DAXX-ATRX-H3.3 Regulation of p53 Chromatin Binding and DNA Damage Response

DAXX and ATRX are tumor suppressor proteins that form a histone H3.3 chaperone complex and are frequently mutated in cancers with the alternative lengthening of telomeres (ALT), such as pediatric glioblastoma. Rapid loss of function of either DAXX or ATRX are not by themselves sufficient to induce the ALT phenotype. However, cells lacking DAXX or ATRX can be readily selected for ALT-like features. Here, we show that DAXX and ATRX null glioblastoma cells with ALT-like features have defects in p53 chromatin binding and DNA damage response regulation. RNA-seq analysis of DAXX or ATRX null U87 glioblastoma cells with ALT-like features revealed that p53 pathway is among perturbed. ALT-selected DAXX and ATRX-null cells had aberrant response to DNA damaging agent etoposide. Both DAXX and ATRX-null ALT cells showed a loss of p53 binding at a subset of response elements. Complementation of DAXX null cells with a wild-type DAXX transgene rescued p53 binding and transcription, while the tumor associated mutation L130R that disrupts ATRX binding was incapable of rescuing p53 chromatin binding. We show that histone H3.3 binding is reduced in DAXX-null cells especially at subtelomeric p53 binding sites and telomere repeats. These findings indicate that DAXX and ATRX function to enable p53 chromatin binding through modulation of histone H3.3 binding, especially at sub-telomeric sites. Both DAXX and ATRX have been implicated in regulation of p53 and DNA damage response pathways. DAXX has been shown to co-purify with p53 in several cell types. A recent study has found that DAXX functions as a protein foldase and its interaction with p53 regulates interaction with MDM2 and p53 functional activity 21 . In a mouse genetic study, heterozygous deletions of DAXX led to an increase sensitivity to low dose radiation and higher incidence of carcinomas, but no overt defect in p53 function or pathway was reported 22 . In a different mouse study, DAXX loss led to the de-repression of endogenous retroviral elements (ERVs) that could trigger DNA damage signaling 23 . In HEK cells, DAXX knock-out had little detectable effect on p53 target genes in response to DNA damage 24 . On the other hand, DAXX knock-out in PanNET cells led to the derepression of oncogenic drivers, including STC2, which was found be regulated by DAXX/H3.3/H3K9me3 pathway 25 . The ability to reverse ALT phenotype by the re-introduction of ATRX and DAXX is also context and cell-type dependent 31,32 . These findings suggest that DAXX may interact with p53 pathway through various mechanisms that depend on cell or cancer type. of G4 quartet structures at p53 response elements. Our study finds that DAXX and ATRX KO cells with ALT-like phenotype have aberrant p53 DNA damage response pathways. These cells have an increase in the constitutive Images were captured with a 63x lens on a Zeiss LSM780 confocal laser scanning microscope (Carl Zeiss). Cells with >5 PML foci colocalizing with telomere DNA foci were scored as APB positive. The quantification was generated from at least three independent Immuno-FISH experiments.


Introduction
DAXX and ATRX have been implicated as tumor suppressor proteins that restrict the alternative lengthening of telomeres (ALT). ALT is a recombination-based mechanisms that arises in 10-15% of cancer through poorly understood genetic and epigenetic changes 5,6,7 . Whole genome sequencing studies have identified loss of function mutations in ATRX, DAXX, and histone H3.3 to be significantly associated with ALT induction in different classes of gliomas [8][9][10] , sarcomas, and pancreatic neuroendocrine tumors 1,2 . Moreover, analysis of 22 immortalized ALT cell lines have also underlined the involvement of ATRX/DAXX complex in ALT activation 11 . ALT cells modulate chromatin and DNA repair pathways that facilitate telomeres elongation through homologous DNA recombination 12 . Telomere recombination is known to be inhibited by telomere repeat binding factors and telomeric heterochromatin 3,4 . Both DAXX or ATRX are known to contribute to telomeric heterochromatin and to limit homologous recombination 5,6,7,8 .
However, loss of DAXX or ATRX is not sufficient to establish telomere recombination and cell proliferation associated with ALT 9, 10 .
DAXX and ATRX are thought to suppress ALT, in part, through their ability to load histone H3.3 into chromatin 5,9,11 . ALT cells have altered telomeric heterochromatin and accumulate high levels of single-strand DNA and DNA damage marked by association with gH2AX and 53BP1 at telomeres 9 . This DNA damage signaling reflects both double strand breaks and homologous recombination repair. The extent and tolerance of DNA damage increases further upon mutation of TP53 (p53) 13 . TCGA pan cancer atlas study of ALT related genes show TP53 to be one of the top mutated genes 14 . p53 is a DNA binding protein and master transcriptional regulator of genome integrity controlling transcription of gene networks for cellular senescence and apoptosis 15,16 . p53 binds to several loci within the cellular genome including promoters, enhancers, transcriptionally inactive regions as well as retrotransposon-like elements in many human subtelomeres 17,18 . The inter-relationship between DAXX-ATRX, and p53 DNA-damage response function in ALT is not completely understood.
Both DAXX and ATRX are multifunctional tumor suppressor proteins 5,12 . ATRX and DAXX form a complex together with the histone variant H3.3 which is responsible for depositing H3.3 at heterochromatin regions and telomeric regions of the genome [22][23][24] . In addition to their function in loading H3.3 onto telomeric chromatin, DAXX and ATRX bind many other proteins, including p53 and MDM2, and can localize with PML-containing nuclear bodies (PML-NBs) that associate with ALT telomeres. ATRX/DAXX/H3.3 and p53 mutations frequently co-occur in ALT cancers 9 , suggesting that these pathways may be additive in cancer cell progression. In a previous study, we demonstrated that loss of ATRX or DAXX is not sufficient for ALT, but that additional genetic and/or epigenetic events can lead to the emergence of clones with ALT-like features 9 . Here, we investigate the relationship between DAXX and ATRX loss of function and p53 DNA damage sensing pathway using sets of syngenic glioblastoma cell lines with CRISPR/Cas9 knock-out of DAXX or ATRX that have acquired ALT-like features relative to parental controls. Our findings suggest that the DAXX-ATRX-H3.3 axis is required for p53 chromatin binding and DNA damage response, and that disruption of these pathways are critical for ALT cell survival.

Deregulation of p53 pathway in DAXX and ATRX knock-out cells with ALT-like features.
We have previously reported the generation of ATRX and DAXX deficient U87-T glioblastoma cell lines that have acquired ALT-like features 9 . In that study, we found that transient depletion of DAXX or ATRX led to an increase in senescent cells, but that continuous passage enabled the isolation of clonal survivors with hallmarks of ALT. To gain further insight into the common features of these ALT-like DAXX_KO and ATRX_KO cells, we performed RNA-seq and transcriptomic analysis. We found overlap of 2724 genes significantly (FDR<5%) affected by both DAXX_KO and ATRX_KO relative to parental U87-T control cells (Fig. 1a). Gene enrichment analysis of the overlap revealed a number of significantly affected regulators (FDR<10 -5 , Z>2) with p53 reaching the most significance and predicted to be more active in the KO cell lines (Fig. 1b). Seven known p53 direct targets among 28 top upregulated genes (FDR<5% in both, >5 fold up regulation) were identified, including TMTC1, PTGES, CSF1, PLAGL1, MRAS, RTN1, and CD70 (Fig. 1c). Several of these gene have been implicated in glioblastoma tumorigenesis 13,14,15,16,17 .
We re-validated that the DAXX_KO cell line had ALT-like features (Fig. S1). Southern blot analysis shows that DAXX knock-out (DAXX_KO) cell lines have more heterogeneous telomere length distribution compared to parental controls (Ctrl) (Fig. S1b). Consistent with our previous publication 18 , the bulk of telomere DNA signals reduced in both Ctrl and DAXX_KO cells treated with topoisomerase inhibitor etoposide. Telomere circles (t-circles) were observed in DAXX_KO cells by 2D agarose gel analysis, and their frequency was reduced by reintroduction of a wild-type (WT) DAXX transgene (Fig. S1c). ALT-associated PML-nuclear bodies (APBs) with enlarged PML-NBs and colocalization with telomere DNA, as measured by IF and FISH respectively, were observed at high-frequency in DAXX_KO relative to WT cells, and this could be suppressed in cells with DAXX WT transgene but not by DAXX L130R mutant transgene ( Fig. S1d and e). DAXX L130R is a tumor derived mutation of DAXX with biochemical defect in ATRX binding and histone H3.3 chaperone function. Furthermore, DAXX_KO cells had elevated telomere C-circles as measured by rolling circle assay, and this was suppressed in cells expressing WT DAXX but not DAXX L130R transgene (Fig. S1f).
These findings are consistent with our previous study showing ALT-like features in these and related cell lines with ATRX and DAXX knock-outs 9 .

Attenuated DNA damage response in ALT-like cells lacking ATRX or DAXX.
Given the transcriptomic alteration of p53 response pathway in ATRX_KO and DAXX_KO cells, we assessed their ability to respond to DNA damage relative to parental control U87-T cell line.
We found that the topoisomerase inhibitor etoposide induced a potent p53-response and corresponding growth inhibition in parental U87-T cells (Fig. 2). Western blot analysis indicated that p53 and phospho-S15 p53 levels were induced in U87-T parental, and to a lesser extent in DAXX_KO (Fig. 2a) and ATRX_KO (Fig 2b) cell lines. In contrast, gH2AX levels were modestly increased in KO cells relative to parental control. These findings suggest that KO cell lines have attenuated p53 response and accumulate higher levels of gH2AX DNA damage in response to etoposide treatment. Cell cycle distribution of each cell type with or without etoposide treatment was assayed by flow cytometry (Fig. 2c and d). We found that DAXX_KO and ATRX_KO cells were enriched in G1 and depleted in S/G2 relative to control under normal growth conditions, while the percent in S phase increased in KO cells relative to control after etoposide treatment. A clonogenicity assay also revealed that DAXX_KO and ATRX_KO had defective growth inhibition relative to parental control in response to etoposide treatment (Fig 2e   and f). Taken together, these findings indicate that DAXX and ATRX KO U87-T cells have changes in cell cycle distribution and cell cycle arrest defects in response to DNA damage.

Altered p53 response transcriptome in U87 ALT-like cells lacking ATRX or DAXX.
We next assayed the RNA-seq (Quantseq) transcriptomic response to DNA damaging agent etoposide in DAXX_KO and ATRX_KO U87 ALT-like cells relative to parental control U87-T cells (Fig. 3a). 933 genes in DAXX_KO and 1562 genes ATRX_KO were found to be less responsive to etoposide (upregulated in WT at FDR<5%, but significantly less upregulated in KO condition at p<0.05). There was an overlap of 512 genes between the two cell lines, a 6.4 fold more than expected by chance alone indicating very similar functional effect between the two KO cell lines. Enrichment analysis showed several the most affected regulators (FDR<10 -3 , Z>2) including p53 to be significantly inhibited in response to etoposide in DAXX_KO or ATRX_KO cells (Fig. 3b). There was a strong correlation of the magnitude of the attenuation effect between DAXX_KO and ATRX_KO (Spearman r=0.34, Fig. 3c). Among the genes with the most reduced response (at least 4 fold, Fig. 3c) there were many well-characterized p53 response genes, including GADD45A, TNFRSF10A (TRAIL), MAFB, TP53INP1, USP2, and TIMP3. RT-qPCR demonstrated that some p53 response genes, such as GADD45A, CYP4F3, and PARDG6 were affected by DAXX-KO and ATRX_KO, while other genes, such as CDKN1A, were not significantly affected in DAXX_KO (Fig. 3d).

Altered p53 chromatin binding in DAXX_KO and ATRX_KO U87 ALT cells.
Since many p53 response genes were affected in both DAXX_KO and ATRX_KO cells, we assayed p53 binding by ChIP-seq in DAXX_KO and control cells with or without etoposide treatment (Fig. 4). After etoposide treatment we observe significant loss of p53 peak signal throughout the genome in DAXX KO cells (Fig. 4a). We identified a subset of p53 binding sites that were more significantly affected as compared to other regions where there was only modest reduction. We highlight sites at the CDNK1A, GADD45A, and miR-34 genes that showed significant disruption in DAXX_KO cells compared to controls. A consensus sequence analysis using JASPAR and HOMER indicated that two subtypes of p53 consensus binding sites were most affected in DAXX_KO cells (Fig. 4b, upper panel). Most p53 binding sites have a consensus sequence, variations of these sequences such as quarter vs half-sites 18 , highbinding cooperativity vs low-binding cooperativity 26 or non-canonical binding sequences have different outcomes in terms of regulating gene expression. We also identified several other consensus binding sites, such as ATF and Forkhead transcription factors Foxf1, FoxL2, and FoxP1 that had weak, but significant enrichments with p53 sites and differential gene expression (Fig. 4b, lower panel). We validated these p53 binding by ChIP-qPCR for several p53 binding sites, including those associated with well-known response genes p21 and GADD45, as well as at sub-telomeric p53 binding sites at chromosomes 18q and 13q (Fig. 4c).
Negative control genomic regions like TMCC1 and CCDC170 demonstrate the specificity of the p53 ChIP assay. Western blot analyses indicate that p53 levels are nearly identical in DAXX_KO and ATRX_KO after etoposide treatment ( Fig. 2a and b), so it is unlikely that total levels of p53 account for these substantial differences in chromatin binding. Further, we assayed another ALT cell line, the osteosarcoma derived U2OS, that is known to have an ATRX null background 19,20 . Although p53 was stabilized and phosphorylated upon etoposide treatment as measured by Western blot (Supplemental Fig. S2a), we observe weak (less than 2-fold) stabilization of p53 at the response elements in the promoters of CDKN1A and GADD45A and no stabilization at the subtelomeric response elements (Supplemental Fig.   S2b). These findings suggest that DAXX or ATRX are necessary for efficient chromatin binding in ALT cells.

Rescue of p53 binding by DAXX requires H3.3 chaperone function.
To more rigorously determine whether the loss of p53 binding is a specific phenotype of DAXX KO and not an off-target effect of CRISPR gene editing, we generated a rescue DAXX KO cell line by expressing full length wild-type YFP-tagged hDAXX (hDAXX) (Fig. 5a). We noticed that DAXX overexpression led to an increase in p53 protein levels in response to etoposide treatment. This suggests p53 protein expression levels may be partly dependent on DAXX. To determine if an interaction between DAXX and p53 may be altered, we assayed this potential interaction by coIP (Supplementary Fig. S3). However, we were unable to demonstrate a stable interaction between p53 and DAXX, although MDM2 was identified in p53 coIP, and ATRX was identified in DAXX co-IP (Supplementary Fig. S3). ChIP-qPCR studies showed that the p53 binding at various p53 response elements was defective in DAXX_KO cells, but was restored in the cells expressing hDAXX (Fig. 5b). This was confirmed at the p21 promoter, GADD45A gene, and subtelomeres 18q and 13q. We next tested whether the DAXX mutant deficient for ATRX binding (L130R) and H3.3 binding histone chaperone function could rescue p53 binding (Fig. 5c-d). In contrast to wild-type hDAXX, L130R mutant failed to rescue p53 binding at each of the response elements tested (Fig. 5d). This suggests that ATRX binding and H3.3 loading are important for p53 chromatin binding. To test this more directly, we assayed H3.3 binding by ChIP-qPCR. We found that DAXX_KO cells were depleted for histone H3.3 at several p53 binding sites, including those at subtelomeres 13q and 18q (Fig. 5e). We further assayed H3.3 loading at telomere repeats by dot blot analysis, and that DAXX_KO cells had reduced H3.3 at telomere repeats relative to controls, and relative to actively transcribing GAPDH gene (Fig. 5h). Taken together, these results suggest that p53 chromatin binding depends on H3.3 loading by DAXX-ATRX in these ALT-like cells.

DISCUSSION
DAXX and ATRX function as tumor suppressors that are frequently mutated in ALT cancers, yet their precise role in regulating ALT is not completely clear. Here, we show that loss of DAXX or ATRX in a U87 derived glioblastoma cell line gives rise to ALT-like phenotype, with corresponding aberrations in the p53-dependent DNA damage response. We found that U87-T cells lacking DAXX or ATRX had increase p53 pathway activation. Paradoxically, these same cells failed to mount a robust p53-response to exogenous DNA damage from etoposide. Our findings suggest that loss of DAXX or ATRX result in a chronic DNA damage tolerance that limits p53 response to further DNA damage. Our data further suggests that this DNA damage tolerance is linked to the failure to load histone H3.3 at p53 response elements and telomere repeat DNA (Fig. 6).
Both DAXX and ATRX have been implicated in regulation of p53 and DNA damage response pathways. DAXX has been shown to co-purify with p53 in several cell types. A recent study has found that DAXX functions as a protein foldase and its interaction with p53 regulates interaction with MDM2 and p53 functional activity 21 . In a mouse genetic study, heterozygous deletions of DAXX led to an increase sensitivity to low dose radiation and higher incidence of carcinomas, but no overt defect in p53 function or pathway was reported 22 . In a different mouse study, DAXX loss led to the de-repression of endogenous retroviral elements (ERVs) that could trigger DNA damage signaling 23 . In HEK cells, DAXX knock-out had little detectable effect on p53 target genes in response to DNA damage 24 . On the other hand, DAXX knock-out in PanNET cells led to the derepression of oncogenic drivers, including STC2, which was found be regulated by DAXX/H3.3/H3K9me3 pathway 25 . The ability to reverse ALT phenotype by the re-introduction of ATRX and DAXX is also context and cell-type dependent 31,32 . These findings suggest that DAXX may interact with p53 pathway through various mechanisms that depend on cell or cancer type.
Mutations in ATRX and DAXX can occur in combination with missense mutations in TP53 in many glioblastomas 26,27 . These missense mutations may have partial and altered activities, and in combination with DAXX or ATRX, result in a further loss of p53 tumor suppressor function. Many TP53 mutations confer gain-of-function activity that affects chromatin accessibility 28 . G-quadruplex DNA was identified as a feature that was differentially recognized by some p53 gain of function mutants. Some gain of function mutations in p53 were found to have differential binding by ChIP-seq and this binding was found dependent on the chromatin modifying complex MLL 29 . ATRX is important for suppression of G4 quartet formation 30 and G4 quartets can inhibit p53 binding 31 . So, it is possible that ATRX deficiency prevent p53 binding due to formation of G4 quartet structures at p53 response elements.
Our study finds that DAXX and ATRX KO cells with ALT-like phenotype have aberrant p53 DNA damage response pathways. These cells have an increase in the constitutive activation of p53 response genes, but a failure of p53 to respond to exogenous damage. We note that p73 and p63 are elevated under these conditions and may partly account for the basal activation of the p53 pathway in the absence of p53 functionality due to ATRX or DAXX loss.
The loss of p53 functionality correlates with reduced binding to consensus sequences throughout the genome with more drastic effects at some subtelomeric sites. We find that the DAXX-dependent defect in p53 correlates with its interaction with ATRX and the assembly of H3.3 on chromatin, including telomeres. We conclude that the DAXX-ATRX-H3.3 axis plays an important role in regulating p53 access to its binding sites and regulation of the p53 DNA damage response. We propose that his aberrant DNA damage response is a key contributor to the acquisition of the ALT phenotype.

Cell lines
U87, U2OS and HEK293T cells used in this study were obtained from American type culture collection (ATCC). U87, U87-T, U87 derived cell lines were described previously 9 . U87-T cells were generated by transducing U87 cells with pBabe-Lox-TERT-Lox. ALT positive single cell clones of DAXX_KO and ATRX_KO U87 cells were previously described and characterized 19 .

Analysis of cell cycle kinetics
U87-T WT shControl, Daxx_KO, and ATRX_KO cells were seeded at 1 x 10 5 cells/well in 6-well plates and exposed to etoposide (10 µM) or DMSO control in biological triplicates per each condition. After 24h, cells were permeabilized with cold, 70% ethanol and resuspended in PBS containing PI (10 mg/mL) and RNAse A solution (100 µg/mL). Flow cytometry was performed on a BD-LSR II (BD Biosciences; Bedford, MA) and FloJo software (Ashland, OR) was used for cell cycle analysis.

RNA-seq
Total RNA was extracted using Trizol following manufacturer's instructions. RNA was treated with Turbo DNA-free kit (life technologies) and quality checked with TapeStation (Agilent).
Libraries were prepared using Quant-seq 3'-mRNA kit (Illumina) and sequencing was performed on Illumina Nextseq500 on high-output mode in a 75 bp single end run. RNA-seq data was aligned using STAR 32 algorithm against hg19 human genome and RSEM v1.2.12 software 33 was used to estimate read counts using gene information from Ensemble transcriptome version GRCh37.p12. Raw counts were used to estimate significance of differential expression difference between any two experimental groups using DESeq2 34 . Overall gene expression changes were considered significant if passed FDR<5 threshold. Gene set enrichment analysis was done using QIAGEN's Ingenuity Pathway Analysis software (IPA, QIAGEN Redwood City,www.qiagen.com/ingenuity) using "Upstream Regulators" options. Most significant regulators (p<0.001, unless stated otherwise) that had a significantly predicted activation state (|Z|>2) were reported.

ChIP-seq
ChIP-seq was performed as previously described 18 with certain modifications. Etoposide (10 µM) treated cells were harvested after 24 hours of treatment, fixed with formaldehyde for 10 min followed by quenching with glycine. 25 million cells per sample were sonicated using Covaris sonicator to obtain fragments of size between 100-500 bp. Chromatin was incubated with 10 mg of p53 antibody (Santa Cruz sc-126X) overnight followed by pull-down using protein-G Dynabeads (life technologies). Beads were washed, eluted and reverse cross-linked overnight.
IP DNA was purified using Sigma PCR purification kit. Quality control was performed on TapeStation (Agilent). Libraries were prepared using Illumina protocol and sequencing 75bp single end on Illumina Nextseq500 on high-output mode. The data was aligned using bowtie 35 against hg19 version of the human genome and HOMER 36 was used to generate bigwig files and call significant peaks vs input using -style factor option. Peaks that passed FDR<5% threshold were considered significant. Normalized signals for significant peaks were derived from bigwig files using bigWigAverageOverBed tool from UCSC toolbox 37 with mean0 option.
Fold differences between samples were then calculated and p53 peaks that reduced signal at least 2-fold were considered as affected by DAXX KO. Genes were associated with peaks based on 3kb from any transcript TSS threshold.

ChIP-qPCR
ChIP-qPCR was performed similar to ChIP-seq with few changes. 1 million cells per sample were used and sonication was performed using Diagenode water bath sonicator at high range 30s on 30s off for a total of ~20 min with intermittent incubation on ice to obtain fragment size between 100-500 bp. qPCR primers were designed using NCBI primer blast and the run was performed on ABI QuantStudio 7. Primers used are listed in Supplemental Table 1.

qRT-PCR
Total RNA was extracted using Trizol following manufacturer's instructions. RNA was treated with Turbo DNA-free kit (life technologies). Primers were designed using NCBI primer blast and the run was performed on ABI QuantStudio 7. Primers used are listed in Supplemental Table 1.

Telomere Restriction Fragment (TRF) Analysis of Telomeric DNA Length
TRF analysis of telomere DNA length was performed as described previously 9 . Briefly, genomic DNA was prepared using Wizard genomic DNA purification kit (Promega), digested with AluI+MboI restriction endonucleases, fractionated in a 0.7% agarose gel, denatured, and

Two-dimensional (2D) gel electrophoresis
Two-dimensional (2D) gel electrophoresis was performed as previously described 9 . Briefly, equal amounts of AluI+MboI digested DNA (15 μg) was subjected to electrophoresis in a 0.4% agarose gel (first dimension) at room temperature and 30 V for 12-14 hr, and then in a 1.2% agarose gel (second dimension) containing 0.3 μg/ml ethidium bromide at 4°C and 150V for 6 hr. The gel was processed as described above for the Southern blot analysis.
Immuno-FISH assay. Indirect immunofluorescence (IF) combined with fluorescence in situ hybridization (FISH) analysis was performed as described previously 9 . Primary antibody to PML (ab96051, Abcam) was prepared in 1:500 dilution in blocking solution. After IF, cells were fixed in 4% paraformaldehyde in 1x PBS for 10 min, washed in 1x PBS, dehydrated in ethanol series (70%, 95%, 100%), and air-dried. Coverslips were denatured for 5 min at 80-85°C in hybridization mix [70% formamide, 10mM Tris-HCl, pH 7.2, and 0.5% blocking solution (Roche)] containing telomeric PNA-Tamra-(CCCTAA)3 probe, and hybridization was continued for 2 hrs at room temperature in the dark moisturized chambers. Coverslips were washed, counterstained with 0.1 µg/ml DAPI in blocking solution, and mounted with VectorShield (Vector Laboratories). Images were captured with a 63x lens on a Zeiss LSM780 confocal laser scanning microscope (Carl Zeiss). Cells with >5 PML foci colocalizing with telomere DNA foci were scored as APB positive. The quantification was generated from at least three independent Immuno-FISH experiments.      ChIP assay for U2OS cells treated as in panel a with either DMSO (grey) or etoposide (red) using antibody to p53 (top row) or igG control (lower row), and assayed by qPCR with primers for p53 binding sites at the p21 promoter, GADD45A, 18q subtelomere, 13q subtelomere, or TMCC1. Figure S3. CoIP of p53 with MDM2 but not DAXX or ATRX in U87 cells. U87T control cells were subject to IP with p53, DAXX, or IgG control, and then assayed by Western blot for DAXX, ATRX, p53 or MDM2. Input (5%) is show for each antibody.