaged M059K cells was first attempted through quantitative stable isotope labeling by amino acids in cell culture (SILAC) analysis using a unit-resolution ion trap mass spectrometer (SI Table 6) (20). A number of peptides that appear in the semiquantitative analysis were confirmed by SILAC analysis. In fact, of the 54 SQ/TQ phosphorylation sites in the SILAC analysis that increased by at least 2-fold upon UV damage, 45 were found in the semiquantitative analysis (marked with a √ symbol in SI Table 6). Of those same 54 phosphorylation sites, 42 sites are previously uncharacterized. Ideally, SILAC analysis is performed with a high-resolution mass spectrometer, but the lower-resolution ion trap instrument we used was sufficient to confirm many of the semiquantitative analysis measurements. SILAC analysis resulted in fewer phosphopeptide identifications than semiquantitative analysis, because part of the instrument duty cycle was spent collecting MS/MS spectra for both the light- and heavy-isotope forms of the same phosphopeptide. The SILAC ratios in SI Table 6 are minimum fold changes, because baseline noise limits ratio measurement when signal intensities are relatively low.

Further confirmation of sites seen in the semiquantitative analysis was performed through Western blotting (Fig. 2A). First, known UV damage-inducible phosphorylation sites seen were confirmed, including S957 of SMC1 and S343 of NBS1. Additionally, phosphorylation of Chk1 and p53 was confirmed, although these peptides were not detected in our system. This may have been due to the specificity of the antibodies used (see SI Fig. 5). An exception is S15 of p53, which has the exact motif preferred by ATM/R substrate motif antibody-1 (LSQE). Because of the high density of N-terminal p53 phosphorylation sites, the tryptic peptide containing p53 S15 may have contained more modifications than the peptide identification software allowed.

To confirm some of the previously undescribed sites described in this study, proteins were immunoprecipitated with the same motif antibodies used for the site discovery studies described above, then blotted with the respective total protein antibody (instead of a phosphorylation site-specific antibody). Fig. 2B shows Western blots for 53BP1, RPA1, Mre11, and BAP1. As seen in Fig. 2B, more of each protein was immunoprecipitated in the UV sample than the control, providing further evidence that these are, in fact, previously undescribed UV-inducible substrates of ATM/R. We cannot exclude the possibility that the motif antibodies immunoprecipitated another phosphorylated protein complexed with one of the proteins shown. Ideally, a reciprocal immunoprecipitation (IP with total antibody, blot with motif antibody) would be performed, but this was not feasible because of poor performance of the motif antibodies as Western blot reagents. That the reported phosphorylation sites on all four proteins are consensus ATM/R substrate sites does give more confidence that these sites are correctly assigned. Differential phosphorylation of two of the proteins shown (53BP1 and BAP1) has also been independently confirmed by quantitative SILAC analysis (see SI Table 6). 53BP1 is a known ATM/R substrate, but the sites we describe here are distinct from the known sites of phosphorylation. We have therefore confirmed differential phosphorylation of both known and previously uncharacterized substrates of ATM/R in response to UV damage of DNA.

ATR-Specific Phosphorylation.

UV damage primarily activates the ATR branch of the DNA damage checkpoint. Seckel syndrome is an autosomal recessive disorder causing dwarfism and mental retardation that has been mapped to mutations in ATR. To further delineate the phosphorylation sites regulated specifically by ATR, we compared the Seckel syndrome fibroblast cell line GM18366 (ATR 2101A>G mutant, http://ccr.coriell.org/nigms) to an age- and race-matched control fibroblast cell line, GM00200. This ATR mutation both decreases ATR protein levels and causes defects in the response to UV damage of DNA (21). Western blots were performed to probe for UV-induced activation of ATM/R-dependent and -independent pathways in this particular cell line. As seen in Fig. 3A, S345 of Chk1 is phosphorylated in response to UV damage in the control but not the Seckel cell line. Total Chk1 levels are lower in the Seckel cell line but not low enough to account for the complete absence of Chk1 phosphorylation. Phosphorylation of p38 MAPK was induced by UV damage in both cell lines. Thus, although this Seckel cell line is deficient in an ATR-dependent response to UV damage (Chk1 S345 phosphorylation), the cells remain proficient in other ATR-independent DNA damage responses (such as p38 MAPK T180/Y182 phosphorylation).

Next, both control and Seckel cell lines were grown, damaged with UV light, and subjected to immunoaffinity purification/LC-MS/MS analysis. SI Table 7 shows a semiquantitative comparison between the control and Seckel syndrome cell lines, listing peptides with intensities at least 2-fold higher in the control than in the Seckel cell line (more abundant when ATR is present and activated). The differentially phosphorylated sites included known ATM/R substrates such as DNA-PK S2612, NBS1 S343, and SMC1 S957. Differential phosphorylation of DNA-PK, NBS1, and VCP was confirmed by Western blotting (Fig. 3B). Twenty-nine of the 39 SQ/TQ phosphorylation sites seen in the semiquantitative analysis are previously uncharacterized. Because two different cell lines were used, we cannot exclude the possibility that some differences seen were due to differences in protein levels between the cell lines. It was, therefore, of interest to find sites for which there was higher confidence in differences seen, such as those differentially phosphorylated in both semiquantitative studies.

The results of this study on phosphorylation in Seckel cells and of the study on UV-damage-responsive phosphorylation in M059K glioblastoma cells provide an overlap that can be used to associate these sites with activation of the ATR protein kinase: These two studies have 19 SQ/TQ sites in common (shown in Table 2). These phosphorylation sites are the most likely to be bona fide ATR substrates phosphorylated in response to UV damage for three reasons: They contain the minimum SQ/TQ motif, they were more highly phosphorylated in UV-damaged than undamaged samples, and their phosphorylation in response to UV damage was decreased in an ATR mutant cell line. These phosphorylation sites are therefore excellent candidates for follow-up hypothesis-driven studies aimed at verifying the kinase–substrate relationships revealed in this work and for determining functional consequences of phosphorylation.

Cell Lines and Reagents.

M059K glioblastoma cells were from the American Type Culture Collection (ATCC). GM18366 Seckel syndrome cells and GM00200-matched control cells were from the Coriell Cell Repository. For culture conditions/UV damage conditions, see SI Text.

Western Blotting.

Detection of total and phosphorylated proteins was performed by using standard methods (see Cell Signaling Technology). Antibodies used were: anti-phospho-ATM/R substrate motif, anti-phospho-Chk2 (T26/S28) (ATM/R substrate motif-2), anti-Chk1, anti-phospho-Chk1 (S345), anti-p38 MAPK, anti-phospho-p38 MAPK (T180/Y182), anti-phospho-SMC1 (S957), anti-phospho-NBS1 (S343), anti-p53, anti-phospho-p53 (S15), anti-beta actin, anti-53BP1, anti-RPA1, and anti-Mre11 (Cell Signaling Technology), and anti-BAP1 (Santa Cruz Biotechnology).

Immunoprecipitation/LC-MS/MS Mass Spectrometry.

Phosphopeptides were immunoprecipitated by using the PhosphoScan Kit (Cell Signaling Technology), with the modifications outlined in SI Text.

LC-MS/MS was performed as described (23), with the modifications outlined in SI Text.

Generation of Frequency Maps.

Redundant peptide lists for M059K cells treated with UV from all experiments were entered into pattern explorer from Scansite (www.scansite.mit.edu) (24). Resulting frequency maps were exported to Excel. Z values (deviation from expected frequency of an amino acid based on amino acid composition of all proteins in the SwissProt database) from frequency map data were input into TIGR MeV V.3.1 for Mac OSX (The Institute for Genomic Research MultiExperiment Viewer) (25). A gradient was created from Z of 0 (white) to Z ≥ 10 (blue).

Peptide Array Blots.

Peptide library arrays (SPOT arrays) were from Massachusetts Institute of Technology Biopolymers Laboratory. The top half of the array contained peptides of the sequence XXLS*QXXX, whereas the bottom half contained peptides of the sequence XXLT*QXXX, with X being an amino acid varied at each position across the blot. The first column (X) contains the minimum motif, whereas subsequent columns test addition of a fixed amino acid at residues relative to the phosphorylation site. Arrays were blotted by using standard Western blotting protocols. The SPOT array blots were digitized by using an Epson 1240 scanner, and intensities were quantified by using Scion Image Quant 1.63 for Mac. Total intensity was found for each row (each position relative to the phosphorylation site), and intensity of each individual spot was expressed as a percentage of the total intensity.

Semiquantitative Data Collection.

Immunoaffinity purification/LC-MS/MS analysis was performed as described. Precursor peptide ion chromatographic peak apex measurements were generated and compared as described in SI Text.

SILAC Data Collection.

SILAC analysis was performed by using standard protocols outlined in SI Text.

Immunoprecipitation with Covalently Coupled Antibody–Protein A Complexes.

ATM/R substrate motif antibodies were coupled to beads and used in immunoprecipitation using standard methods as detailed in SI Text.