GSE1 links the HDAC1/CoREST co-repressor complex to DNA damage

Abstract Post-translational modifications of histones are important regulators of the DNA damage response (DDR). By using affinity purification mass spectrometry (AP-MS) we discovered that genetic suppressor element 1 (GSE1) forms a complex with the HDAC1/CoREST deacetylase/demethylase co-repressor complex. In-depth phosphorylome analysis revealed that loss of GSE1 results in impaired DDR, ATR signalling and γH2AX formation upon DNA damage induction. Altered profiles of ATR target serine-glutamine motifs (SQ) on DDR-related hallmark proteins point to a defect in DNA damage sensing. In addition, GSE1 knock-out cells show hampered DNA damage-induced phosphorylation on SQ motifs of regulators of histone post-translational modifications, suggesting altered histone modification. While loss of GSE1 does not affect the histone deacetylation activity of CoREST, GSE1 appears to be essential for binding of the deubiquitinase USP22 to CoREST and for the deubiquitination of H2B K120 in response to DNA damage. The combination of deacetylase, demethylase, and deubiquitinase activity makes the USP22-GSE1-CoREST subcomplex a multi-enzymatic eraser that seems to play an important role during DDR. Since GSE1 has been previously associated with cancer progression and survival our findings are potentially of high medical relevance.


Introduction
DNA damage represents a challenge to the genomic integrity on a daily basis.To address this constant threat, the eukaryotic cell developed a system integrating multiple pathways that act synergistically to promote cell survival in response to DNA damage.Once DNA damage is sensed, a cell cycle arrest accompanied by DNA repair signalling is initiated.Impaired repair mechanisms have a severe impact on fitness of the organ-ism, leading to the onset of diseases such as cancer ( reviewed in ( 1 ) ) .
Post-translational modifications ( PTMs ) of histones such as phosphorylation, methylation, monoubiquitination and acetylation were shown to be important regulators of the DNA damage response ( DDR ) ( reviewed in ( 2 ) ) .For instance, protein kinases ATM ( ataxia telangiectasia mutated ) , DNA-PK ( DNA-dependent protein kinase ) or A TR ( A TM and Rad3 related ) phosphorylate histone variant H2AX at serine 139 ( γH2AX ) , which is one of the earliest markers of DNA damage and crucial for recruitment of downstream effector proteins such as MDC1, 53BP1, RAD50, RAD51 or BRCA1 to sites of DNA lesions (3)(4)(5) .In addition, histone H2A is ubiquitinated at the site of these lesions by the E3 ubiquitin ligases RNF8 / RNF168, and histone H2B is monoubiquitinated by the RNF20 / RNF40 ubiquitin-protein ligase complex.Ubiquitination of the two histones then enables the assembly of the DNA repair machinery (6)(7)(8)(9)(10) .However, formation of γH2AX may require dynamic regulation of H2B monoubiquitination, as loss of the H2B-specific deubiquitinase USP22 led to reduced levels of γH2AX ( 11 ) .The lysine-specific demethylase LSD1 / KDM1A can be recruited to sites of DNA damage by physical interaction with RNF168, catalysing demethylation of dimethylated lysine 4 of histone H3 ( H3K4me2 ) ( 12 ) .Alongside, H4K16 acetylation becomes increased upon ionising radiation ( 13 ) and deacetylation of H3K9 and H2K56 upon genotoxic stress were reported ( 14 ) .Experiments in diverse cancer cells showed that treatment with HDAC inhibitors can impair DNA damage repair.Based on these results it was proposed that specific histone deacetylation is necessary for the repair and for the restoration of the local chromatin environment (15)(16)(17)(18)(19) .In addition, PTMs of non-histone proteins play a role in DNA damage signalling (20)(21)(22) .Acetylation of ATM in response to DNA damage was reported to activate the kinase activity of the checkpoint kinase ( 23 ) .Furthermore, pharmacological inhibition of class I HDACs during replicative stress has been shown to promote mitotic catastrophe by affecting phosphorylation and acetylation of key regulators during replicative stress ( 24 ) .
There is increasing evidence that all these PTMs must be orchestrated sequentially and many of them are interdependent for proper signal propagation (25)(26)(27)(28)(29)(30) .Moreover, to resolve DNA damage signalling, these modifications must also be removed consecutively, which requires coordinated cooperation of respective eraser enzymes.Nevertheless, it is not understood how these enzymes uphold their activities and whether they function as a united entity within one or multiple specific complexes.Moreover, their capacity lies beyond histones since PTMs on non-histone proteins are being vastly identified and functionally annotated to the DDR in recent years (31)(32)(33)(34) .
In this study, we propose a potential mechanism for regulating the DDR by identifying a multi-enzymatic eraser subcomplex of HDAC1 / CoREST ( co-repressor of Repressor Element 1 Silencing Transcription Factor ) .This sub-complex combines the activities of the deacetylase HDAC1 and the demethylase activity LSD1 with the deubiquitinase activity of USP22.Using affinity-purification mass spectrometry ( AP-MS ) we identified GSE1 ( genetic suppressor element 1 ) as a prominent interaction partner of HDAC1, and more specifically, as an interactor of the CoREST complex.While loss of GSE1 did not affect the deacetylase activity of the CoREST complex, it resulted in reduced γH2AX formation and an impaired signalling response upon DNA damage.Further, by indepth quantitative mass spectrometry, we determine common and individual effects caused by loss of GSE1 or HDAC1 activity.We show that GSE1 acts at an initial step of the DDR.Ultimately, we demonstrate that GSE1 is required for the recruitment of the deubiquitinase USP22 to the CoREST complex, potentially serving as a scaffold, and that loss of GSE1 inhibits deubiquitination of histone H2B at lysine 120, which was previously shown to facilitate γH2AX formation ( 11 ,35 ) .
Through biochemical and phenotypic characterisation of GSE1, we unravel its role in response to DNA damage as a part of multi-enzymatic eraser complex together with HDAC1 / 2, USP22 and LSD1, which maintains proper chromatin environment and downstream DNA repair signalling.Hence, we emphasise the importance of understanding the cooperation of not only writers but also eraser enzymes within the kinetic window of DNA damage and the chromatin challenges that have to be resolved during DNA repair.

CRISPR-Cas9 gene editing
The GSE1 KO ( HZGHC007762c010 ) and 53BP1 KO ( HZGHC000413c010 ) HAP1 cell lines were obtained from Horizon Genomics ( now Horizon Discovery ) .To tag endogenous human Gse1 in HAP1 cells with a V5-tag, Precise Integration into target chromosome ( Crispr PITCh ) was used as described before ( 31 ) .The sequence of guide RNA was designed using an online web tool ( https://eu.idtdna.com/ site/ order/ designtool/ index/ CRISPR _ SEQUENCE ) of the IDTDNA company.The RNA duplex complex was assembled by adding 320 pmol guide RNA and 320 pmol tracr-RNA (ID TDNA) in ID TE buffer and annealed in the following reaction: 94 • C for 5 min, twelve times 93 • C for 20 s with -1 • C / cycle, 80 • C for 4 min, three times 79 • C for 20 s with -1 • C / cycle, 75 • C for 4 min, three times 74 • C for 20 s with -1 • C / cycle and finally at 70 • C for 4 min.Next, the RNP complex was prepared using RNA duplex with 5 μg Cas9 protein in 2 μl cleavage buffer (20 mM Hepes, pH 7.5, 0.15 M KCl, 0.5 mM DTT, 0.1 M EDTA) and incubated for 10 min at room temperature.After incubation, 2 × 10 6 of HAP1 cells were electroporated using The Neon Transfection System™ (MPK5000) set to 3 pulses, 10 ms width and 1575 V.The obtained single clones were cultured and screened for V5-tag integration by PCR (fw: 5 -GGT TTA CCC TTC CAT CC A GC A-3 ; rv: 5 -A GG A GA GGG TTA GGG A T A GGC-3 ).Guide RNA sequence: GCT ACC TGA AGG GA T A TC CC.Homology arm sequence: 5 -CTG GAC CAC TTA CGA AAG TGC CTT GCC TTG CCT GCA ATG CAC TGG CCT A GG GGC TA C CTG AA G GGA T A T CCC AGG GGT AAG CCT ATC CCT AAC CCT CTC CTC GGT CTC GAT TCT ACG TGA CGG TTT CCC TTG CAC TAG GCC GAA CCT

Cell cycle analysis
To monitor changes in cell cycle, fluorescent ubiquitination cell cycle indicator system (FUCCI) was used as described before ( 37 ).The 293T HEK cells were transfected with Lipofectamine TM 3000 (Thermo Fisher, L3000075) to produce lentiviral particles containing pBOB-EF1-FastFUCCI-PURO according to the manufacturer's protocol.48 h post transfection, media containing virus was collected, filtered, and added onto HAP1 cells.After 48 h, the positive cells were selected by addition of 1 μg / ml puromycin for two days.FUCCI positive cells were cultured and washed once with phosphate-buffered saline (PBS, 300 × g / 5 min) and fixed with 4% methanol-free formaldehyde (Thermo Fisher Scientific, 28906) for 20 min at room temperature.Afterwards, the samples were washed twice with PBS and resuspended in the final volume of 200 μl PBS and taken for measurement.The samples were measured with Cytoflex Flow Cytometer (Beckman Coulter Life Sciences) and analysed with CytExpert software (Beckman Coulter Life Sciences) or FLowJo (TreeStar Inc.).

Western blot analysis
To generate whole-cell lysates, cells were washed in 1 ml PBS on ice and centrifuged at 300 × g / 5 min.The pellets were lysed in Hunt buffer (20 mM Tris / HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% NP-40) supplemented with Complete Protease inhibitor cocktail (Roche 11697498001), Complete Phosphatase inhibitor cocktail (Roche, 11836145001), 10 mM sodium fluoride, 10 mM β-glycerophosphate, 0.1 mM sodium molybdate and 0.1 mM PMSF.The samples were frozen and thawed three times in liquid nitrogen and centrifuged (21 000 × g / 30 min, 4 • C).The supernatants were transferred, and protein lysates concentration measured by Bradford assay.Detection of proteins was performed by SDS-PAGE and western blots using nitrocellulose membranes and chemiluminescence.The blots were developed using ECL Western Blot detection reagents (Ge Healthcare, RPN2106) and visualised via chemiluminescence using Fusion FX7-Spectra machine.

Indirect immunofluorescence analysis
The cells were grown on coverslips.The cells were preextracted with 0.1% Tween-20 in PBS (PBST) for 1 min and then fixed with 4% methanol-free formaldehyde (Thermo Fisher Scientific, 28906) for 20 min at room temperature.The samples were washed two times with PBS and permeabilized with 0.5% Triton-X in PBS for 10 min at room temperature.After three times washing with PBS, the samples were blocked with 5% bovine serum in PBST for 1 h at room temperature.After blocking, the primary antibody was added in blocking solution and incubated overnight at 4 • C. The samples were washed three times with 3% bovine serum in PBS for 5 min and stained with secondary antibody in blocking solution for 1 h at room temperature in the dark.Next, samples were washed three times with 3% bovine serum in PBS for 5 min, incubated with DAPI in PBS for 10 min at room temperature and washed again two times with PBS.Afterwards, the slides were mounted using DAKO mounting medium and dried.Antibodies: DAPI (Sigma Aldrich, D9542-10MG,1:10000), γH2AX (Millipore, JBW301,1:1000), H3Ser10ph (Santa Cruz, 8656, 1:2000), Alexa Fluor 546 (Thermo Scientific, A11003,1:500), Alexa Fluor 488 (Thermo Scientific, A32731,1:500).
The staining was monitored by Olympus Confocal Microscope and the signals were quantified using ImageJ software.Either number of foci was counted and compared, or signal strength (integrated density) was measured.For quantification of G2 and M phase positive cells stained with H3Ser10ph antibody.As described in De La Barre et al. ( 39 ) the cells were counted as G2 positive when pericentromeric chromatin was visible as foci and as M positive when the signal was fully saturated across the chromatin.

Co-immunoprecipitation assays
Whole-cell lysates were prepared as described above (Protein lysates, western blot).40 μl Protein A or Protein G Dynabeads (Thermo Fisher Scientific, 1002D / 1004D) per sample were washed three times with Hunt buffer, resuspended in Hunt buffer containing 10% bovine serum albumin and incubated for 1 h rotating at 4 • C for blocking.At the same time, the whole-cell extract (0.5-1 mg protein) was incubated with an antibody mix suitable for immunopurification of the protein of interest for 1 h rotating at 4 • C.After blocking, the beads were added to the cell lysate-antibody mix and incubated overnight at 4 • C rotating.Next day, the beads were washed twice with TBS for 5 min and once for 15 min at 4 • C. Subsequently, the purified proteins were eluted from the beads in 1 × SDS loading dye by heat-incubation at 95

Cell extract and mass spectrometry sample preparation
Whole cell extracts were prepared as follows: cell pellets were resuspended in Hunt buffer (20 mM Tris / HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% NP-40), sonicated twice for 3 s (Bandelin Sonoplus, power of 45-50%), and centrifuged at 21 000 × g for 15 min at 4 • C. Protein concentration of supernatants was determined via Bradford.1 mg of protein was used for each sample.
80 μl of V5-beads (MBL International, M167-10) per sample were used for isolation of Gse1-V5, 120 μl of crosslinked M2-FLAG beads (Sigma Aldrich, M8823-5ML) were used for purification of HDA C1-FLA G. Briefly, for antibody crosslinking the bead slurry was equilibrated with sodium borate buffer (200 mM sodium borate pH 9.0), resuspended in 20 mM dimethyl pimelimidate in sodium borate buffer, incubated rotating for 1 h at room temperature, washed and incubated with 200 mM ethanolamine pH 8.0 for 2 h, washed with 100 mM glycine and subsequently stored in 1 × TBS.Beads were washed three times with Hunt buffer, resuspended in whole cell extracts and incubated overnight rotating at 4 • C. Next day, the beads were washed twice with TBS for 5 min and once for 15 min rotating at 4 • C. Beads were transferred to a 0.2 ml PCR tube, re-buffered to and washed once with ammonium bicarbonate (ABC) (Sigma Aldrich, 09830-500G) buffer, pelleted by gentle centrifugation, and resuspended in 50 μl 2 M urea (VWR, 0568-500G), 50 mM ABC.
Predigestion of bound proteins was performed by addition of 200 ng / μl of a 1:1 LysC (FUJIFILM Wako Pure Chemical Corporation, 125-02543) trypsin (Trypsin Gold, Mass Spec Grade, Promega, V5280) mix, followed by an incubation at RT for 90 min in the dark.The eluate containing predigested proteins and peptides was separated from the beads, and beads were subsequently rinsed with 50 μl ABC.The eluates were united.

Mass spectrometry measurements
MS measurements were performed on a Q Exactive HF Orbitrap (with HCD , higher -energy collisional dissociation mode) mass spectrometer (Thermo Fisher), operated in datadependent acquisition (DDA) mode with dynamic exclusion enabled.For peptide separation on the HPLC, the concentration of organic solvent (acetonitrile, VWR, 83639.320)was increased from 1.6% to 32% in 0.1% formic acid at a flow rate of 230 nl / min, using a 2 h gradient time.Peptides were ionised with a spray voltage of 1,9 kV.MS1 survey scans were recorded with the following parameters: resolution 120 000, scan range 375-1500 m / z , automatic gain control (AGC) target = 3e6, and maximum injection time (IT) = 60 ms.MS2 analysis was performed with the following parameters: resolution 30 000, AGC target = 1e5, maximum IT = 250 ms, isolation window 1.6 m / z .The mass spectrometer was configured to pick the eight most intense precursor ions for datadependent MS2 scans, applying HCD for fragmentation with a normalised collision energy (NCE) of 28.Dynamic exclusion time was set to 30 s.

Mass spectrometry data analysis with MaxQuant
MS raw files were searched using MaxQuant (version 2.1.4.0) against the Homo sapiens (human) (HAP1 samples) or the Mus musculus (mouse) (MEF samples) Uniprot database (release 2022.02) and a database of common laboratory contaminants.Enzyme specificity was set to 'Trypsin / P'.Carbamidomethylation of cysteine was searched as a fixed modification.Protein N-term acetylation and Oxidation (M) were set as variable modifications.The data were filtered at 1% FDR (PSM and protein), reverse hits were removed.MaxLFQ and match-between-runs were activated.MaxQuant output was further processed in R using Cassiopeia (DOI: 10.5281 / zenodo.5758975).After removing contaminants and proteins with less than two razor & unique peptides, missing LFQ protein intensities were imputed if at least two valid values were present in one experimental group.Imputation was performed in log2 space by drawing random values from a normal distribution ( μ = mean sample LFQ intensity -1.8 standard deviations of the sample LFQ intensities, σ = 0.3 × standard deviation of the sample LFQ intensities).LIMMA was used to determine differentially enriched proteins at 1% FDR (Benjamini-Hochberg).
Inspection of the HDA C1-FLA G IP data obtained from HAP1 cell extracts (Figure 1 A) revealed that MaxLFQ did not properly normalise the data due to the high number of enriched proteins, violating the assumption of MaxLFQ that the majority of proteins is not regulated.For this reason, in a first analysis Cassiopeia was used to determine a subset of background proteins (determined by k-means clustering in Cassiopeia) for the respective MQ output files.Then the Cassiopeia script was run again on the original proteins-Groups.txtfile but specifying the background protein cluster to be used for normalisation in the Cassiopeia script.In contrast to the original Cassiopeia script LFQ values were used for renormalisation instead of raw intensities.

Preparation of cells
Three biological replicates of each experimental condition were prepared.To induce DDR cells were treated with 0.1 μM etoposide (Sigma Aldrich, E1383-25MG) for 6 h.Cells were washed twice in 1 × PBS, supplemented with 10 mM sodium butyrate (Sigma Aldrich, 303410) and 3 mM nicotinamide (NAM) (Sigma Aldrich, 72340), and pelleted prior to lysis.

Acetylated peptide enrichment
Acetylated peptides were enriched, washed, eluted and desalted as described in ( 36 ), using the PTMScan Acetyl-Lysine Motif Kit (Cell Signaling Technology, 11663V1).The flowthrough of the Acetyl-Lysine immunoprecipitation was kept for phosphorylated peptide enrichment.Eluted peptides were dried to completeness in a SpeedVac concentrator (Eppendorf) and subsequently resuspended in 0.1% TFA.

Mass spectrometry measurements
DDA-SPS-MS3 method with online real-time database search (RTS) (for proteome analysis) was performed on an UltiMate 3000 Dual LC nano-HPLC System (Dionex, Thermo Fisher Scientific), containing both a trapping column for peptide concentration (PepMap C18, 5 × 0.3 mm, 5 μm particle size) and an analytical column (PepMap C18, 500 × 0.075 μm, 2 μm particle size, Thermo Fisher Scientific), coupled to a Orbitrap Eclipse mass spectrometer (Thermo Fisher) via a FAIMS Pro Duo ion source (Thermo Fisher).The instruments were operated in data-dependent acquisition (DDA) mode with dynamic exclusion enabled.For peptide separation on the HPLC, the concentration of organic solvent (acetonitrile, VWR, 83639.320)was increased from 1.6% to 32% in 0.1% formic acid at a flow rate of 230 nl / min, using a 2 h gradient time for proteome analysis.Peptides were ionised with a spray voltage of 2.4 kV.The instrument method included Orbitrap MS1 scans (resolution of 120 000; mass range 375 −1500 m / z ; automatic gain control (AGC) target 4e5, max injection time of 50 ms, FAIMS CV −40 V, −55 V and −70 V, dynamic exclusion 45 s), and MS2 scans (CID collision energy of 30%; AGC target 1e4; rapid scan mode; max injection time of 50 ms, isolation window 1.2 m / z ).RTS was enabled with full trypsin specificity, max. 2 missed cleavages, max.search time of 40 ms, max. 2 variable modifications, max. 10 peptides per protein, and considering the following modifications: Carbamidomethyl on cysteine and TMT-6-plex on Lys and peptide N-terminus as static, oxidation on Met and deamidation on Asn and Gln as variable modifications.The scoring thresholds were applied (1.4 for Xcorr, 0.1 for dCn, 10 ppm precursor tolerance) and the Homo sapiens Uniprot database (release 2020.01) was used for the search.Quantitative SPS-MS3 scans were performed in the Orbitrap with the following settings: HCD normalised collision energy 50%, resolution 50 000; AGC target 1.5e5; isolation window 1.2 m / z , MS2 isolation window 3 m / z , 10 notches, max injection time of 150 ms.Total cycle time for each CV was set to 1.2 s.
DDA-MS2 method (acetylome and phosphorylome analysis) was performed on an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher) and similar HPLC and nanospray settings as described above, with the exception of a FAIMS Pro ion source and that 3 h gradient time was applied.MS1 survey scans were recorded with the following parameters: resolution 60 000, scan range 375-1500 m / z , automatic gain control (AGC) target = Custom, and maximum injection time (IT) mode = Auto, FAIMS voltages on, FAIMS CV −40 V, −55 V and −70 V, dynamic exclusion was set to 45 s.MS2 analysis was performed with the following parameters: resolution 45 000, normalised AGC Target 200%, AGC target = Custom, maximum IT mode = Custom, isolation window 0.7 m / z , and HCD normalised collision energy of 34%.

Mass spectrometry data analysis with MaxQuant
MS raw files were split according to CVs ( −40 V, −55 V, −70 V) using FreeStyle 1.7 software (Thermo Scientific).All raw MS data were commonly analysed using MaxQuant ( 42 ) software version 1.6.17.0, using default parameters with the following modifications.Proteome (MS3), acetylome and phosphorylome (MS2) raw data were included in the same search but in separate parameter groups.MS2 and MS3 spectra were searched against the Homo sapiens (human) Uniprot database (release 2020.01; with isoforms) and a database of common laboratory contaminants.Enzyme specificity was set to 'Trypsin / P', the minimal peptide length was set to 7 and the maximum number of missed cleavages was set to 2 for proteome and phosphorylome data and to 4 for the acetylome data.Carbamidomethylation of cysteine was searched as a fixed modification.'Acetyl (Protein N-term)', 'Oxidation (M)' were set as variable modifications.A maximum of 5 variable modifications per peptide was allowed.For search groups containing acetylome and phosphorylome files 'Acetyl (K)' and 'Phospho (STY)' were, respectively, set as variable modification.The identification and quantification information of acetylated peptides, sites, and proteins was obtained from the MaxQuant 'Acetyl (K) Sites', 'Phospho (STY) Sites', 'ModificationSpecificPeptides', and 'ProteinGroups' tables.
Data were analysed in R (4.1.0)using custom scripts (Supplemental Material).The analysis procedure covered: correction for isotopic impurities of labels, within-plex median normalisation, between-plex IRS-normalisation ( 43 ) and statistical between-group comparisons using LIMMA (3.50).To account for differences of protein abundance on lysine acetylation-site level, site intensities were normalised to protein intensities before group comparisons with LIMMA.S3 and S4 were created in Perseus (version 1.6.15.0), for clustering the k -means algorithm was used.TMT reporter ion intensities were log 2 transformed and z-scored.Mapping of human to mouse proteins was done using the ortholog database from OrthoInspector 3.0 (database version 2019) ( 45 ), proteins with no annotated ortholog between human and mouse were excluded from the analysis (Supplementary Figure S1).For bar plots shown in Figures 4 C and 5 B TMT reporter ion intensities were log 2 transformed and z -scored using inhouse Python scripts.Bands of Western blot images were cut out using Adobe Photoshop version 10.0 and figures were prepared using Adobe Illustrator version 16.0.0.

Integration of published mass spectrometry datasets
For integration of published datasets, we used in-house Python scripts.The list of previously described HDAC1 interactors in CEM T-cells was obtained from Joshi et al. ( 46 ) 'Supplementary Table S2', by selecting entries with a value > 0.9 in the 'HDAC1_Prob' column, as described by the authors.
SILAC ratios of acetylome and phosphorylome changes in response to 10 μM etoposide treatment for 24 h in U2OS cells were obtained from Beli et al .( 47 ).Phosphorylation, acetylation, and protein SILAC ratios were extracted from the 'Etoposide' sheet in 'T able S1', 'T able S2' and 'Table S3' using the column 'Ratio H / L Etoposide' or 'Ratio H / L Etoposide Combined'.SILAC ratios of phosphorylation and acetylation sites were normalised to protein SILAC ratios, sites of not quantified proteins were discarded.Phosphorylation and acetylation sites were mapped to the Homo sapiens (human) Uniprot database (release 2020.01) by using the 'Gene Names', 'Position' and 'Sequence Window' columns, and sites with ambiguous mapping were removed.Normalised SILAC ratios were then integrated into Supplementary Tables S3 and  S4 as column 'Ratio H / L etoposide [log2]'.

Protein network analysis
Protein-protein interaction networks were generated using the STRING database v11.5 ( 48 ), network parameters were exported and modified as follows: column 'node1' was defined as 'source', column node 2 was defined as 'target' and 'combined STRING score' was defined as weight of the network edge.Network type was undirected.The modified table was imported into Gephi v0.9.7 ( 49 ) for network visualisation.Node sizes were set proportional to the degree value and edge thickness to the combined interaction score.Network was visualised using the Fruchteman Reingold function with default settings and ForceAtlas 2 function ( 50 ) using balanced scaling and gravity settings, and the 'prevent overlaps' option being activated.All other settings were default.Nodes were either coloured according to the classification presented in ( 46 )

GSE1 is a prominent interactor of HD A C1
The class I histone deacetylase HDAC1 is the catalytic component of several multiprotein co-repressor complexes, present in virtually all mammalian cells.To gain a deeper understanding of the functional landscape of HDAC1, we examined the composition of HDAC1-associated proteins in different cellular backgrounds.First, we affinity purified (AP) C-terminally FLA G-tagged HDA C1 from the nearly haploid human cell line HAP1 ( 36 ,51 ) and analysed the interactome using quantitative mass spectrometry (MS).In total, we identified 161 proteins as potential HDAC1 interaction partners (fold change [log 2 ] > 1.5, adjusted P -value < 0.05), thereby providing a comprehensive analysis of the interactions of the enzyme (Figure 1 A, Supplementary Table S1).
The presence of well-known subunits of co-repressor complexes such as NuRD, Sin3-related, MiDAC or CoREST ( 46 ) in the HDAC1 interactome indicated that our analysis adequately covered anticipated binding partners of the deacetylase.In addition, we found the HDAC1 substrates H2AFV, H2AFX ( 52 ) and the prefoldin / CTT complex ( 53 ) among the co-precipitated proteins.We also confirmed interaction between HDAC1 and a group of importins, zinc finger proteins or transcriptional regulators possessing an ELM-SANT domain (Figure 1 B).
When comparing our results to those obtained from a similar approach in human CEM T-cells ( 46 ), we observed a high level of overlap in the identified binding partners of HDAC1 (Supplementary Figure S1A).78 of the 91 HDAC1-interacting proteins reported by Joshi et al. ( 46 ) were also identified in the HAP1 dataset.This suggests stable HDAC1-complex composition across different cell types.However, we have covered additional 83 proteins as HDAC1 interactors that were not reported by Joshi et al., providing an in-depth view on potential HDAC1 binders and cell type specific differences.Among these we found many transcriptional regulators such as TFDP1, MAX, ZEB2, CTBP2 and MECOM, and negative regulators of ubiquitination BAG2, OGT or SQSTM1 (Figure 1 B).
To examine cross-species conservation in the composition of HDAC1 complexes we additionally analysed HDA C1-FLA G purifications from mouse embryonic fibroblasts (MEFs) (Figure 1 C and Supplementary Figure S1B, Supplementary Table S2).Again, we observed a high level of overlap (68 proteins) in the identified interaction partners.Those in common contained mainly known corepressors subunits, as expected.Interestingly, a wide fraction of MEFspecific HDAC1 binders contained proteins involved in proteasomal degradation, transport or mitochondria related processes (Supplementary Figures S1C and S1D).We thus recovered generally accepted hallmarks of the HDAC1 interactome and demonstrate data consistency with previous studies.Additionally, we provide a view on the HDAC1 interactome across different species and cell types showing a high level of conservation among core corepressor complex subunits as well as variations specific for species or cell type.It is worth mentioning that our findings align closely with those of a comparable study on the HDAC1 interactome in Th1 T-cells (Ci, Stolz et al ., doi: https:// doi.org/ 10.1101/ 2023.04.14.536700 , manuscript in preparation).
Remarkably, a previously reported interaction between HDAC1 and Genetic suppressor element 1 (GSE1) ( 46 ,54-57 ), that has been not fully appreciated so far, was found across all experimental conditions tested (fold change [log 2 ] = 12.37, adjusted P -value < 0.001, Figure 1 A, B Supplementary Table S1, Supplementary Table S2).We additionally confirmed this interaction by co-IP experiments in combination with western blot analysis in murine Th1 T-cells (Figure 1 D), thereby also confirming cross-species conservation of a GSE1-HDAC1 complex.Since the binding of GSE1 to HDAC1 appeared prominent we decided to characterise this interaction in more detail.
We first investigated whether GSE1 binding possibly affects the enzymatic activity of HDAC complexes.To this end, we employed and expanded a toolbox of transgenic human tumour cell lines ( 51 ), previously created in our research group ( 36 ).In particular, we used HAP1 cells expressing either C-terminally FLAG-tagged wildtype (WT) or catalytic inactive (CI) mutants of HDAC1 or its paralog HDAC2 in the respective knockout background (Supplementary Figures S2A and S2B).Using a GSE1-specific antibody, we purified GSE1-containing HDAC complexes from transgenic HAP1 cells and determined the associated deacetylase activity using H 3 -acetylated histones as substrate ( 36 ,38 ).We observed a robust deacetylase activity in cells expressing wildtype HDAC1 (HDAC1 WT) as well as HDAC2 (HDAC2 WT), demonstrat- ing that GSE1-containing HDAC complexes are active (Figure 2 A).Moreover, this activity was significantly reduced in cells expressing catalytically inactive HDAC1 (HDAC1 H141A, HDAC1 CI), but not in cells expressing catalytically inactive HDAC2 (HDAC2 H142A, HDAC2 CI).HDAC1 KO cells showed only mild reduction in deacetylase activity, whereas HDAC2 KO showed increased activity when compared to HDAC2 WT (Figures 2 B and C).This is most likely due to a compensatory effect of upregulated HDAC1 in the absence of HDAC2 ( 36 ).In summary, these data suggest that HDAC1 is the main deacetylase of GSE1-containing complexes.Next, we used GSE1 KO cells to test whether loss of GSE1 affects HDAC1 and HDAC2 deacetylase activity.Activity assays of HDAC1-or HDAC2-associated complexes, isolated from wildtype and GSE1 KO cells, showed no significant difference (Figures 2 D and E), suggesting that deacetylase activity is not dependent on GSE1.As expected, an HDAC1 KO had no effect on the activity of HDAC2-associated complexes whereas HDAC1-associated complex activity was abrogated in HDAC1 KO cells (Figures 2 D and E).
In summary, we find that HDAC1 interacts with known co-repressor complexes and various other proteins, among which GSE1 was identified as a prominent binding partner within different cell types (human HAP1, MEFs and mouse Th1 T-cells).In addition, we demonstrate that GSE1-HD AC1 / HD AC2 complexes are catalytically active, and their deacetylase potential is dependent on HDAC1 and HDAC2 activity.Moreover, GSE1 ablation has no impact on deacetylase activity of HDAC1 and HDAC2.

GSE1 KO leads to reduced response to DNA damage
A genetic screen performed by Gudkov et al. was first to identify GSE1 as a potential mediator of cellular sensitivity to the topoisomerase II inhibitor etoposide ( 58 ).However, this study did not investigate the molecular function of GSE1.To test whether GSE1 plays a role in the DDR, we treated HAP1 wildtype and GSE1 knock-out (KO) cells with etoposide for 24 h and determined the level of phosphorylation at S139 of histone H2AX (designated as γH2AX), a marker of DNA double strand breaks.Using indirect immunofluorescence microscopy and western blot analysis, we observed an increase in γH2AX in wildtype cells upon etoposide treatment.This increase, however, was severely reduced in GSE1-negative cells (Figures 3 A and B).To corroborate this finding, we counted the number of DNA damage foci (designated as γH2AX foci).The analysis revealed a decreased number of foci upon etoposide treatment in GSE1 KO cells when compared to wildtype (Supplementary Figure S3A).In addition, we followed phosphorylation of serines 4 and 8 of Replication protein A 32 (RPA32), a known marker of DNA damage and replication stress ( 59 ).Consistent with the observed impairment in γH2AX formation, we detected a reduced phosphorylation of RPA32 in response to etoposide treatment in GSE1-negative cells compared to the wildtype (Supplementary Figure S3B).Next, we monitored caspase 3 cleavage to determine the level of apoptotic cells by western blot analysis in cells treated with etoposide for 12, 18 and 24 h.We detected reduced cleavage of caspase 3 in GSE1 KO cells compared to the control, indicating a reduced level of DNA damage induced apoptosis in GSE1 KO cells (Figure 3 B).Moreover, complementation of GSE1 fully restored etoposide-induced γH2AX formation upon 24 h of etoposide treatment (Figure 3 C).Furthermore, the application of the caspase inhibitor z-VAD-FMK led to an additional decrease in etoposide-induced RPA32 phosphorylation at serines 4 and 8 in GSE1 KO cells.This suggests that, akin to the wildtype, the DNA damage response in GSE1deficient cells is partly dependent on apoptosis (Supplementary Figure S3B).
Etoposide is known to form a complex with topoisomerase II that inhibits DNA synthesis and in turn induces DNA damage via double strand breaks ( 60 ).There are, however, various types of DNA damage that can be repaired by different mechanisms during different stages of the cell cycle ( 61 ).To test whether deletion of GSE1 also affects other repair mechanisms that are not induced by DNA double strand breaks, we used the alkylation reagent mitomycin C, as well as the DNA synthesis inhibitor hydroxyurea as alternative DNA damage inducers.Similar to the effect of etoposide treatment, we observed a significant reduction in γH2AX signal, γH2AX foci formation and caspase 3 cleavage in GSE1 KO cells upon treatment with mitomycin C or hydroxyurea (Figures 3 A and  B and Supplementary Figure S3A).These data strongly suggest that GSE1 plays a central role in γH2AX formation independent of the type of DNA damage.
Next, we inquired whether DNA damage induced by etoposide, hydroxyurea or mitomycin C results in a higher occurrence of single or double strand breaks in GSE1 KO cells.To access this, we monitored formation of Replication protein A (RPA) and p53 binding protein 1 (53BP1) foci, which are indicative for single and double strand breaks respectively.Indirect immunofluorescence microscopy revealed a moderate increase in the number of 53BP1 foci in untreated conditions in GSE1 KO cells in comparison to wildtype, which is even more prominent in response to DNA damage treatment (Supplementary Figures S3C and S3D).Notably, we also noticed a subtle increase in the number of γH2AX foci under untreated conditions in GSE1 KO cells (Supplementary Figure S3A), which is consistent with this observation and suggests a partial constitutive activation of the DDR in GSE1 KO cells.In summary, deletion of GSE1 results in a higher number of double strand breaks.

DNA damage does not impact GSE1-HD A C1 complex deacetylase activity
Increased sensitization of cancer cells with decreased homologous recombination repair capacity was observed upon HDAC inhibition in combination with antitumor drugs ( 18 ,62-64 ).Specifically, inhibition of class I HDACs by high dosage of Entinostat (MS-275) induced γH2AX, significantly inhibited cellular growth and triggered apoptosis ( 65-67 ).However, effects of higher MS-275 concentrations seem to be cell type dependent and did not always cause DNA damage ( 24 ,68 ).To avoid cytotoxic side effects caused by the high dose of HDAC1 inhibitor ( 64 ), we decided to apply a mild dose of MS-275 (1 μM) that functionally inhibits class I HDACs yet does not cause detrimental DNA damage.We tested whether HDAC inhibition has a synergistic effect with etoposide on the DDR in our cellular system and whether this response is altered by loss of GSE1.As expected, we could not detect a significant induction of γH2AX with MS-275 alone in HAP1 WT or GSE1 KO cells (Figure 4 A).However, in combination with etoposide, γH2AX signal became increased in both cell lines when compared to etoposide alone pointing   towards synergistic influence on DDR induction.Interestingly, these data show that inactivation of HDACs by MS-275 partly restores γH2AX signalling in GSE1 KO cells.A possible explanation for this is that GSE1 may negatively regulate the activity of HDAC1 in the context of DNA damage.To address this, we treated HAP1 wildtype cells for 6 h with etoposide, immunoprecipitated GSE1-HDAC1 complexes and measured deacetylase activity.We did not detect any changes in activity of GSE1-HDAC1 complexes upon etoposide treatment (Supplementary Figure S3E).This suggests that the enzymatic activity of GSE1-HDAC1 complexes is rather stable upon DDR induction, and that GSE1 possibly affects γH2AX formation independent of changes in HDAC1 activity.

GSE1 does not regulate the DNA damage response via acetylation
DDR is vastly orchestrated by a sequential cascade of multiple signalling events that are tightly regulated.In addition, a functional role between phosphorylation and acetylation crosstalk was established in cardiomyocytes where inhibition of HDACs can lead to increased phosphorylation of proteins such as p38, ERK1, and AKT (reviewed in ( 69)).
To further investigate how HDAC1 and GSE1 can regulate DDR and whether their mode of action might be overlapping, we next performed a proteomic analysis using quantitative mass spectrometry based on the recently developed TMTpro 16-plex isobaric labelling technology ( 70 ).We determined global proteome, acetylome and phosphorylome profiles in HAP1 WT, GSE1 KO and HDAC1 CI cells and compared patterns between untreated and etoposide-treated (0.1 μM) cells (Supplementary Figure S4A, Supplementary Tables S3 and S4).To avoid adaptive responses and bystander effects, etoposide treatment was limited to 6 h.In total, we covered 9051 proteins, 15486 acetylation sites and 24602 phosphorylation sites in our quantitative analysis.Changes in phosphorylation and acetylation site abundance were further normalised to the corresponding protein levels.Principal component analysis of the proteome, acetylome and phosphorylome separated out distinct clusters for wildtype, GSE1 KO and HDAC1 CI cells, pointing towards quantitative differences between the genetic backgrounds and indicating a high level of reproducibility between replicates (Supplementary Figure S4B).Phosphorylation sites showing a fold change of ≥1.5 over wildtype cells and an adjusted P -value of ≤ 0.05 were considered as significantly regulated.However, a higher level of variance was observed in untreated acetylome samples.We therefore used more stringent filtering for the acetylome dataset (fold change ≥ 2 over wildtype cells, adjusted P -value ≤ 0.01).
First, we investigated etoposide-induced acetylome changes in HAP1 WT, GSE1 KO and HDAC1 CI cells.We detected a rather small population of increased lysine acetylation of sites in HAP1 WT upon etoposide treatment, which is in accordance with the previous studies, reporting only minor changes in acetylation levels upon 24 h of etoposide treatment ( 47 ) or in response to irradiation ( 71 ).The number of acetylation sites affected by etoposide treatment was reduced in both HDAC1 CI (19 sites) and GSE1 KO (7 sites) cells (Figure 4 B and Supplementary Figure S4C).
Next, we analysed how the mutants affect the set of DNA damage-induced acetylation sites.We selected the set of 46 sites induced in the wildtype and studied their behaviour at basal level and after etoposide treatment.After etoposide treatment, acetylation of those sites reached a similar level in the mutants as in the wildtype.When looking at untreated conditions, however, this set of sites displayed elevated acetylation levels in both, HDAC1 CI as well as in GSE1 KO cells, (Figure 4 C), indicating that both mutants cause dysregulation in DDR signalling.Therefore, we tested whether both mutants had the same effect on the global acetylome.When comparing global acetylation patterns of HDAC1 CI to GSE1 KO cells, we found 102 sites of different abundance (fold change ≥ 2, adjusted P -value ≤ 0.01), suggesting that the mutants have to a certain extent individual effects on the acetylome (Supplementary Figure S4D).In line with this, recently described substrate sites of HDAC1 ( 36 ) show higher acetylation levels in the HDAC1 CI mutant (Supplementary Figure S4D), which also supports our observation that GSE1 does not affect HDAC activity (Figure 2 D).We next wanted to get an insight into DNA damage-induced processes that are commonly dysregulated in GSE1 KO and HDAC1 CI.We focussed on acetylation sites that are etoposide-responsive (fold change ≥ 2, adjusted P -value ≤ 0.01, in wildtype cells) in wildtype cells but remain unaffected in both mutants (Figure 4 D and E).The set of corresponding proteins was submitted to a protein network analysis using the STRING database tool ( 48 ).We found that proteins harbouring these acetylation sites are part of cell cycle control networks (Figure 4 F), pointing to an affected cell cycle regulation in these mutants.Based on this finding, we asked whether protein levels of cyclins and cyclin dependent kinases are altered in the mutants.Indeed, in both mutants we observed strong differences in the protein abundance of several cyclins known to regulate cell cycle transition points ( 72-75 ) (Figure 4 G).Decreased abundance of cyclin dependent kinases CDK1, CDK2, and CDK6 was observed in GSE1 KO as well as HDAC1 CI cells (Supplementary Figure S4E).To monitor potential changes in cell cycle distribution, we used the fluorescent ubiquitination cell cycle indicator (FUCCI) system developed by Sakaue-Sawano et al. ( 37 ).We observed a similar cell cycle phase distribution between GSE1 KO and HDAC1 CI cells that was different from HAP1 WT cells (Figure 4 H and Supplementary Figure S4F).GSE1 KO and HDAC1 CI showed an increased fraction of cells in S / G2 / M and G1 / G0 phase and a reduced number of early G1 cells compared to HAP1 WT (Figure 4 H).In response to treatment with etoposide, the wildtype and both mutants showed an increase in S / G2 / M phase (Supplementary Figure S4F), which is in line with previous reports (76)(77)(78).Based on these findings, we wanted to clarify whether deletion of GSE1 leads to an accumulation of G2 cells, which could indicate DNA repair incompetence.Cells were stained for phosphorylation of histone H3 at Ser10 that is known to correlate with chromosome condensation during cell division.Quantifying the percentage of cells in G2 and M phase revealed a significantly increased number of GSE1 KO cells in G2 phase when compared to HAP1 WT (Supplementary Figure S4G).These data indicate that the incompetent DNA repair in GSE1 KO cells causes a G2 / M arrest.We also observed a decreased proliferation rate in the GSE1 KO compared to HAP1 WT (Figure 4 I).In addition, when cultivating the cells in etoposide containing media, GSE1 KO cells acquired etoposide resistance and restored their growth after 96 h in presence of the drug, whereas the vast majority of the control cells were dead (Figure 4 J).These findings are consistent with the previous report from Gudkov et al. ( 58 ).
Taken together, we find that the overall impact of the DDR on the acetylome seems to be rather neglectable.Etoposideresponsive sites that are commonly affected in both mutants are most likely caused by cell cycle defects.Therefore, we conclude that the mode of action of GSE1 in regulation of DDR is independent of the changes in the acetylome.
Loss of GSE1 affects early events of DDR and attenuates downstream DNA repair signalling Since GSE1's role in response to DNA damage does not depend on protein acetylation, we next examined DNA damageinduced phosphorylation changes after 6 h of etoposide treatment.In total, we identified 657 (2.7%) up-and 1859 (7.6%) down-regulated phosphorylation sites in HAP1 WT cells (fold change ≥ 1.5, adjusted P -value ≤ 0.05) in response to DNA damage (Figure 5 A).We observed a high level of correlation between our results and those of Beli et al. ( 42 ), monitoring etoposide-induced phosphorylation patterns 24 h after induction with etoposide in osteosarcoma U2OS cells (Supplementary Figure S5A).Notably, 53.6% of commonly induced sites constitute serine glutamine (SQ) motifs, which are putative A TR / A TM substrates and a hallmark of the DDR ( 79 ).About a quarter of the commonly affected SQ sites are found on DDR-related proteins such as BRCA1, FANCI, MCM6, MLH1, NBN and SMC1A.We also confirmed the etoposideinduced phosphorylation of THRAP3 serine 211, which was identified as a novel DDR regulator by Beli et al. ( 47 ).Phosphorylation sites that were only affected after 24 h of etoposide treatment might correspond to downstream effects of the DDR or secondary effects caused by prolonged DNA damage.
Next, we focused on how a GSE1 knock-out or loss of HDAC1 catalytic activity affects DNA damage-induced phospho-signalling.We observed a reduced number of etoposide-affected phosphorylation sites in GSE1 KO cells (41.4% upregulated and 35.2% downregulated sites compared to wildtype cells) and to a lesser extent also in the HDAC1 CI mutant (62.4% upregulated, 78.8% downregulated sites compared to wildtype cells) (Figure 5 A).Notably, the set of 657 sites that become induced by etoposide treatment in wildtype cells were already slightly elevated in phosphorylation at basal conditions in GSE1 KO and HDAC1 CI cells (Figure 5 B), pointing to a partial constitutive activation of the DDR as previously suggested (Figure 4 B, Supplementary Figures S3A, S3B and S3C).However, when the mutants were exposed to etoposide, these sites were not phosphorylated to the same magnitude as in the wildtype.Moreover, both mutants showed a reduced etoposide-induced phosphorylation at several hallmark sites of the DDR, such as γH2AX, pT1989 of ATR, pS317 of CHEK1, when compared to wildtype, indicating that DNA damage-signalling is hampered (Figure 5 C and Supplementary Figure S5B).On the other hand, activating autophosphorylation at S1981 of ATM kinase was induced upon etoposide in HAP1 WT and HDAC1 CI but to a lesser extent in GSE1 KO (Figures 5 A and C).Therefore, we next asked whether the GSE1 KO or HDAC1 CI mutant affects the entire DNA damage-induced phosphorylation response or just a specific set of sites indicative for individual pathways.In fact, the GSE1 KO mutant showed generally hampered phosphorylation at SQ motifs (r-squared = 0.587) and non-SQ motifs in response to etoposide treatment, including the entire set of targets of ATM / ATR kinases (Figure 5 D).On a side note, HDAC1 CI showed similar results, albeit to a lesser degree (R-squared = 0.784) and without dysregulation of A TM / A TR kinase regulatory sites (Supplementary Figure S5C).In conclusion, GSE1 appears to play a role in the initial stages of DNA damage detection and recognition, which are essential for full activation of phosphorylation cascades downstream.
To identify such upstream regulators that could be affected by GSE1 we created a STRING database-based protein network of the etoposide phosphorylation response (proteins harbouring SQ motifs, FC ≥ 1.5, P adj-value of ≤0.05, in wildtype cells).Analysis of the networks revealed proteins mostly associated with DDR and DNA repair, but also with spliceosome, ubiquitination machinery, and histone modifications (Figure 5 E).The latter seemed particularly relevant in the context of DNA damage sensing ( 80 ), encompassing factors such as EYA3, associated with dephosphorylation of histone H2AX Y142 ( 81 ), EHMT2 and TRIM28, both associated with methylation of K9 of histone H3 ( 82 ,83 ), UIMC1, selectively binding K63 of histones H2A and H2AX ( 84 ), and RNF20 mediating monoubiquitination of K120 of histone H2B ( 85 ).We speculated that ablation of GSE1 might potentially negatively affect the activity of one of these factors or its associated mechanisms.Notably, phosphorylation at S533 of RNF20 was similarly reduced in both GSE1 KO and HDAC CI compared to wildtype, indicating that a fully assembled GSE1-HDAC1 complex is required.Since monoubiquitination / deubiquitination of H2B K120 was previously described as a prerequisite for DDR induction ( 85 ), we turned our attention to this particular lead.

GSE1 forms a multi-enzymatic COREST subcomplex with deubiquitinases
To test whether H2B K120 monoubiquitination (H2Bub) levels are affected in GSE1 KO cells, we followed this modification in a time-resolved manner using western blotting in combination with an anti-H2Bub antibody.In wildtype cells we observed a prominent and continuous decline in H2Bub levels from the 6 h to the 24 h time point of etoposide treatment.Interestingly, in GSE1 KO cells H2Bub levels remained constant throughout the same timeframe (Figure 6 A), demonstrating that deubiquitination of H2B K120 is compromised by loss of GSE1.In light of previous findings that deubiquitination of H2B K120 is crucial for DDR induction ( 11 ,35 ), a defect in this step could explain the observed phenotype of GSE1 KO cells.Following this observation, our next goal was to understand if GSE1 directly regulates H2B deubiquitination.To this end, we generated by CRISPR / Cas-9 mediated knock-in a HAP1 cell line expressing a V5-tagged version of the endogenous GSE1 protein and used AP-MS to determine the GSE1 interactome.We purified GSE1 from untreated and 6 h etoposide treated cells and found the entire set of CoREST complex members to be co-purified (Figure 6 B).
Interestingly, GSE1 seems to exclusively bind to the CoR-EST complex as subunits of other known corepressor complexes such as SIN3, NURD or MiDAC were not enriched by the GSE1 pull down (Supplementary Figure S6A).In addition to CoREST complex members, the proteasome subunits PSMA2 and PSMB4 / 6, and the splicing factors DDX46 and SLU7 were found to co-purify with GSE1 (Figure 6 C, Supplementary Table S5).Notably, we could not detect differences in GSE1-CoREST complex composition in response to etoposide treatment, indicating a stable  To sustain proper signalling resulting in efficient DNA repair, a chromatin environment suitable for the recruitment of effector molecules of DNA damage signalling pathway is one of the essential requirements.Here we show a novel role of a neglected protein GSE1, a subunit of a specific CoREST / HD A C1 complex, in DDR.We propose that during DNA damage, GSE1 / CoREST / HD A C1 complex can bind USP22 that can in return deubiquitinate histone H2B at lysine 120.This eraser event is necessary for sufficient γH2AX signalling as reported previously ( 11 , 30 ).In the absence of GSE1, USP22 cannot bind to GSE1 / CoREST / HD A C1 complex and H2B lysine 120 remains ubiquitinated.This inadequate removal of the ubiquitin mark might interfere with γH2AX formation leading to reduced γH2AX signal and downstream signalling.
composition of the complex core that is most likely not affected by DDR induction (Supplementary Figure S6B).As GSE1 is associated with CoREST, we wanted to know whether its deletion affects CoREST-associated deacetylase activity.As expected, we detected no significant change in enzymatic activity associated with CoREST in GSE1 KO cells when compared to wildtype, while HDAC1 ablation significantly reduced CoREST-associated deacetylase activity (Supplementary Figure S6C).Therefore, DNA damage-mediated deubiquitination of H2Bub is not regulated via etoposide-induced changes in GSE1-CoREST complex composition or deacetylation activity.
We did, however, detect traces of UCH proteinases USP22 and USP15 in the affinity purification samples of GSE1-V5, but not in the background controls (see MS / MS counts of USP22 and USP15 in Supplementary Table S5).These low confidence hits did not pass the applied cut-offs.However, in particular USP22 has been shown to remove monoubiquitin specifically from K120 of H2B during DNA damage ( 11 ).We therefore wanted to independently test whether UCH proteinases bind to GSE1 or even to the GSE1-CoREST complex.Indeed, co-IP experiments using USP22 as bait resulted in successful co-isolation of GSE1, HDAC1, HDAC2 and CoREST upon basal conditions and in response to etoposide treatment (Figure 6 D).Binding of USP22 to CoREST was independent of HDAC1 (Figure 6 D) and also not affected by treatment with the HDAC inhibitor trichostatin A (TSA) (Supplementary Figure S6D).Additionally, the interaction properties of USP22-CoREST remained unaltered upon DDR induction by etoposide and intervention in DNA repair processes, as evidenced by the USP22 copurification experiments conducted in 53BP1 KO cells (Supplementary Figure S6E).However, none of these interactions were observed in GSE1 KO cells, showing that GSE1 is indeed required for the binding of USP22 to the CoREST complex (Figure 6 D).We therefore show that GSE1-CoREST complex ultimately combines several enzymatic properties essential to the DDR, such as deacetylation, demethylation, and deubiquitination.
In summary, we propose that GSE1 functions as a scaffolder for assembly of a multi-eraser CoREST complex, harbouring the activities of HDAC1 / 2, LSD1 and USP22.Presence of USP22 might be key for K120 deubiquitination upon induction of DDR, a primary event in complex PTM rearrangements of histones required for DDR (Figure 6 E).We thereby provide a first insight into a mechanistic rationale of how the GSE1-HDAC1-CoREST complex modulates the response to DNA damage.

Discussion
We identified GSE1 as an important interactor of HDAC1 as a part of a novel subvariant of the CoREST complex that is involved in regulating the initiation of the DDR.The GSE1 / CoREST sub-complex comprises a multi enzymatic eraser combining the deacetylation activity of HDAC1, demethylation activity of LSD1, and deubiquitination activity of USP22.Notably, GSE1 plays a central role in this process, as it constitutively binds to CoREST and is required for USP22 binding.Classical CoREST is primarily associated with silencing of neural genes and was implicated to govern stem cell differentiation and development ( 86 ).These processes are primarily controlled by the enzymatic activities of its known catalytically active subunits HDAC1 and LSD1 ( 87 ).Their dysregulation, together with additional CoREST subunits, was associated with various neurodegenerative diseases as well as cancer (88)(89)(90).However, to our knowledge, CoR-EST has hardly been associated with DNA damage ( 12 ) and therefore represents a potentially highly interesting but until now mostly neglected target in the field of medical research.
Despite being largely overlooked in previous research, GSE1 caught our attention due to its robust interaction with HDAC1 and its previous association with cancer.GSE1 was identified as a putative subunit of the CoREST and BRAF-HDAC (BHC) complexes ( 46 ,54-57 ).Here, we determine the interactome of GSE1 in human HAP1 cells by direct pulldowns and show that GSE1 interacts with core CoREST complex components, including HMB20A / B, HDAC1, HDAC2, RCOR1 / 2 / 3 and LSD1, however, we did not find components of other HDAC1 subcomplexes.Of note, we observed that GSE1 is essential for the binding of USP22 to CoREST, suggesting a role as a scaffolder.
Why has this multi-eraser complex gone unnoticed in previous research?We believe that USP22-GSE1-CoREST represents only a small fraction of CoREST complexes.Our IBAQ values (Supplementary Material), providing an estimate of protein abundance, as well as data provided at ProteomicsDB ( 91 ), ImmProt ( 92 ), and ImmPress ( 93 ) report very distinct protein abundance levels for HDAC1, RCOR1, GSE1, and USP22, indicating that these components are not expressed in equimolar levels.This is in agreement with findings of Macinkovic et al., who demonstrated the heterogeneous composition and functional diversity of cellular CoREST complexes ( 94 ).Consistent with this, we report that the deacetylase activity associated with CoREST is not dependent on the presence of GSE1, implying that GSE1 may not be a constitutive subunit of all CoREST complexes.The GSE1-USP22-CoREST complex seems to be stable, as we do not see changes in its composition following DNA damage induction.These results align with the findings of Barnes et al ., who observed stable GSE1-CoREST interaction during stem cell differentiation ( 54 ).
What is the biological function of the GSE1-USP22-CoREST complex?Here we show that loss of GSE1 results in hampered γH2AX formation, affects the abundance of HDR as well as NHJE-related proteins (Supplemental Material), shows altered impaired phosphorylation of key sites in DNA damage signalling and an overall compromised DDR.These findings point to a connection between GSE1-USP22-CoREST and the regulation of DNA damage initiation.One of key requirements in this process seems to be the GSE1-dependent binding of USP22 to CoREST.We hypothesise that during DNA damage USP22 is essential for removal of H2B K120ub, which seems to be a prerequisite signal for γH2AX formation (Figure 6 E).It was previously reported that B-cells lacking USP22 showed impaired removal of H2Bub which caused reduced A TM / A TR and DNA-PK-induced γH2AX formation in antibody class-switch recombination ( 11 ,35 ).In agreement with this we also do not see deubiquitination of H2Bub upon DNA damage in GSE1 KO cells.We suspect that the inefficient removal of H2B K120ub might lead to impaired kinetics of DNA damage-specific chromatin establishment necessary for recruitment of downstream effector molecules.Multiple mechanisms have been discussed in this context.For example, persistent presence of H2Bub could lead to extensive H2A Y142 phosphorylation.Phosphorylation of H2A Y142 in turn interferes with recruitment of MDC1 which is known to shield γH2AX signal from phosphatases ( 10 , 11 , 91 ).Additionally, it was postulated that H2B K120 undergoes a monoubiquitination-acetylation switch at DNA double strand breaks (DSBs) loci.This switch is needed for relaxation of the chromatin, facilitating resection and accessibility of DNA repair factors.Hence, removal of H2B monoubiquitination is inevitable for proper DDR ( 80 ).Notably, our study shows that the function of GSE1 is not limited to DSBs since we observed in the absence of GSE1 reduced γH2AX signal upon treatment with various agents including mitomycin C or hydroxyurea (Figure 3 A and B).
It also should be mentioned, however, that other studies reported an increase of H2B monoubiquitination levels in response to DNA damage ( 6-10 ).We could not observe such an effect in our experiments.It is possible that such discrepancies might be due to differential doses of DNA damaging agents used, different kinetic windows chosen or regions of chromatin affected.It could simply be that we missed such an increase in H2Bub levels with our experimental settings.
Our identification of the GSE1-HDAC1-USP22 multieraser complex and its function in DNA damage might be of relevance for cancer research.Various studies recently connected GSE1 function to tumour growth in distinct types of cancer and highlighted its potential oncogenic and tumour suppressive functions (95)(96)(97)(98)(99)(100)(101).Overexpression of GSE1 was shown to promote proliferation of human breast cancer cells ( 96 ) whereas its downregulation positively correlated with decreased tumour invasion and metastasis in lung ( 97 ) or gastric cancer ( 101 ).Moreover, mutations in GSE1 were linked to a specific subtype of medulloblastomas accelerating their growth ( 98 ).Class I HDAC inhibitors particularly in combination with other chemotherapeutic agents have been shown to exert antineoplastic activities by induction of replication stress and DNA damage ( 24 , 68 , 102-104 ).Notably, the dual-specificity inhibitor domatinostat (4SC-202) targeting both HD AC1 / HD AC2 and LSD1 catalytic activities exhibits anti-tumorigenic effects in various cancer models ( 105 ,106 ).
In our study, we show that GSE1 is essential to integrate different enzymatic activities.It is, however, still an open question how these activities are regulated in response to DNA damage.We can rule out changes in deacetylase activity of HDAC1.We do not know how and whether the activity of USP22 is regulated, or whether the complex becomes localised to specific sites of damaged chromatin.However, we also dismiss DNA damage induced changes in complex composition.Future work will certainly focus on these aspects.The fact that inhibitors of separate enzyme activities combined in the GSE1-HDAC1-USP22 complex are currently applied and tested in the clinics for cancer patients is potentially putting our findings into the context of clinical relevance ( 107 ,108 ).
Perseus (version 1.6.15.0,( 44 )) Python (version 3.9) and R (version 4.0.4) were used for statistical analysis and visualisation of graphs and plots.GraphPad Prism version 8.0.1 for Windows (GraphPad Software, San Diego, California, USA) was used to generate bar graphs and viability curves in Figures 2 and 3 and Supplemental Figures S2 and S5.In case of histone deacetylase assays, immunofluorescence and western blot quantification or viability, unpaired parametric t-test was always used for comparison of one sample to wildtype.To quantify western blot signal Fusion FX EvolutionCapt (Vilber) software was used.Immunofluorescence foci counting was performed in Fiji software (version 2.1.0).For OMICS data, the individual statistical analysis is indicated in corresponding method parts.Heatmaps shown in Figures 4 and 5 and Supplementary Figures Figure 1 B and Supplementary Figure S1B) or according to modularity class (Figures 5 E and 6 C).

5 HDAC1 2 HDAC1Figure 1 .Figure 2 .
Figure 1.GSE1 as a prominent interactor of HD A C1 across species.(A) Volcano plot displaying the interactome of FLAG-tagged HDAC1 in HAP1 cells determined by AP-MS.Significant interactors ( ≥ 1.5-fold change [FC] over HD A C1 KO [Ctrl], adjusted P -value ≤ 0.05) in HD A C1-Flag cells are shown with red dots.Dashed line indicates fold change cut-off.Gene names of known subunits of the HD A C1 corepressor complexes are highlighted, such as CoREST, SIN3-related or NuRD.Among the binding partners GSE1 is highlighted with bold red text.(B) STRING DB-based protein interaction network of the HD A C1 interactome in HAP1 cells.High-confidence interactions (S TRING confidence score ≥ 0.7) w ere selected.Disconnected nodes are not sho wn.T he interactors w ere classified into functional protein groups in accordance with J oshi et al. ( 46 ).Strength of the associations is represented b y the thickness of the edges.(C) Volcano plot displaying the interactome of FLAG-tagged HDAC1 in mouse embryonic fibroblasts (MEFs) cells determined by AP-MS.Significant interactors ( ≥1.5-fold change [FC] over wildtype cells [Ctrl], adjusted P -value ≤ 0.05) in HD A C1-Flag are shown with red dots.Dashed line indicates fold change cut-off.Gene names of known subunits of the HD A C1 corepressor complexes are highlighted, such as CoREST, SIN3-related or NuRD.Among the binding partners GSE1 is highlighted with bold red text.(D) Co-IP / western blot analysis illustrating HD A C1-GSE1 interaction in HAP1, MEFs and murine Th1 T-cells.HD A C1 KO HAP1 cells were used as negative control.Antibodies used for detection are depicted on the right.

Figure 3 .
Figure 3. Loss of GSE1 leads to impaired DNA damage response signalling.(A) Phosphorylation of H2AX histone ( γH2AX) was analysed by immunofluorescent staining for γH2AX.The nucleus was stained with DAPI.The arrow indicates the selected enlarged cell.Relative fluorescent intensity was plotted, and significance determined by unpaired parametric t-test by comparison to wildtype cells.* P < 0.05, ** P < 0.01, *** P < 0.001.HAP1 WT cells were set to 100% as a control.(B) Downregulation of γH2AX in GSE1 KO was confirmed by western blot analysis upon treatment for 12, 18 and 24 h with etoposide, mitom y cin C or h y dro xyurea.Upper panel: original western blot images.Apoptosis was determined by cleavage of caspase 3 and actin was used as loading control.Lower panel: quantification of γH2AX signal normalised to loading control ( β-actin).Error bars indicate standard deviation of three biological replicates.(C) Phosphorylation of H2AX histone ( γH2AX) was analysed by western blot upon re-introduction of GSE1 in untreated or 24 h etoposide treated HAP1 WT and GSE1 KO cells.Apoptosis was determined by cleavage of caspase 3 and actin was used as loading control.GSE1 expression was analysed using GSE1-specific antibody.

Figure 4 .
Figure 4. GSE1 KO and HD A C1 CI show distinct acetylation and protein signatures affecting cell cycle.(A) Western blot illustrating synergistic effect of MS-275 and etoposide on DNA damage induction ( γH2AX) in HAP1 WT and GSE1 KO cells.Actin was used as a loading control.(B) Volcano plots displaying acetylome changes in HAP1 WT, HD A C1 CI and GSE1 KO cells upon 6 h treatment with 0.1 μM etoposide.Significantly upregulated and downregulated acetylation sites are shown in red ( ≥2-fold change over untreated control, adjusted P -value ≤ 0.01).(C) Boxplots of z-scored acetylation site intensities in different cell lines and conditions of sites that are etoposide induced in HAP1 WT cells.Coloured horizontal lines represent medians, bo x es represent the 25th and 75th percentiles, whiskers expand to the 5th and 95th percentile, outliers are not shown.(D, E) Heatmaps showing acetylation profiles of selected downregulated and upregulated sites in HAP1 WT upon etoposide treatment that are not responsive in GSE1 KO and HD A C1 CI mutants.Corresponding gene names and positions of acetyl-lysine are indicated on the left.* and ** mark quantification of acetylation sites arising from *dual or **triple acetylated peptides where only a single site was confidently assigned.(F) STRING DB-based network of proteins associated with etoposide-responsive acetylation sites in HAP1 WT cells that are stable in GSE1 KO and HD A C1 CI.High-confidence interactions (confidence score ≥ 0.7) were selected.Disconnected nodes are not shown.Colouring of nodes is according to modularity classes.Proteins functionally associated with cell cycle regulation are highlighted in red.Strength of the associations is represented by the thickness of the edges.(G) Heatmap illustrating protein abundance of selected cyclins in HAP1 WT, GSE1 KO and HD A C1 CI.Corresponding gene names are indicated on the left.(H) Quantified cell cycle phase distribution of HAP1 WT, HD A C1 CI and GSE1 KO FUCCI cells using flow cytometry.The data represent the percentage of cells in each cell cycle phase ± standard deviation (SD) of three biological replicates.Significance is determined by unpaired parametric t -test by comparison to wildtype.* P < 0.05, ** P < 0.01, *** P < 0.001.(I) Cell viability was calculated by counting of the cells every 24 h cultured in the presence of etoposide up to 96 h.Significance is determined by unpaired parametric t-test by comparison to wildtype.* P < 0.05, ** P < 0.01, *** P < 0.001.(J) Cell proliferation was calculated by counting of the cells every 24 h cultured in full media.Significance is determined by unpaired parametric t -test by comparison to wildtype.* P < 0.05, ** P < 0.01, *** P < 0.001.

CFigure 6 .
Figure 6.Identification of specific GSE1 / CoREST complex with multi-enzymatic eraser activities.(A) Histone western blot detecting levels of ubiquitination of H2B at K120 upon treatment with etoposide for 6, 12 and 24 h.C-terminal histone H3 was used as a loading control.(B) Volcano plot displaying the interactome of GSE1 in HAP1 cells determined by AP-MS.Significant interactors ( ≥1.5-fold change [FC] over wildtype cells [dashed line], adjusted P -value ≤ 0.05) in GSE1 V5-tagged cells over HAP1 WT cells are shown with red dots.Gene names of known subunits of the HD A C1 / CoREST complex are highlighted, GSE1 is highlighted with bold red text.(C) STRING DB-based protein interaction network of the GSE1 interactome.High-confidence interactions (confidence score ≥ 0.7) were selected.Strength of the associations is represented by the thickness of the edges.(D) Western blot depicting immunoprecipitated endogenous USP22 in HAP1 WT, HD A C1 KO and GSE1 KO cells.Antibodies used for detection are depicted on the right.(E) Potential mechanism of DNA damage regulation by GSE1.Proper initiation and orchestration of downstream signalling during DNA damage is highly dependent on the induction of Serine 139 phosphorylation of histone variant H2AX.To sustain proper signalling resulting in efficient DNA repair, a chromatin environment suitable for the recruitment of effector molecules of DNA damage signalling pathway is one of the essential requirements.Here we show a novel role of a neglected protein GSE1, a subunit of a specific CoREST / HD A C1 complex, in DDR.We propose that during DNA damage, GSE1 / CoREST / HD A C1 complex can bind USP22 that can in return deubiquitinate histone H2B at lysine 120.This eraser event is • C).The supernatants were transferred into a new tube with 10 times volume of acetone, vortexed for 1 min and precipitated o / n at −20 • C. The precipitates were collected by centrifugation (18 400 × g / 15 min, 4 • C), air-dried and resuspended in H 2 O. 2 μg of histones were used for SDS-PAGE and western blot.