GCN5 is recruited to DNA Double Strand Breaks post genotoxic stress

Previously, we have shown an atypical association of histone acetyltransferase, GCN5 with ATM kinase in AML [27]. Here we interrogated the detailed mechanism by which GCN5 can regulate DSB repair. First, we wanted to check whether GCN5 is recruited to DSBs post genotoxic stress. For this, we induced DSBs in GBM (U87MG and LN229), Lung Adenocarcinoma (PC9), Osteosarcoma (U2OS) cells by subjecting them to 8 Gy (GBM cells) and 4 Gy (PC9 and U2OS) radiation doses (sub-lethal doses, that sufficiently induced DSBs) [21, 26, 27]. Next, we checked GCN5 recruitment to the DSBs by analyzing the RIF (radiation induced foci) of GCN5 and their colocalization with ɣH2AX (S139), phosphorylated histone variant which marks the damage site. GCN5 was found to be recruited to DSBs as observed by positive Pearson’s Correlation coefficient (U87MG-0.67; LN229-0.66; PC9: 0.7; U2OS:0.53), both at 0.5 and 3 h post IR and foci quantification (Fig. 1A–B, S1A–B). Similarly, treatment with lower dose of 2 Gy IR in U87MG and LN229 also facilitated GCN5 RIF formation, immediately at 1 min post IR and reached maximum at 0.5 h post IR (Fig S1C). Additionally, to monitor recruitment of GCN5 to DSBs induced by sources other than radiation, we used genetically modified DIvA U2OS cell line, where the gene encoding for AsiSI restriction enzyme is fused to the ligand binding domain of Oestrogen receptor. Upon 4 hydroxytamoxifen (4-OHT) treatment, cytoplasm to nuclear translocation of AsiSI takes place which lead to DSBs at 100 positions across the genome [28]. As shown in Fig. 1C, at 2 h post 4-OHT treatment, GCN5 was recruited to DNA DSBs in DIvA cells (R value: 0.46). In all the cell lines tested, post DSB induction, GCN5 was enriched at DSBs, along with being localized to other areas within the nucleus. Given the fact that GCN5 is an essential HAT, which functions as a transcriptional coactivator, the localization of GCN5 at sites other than DSBs is expected, to execute other vital cellular processes, especially transcription, that happens in parallel to repair [29].

Fig. 1

GCN5 is recruited to DNA DSBs post IR and ensures efficient DSB repair kinetics. A) Representative cropped immunofluorescence images of 8 Gy IR treated U87MG and LN229 cells at the indicated time points. Cells were stained for GCN5 (green) and γH2AX (red). Nuclei were stained with DAPI (blue). B) Colocalization between GCN5 and ɣH2AX is shown by calculation of Pearson’s Correlation Coefficient after drawing nuclear ROIs using Coloc 2 Plugin of Fiji (ImageJ). Foci number of each nucleus is calculated using particle counter of Fiji (ImageJ). Significance of each time point post IR was calculated w.r.t no IR. Mean ± SD of Pearson’s R value was plotted (N = 3; ****P < 0.0001; *** P < 0.001; Unpaired Student’s t test). Mean ± SEM of average foci number/nucleus have been plotted (N = 3; ns: non-significant; *P < 0.05; **P < 0.01; ****P < 0.0001). C) Representative cropped immunofluorescence images of untreated and 2 h 4-OHT treated DIvA U2OS cells stained for GCN5 (green) and γH2AX (red). Bar graph represents Mean ± SD of Pearson’s R value (N = 3; P < 0.0001; Unpaired Student’s t test). D) Representative cropped immunofluorescence images of scrambled and GCN5 Knockdown U87MG and LN229 cells at indicated time points post IR stained for γH2AX (red). Images were acquired using Nikon AX confocal microscope. E) Line graphs depicting average nuclear intensities of scrambled (Scr) and GCN5 knockdown (shGCN5) U87MG and LN229 cells at the indicated time points. Mean ± SD were plotted (N = 3; *P < 0.05; *P < 0.01, ***P < 0.001; ****P < 0.0001; Multiple two-tailed t test). F) Neutral comet assays were performed in U87MG and LN229 at indicated time points post 8 Gy IR. Representative images were obtained using Zeiss Upright Microscope at 20X magnification. G) 100 comet tails from each group were taken for quantification of neutral comet assay. Bar graphs depict the average (N = 2) percent tail moment of scrambled (Scr) and GCN5 knockdown (shGCN5) U87MG and LN229 cells at the indicated time points. Mean ± SD were plotted. (**P < 0.001; *P < 0.05; ns: non-significant; Multiple T test)

GCN5 facilitates resolution of DNA DSBs

To understand if recruitment of GCN5 to DNA DSBs have any role in maintaining DSB repair kinetics, we knocked down GCN5 by shRNA in U87MG and LN229 cells. Loss of GCN5 expression post shRNA transfection confirmed GCN5 knockdown (Fig. S2C). We then subjected scrambled, U87MG-GCN5KD and LN229-GCN5KD cells to 8 and 2 Gy IR and monitored ɣ-H2AX intensities and foci number at various time points post IR. As shown in Fig. 1D–E; S2A, B in control U87MG and LN229, ɣH2AX foci started to form as early as 30 min post IR, peaked at 1 h and gradually got resolved completely by 24 h. On the contrary, in both U87MG-GCN5KD and LN229-GCN5KD cells there was significant ɣ-H2AX foci retention till 24 h, indicative of damage accumulation and failure of repair (Fig. 1D, E; S2A, B). To confirm this further, we performed Neutral Comet Assay at 0.5 and 24 h post IR in GCN5 knockdown U87MG and LN229 cells (Fig. 1F, G). In both scrambled control and GCN5 knockdown cells, induction of damage led to significant accumulation of DSBs at 30 min post IR, compared to non-radiated condition. However, at 24 h, the tail moment of control cells returned to the basal level, indicative of efficient resolution of DSBs, whereas that of GCN5 knockdown cells persisted, confirming the DSB accumulation. Interestingly, we found that compared to scrambled control, GCN5 knockdown alone also led to accumulation of DSBs as observed by increase in tail moment. A plausible explanation for this observation could be that GCN5 helps in resolution of DSBs that are intrinsically present in U87MG and LN229, owing to the fact that GBM cells possess high genomic instability [30]. Similar results were found in 293FT-GCN5KD cells where GCN5 knockdown alone as well as in combination with IR, showed increased DSBs, persistent till 24 h (Fig. S2D).

PARylation of GCN5 by PARP1 is essential for its recruitment to DNA DSBs

Next, we sought to identify the molecule which could be responsible for GCN5 recruitment to DSBs. We pharmacologically and genetically inhibited crucial enzymes of sensor complexes and examined their effect on the recruitment of GCN5 to DSBs. Inhibition of exonuclease Mre11 by mirin (10 µM) and DNA-PKcs by Nu7026 (10 µM) was confirmed by significantly reduced phosphorylation of ATM and DNA-PK, the downstream targets for MRN and DNA-PKcs respectively in U87MG cells (Fig. S3A–B). We found no significant change in the recruitment of GCN5 to DSBs in the presence or absence of both the inhibitors at 0.5 h post IR as shown by comparable GCN5-ɣH2AX Pearson’s correlation coefficient post Mre11 or DNA-PK inhibition (Fig. 2A–B). Similarly genetic knockdown of DNA-PKcs (Fig. 2C) also did not show any differential effect on the recruitment of GCN5 to DSBs post IR. Next, we inhibited PARP1 by 5 µM olaparib for 12 h in U87MG, LN229 and U2OS cells. Decreased Poly ADP ribosylation (PAR) at DSBs (Fig. S4A) and decrease in total PARylated cellular proteins (Fig. S4B) upon olaparib treatment confirmed PARP1 inhibition. PARP1 inhibition significantly reduced GCN5 recruitment to DSBs at 0.5 h post radiation or 2 h post doxorubicin treatment as shown by significant decrease in Pearson’s R value in U87MG, LN229 and U2OS cells (Fig. 2D–E, S4C, D). Similarly, olaparib when treated with lower dose of 2 Gy IR inhibited GCN5 foci formation (Figure S4E). We also genetically depleted PARP1 in U87MG and LN229 by shRNA, which was confirmed by loss of PARP1 expression by immunoblotting and immunofluorescence (Fig. S5A, B). Similar to pharmacological inhibition, PARP1 knockdown also impaired GCN5 recruitment to DSBs as shown by the decrease in Pearson’s R value of 0.32 in U87MG and 0.39 in LN229 cells at 0.5 h post IR (Fig. 2F and S5C). Further, confirmation of GCN5 DSB recruitment was done by isolation of chromatin fraction from U87MG and LN229. Expectedly, increased recruitment of GCN5 to chromatin in U87MG and LN229 was observed at 30 min post IR, compared to non-radiated condition while olaparib treatment prevented GCN5 recruitment to chromatin (Fig. S5D). Together these data confirm that PARP1 mediates GCN5 recruitment to DNA DSBs post genotoxic stress. To rule out the possibility that the reduced GCN5 recruitment to DSBs is due to decreased expression of GCN5, we performed immunoblotting post PARP1 inhibition and knockdown in U87MG and LN229. As seen from Fig. S5E and F, PARP1 inhibition did not change expression of GCN5 in untreated or IR treated whole cell lysates of U87MG and LN229, implying that PARP1 do not regulate GCN5 expression.

Fig. 2

PARP1 is required for GCN5 recruitment to DNA DSBs. A-F) Representative cropped immunofluorescence images of DMSO and 10 µm Mirin (A), 10 µm NU7026 (B) treated, scrambled (Scr) and DNA-PKcs knockdown (siDNA-PKcs) (C), DMSO and 5 µm olaparib treated with IR (D) and 5 µM doxorubicin (E), scrambled (Scr) and PARP1 knockdown (shPARP1) (F) U87MG with/without 8 Gy IR stained for GCN5 (green) and γ-H2AX (red). For A-F, images were acquired at 63X objective magnification and Pearson’s R value of each nucleus was calculated using Coloc2 plugin of Fiji (ImageJ). Minimum 50 cells for each group were taken for analysis. Data is represented as Mean ± SD (N = 3; ns: non-significant; *p < 0.05, **P < 0.01; Unpaired two-tailed t test with Welch’s correction)

Next to find the mechanism behind PARP1 mediated GCN5 recruitment to DSBs, we first checked if GCN5 and PARP1 physically associate with each other. For this we performed co-immunoprecipitation with GCN5 antibody in control, IR and doxorubicin treated U87MG, LN229 and 293FT cells. GCN5 was found to interact with PARP1 in control as well as in DSB induced conditions in all the three cell lines (Fig. 3A, S6C). Additionally, significant increase in foci count by proximity ligation assay (PLA) confirmed GCN5-PARP1 interaction post genotoxic stress induced by IR and doxorubicin in U87MG and LN229 (Fig. 3B). Since we found PARP1 binding to GCN5, we asked if GCN5 was getting PARylated. PARylation of GCN5 was checked in U87MG and LN229 control, IR and doxorubicin treated cells by immunoprecipitation with GCN5 antibody and immunoblotting with Poly ADP Ribose (PAR) antibody. Indeed, GCN5 was found to get PARylated (Fig. 3C). To further confirm GCN5 PARylation, we performed reverse IP with PAR antibody followed by immunoblotting for GCN5. As seen in Fig. 3D, GCN5 was pulled-down with PAR antibody confirming that GCN5 was PARylated. We did notice that in case of LN229 cells, most prominent GCN5 PARylation was observed only post genotoxic stress. Furthermore, the presence of PARylated GCN5 in control (no IR) U87MG is suggestive of its role in the repair of DSBs that are intrinsically present in the cell line. To confirm that GCN5 PARylation was indeed mediated by PARP1, we pharmacologically and genetically perturbed PARP1 activity using olaparib and shRNA respectively in U87MG cells. These cells were then subjected to IR followed by GCN5 IP at 0.5 h. Decreased GCN5 PARylation was observed post PARP1 inhibition as shown by immunoblotting with GCN5 in U87MG (Fig. 3E–F). Altogether, these results confirm that GCN5 gets PARylated by PARP1 and that PARylation is essential for its recruitment to DSBs.

Fig. 3

PARP1 PARylates GCN5 post genotoxic stress. A) Western blots showing GCN5 and PARP1 after immunoprecipitation with GCN5 and IgG in U87MG and LN229 control, 30 min post 8 Gy IR and 2 h post 5 µM doxorubicin treated cells. Vinculin was used as loading control. B) GCN5-PARP1 interaction foci are shown using Duolink insitu detection reagent red (DU092008) post proximity ligation assay probed with GCN5 (anti-mouse) and PARP1 (anti-rabbit) antibodies in U87MG and LN229 control, IR and doxorubicin treated conditions. Nuclei are stained with DAPI and images are acquired using 63X objective. Mean ± SEM of 60 cells (N = 2) are plotted and unpaired t -test with Welch’s correction has been performed****P < 0.0001. C, D) Western blots showing GCN5 and Poly-ADP Ribose (PAR) following immunoprecipitation with GCN5 and IgG (C), PAR and IgG (D) in U87MG and LN229 control, 30 min post 8 Gy IR and 2 h post 5 µM doxorubicin treated cells. Actin and vinculin were used as control for U87MG and LN229. E) Western blots of GCN5 and Poly-ADP Ribose (PAR) following immunoprecipitation with GCN5 and IgG in U87MG control and 30 min post 8 Gy IR, in combination with 5 µM olaparib. Actin was used as loading control. F) Western blots of PARP1 and Poly-ADP Ribose (PAR) following immunoprecipitation with GCN5 and IgG in U87MG scrambled and PARP1 knockdown (shPARP1) cells. Actin was used as loading control. The smears obtained in PAR inputs (C-F) indicate all PARylated proteins within the mentioned molecular weight range

GCN5 interacts with DNA-PKcs and Ku80

To further explore the function of GCN5 at the DSB site, we took a global approach of IP-MS (Immunoprecipitation followed by Mass spectrometry) to identify all the DNA DSB repair proteins that interact with GCN5 upon DSB induction. We performed immunoprecipitation with GCN5 and Isotype (IgG) control followed by mass spectrometry of control, IR and doxorubicin treated U87MG cells in 2 biological replicates. A total of 1546 [control (549), IR (647), doxorubicin (350)] and 1006 [control (84), IR (520), doxorubicin (402)] proteins were obtained in the two replicates respectively (S6A). On overlapping the GCN5 interactors obtained exclusively post IR or doxorubicin treatment, and not present in control, with the 594 DSB response and Repair proteins curated from Uniprot, 7 proteins (Prelamin A/C, DNA-PKcs, Ku80, DHX9, FUS, SFPQ, HMGB1) were found as common between the two biological replicates (Fig. S6B). The peptide counts for each protein tabulated in Fig. 4A are combination of IR and doxorubicin treated samples. Out of these, DNA-PKcs and Ku80 were the only two proteins known to directly take part in NHEJ Repair. We validated GCN5-DNA-PK physical interaction in U87MG, LN229 and 293FT cells by GCN5 immunoprecipitation followed by immunoblotting with DNA-PKcs, Ku80 and GCN5. Indeed, GCN5 was found to interact with DNA-PKcs post genotoxic stress, in U87MG and LN229 cells (Fig. 4B–C). Interestingly, in 293FT, GCN5-DNA-PKcs interaction was found without DNA DSB induction in control cells as well, indicative of their interaction to be necessary for other cellular processes (Fig. S6C). However, GCN5 interaction with Ku80 was found to be independent of genotoxic stress in all three cell lines. These data suggest that in GBM cells (U87MG and LN229) GCN5 interaction with DNA-PKcs is dependent on induction of DSBs, wherein the DNA-PK assembly takes place whereas GCN5-Ku80 association may be pivotal for other cellular processes as well.

Fig. 4

GCN5 interacts with and acetylates DNA-PKcs required for its phosphorylation. A) Table showing the total combined number of peptides corresponding to each interactor of GCN5 following mass spectrometry post IR and doxorubicin treatment. B, C) Representative western blots of DNA-PKcs, GCN5 and Ku80 following immunoprecipitation with GCN5 and IgG in control, 30 min post 8 Gy IR and 2 h post 5 µM doxorubicin treated U87MG (B) and LN229 (C) cells. Vinculin was used as loading control. D) In vitro acetyltransferase assay was performed using 25 ng purified GCN5 and DNA-PKcs concentrations as indicated. Each line represents the % Histone acetyltransferase (HAT) activity with different DNA-PKcs concentrations at the indicated time points. E) Table shows the amino acid sequences of DNA-PKcs peptides used for acetyltransferase assay. The amino acids highlighted in yellow are the ones that are mutated. In vitro acetyltransferase assay was performed using 25 ng purified GCN5 and 0.1 mM of indicated DNA-PKcs peptides. Each line represents the % Histone acetyltransferase (HAT) activity of each mentioned DNA-PKcs peptide at the indicated time points. F) Western blots showing expression of pDNA-PKcs, DNA-PKcs, GCN5, H3K27Ac in U87MG and LN229 cells transfected with empty (Control) and GCN5 acetyltransferase mutant (Y621A/P622A) vectors, with or without IR. Tubulin and total H3 were used as loading control. Quantification of H3K27Ac w.r.t total H3 has been shown, calculated using ImageJ. Images are representative of 2 biological replicates. G) Representative cropped immunofluorescence images of YFPDNA-PKcs expressing U2OS cells transfected with scrambled (Scr) and 2 shRNAs (shGCN5-1 and shGCN5-2) targeting GCN5 post 4 Gy IR. pDNA-PKcs (red) was stained with Alexa-Fluor (594) and YFP DNA-PKcs imaging was done at 488 nm excitation channel using Nikon AX Confocal microscope at 63X magnification. Fluorescence intensities of mentioned proteins were calculated by drawing nuclear ROIs for minimum 100 cells using Fiji (ImageJ). Data is represented as Mean ± SD (N = 3; Unpaired t test with Welch’s correction was performed; ***P < 0.0001)

Since we found GCN5 binding to DNA-PKcs, we asked if DNA-PKcs kinase activity was required for GCN5 function as acetyltransferase. For this, Acetyl transferase activity of GCN5 was checked post IR + DNA-PKcs inhibition with Nu7026. Comparable nuclear intensities of H3K9Ac/H3K27Ac (known targets of GCN5) post Nu7026 treatment, with or without IR indicated no change in GCN5 acetyltransferase activity in U87MG (Fig. S6D) and LN229 (Fig. S6E) upon DNA-PKcs inhibition. These data confirmed that GCN5 directly interacts with DNA-PKcs but its function is not dependent on DNA-PKcs kinase activity.

GCN5 predominantly acetylates lysine 3241 of DNA-PKcs

Given that GCN5 interacts with DNA-PKcs, we hypothesized that GCN5 acetylates DNA-PKcs. To test our hypothesis, we performed chemiluminescence based in vitro histone acetyltransferase (HAT) assay using recombinant GCN5 and DNA-PK proteins, wherein absorbances recorded at 440 nm wavelength served as the readout for GCN5 HAT activity. HAT assay functionality was first checked using Histone H3, a known GCN5 target that gets acetylated at lysines 9, 14 and 27 by GCN5 [31, 32] and Fibroblast Growth Factor (FGF) was taken as a negative control (Fig. S6F). Following kinetics studies, sigmoidal curves were obtained for H3, in contrast to no change in absorbance for FGF indicating H3 acetylation and ensured proper functioning of the assay. Optimum concentration of GCN5 was identified as 25 ng that showed maximum acetyltransferase activity with 32 units of DNA-PKcs as previously described [22] (Fig. S6G). Next, we used varying concentrations of DNA-PK (8, 16, 24, 32, 40 units) with 25 ng of GCN5 for the assay. As shown in Fig. 4D, increase in % histone acetyltransferase activity of GCN5 was observed with increasing DNA-PKcs concentration, from 8 to 16 units, followed by subsequent decrease with 32 and 40 units at all the time points. The concentration and time dependent sigmoidal curves obtained confirmed GCN5 mediated DNA-PKcs acetylation. We then wanted to know which residue of DNA-PKcs was acetylated by GCN5. A previous study has shown that acetylation of DNA-PKcs lysine residues 3241 and 3260 mediate radio resistance. However, the acetyltransferase involved in this critical process remained to be elucidated [33]. Thus, we proceeded to check whether GCN5 mediates the acetylation of DNA-PKcs K3241 and K3260 residues. In vitro acetyltransferase assay was carried out in presence of 25 ng GCN5 and 0.1 µM of four DNA-PKcs peptides of 25 amino acids with a) both wild type K3241 and K3260; b) wild type K3241 and mutant K3260A; c) mutant K3241A and wild type K3260; d) K3241A + K3260A double mutant. As shown in Fig. 4E, % HAT activity of GCN5 in presence of K3260A mutant was same as that of wild type and decrease in HAT activity was observed upon K3241A as well as K3241A + K3260A mutations in the peptides. The comparable reduction in double mutant acetylation kinetics to that of K3241A mutation indicated DNA-PKcs K3241 residue was predominantly acetylated by GCN5.

GCN5 mediated acetylation of DNA-PKcs is prerequisite for its phosphorylation

We then wanted to check if the acetylation of DNA-PKcs by GCN5 is required for its phosphorylation at S2056 and recruitment to DSBs. Interestingly, DNA-PKcs activation as evidenced by pDNA-PKcs RIFs peaked at 30 min post IR, correlating with maximum GCN5 recruitment (Fig. S7A). Thus, we checked DNA-PKcs phosphorylation (S2056) post IR in U87MG and LN229 cells overexpressing catalytically inactive GCN5 acetyltransferase mutant (Y621A/P622A). GCN5 acetyltransferase activity inhibition was confirmed by loss of H3K27 acetylation (Fig. 4F). Indeed, overexpression of catalytically inactive GCN5 completely abrogated DNA-PKcs expression and inhibited DNA-PKcs phosphorylation at Ser2056 as observed by immunoblotting (Fig. 4F). Since we found complete loss of endogenous DNA-PKcs protein with catalytically inactive GCN5, we could not assess if acetyltransferase activity of GCN5 was solely required for DNA-PKcs phosphorylation. Thus, to further examine that the post-translational acetylation of DNA-PKcs by GCN5 is essential for its phosphorylation at S2056, we performed shRNA mediated GCN5 knockdown in U2OS cells that constitutively overexpressed YFP tagged DNA-PKcs. Remarkably, at 0.5 h post IR, GCN5 knockdown inhibited only DNA-PKcs activation, as seen by the significant decrease in nuclear intensity of pDNA-PKcs, without hampering its expression (Fig. 4G). Together, these results confirm that GCN5 mediated DNA-PKcs acetylation is critical for its expression as well as activation by S2056 phosphorylation.

GCN5 inhibits ATM and BRCA1 activation

Since histone acetylation at DSBs results in chromatin relaxation, favoring activation of many DDR proteins, we asked if the recruitment and activation of any of these proteins (other than DNA-PKcs) was regulated by GCN5. For this, we performed siRNA mediated knockdown of GCN5 in U87MG and LN229 cells followed by DSB induction by radiation and evaluated the recruitment and activation of DDR pathway sensors (MRE11, KU80, PARP1) and mediators (pATM, pATR, pBRCA1) by assessing their radiation induced foci (RIF) formation. We found that recruitment of sensor proteins was not affected in the absence of GCN5 (Fig. 5A–C). Further, of all the mediators analyzed (Fig. 5D–F), activation of pATM and its downstream target pBRCA1 were affected post GCN5 knockdown as shown by significant decrease in their average foci number/nucleus in both U87MG and LN229 cells (Fig. 5D–F). ATM activation kinetics also correlated well with GCN5 DSB recruitment, peaking at 30 min post IR (Fig. S7B). Furthermore, we checked whether GCN5 regulated the expression and DSB recruitment of total ATM and BRCA1 by immunoblotting and immunostaining respectively. As shown in figure S7C, GCN5 knockdown did not decrease the total protein level of ATM and BRCA1, nor did it regulate their DSB recruitment (Figure S7D–E). This confirms that GCN5 specifically regulates ATM and BRCA1 activation at DSBs. This is consistent with our previous findings in AML cells wherein GCN5 was shown to regulate ATM and downstream signaling [19].

Fig. 5

GCN5 is responsible for activation of p-ATM and p-BRCA1 at DSBs in U87MG and LN229 cells. A-F) Representative cropped immunofluorescence images of scrambled (Scr) and GCN5 knockdown (siGCN5) U87MG and LN229 cells at 30 min post 8 Gy IR treatment stained for Mre11 (A), PARP1 (B), Ku80 (C), pATM (D), pBRCA1 (E), pATR (F) in green. Nuclei were stained with DAPI (blue). 100 cells for each group were taken for analysis. Average foci/nucleus were calculated using particle counter of Fiji (ImageJ). Data is represented as Mean ± SEM (N = 2; Unpaired T test with Welch’s correction; * P < 0.05, **P < 0.01, ***P < 0.001)

p300 overexpression does not rescue GCN5 mediated ATM and DNA-PKcs activation

To analyze if the regulation of ATM, DNA-PKcs activation is specifically regulated by GCN5 or is redundant with any other HAT activity, we overexpressed p300 in GCN5 knockdown U87MG and LN229 cells. Decreased GCN5 intensity of shGCN5 transfected cells as compared to scrambled and increased p300 intensity post immunostaining confirmed GCN5 knockdown and p300 overexpression (Fig. S8A). Next, we monitored pDNA-PKcs and pATM foci formation in scrambled, GCN5 knockdown and GCN5 knockdown + P300 overexpressing U87MG and LN229 cells at 30 min post 8 Gy IR. As shown in Figure S8B-C, pDNA-PKcs and pATM foci formation at 30 min post IR decreased in GCN5 knockdown U87MG and LN229. Interestingly, their foci were not restored upon p300 overexpression (Fig. S8B, C). This confirms that ATM and DNA-PKcs activation is solely regulated by GCN5 and not by p300.

GCN5 regulates transcription of essential DDR genes

Since GCN5 is known as an important protein of ATAC-STAGA transcriptional coactivator complex, we reasoned that in addition to directly regulating the activation of key kinases (DNA-PKcs and ATM) at DSBs, GCN5 can also regulate the expression of DDR genes, either as a transcription factor or via histone acetylation. To assess this, we genetically perturbed GCN5 using shRNA in U87MG and LN229 cells and subjected the cells to 8 Gy IR. At 0.5 h post IR, we screened for the expression of pivotal Homologous Recombination (MRE11, RAD50, NBN1, ATR, ATM, CHEK1, CHEK2, BRCA1, BRCA2, RPA2, RAD52) and Non-Homologous End Joining (PRKDC, PARP1, XRCC5, XRCC6, TP53BP1, NHEJ1, DCLRE1C, XRCC4, ERCC4, LIG4, POL-L, POL-M) repair pathway genes by real-time quantitative PCR. In LN229 cell line, NBN1, CHEK1, PRKDC, PARP1, TP53BP1, DCLRE1C, XRCC4, LIG4, POL-L, POL-M were significantly downregulated post GCN5 ablation (Fig. 6A). In U87MG, GCN5 knockdown downregulated the expression of MRE11, NBN1, CHEK1, CHEK2, BRCA2, RPA2, PRKDC, XRCC6, TP53BP1, POL-L (Fig. 6B). Combining data from both LN229 and U87MG cell lines, significant decrease in 2 HR genes- NBN1 (0.3-fold) and CHEK1 (0.2-fold) and 3 NHEJ genes-PRKDC (0.06-fold), 53BP1 (0.6-fold), POL-L (0.4-fold) were obtained post GCN5 knockdown (Fig. 6A, B). Further, we performed immunoblotting to check the expression of the proteins encoded by these 5 genes post GCN5 knockdown in U87MG and LN229. As seen from Fig. S9A, expression of Chk1 and Nbs1 proteins did not significantly decrease post GCN5 knockdown, despite their transcript downregulation. This may be due to efficient translation of the residual transcripts or stabilization of the proteins post GCN5 knockdown. Significant downregulation of DNA-PKcs, 53BP1, Polymerase Lambda encoded by, PRKDC, TP53BP1, and POL-L genes respectively were obtained in both U87MG and LN229 (Fig. S9A). Together these data show that GCN5 regulates the transcription of specific DDR genes in GBM cells post irradiation.

Fig. 6

GCN5 regulates PRKDC expression by acetylating H3K27 in its promoter. A, B) Bar graphs show qPCR analysis of HR and NHEJ pathway associated genes in GCN5 knockdown and scrambled U87MG(A) and LN229(B) cells at 30 min post IR. Each bar represents the average fold change in transcripts of the respective genes post GCN5 knockdown w.r.t. scrambled. Data is represented as mean ± SEM (N = 3; Multiple two-tailed t-test ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, ns- not significant). C) Schematic of PRKDC gene regulatory region coordinates that were amplified by qPCR from immunoprecipitated DNA following ChIP by GCN5, H3K27ac and IgG in LN229 cells at 30 min post IR. D-F) ChIP with GCN5 + IgG (D), H3K27Ac + IgG (E), GCN5 followed by H3K27Ac + IgG (F) were performed in LN229 cells at 30 min post IR treatment followed by qPCR using the primer sets(S1-S8) corresponding to PRKDC promoter sequences. The bar plots represent the Mean ± SEM of % input enrichment (n = 2; Multiple t test; *P < 0.05). G) Luciferase reporter-based promoter assay result depicted by bar plots representing the fold change of PRKDC promoter (-1273 bp to + 52 w.r.t TSS) w.r.t. luciferase control vector in LN229 cells transfected with GCN5 overexpression construct, GCN5 Acetyltransferase mutant construct (GCN5 Acmutant), GCN5 shRNA. Fold change of shGCN5 has been calculated w.r.t pLKO scrambled and both GCN5 wt and Acmutant w.r.t pcDNA3-EGFP empty vector. Mean ± SEM from 3 replicates have been plotted (Multiple t test; ****P < 0.0001; *P < 0.05)

GCN5 regulates DNA-PKcs transcription via H3K27 acetylation

Of the 3 NHEJ factors, DNA-PKcs, 53BP1 and Polymerase Lambda, which were most severely affected post GCN5 knockdown in GBM cells, we found direct binding of GCN5 only on PRKDC promoter from the analysis of available ChIP sequencing data from ENCODE and ChIP Atlas and recently reported GCN5 mediated transcriptional regulation of PRKDC in AML [22]. Thus, we went ahead to find the detailed molecular mechanism of PRKDC regulation by GCN5 post radiation. To locate the exact GCN5 binding site on PRKDC promoter, GCN5 and IgG ChIP was performed in LN229 cells at 0.5 h post IR followed by qPCR using 8 primers (Fig. 6C) spanning the PRKDC promoter (− 6 to − 1600 bp). As compared to IgG, significant GCN5 binding was observed in sequence 3 (fold change: 17) and sequence 5 (fold change: 25), corresponding to − 1050 to − 879 and − 710 to − 554 regions respectively (Fig. 6D). GCN5 is known to regulate gene expression via histone acetylation [11, 34, 35]. Therefore, we asked if GCN5 regulate the expression of DNA-PK by acetylating histones in its promotor. Since the predominant histone, associated with transcriptional activation, acetylated by GCN5 is H3K27, we performed ChIP with H3K27Ac post IR which showed maximum fold enrichment of 18 in − 710 to − 554, corresponding to S5 (Fig. 6E). Since both GCN5 and H3K27Ac enrichment was found at the same promoter sequence of PRKDC independently, to confirm their co-occupancy, we proceeded with sequential ChIP, first with GCN5 followed by H3K27Ac in LN229 IR treated cells. Sequential ChIP also revealed maximum H3K27Ac binding, mean fold change of 6 at the same PRKDC promoter sequence (− 710 to − 554 bp) as that of GCN5 (Fig. 6F). The results strongly suggest that GCN5 regulates PRKDC expression post IR in GBM by acetylating H3K27 in its promotor region.

Next, to check if GCN5 can also regulate DNA-PKcs transcription independent of histone acetylation, we performed luciferase-based promoter assay for DNA-PK in LN229 cell lines having shRNA mediated GCN5 knockdown, overexpression of wild type GCN5 and overexpression of catalytically inactive GCN5 acetyltransferase mutant. Decreased GCN5 expression post knockdown, increased GCN5 expression post transfection of GCN5 acetyltransferase mutant and wild type plasmids, along with loss of H3K27Ac post acetyltransferase mutant overexpression confirmed the genetically altered LN229 cell lines (Fig. S9B). As shown in Fig. 6G, GCN5 knockdown drastically reduced (0.05-fold) while wild type GCN5 overexpression increased the luciferase activity (3.56-fold). However, only a slight increase in luciferase activity was found on overexpression of GCN5 acetyltransferase mutant (1.36-fold), albeit to a much lesser extent than that of wild type indicating that the direct transcriptional regulation of PRKDC by GCN5 requires its acetyltransferase activity. Altogether, ChIP-qPCR and luciferase assay results in LN229 cells revealed that GCN5 directly, being a part of the transcription factor complex as well as via H3K27 acetylation regulates PRKDC expression.

PARP1 and GCN5 perturbation impairs HR and NHEJ repair and radio sensitizes GBM cells

We reasoned if PARP1-GCN5-DNA-PKcs and ATM axis is important for genotoxic stress mediated DNA repair, then inhibition of PARP1 should affect recruitment of downstream repair proteins to DSBs and subsequent DNA repair. Thus, we pharmacologically inhibited and genetically silenced PARP1 and checked for activation of GCN5 targets, H3K9Ac, H3K27Ac, pATM and pDNA-PKcs. In U87MG and LN229 cells, inhibition and PARP1 knockdown significantly decreased H3K9 and H3K27 acetylation and activation of ATM and DNA-PKcs post IR treatment (Fig. 7A–H, S10A–H). Similar results were seen in U2OS cells, where PARP1 inhibition decreased activation of GCN5 targets (H3K27Ac, pATM, pDNA-PKcs) post radiation (Fig. S11A–C).

Fig. 7

Pharmacological and genetic perturbation of PARP1 inhibits GCN5 repair activity in U87MG. A-D) Representative cropped immunofluorescence images of DMSO and Olaparib treated U87MG, with or without 8 Gy IR stained for H3K9Ac (A), H3K27Ac (B), pDNA-PKcs (C), pATM (D) (green) and γ-H2AX (red). E–H) Representative cropped immunofluorescence images of scrambled (Scr) and PARP1 knockdown (shPARP1) U87MG cells, with or without 8 Gy IR stained for H3K9Ac (E), H3K27Ac (F), pDNA-PKcs (G), pATM (H) (green) and γ-H2AX (red). All the images of A-H were acquired at 63X objective magnification and mean fluorescence intensities of mentioned proteins were calculated for individual cells after drawing nuclear ROIs in FIJI (ImageJ). For each group, minimum 60 cells were taken for analysis. Data is represented as Mean ± SD (N = 3; ***p < 0.001, ****p < 0.0001, ns: non-significant, Unpaired two-tailed t test with Welch’s correction)

Since ATM and DNA-PKcs are the key proteins of HR and NHEJ and their activation decreased upon PARP1 and GCN5 inhibition, we performed in-vivo HR and NHEJ reporter assay, as previously described [34] (Fig. 8A) post shRNA mediated PARP1 and GCN5 ablations. Indeed, both HR and NHEJ repair efficiencies were significantly reduced upon PARP1 and GCN5 knockdown in U87MG (Fig. 8B–C). Since there was impaired HR and NHEJ upon GCN5 and PARP1 knockdown, we performed clonogenic radiation survival assay to evaluate whether GCN5 knockdown and olaparib treatment leads to radio sensitization of GBM cells. U87MG and LN229 cells were treated with shGCN5 or 5 µM olaparib followed by variable radiation doses (2, 4, 6, 8, 10 Gy). Colonies were counted at day 7. Indeed, GCN5 knockdown and olaparib treatment led to radio-sensitization of GBM cells (Fig. 8D–E and S11D). Conversely, GCN5 overexpression in LN229 increased radio-resistance (Fig. 8F). We then monitored repair kinetics by ɣ-H2AX immunostaining at 30 min, 3 h, 12 h, 24 h post 4 Gy LN229 cells, transfected with empty vector and GCN5 overexpression construct. GCN5 overexpression hastened resolution of DSBs as observed by decreased ɣ-H2AX at all the time points post IR (Figure S11E).

Fig. 8