DAXX-ATRX 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). Here, we show that DAXX and ATRX knock-out (KO) U87-T cells that have acquired ALT-like features have defects in p53 chromatin binding and DNA damage response. RNA-seq analysis revealed that p53 pathway is among the most perturbed. ChIP-seq and ATAC-seq revealed a genome-wide reduction in p53 DNA-binding and corresponding loss of chromatin accessibility at many p53 response elements across the genome. Both DAXX and ATRX null cells showed a depletion of histone H3.3 and accumulation of γH2AX at many p53 sites, including subtelomeres. These findings indicate that loss of DAXX or ATRX can compromise p53 chromatin binding and p53 DNA damage response in ALT-like cells, providing a link between histone composition, chromatin accessibility and tumor suppressor function of p53.

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). Here, we show that DAXX and ATRX knockout (KO) U87-T cells that have acquired ALT-like features have defects in p53 chromatin binding and DNA damage response. RNA-seq analysis revealed that p53 pathway is among the most perturbed. ChIP-seq and ATAC-seq revealed a genome-wide reduction in p53 DNA-binding and corresponding loss of chromatin accessibility at many p53 response elements across the genome. Both DAXX and ATRX null cells showed a depletion of histone H3.3 and accumulation of γH2AX at many p53 sites, including subtelomeres. These findings indicate that loss of DAXX or ATRX can compromise p53 chromatin binding and p53 DNA damage response in ALT-like cells, providing a link between histone composition, chromatin accessibility and tumor suppressor function of p53. DAXX and ATRX have been implicated as tumor-suppressor proteins that restrict the alternative lengthening of telomeres (ALT) [1][2][3][4] . ALT is a recombination-based mechanism that arises in 10-15% of cancer through poorly understood genetic and epigenetic changes [5][6][7][8] . 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 [9][10][11] , sarcomas 12,13 , and pancreatic neuroendocrine (PanNET) tumors 4,[14][15][16] . Moreover, analysis of 22 immortalized ALT cell lines underscored the involvement of the ATRX/DAXX complex in ALT activation 17 . ALT cells modulate chromatin and DNA repair pathways that facilitate telomere elongation through homologous DNA recombination 10,18 . Telomere recombination is known to be inhibited by telomere repeat binding factors and telomeric heterochromatin 19,20 . Both DAXX or ATRX are known to contribute to telomeric heterochromatin and to limit homologous recombination [1][2][3]9 . However, loss of DAXX or ATRX is not sufficient to establish telomere recombination and cell proliferation associated with ALT 10,11 .
DAXX and ATRX are thought to suppress ALT, in part, through their ability to load histone H3.3 into chromatin 1,10,21 . ALT cells have more accessible telomeric heterochromatin and accumulate high levels of single-strand DNA and DNA damage marked by association with γH2AX and 53BP1 at telomeres 10 . 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 ALTrelated genes shows 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 1,4 . ATRX and DAXX form a complex together with the histone variant H3.3 which is responsible for depositing H3.3 at heterochromatic regions of the genome, including telomeres [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 10 . 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 knockout of DAXX or ATRX that have acquired ALT-like features relative to parental controls. Our findings suggest that the DAXX-ATRX regulates p53 chromatin accessibility and DNA-damage response, that H3.3 may also be involved in this mechanism, and that disruption of this pathway is critical for ALT cell survival.

Deregulation of p53 pathway in DAXX and ATRX-knockout 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 10 . 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 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 DAXX_KO and ATRX_KO, > fivefold upregulation) were identified, including TMTC1, PTGES, CSF1, PLAGL1, MRAS, RTN1, and CD70 (Fig. 1c). Several of these genes have been implicated in glioblastoma tumorigenesis [22][23][24][25][26] . An analysis of TCGA GBM dataset showed that the TP53 pathway is significantly downregulated in ATRX mutated GBM ( Supplementary Fig. 1). The analysis also revealed that all of the ATRX mutated GBMs in TCGA had a corresponding loss of function mutations in TP53, consistent with other reports 27,28 , and suggesting these two pathways are mechanistically linked during GBM oncogenesis.
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, but to a lesser extent in DAXX_KO (Fig. 2a) and ATRX_KO (Fig. 2b)  without etoposide treatment was assayed by flow cytometry (Fig. 2c-e). 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 percentage in the S phase increased in KO cells relative to control after etoposide treatment. A clonogenicity assay also revealed that DAXX_KO and ATRX_KO had defects in growth inhibition relative to parental control in response to etoposide treatment (Fig. 2f, g). Taken together, these findings indicate that DAXX and ATRX KO U87-T cells have changes in cell cycle distribution and defects in cell cycle arrest 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 ALT-like cells relative to parental control U87-T cells using RNA-seq (Quantseq) (Fig. 3a). In total, 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 502 genes between the two knockout conditions, a 6.4-fold more than expected by chance alone (P < 10 −10 , Fisher exact test), indicating very similar functional effect between the two KO cell lines. Enrichment analysis showed several of 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, P < 10 −10 ) (Fig. 3c). Among the genes with the most reduced response (at least fourfold, Fig. 3c) were many wellcharacterized 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). RNA-seq analysis detected no mutations in p53, indicating the reduced p53 response is not due to acquired mutation in the p53 proteincoding sequence.
Altered p53 chromatin binding in DAXX_KO and ATRX_KO U87 ALT-like 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 etoposide treatment (Fig. 4). p53 ChIP-seq peaks were distributed throughout the genome, with the largest distribution occurring >10 K from the transcription start sites (TSS) or at intergenic loci (Fig. 4a). DAXX_KO led to changes in p53 ChIP-seq peaks in both positive and negative directions, with the large majority of strong signal p53 peaks being downregulated in DAXX_KO (Fig. 4b). The mean intensity and overall distribution of strong signal p53 ChIP-seq peaks were reduced in DAXX_KO relative to DAXX_WT (Fig. 4c, d). Integration of ChIP-seq with RNA-Seq identified specific genes co-regulated with changes in neighboring p53-binding sites ( Supplementary Fig. 2). Among those downregulated were LIF, LMNA, and WDR43, while upregulated genes included EDF1, METTL7B, and ARHGAP6. We show specific examples of p53-binding sites decreasing at the CDKN1A and GADD45A (Fig. 4e), compared to not affected (MYOB3) or increasing (EDF1) in DAXX_KO cells relative to control ( Supplementary Fig. 2c). A consensus motif sequence analysis using JASPAR and HOMER indicated that two subtypes of p53 consensus binding sites were most affected in DAXX_KO cells (Fig. 4f, upper panel). Most p53-binding sites have a consensus sequence, variations of these sequences such as quarter vs half-sites 18 , high-binding cooperativity vs low-binding cooperativity 26 or non-canonical binding sequences have different outcomes in terms of regulating gene expression. We identified several other consensus binding sites, such as ATF and Forkhead transcription factors Foxf1, FoxL2, and FoxP1 that had weak, but significant enrichments within p53 sites associated with differential gene expression (Fig. 4f, lower panel). We also correlated p53-binding sites with all available ChIP-seq datasets from ENCODE consortium to identify potential overlaps with known transcription factors or histone modifications ( Supplementary Fig. 3). Each set of ENCODE peaks was overlapped with p53 peaks and enrichment of overlap within downregulated p53-binding sites was calculated and tested using Fisher exact test. Among the most significantly associated with transcription factors were cFOS, STAT3, cMYC, and RNA PolII, and the most correlated histone modifications were H3K4me1 and H3K4me2 (FDR < 5%, Fisher exact test). We validated by ChIP-qPCR some of these changes in p53 binding in DAXX_KO and control cells in response to etoposide treatment (Fig. 4g). We found that the induction of p53 binding to p21 promoter (CDKN1A), GADD45A, as well as subtelomeric sites at 18q and 13q were abrogated in DAXX_KO cells treated with etoposide relative to its parental U87-T cells. ChIP-qPCR analysis revealed a similar reduction of p53 binding to the target genes after etoposide treatment in ATRX_KO cells relative to U87-T control cells (Fig. 4h, IgG controls shown in Supplementary Fig. 4). Negative control genomic regions like TMCC1 and CCDC170 demonstrate the specificity of the p53 ChIP assay (Fig. 4g, h and Supplementary Fig. 4).
To investigate if this effect on p53 binding is observed in any cancer-derived ALT cells, we assayed U2OS ( Supplementary Fig. 5a, b) and GM847 and VA13 ( Supplementary Fig. 5c, d), known to have WT p53 protein and an ATRX null background 29,30 . Western blot revealed that p53 protein was stabilized and phosphorylated upon etoposide treatment in U2OS ( Supplementary Fig. 5a) and constitutively elevated in GM846 and VA13 ( Supplementary Fig. 5c). However, ChIP-qPCR revealed weak (less than twofold) binding of p53 at the response elements in the promoters of CDKN1A and GADD45A and no binding at the subtelomeric response elements in U2OS and GM847 (Supplementary Fig. 5b, d), while VA13 showed some p53 binding to subtelomere repeats ( Supplementary Fig. 5d). These data suggest that several cancer-derived ALT cells with ATRX mutation are also compromised for p53 binding to its target sites in response to DNA-damage agents.

DAXX and ATRX-dependent chromatin accessibility regulates p53-binding patterns
To determine if ATRX and DAXX regulate p53 binding through chromatin accessibility, we performed ATAC-seq in WT, DAXX_KO, and ATRX_KO cells treated with etoposide or DMSO control ( Fig. 5 and Supplementary Figs. 6 and 7). We found a strong correlation between the reduction in p53 binding with the loss of ATAC-seq peaks (Fig. 5a, b and Supplementary Fig. 6a). ATAC-seq peaks that directly overlap with p53 ChIP-seq peaks followed the same trend of reduced peak intensity in both DAXX_KO and ATRX_KO relative to WT cells (Fig. 5c). We highlight several examples of p53 ChIP-seq and ATAC-seq changes ( Fig. 5d-f and Supplementary Fig 6a). At PIEZO2 gene locus, p53binding site overlaps with ATAC-seq peak in control cells with or without etoposide treatment, and each of these peaks is reduced in DAXX_KO and ATRX_KO (Fig. 5d). We also observed examples of p53binding loss, with loss of ATAC-seq peak occurring at neighboring sites, with no change to the p53 overlapping peaks, as seen at the SLC1A3 locus (Fig. 5e). We found examples of large chromosome domains (>100 kb) containing p53-binding sites that were uniformly reduced in ATAC-seq peaks, as was observed for a PIEZO2 and ANKFN1 loci ( Supplementary Fig 6c, d). We also observed changes at subtelomeric p53-binding sites ( Fig. 5f and Supplementary Fig. 7), with the cells. e Percent increase in S phase populations in Etoposide (10 µM) treated cells compared to DMSO treated controls. (ANOVA P = 0.0079; Tukey's multiple comparisons test P values indicated). f Clonogenicity assay with U87-T parental control, DAXX_KO, or ATRX_KO cells treated with DMSO or 50 μM etoposide followed by replating and staining with crystal violet. g Quantification of clonogenicity assay shown in panel f Error bars are sdm and P values determined by student t test for three biological replicates. loss of p53 binding corresponding to changes in ATAC-seq, including an increase in peak score for DAXX_KO, but not ATRX_KO. Interestingly, we observed that ATAC-seq peaks in DAXX_KO and particularly ATRX_KO were increased at telomere repeats containing transitions between perfect (TTAGGG)n and imperfect telomere repeat elements ( Fig. 5f and Supplementary Fig. 7c). Integration of ATAC-seq with RNAseq showed a general direct correlation between ATAC-seq peak changes and changes in transcription in DAXX_KO and ATRX_KO cells ( Supplementary Fig. 8).

Rescue of p53 binding by DAXX requires ATRX interaction
To investigate whether the loss of p53 binding is a specific phenotype of DAXX KO and not an off-target effect of CRISPR gene editing, we utilized a rescue DAXX KO cell line expressing full-length wild-type YFP-tagged hDAXX (hDAXX) 10 (Fig. 6a). 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. 9). 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 coIP (Supplementary Fig. 9). 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. 6b). This was confirmed at the p21 promoter, GADD45A gene, and subtelomeres 18q and 13q. We next tested whether p53 ChIP binding could be rescued by a cancer-associated DAXX mutant previously shown to be deficient for interaction with ATRX (L130R) 10,31 (Fig. 6c, d). In contrast to wild-type hDAXX, L130R mutant failed to rescue p53 binding at each of the response elements tested (Fig. 6d). We also confirmed that DAXX_KO cells demonstrated ALT-associated PML-nuclear bodies (APBs) with enlarged PML-NBs and colocalization with telomere DNA, as measured by IF and FISH respectively, and this could be suppressed in cells with DAXX WT transgene but not by DAXX L130R mutant transgene (Fig. 6e-g). DAXX_KO cells had elevated telomere C-circles as measured by rolling circle assay, and this was suppressed in cells lower quartile. e ChIP-seq tracks for p53 in control (blue) or DAXX_KO (red) at gene loci for CDKN1A (top) and GADD45A (middle). f Consensus motifs for p53 ChIP-seq sites with most differential binding between DAXX_KO and WT (top) and for overlapping sites with known factors other than p53. g ChIP-qPCR for p53 in DAXX_KO and WT control cells treated with DMSO or 50 μM etoposide (red) for binding sites at the p21 promoter, GADD45A, 18q subtelomere, 13q subtelomere, TMCC1 or CCDC170 gene loci. ChIP-qPCR with IgG control is shown in lower panels. h Same as in panel g, except for ATRX_KO cells. Error bars are sdm and P valued determined by student t test for three biological replicates. . d-f UCSC browser visualization of p53 ChIP-seq (top) and ATAC-seq data (lower panels) at PIEZO2 (d) and SLC1A3 (e) and subtelomeric p53 site at telomere 5p (f). Telomere repeat sequence is indicated for ATAC-seq peaks at telomere 5p. expressing WT DAXX but not DAXX L130R transgene ( Supplementary  Fig. 10). These dependencies are consistent with our previous study showing ALT-like features in these and related cell lines with DAXX, as well as ATRX knockouts 10 .

DAXX and ATRX deficiency alter histone composition at affected p53-binding sites
To test if DAXX and ATRX KO alter histone composition at p53-binding sites, we first 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. 7a). 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. 7b), consistent with our previous finding that ATRX_KO led to a reduction of H3.3 occupancy at telomeres 10 . We next assessed whether total histone H3 was also affected at these sites (Fig. 7c). We found that total H3 was also reduced in ATRX_KO and DAXX_KO cells at all sites tested, including p53-binding sites at p21, GADD45A, and subtelomere 18q. A similar loss occurred at a non-p53 bound control region TMCC1, suggesting this effect is not restricted to p53-binding sites. Interestingly, etoposide treatment led to a decrease in total H3 in control cells, but not in ATRX_KO or DAXX_KO cells (Fig. 7c). In contrast, we found that these same sites had an increase in γH2AX binding, correspondingly inverse to that observed with total H3 (Fig. 7d). γH2AX was elevated in basal condition in ATRX_KO and DAXX_KO at all sites tested, and had diminished fold increases after etoposide treatment relative to parental control cells (Fig. 7d). No significant background binding was observed with isotype control IgG (Fig. 7e). These findings suggest that DAXX_KO and ATRX_KO ATL-like cells have altered histone composition with diminished histone total H3 and H3.3, and elevated γH2AX at many genomic sites where p53 binding has been compromised.

Discussion
DAXX and ATRX function as tumor suppressors that are frequently mutated in ALT cancers, yet their precise role in regulating ALT is not fully elucidated. Here, we show that loss of DAXX or ATRX in a U87derived glioblastoma cell line gives rise to ALT-like phenotype with corresponding aberrations in the p53-dependent DNA-damage response. We found that ALT-like U87-T cells lacking DAXX or ATRX had increased p53 pathway signaling in the absence of exogenous DNA damaging agents. Paradoxically, these same cells failed to mount a robust p53 response to DNA damage after treatment with etoposide. ATAC-seq studies revealed that loss of ATRX and DAXX alter the landscape of accessible chromatin at many sites that overlapped with p53-response elements, including subtelomeric p53-binding sites and telomere repeat elements themselves. Our findings suggest that loss of DAXX or ATRX result in a loss of chromatin accessibility at p53-binding sites that attenuates the p53 response. Further, loss of DAXX and ATRX alter chromatin structure at many sites, and these changes correlate with reduced total H3 and H3.3 assembly and increased DNA-damageassociated γH2AX formation at numerous genomic loci, including p53response elements and telomere repeat DNA. We propose that failure of DAXX and ATRX to load H3.3 leads to a global change in chromatin structure, the formation of chronic DNA-damage signaling, attenuated p53 chromatin binding at key sites, and changes in telomeric and subtelomeric chromatin accessibility (Fig. 8). These changes are likely to contribute to the emergence of the ALT cancer pathway.
Both DAXX and ATRX have been implicated in the regulation of p53 and DNA-damage response pathways. DAXX has been shown to copurify with p53 32 , or associate with p53-regulatory proteins, such as MDM2 [33][34][35] . A recent study has found that DAXX functions as a protein foldase and its interaction with MDM2 regulates p53 functional activity 36 . 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 37 . In a different mouse study, DAXX loss led to the derepression of endogenous retroviral elements (ERVs) that could trigger DNA-damage signaling 38 . In HEK cells, DAXX knockout had little detectable effect on p53 target genes in response to DNA damage 39 . On the other hand, DAXX knockout in PanNET cells led to the derepression of oncogenic drivers, including STC2, which was found be regulated by the DAXX/H3.3/H3K9me3 pathway 40 . The ability to reverse ALT phenotype by the re-introduction of ATRX and DAXX is also context and cell-type dependent 9,10,41 . Furthermore, a recent study found that ATRX chromatin accessibility is cell-type specific and stochastic in single-cell studies 42 . These findings suggest that both DAXX and ATRX regulate the p53 pathway through multiple epigenetic mechanisms.
Our findings indicate that both DAXX and ATRX have related, but non-redundant effects on chromatin accessibility. CRISPR knockout of either DAXX or ATRX U87-T cells had highly concordant effects on chromatin accessibility at p53-binding sites throughout the genome. However, the effects on DAXX and ATRX knockout on chromatin accessibility were complex, with many changes occurring in the absence of exogenous DNA damage, at non-p53-binding sites, and sometimes different for ATRX and DAXX. While chromatin accessibility was typically decreased at strong p53 sites, we found several other sites, including imperfect telomere repeats with increase chromatin accessibility. This is consistent with a proposed role for DAXX and ATRX in telomere heterochromatin and restriction of DNA accessible to ALT-associated DNA damage and recombination. We also observed that DAXX and ATRX KO had common changes in chromatin accessibility associated with large (>100 kb) chromatin domains that often contained p53-binding sites, among other factors and gene transcripts. We do not yet know the mechanism or significance of these large chromatin domain effects, but suggest they may account for some of the tumor-suppressor functions of DAXX and ATRX. There are several limitations to our study. Most of the work focuses on a single-cell line (U87-T) and CRISPR-deletions generated in this background where ALT phenotypes were selected. Despite this, we did observe many overlapping chromatin and p53 phenotypes in the independent ATRX and DAXX CRISPR-knockout cell lines. We also found similar attenuation of the p53 pathway in several tumor-derived ALT cell lines, including U2OS, GM847, and VA13. Our study can not exclude that non-chromatin-dependent activities may also regulate p53 stability and DNA-binding activity, as we did observe a general decrease in p53 protein abundance in ATRX and DAXX KO cell lines. However, p53 binding was found to be elevated at a small fraction of chromosomal sites (Fig. 4b and Supplementary Fig 2c) suggests that chromosome structure and accessibility, rather than p53 protein stability, are primary regulators of p53 binding. While DAXX-ATRX has been a well-established chaperone for H3.3, and we demonstrate changes in chromatin accessibility and H3.3 loading at many p53 sites, we also observed a similar loss of total H3 and an increase in γH2AX. This suggests that DAXX and ATRX loss leads to a more general defect in histone assembly. The nature of the chromatin structure that blocks p53 binding is not yet known. R-loops and G-quadruplex DNA have been implicated as potential sources of DNA damage in ALT cells 43 . ATRX is important for the suppression of R-loops 44,45 and G4 quartet formation 46 and these structures are likely to interfere with p53 binding 47 . In this respect, ATRX and DAXX may regulate additional processes related to chromatin structure and DNA-damage control. Others have shown that mutations in ATRX and DAXX can occur in combination with missense mutations in TP53 in many glioblastomas 48,49 . Thus, it should not be surprising that mutations in p53 accelerate tumorigenesis even in the context of DAXX or ATRX defects that compromised chromatin binding. Furthermore, many TP53 mutations confer gain-of-function activity that affects chromatin accessibility 50 , suggesting that barriers to p53 chromosome binding may be a common mechanism of tumorigenesis.
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 family of histones on chromatin, including telomeres. We conclude that the DAXX-ATRX complex plays an important role in regulating p53 access to its binding sites and regulation of the p53 DNA-damage response. We propose that this aberrant DNA-damage response is a key contributor to the acquisition of the ALT-like phenotype.

Cell lines
U87 and U2OS cells used in this study were obtained from American Type Culture Collection (ATCC). ALT cell lines GM847 and VA13 were described as p53+ ALT cells previously 17 and obtained as gift from Roger Greenberg (University of Pennsylvania). U87, U87-T, U87derived cell lines were described previously 10 . U87-T cells were generated by transducing U87 cells with pBabe-Lox-TERT-Lox. ALTpositive single-cell clones of DAXX_KO and ATRX_KO U87 cells were previously described and characterized 19 . DAXX_KO cells is previously . This correlated with a loss of H3.3 binding and accumulation of γH2AX at many loci, including p53-response elements and subtelomeres. Chromatin accessibility was decreased at p53 sites, but increased at telomere repeat DNA, along with loss of telomere T-loops (C-circle formation) and DNA-damage-associated telomere foci (APBs). referred to as U87-T 4-17 10 . ATRX_KO is previously referred to as U87-T A-d#1 10 . DAXX_KO cells is generated as previously described 10 . Briefly, U87-T cells were co-transfected with sgRNA pairs targeting hDAXX and a Puro r containing vector, selected with puromycin for 3-5 days, followed by removal of puromycin. Cells were recovered in a regular growth medium, single clones were expanded and screened by immunostaining and Western blot. U87-T sgCon (referred to a control) has empty Cas9 lentivirus vector with TERT transgene. U87-T derived and ALT cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS and penicillin/streptomycin. U2OS cells were cultured in McCoy's 5 A medium (Life Technologies) supplemented with 10% FBS and penicillin/streptomycin.

Analysis of cell cycle kinetics
U87-T WT sgControl, DAXX_KO, and ATRX_KO cells were seeded at 1 × 10 5 cells/well in six-well plates and exposed to etoposide (10 μM) or DMSO control in biological triplicates per each condition. After 24 h, 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 Quantseq 3'-mRNA kit (Illumina), and sequencing was performed on Illumina Nextseq500 on the high-output mode in a 75 bp single-end run. RNA-seq data were aligned using STAR 51 algorithm against hg19 human genome and RSEM v1.2.12 software 52 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 53 . 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 (FDR < 5%, unless stated otherwise) that had a significantly predicted activation state (|Z| > 2) were reported. Known p53 target list was derived from a published study 54 .

ChIP-seq
ChIP-seq was performed as previously described 18 with certain modifications. Etoposide-(50 μM) treated cells were harvested after 24 h of treatment, fixed with formaldehyde for 10 min followed by quenching with glycine. In total, 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 crosslinked overnight. IP DNA was purified using a Sigma PCR purification kit. Quality control was performed on TapeStation (Agilent). Libraries were prepared using Illumina protocol and sequencing 75 bp single end on Illumina Nextseq500 on high-output mode. The data were aligned using bowtie 55 against hg19 version of the human genome, and HOMER 56 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 57 with mean0 option. Fold differences between samples were then calculated and p53 peaks that reduced signal at least twofold were considered as affected by DAXX KO. Genes were associated with peaks based on 3 kb from any transcript TSS threshold.

ATAC-seq
ATAC-seq was performed essentially as previously described 58,59 . WT, DAXX_KO, and ATRX_KO U87-T cells were treated with DMSO or 50 μM etoposide for 24 h. Then cells were harvested, and ATAC-seq was performed in two biological replicates according to the Omni-ATACseq protocol with modifications. Briefly, 1 × 10 5 cells (>95% viability) were washed in 50 μl cold PBS, spun down at 500 × g and 4°C for 5 min, and resuspended in 50 μl cold ATAC-Resuspension Buffer (RSB) (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 ) containing 0.1% IGEPAL CA-630, 0.1% Tween-20 and 0.01% Digitonin. Resuspended cells were kept on ice for 3 min, then washed with 1 ml cold ATAC-RSB containing 0.1% Tween-20 but no IGEPAL CA-630 or Digitonin. Nuclei were pelleted at 500 × g and 4°C for 10 min, resuspended in a 50 μl Tn5 transposase reaction mixture following the manufacturer's protocol (Illumina Tagment DNA Enzyme and Buffer, Illumina), and incubated at 37°C for 30 min in a thermomixer with 300 rpm mixing. DNA was purified using a MinElute PCR purification kit (Qiagen) and eluted in 10 μl Elution Buffer for library amplification. Tagmented DNA was PCR amplified using the NEBNext HiFi PCR mastermix (New England Bioloabs) with a universal forward and sample-specific reverse oligo for sample barcoding using the following PCR conditions: initial incubations of 72°C for 5 min and 98°C for 30 s, followed by five cycles of 98°C for 10 s, 63°C for 30 s, and 72°C for 1 min. Additional number of cycles was determined for each sample through a "side" qPCR reaction using an aliquot of the PCR as a template to determine the number of cycles needed to reach 1/3 of the max fluorescence. Total cycle numbers ranged from seven to ten cycles. PCR products were run on a 1% agarose gel, regions from~50 bp to~1 kb were excised, and DNA was extracted using a gel extraction kit (Qiagen). Purified DNA was submitted to the Wistar Institute Genomics core facility for quality analysis and sequencing. All samples were sequenced on NextSeq500 (Illumina) to generate paired-end 2 × 42 bp reads. ATAC-Seq raw reads were aligned using bowtie 55 against the hg19 version of the human genome, followed by HOMER 56 to generate bigwig files and call significant peaks using the "-style dnase" option. Normalized ATAC signals within p53 sites were derived from bigwig files using bigWigAverageOverBed tool from the UCSC toolbox 57 with mean0 option. Unique ATAC-Seq sites with at least one peak that passed an FDR < 5% threshold in both replicates were considered significant. The significance of pair-wise differences between conditions was estimated using DESeq2 53 . Large chromosomal domains with uniform overall ATAC signal changes were defined as any >100 kb regions with at least 5 significant ATAC sites with at least 70% of those significantly changed by both DAXX and ATRX KO (FDR < 5%, fold > 2).

ENCODE
All available transcription factor and histone modification data peaks were downloaded from ENCODE and overlapped with the set of p53 peaks. The difference in proportions of overlapped peaks between p53 sites downregulated in DAXX KO vs not downregulated was tested using Fisher exact test and Benjamin-Hochberg procedure was used to correct for multiple testing. ENCODE factors and histones that passed FDR < 5% were clustered using binary distance with average linkage.

ChIP-qPCR
ChIP-qPCR was performed similar to ChIP-seq with few changes. One million cells per sample were used, and sonication was performed using Diagenode water bath sonicator at a high range 30 s on 30 s 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 Quant-Studio 7. Primers used are listed in Supplementary Data 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 Data 1.

Immuno-FISH assay
Indirect immunofluorescence (IF) combined with fluorescence in situ hybridization (FISH) analysis was performed as described previously 10 . Primary antibody to PML (ab96051, Abcam) was prepared in 1:500 dilution in blocking solution. After IF, cells were fixed in 4% paraformaldehyde in 1× PBS for 10 min, washed in 1× PBS, dehydrated in ethanol series (70%, 95%, 100%), and air-dried. Coverslips were denatured for 5 min at 80-85°C in a hybridization mix [70% formamide, 10 mM Tris-HCl, pH 7.2, and 0.5% blocking solution (Roche)] containing telomeric PNA-Tamra-(CCCTAA) 3 probe, and hybridization was continued for 2 h 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 ×63 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.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data supporting the findings of this study are included within the article and its Supplementary Information files. All Illumina-based sequencing data are available from the NCBI Sequence Read Archive (SRA) under accession number code PRJNA773790 with the specific accession code PRJNA773796 for RNA-seq, PRJNA773795 for ChIP-seq, and PRJNA853364 for ATAC-seq datasets. Source data are provided with this paper.