Chemo-Phosphoproteomic Profiling with ATR Inhibitors Berzosertib and Gartisertib Uncovers New Biomarkers and DNA Damage Response Regulators

The ATR kinase protects cells against DNA damage and replication stress and represents a promising anti-cancer drug target. The ATR inhibitors (ATRi) berzosertib and gartisertib are both in clinical trials for the treatment of advanced solid tumors as monotherapy or in combination with genotoxic agents. We carried out quantitative phospho-proteomic screening for ATR biomarkers that are highly sensitive to berzosertib and gartisertib, using an optimized mass spectrometry pipeline. Screening identified a range of novel ATR-dependent phosphorylation events, which were grouped into three broad classes: (i) targets whose phosphorylation is highly sensitive to ATRi and which could be the next generation of ATR biomarkers; (ii) proteins with known genome maintenance roles not previously known to be regulated by ATR; (iii) novel targets whose cellular roles are unclear. Class iii targets represent candidate DNA damage response proteins and, with this in mind, proteins in this class were subjected to secondary screening for recruitment to DNA damage sites. We show that one of the proteins recruited, SCAF1, interacts with RNAPII in a phospho-dependent manner and recruitment requires PARP activity and interaction with RNAPII. We also show that SCAF1 deficiency partly rescues RAD51 loading in cells lacking the BRCA1 tumor suppressor. Taken together these data reveal potential new ATR biomarkers and new genome maintenance factors.

Table S6.New ATR targets not previously implicated in cell responses to DNA damage or replication stress.Table S7.Phosphorylation sites higher in abundance after ATR inhibitor treatment Table S8.Proteins with phosphorylation sites higher in abundance after ATR inhibitor treatment Table S9.Lists of reagents used in this study

Sample preparation for phosphoproteomic analysis
Five 150 mm plates were seeded with U-2 OS cells at 30% confluency and synchronised to enrich cells in S phase as described in the method section.Cells were mock treated or treated with 1 µM ATR inhibitors (berzosertib or gartisertib) followed by treatment with 1 mM hydroxyurea for 30 min.Post drug treatment, cells were harvested by trypsinisation and washed once with ice cold PBS.To the cell pellets 1.5 mL ice cold TCA-Acetone solution (20 % (v/v) TCA, 80% (v/v) acetone, 0.2% DTT (wt/v)) was added, cells were vortexed vigorously to disperse the pellet and kept overnight at -20°C.After 24 h, samples were centrifuged at 20,000 g at -9°C for 20 min, and the supernatants was discarded.Pellets were washed once with 2 mL ice cold 80% (v/v) acetone, vortexed vigorously, centrifuged again at 20,000 g at -9°C for 20 min.Carefully removed the supernatant and pellets were air dried for 10 min at room temperature.Cell pellets from each plate were then lysed in 350 µL of 8 M urea, 50 mM AmBiC, 1% (v/v) phosphatase inhibitor cocktail-2, 0.1% (v/v) microcystin-LR, pH 8. Samples were mixed thoroughly by pipetting up and down and incubated at room temperature for 15 min.Post incubation samples were sonicated using a Bioruptor sonicator at high amplitude for 10 cycles (30 sec on/30 sec off).Lysates were clarified by centrifugation at 17,000 g for 10 min and samples were stored at -80°C until further use for mass spectrometric analysis.Five independent biological replicates per condition were prepared on different days using the two different ATR inhibitors berzosertib and gartisertib.Protein concentrations were determined using BCA assay.
A total of 3.5 mg total protein from each sample was taken and volume was equalized using 8M urea (Cat.No. 1.08487.0500,Merck, Darmstadt, Germany), 50mM ammonium bicarbonate
Subsequently, the phosphopeptides were labelled via tandem mass tags (TMT10plex, Cat.No. 90113, Thermo Fisher Scientific, Waltham, Massachusetts, US).Freeze dried peptides were reconstituted in 100mM Triethylammonium bicarbonate (TEAB, Cat.No. 90114, Thermo Fisher Scientific).TMT labels were resuspended in anhydrous ACN (Cat.No. 271004, MilliporeSigma), added to the respective samples and incubated 2h at RT.In case of the (HU vs. HU + gartisertib) experimental series, the HU group was labelled with TMT126-TMT128C, the inhibitor treatment group was labelled with TMT129N-TMT131.In the experimental series (HU vs HU + berzosertib), the inhibitor treatment group was labelled with TMT126-TMT128C, while the HU group was labelled with TMT129N-TMT131.An aliquot was taken (the remainder samples were interim stored at -80°C) and labelling efficiency was checked via LC-MS/MS.Peptide N-termini were detected as labelled in 96.8% (gartisertib) or 87.2% (berzosertib) of all peptides, and lysines were detected as labelled in 97.8% (gartisertib) or 85.1% (berzosertib) of all cases.The labelling reaction of the remainder samples was stopped using hydroxylamine (Cat.No. 467804, MilliporeSigma).Afterwards, the samples were pooled, snap frozen and freeze dried.

Sample prefractionation using high pH reverse phase liquid chromatography
The pooled sample was fractionated using high pH reverse phase liquid chromatography.
Peptides were reconstituted in 200 µL 10mM ammonium formate (Cat.No. 70221, MilliporeSigma) pH 10 and separated on a C18 column (4.6 x 250 mm, 3.5µm particle size, Cat. No. 186003570, Waters) using a two-component buffer system.Buffer A consisted of 10mM ammonium formate, pH 10 while Buffer B consisted of 80% acetonitrile and 10mM ammonium formate, pH10.Peptide separation was achieved using a 75min gradient with following settings:

LC-MS/MS
Before LC-MS/MS analysis, samples were reconstituted in 2% (v/v) ACN (Merck), 0.1% formic acid (FA, Cat.No. 5.33002.0050,Merck).LC-MS/MS was performed using a Thermo Dionex Ultimate 3000RLSC system coupled to an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher).Samples were injected in triplicates onto a C18 trap column (Acclaim TM PepMap TM 100, Cat.No. DX164564, Thermo Fisher) and washed for 5 minutes using 3% (v/v) ACN, 0.5% (v/v) TFA at a flowrate of 5 µL/min.The column oven was set to 45°C.Peptides were separated on a C18 column (EASY-Spray TM , 50 cm length, Cat.No. 03-251-874, Thermo Fisher) using a segmented linear gradient applying a two-buffer system with a total run-time of 180 minutes at a flowrate of 300 nL/min.Buffer A consisted of 3% (v/v) ACN, 0.1% (v/v) FA, Buffer B of 80% ACN, 0.08% (v/v) FA.During the first 145 minutes Buffer B percentage increased from 5 to 25%, followed by an increase to 35% B until minute 155, then increase to 95% B until minute 160 with isocratic elution until minute 165, this was followed by a decrease in Buffer B to 3% over 30 seconds with isocratic column wash until minute 180.Eluted peptides were injected into the Orbitrap MS.The ion source spray voltage was operated in positive mode and set to 2000 V. Ion transfer tube was heated to 275 °C.The mass spectrometer was operated in datadependent Top20 mode.MS full scan was performed using the Orbitrap detector, detecting positively charged ions between 375-1500 m/z at a resolution of 120K at 200 m/z.AGC target was set to 4 x 10 5 with a maximal ion injection time of 50 ms.For MS2 scan, precursor ions had to exhibit the following properties: Intensity above 2 x 10 4 , charge states between 2 and 7, while exhibiting an isotopic distribution expected for a peptide.Precursor ions were isolated using a width of 1.6 m/z and fragmented by higher-energy collisional dissociation (HCD) with 37% normalized collision energy.AGC target was set to 5 x 10 4 .In case of samples from gartisertib experimental series the maximum injection time was set to 120 ms, while for samples from the berzosertib series this was set to 200 ms.MS/MS spectra were recorded in the Orbitrap detector at a resolution of 50K (at 200 m/z).To avoid repeated fragmentation of the same precursor ion species, dynamic exclusion was set to 40 seconds with a mass tolerance of 10 ppm.

Data analysis global phosphoproteomics
Mass spectrometry raw data was searched against a homo sapiens FASTA file (42,326 entries including protein isoforms, downloaded from www.uniprot.orgon 5 th April 2018) (1) using MaxQuant (Version 1.6.2.3) (2).Default search settings were applied, as digestion enzyme Trypsin/P was set and a maximum 2 missed cleavages was allowed.As fixed modifications the carbamidomethylation of cysteine and as variable modifications the oxidation of methionine, acetylation of the protein N-terminus was set.Additionally, the following non-default variable peptide modifications were set: Phosphorylation of serine, threonine and tyrosine, deamidation of asparagine.Mass tolerance for the precursor ions for the first search was set to 20 ppm, the main search tolerance was 4.5ppm.The MS/MS tolerance was set to 20 ppm.Peptide false discovery rate (FDR) and protein FDR was set to 5%.All mass spectrometry raw data, the MaxQuant search parameters and output files were uploaded to jPOSTrepo (3) and can be downloaded via ProteomeXchange (4) with the identifier PXD040469.Results were analysed using an in-house developed R (version 4.1.1)(5) computational pipeline, based on modified scripts applied in (6).In brief, within each fraction the peptide intensities were averaged.Data were normalized using variance stabilizing normalization (VSN) (7,8) and statistically tested using linear models for microarrays (limma) (9,10).A peptide was regarded as differentially abundant between the two conditions if the adjusted p-value was ≤ 0.05.Singly phosphorylated peptides, lower abundant under inhibitor treatment, exhibiting a phosphorylation PTM score probability ≥ 0.994 (corresponding to a false localization rate of 1% ( 11)) underwent phosphorylation sequence motif analysis using rmotifx (12).GO term analysis was based on a modified version of a previous R script file (6), applying a Fisher's exact test.For this, the unique leading razor protein names connected to phosphopeptides lower abundant under inhibitor treatment were used as foreground.The unique leading razor protein names of the whole dataset were used as background.Resulting p-values were converted into q-values (13) and GO terms with q ≤ 0.01 and at least 10 protein hits were regarded as enriched in the foreground.
The script files for the analysis of the phosphorylation sites higher abundant under ATR inhibitor treatment can be downloaded from Zenodo via https://doi.org/10.5281/zenodo.10850046.

Extracted ion chromatogram (XIC) analysis
Protein extraction from SDS-PAGE bands, digestion with Trypsin/LysC followed the protocol described in (46), deviating in the usage of 5% hydroxylamine to stop the TMT labelling reaction.The samples were analysed on an Orbitrap Lumos (Thermo Fisher), connected to a Thermo Dionex UltiMate 3000RLSC system.The identical MS instrument settings and LC gradient as detailed in (6) were applied.Mass spectrometry raw data was searched using MaxQuant (Version 2.1.3.0) using the same FASTA file as described under "Data analysis global phosphoproteomics".Standard MaxQuant search settings were applied with the additional variable modification of phosphorylation of serine, threonine and tyrosine.Peptide and protein FDR threshold was set to 5%.The TMT intensity data for the LUZP1 phosphopeptide AIGALASpSR (Spectrum #17044 in file LUM190801-16_FW052_TL-01_FW.raw) had to be extracted manually using Xcalibur QualBrowser (Version 2.2 SP1.48, Thermo Fisher) as MaxQuant did not detect these signals although they are clearly present in the raw data.Data were analysed using modified versions of the R scripts used in (6,46).In summary, data were normalized using VSN and the TMT reporter intensities were statistically tested using t-tests under application of Bonferroni correction of the significance threshold to α = 0.05 / 3 = 0.0167 (3 t-tests).All mass spectrometry raw data, MaxQuant search settings and output files were uploaded to jPOSTrepo and can be downloaded via ProteomeXchange with the identifier PXD040476.Data analysis scripts and annotated spectra (44) can be downloaded from Zenodo under https://doi.org/10.5281/zenodo.10581706.
Samples from DHX9 immunoprecipitation were analysed after SDS-PAGE, digestion and peptide extraction (see main manuscript text) as follows: Peptides were resuspended in 5% formic acid in water and injected on an UltiMate 3000 RSLCnano System coupled to an Orbitrap Exploris 240 (Thermo Scientific).Peptides were loaded on an Acclaim PepMap trap column (Thermo Scientific #164750) prior analysis on a PepMap RSLC C18 analytical column (Thermo Scientific #ES903) and eluted on a 60 min linear gradient from 3 to 35% Buffer B (Buffer A: 0.1% formic acid in water, Buffer B: 0.08% formic acid in 80:20 acetonitrile:water (v:v)).Eluted peptides were then analysed by the mass spectrometer operating in data dependent acquisition mode using a cycle time of 2 s.MS1 were acquired at a resolution of 60000 with an AGC target of 300% and a maximum injection time of 25 ms.Peptides were then selected for HCD fragmentation using an isolation width of 1.2 Th, NCE of 30%, AGC of 100% and maximum injection time of 100 ms and MS2 were acquired at a resolution of 15000.Dynamic exclusion was set to 30 s with a tolerance of +/-10 ppm.Mass spectrometry raw data was searched using MaxQuant (Version 2.2.0.0) against a Uniprot homo sapiens FASTA (42,397 entries inclusive isoforms, downloaded 2 nd January 2023 from www.uniprot.org).In addition to the standard MaxQuant search parameter, deamidation of N and Q, and phosphorylation of S, T and Y was added as variable modifications.Intensities of peptides from DHX9 which were repeatedly measured were averaged using their median intensity and then all DHX9 peptides from the different samples were normalized via VSN.Data was then statistically tested using limma.The mass spectrometry raw data, MaxQuant search settings and output files were uploaded to PRIDE (45) and can be accessed via ProteomeXchange (PXD041250).Data analysis scripts and annotated spectra (44) were uploaded to Zenodo and can be downloaded via https://doi.org/10.5281/zenodo.10882997.In case of DHX9 from HeLa and HEK293 cells, the immunoprecipitated samples were separated via SDS-PAGE, the relevant protein bands were cut-out and digested and peptides extracted as described in the main manuscript text.
Afterwards, the peptides were resuspended in 0.1% (v/v) formic acid and 0.015% (w/v) n-Dodecyl-beta-maltoside.Samples were analysed using an UltiMate 3000 RSLCnano System coupled to an Orbitrap Exploris 240 (Thermo Fisher Scientific).Peptides were trapped on an Acclaim PepMap trap column (Thermo Fisher Scientific #164750) and separated on a PepMap RSLC C18 analytical column (Thermo Fisher Scientific #ES903).Peptides were eluted using a 60 min linear gradient from 3 to 35% Buffer B (Buffer A: 0.1% (v/v) formic acid in H2O, Buffer B: 0.08% (v/v) formic acid in 80:20 acetonitrile:H2O (v:v)) and analysed using parallel reaction monitoring mass spectrometry.In brief, the mass spectrometer conducted two consecutive MS1 scans spanning a width of 2 Da, centring at 538.2802 m/z or 578.2633 m/z, respectively (60,000 resolution).These two masses correspond to the non-phosphorylated and phosphorylated forms of the doubly charged peptide LAQFEPSQR (Ser321 of DHX9).This was followed by two MS/MS scans using the same MS1 mass windows at a resolution of 15,000 with a normalised collision energy of 30.Automatic gain control was set to "Standard" and maximum injection time mode was set to "Auto".As first step the elution time window of the phospho-peptide was established.Fort this, the raw data was searched using MaxQuant (version 2.2.0.0) as stated above, with variable modifications set to oxidation of methionine, acetylation of protein Nterminus and phosphorylation of serine and threonine.Raw data was then converted to mzML using the MSConvertGUI tool of ProteoWizard (version 3.0.23301-4fd35e2)(46), enabling peak picking (Vendor algorithm) on MS levels 1-2 and enabling the "SIM as spectra" option.Data was then analysed in R (version 4.3.1)for the coincidence of the precursor mass (578.2633m/z) and the highest abundant fragment ion of the phospho-peptide (neutral loss of y4 (y4*): 469.2518 m/z) (see representative spectrum in Figure S9).This coincidence appeared with increased density during an elution time window which overlapped with the elution time window as established by the MaxQuant analysis.By this the elution time window of the phospho-peptide in all samples was established for subsequent analysis in Skyline (v23.1.0.380 (cf25ad847)) (47,48) monitoring b1-8 and y1-8 ions and manual integration of the corresponding peaks.The Skyline transition data was exported as comma separated values.This file was processed in R (version 4.3.1),excluding the fragment ions y1-3 and b1-3 of both the phospho and nonphosphopeptide.First, the respective background area was subtracted from the area of each individual detected fragment ion.The background corrected areas were then summed up and standardized by dividing the area of the phospho-peptide by the sum of the area of the nonphosphopeptide and phospho-peptide.This ratio was then log2 transformed and statistically tested using a two-tailed student's t-test under assumption of equal variance between the two tested groups.A p-value ≤ 0.05 was regarded as evidence for a difference in the mean between the tested groups.The analysis scripts for the elution time window establishment of the phospho-peptide, the Skyline analysis raw data and all annotated spectra can be downloaded from Zenodo via https://doi.org/10.5281/zenodo.10882997.The raw files and MaxQuant search results are available under jPOSTrepo (HeLa: PXD050953, HEK-293: PXD050954).       A. Phosphorylation sites as called by MaxQuant and is here not FLR controlled.B. Sequence motifs detected in higher abundant phosphopeptides occurring common under both ATR inhibitor treatment regimes.Only peptides with single phosphorylation sites exhibiting a ≤ 1% FLR were taken into account.C.Gene Ontology terms of proteins with higher abundant phosphorylation sites common to gartisertib and berzosertib.The enrichment factor is calculated as the ratio of the portion of GO term occurrences within the set of proteins with higher abundant phosphorylation sites under inhibitor treatment and the portion of the GO term occurring in the complete protein set.Table S9.Lists of the reagents, antibodies, plasmid constructs, siRNA sequences, peptide sequences and primer sequences used in this study.Datasheets for each plasmid used in this study will be available on a dedicated page of our Reagents website upon publication.

Fig. S5
Fig. S5 Protein interaction network relating to phosphopeptides downregulated after gartisertib

Figure S2 .
Figure S2.Analysis of Gene Ontology and protein recruitment.(A).Gene Ontology terms

Figure S5 .
Figure S5.Protein interaction network relating to phosphopeptides downregulated after

Figure S6 .
Figure S6.Protein interaction network relating to phosphopeptides downregulated after

Figure S7 .
Figure S7.Proteins with phosphorylation sites upregulated upon ATR inhibition.

Fig. S8
Fig. S8 Protein interaction network relating to phosphopeptides commonly upregulated