The Transcriptional Repressor hDaxx Potentiates p53-dependent Apoptosis*

p53 and its homologues p73 and p63 are transcription factors that play an essential role in modulating cell cycle arrest and cell death in response to several envi-ronmental stresses. The type and intensity of these responses, which can be different depending on the inducing stimulus and on the overall cellular context, are believed to rely on the activation of defined subsets of target genes. The proper activation of p53 family members requires the coordinated action of post-transla-tional modifications and interaction with several cofactors. In this study, we demonstrate that the multifunctional protein hDaxx interacts with p53 and its homologues, both in vitro and in vivo , and modulates their transcriptional activity. Moreover, we show that hDaxx, which has been implicated in several apoptotic pathways, increases the sensitivity to DNA damage-induced cell death and that this effect requires the presence of mouse monoclonal anti-Mdm2 (kindly provided by M. Oren); rabbit polyclonal anti-PIG3 (Oncogene Science); C19, rabbit polyclonal anti-p21 (SantaCruz Biotechnology); and N20, rabbit polyclonal anti-Bax (Santa Cruz Biotechnology). For detection of endogenous Daxx protein in RNA interference experiments, we used an affinity-purified rabbit anti-Daxx polyclonal antibody kindly provided by J. D. Chen (13). Western blot was performed with standard procedures, using horse- radish peroxidase-conjugated secondary antibodies (Sigma). Bound antibodies were visualized by enhanced chemiluminescence (Pierce). For immunofluorescence analysis, cells seeded on coverslips were fixed with 3% paraformaldehyde, permeabilized by treatment with 0.1% Triton X-100 in PBS, and then incubated fo r 1 h at 37 °Cwith the indicated primary antibodies. Primary antibodies were revealed by a 30-min incubation with goat anti-mouse tetramethylrhodamine B iso-thiocyanate- and anti-rabbit fluorescein isothiocyanate-conjugated sec- ondary antibodies (Sigma). Images were obtained with the use of a Leica DMLB microscope and Photometrics Coolsnap camera. Yeast Two-hybrid Screening— The yeast two-hybrid screening with LexA-p53wt and LexA-p73 (cid:2) strains was performed as described previ- ously (27) using a human fetal brain cDNA library cloned into the galactose-inducible expression vector pJG4–5. At least 2 million pri- mary clones were analyzed in each screening. Positive interaction between bait and fish protein resulted in the transcription of two different reporters, lacZ and Leu2, and was evaluated by blue staining on X-Gal-containing plates and growth on medium lacking leucine. For Western blot analysis, cells grown in medium containing either glucose or ga-lactose and raffinose were subjected to mechanical lysis with glass beads in SDS-containing sample buffer.

The p53 protein is one of the most important cellular tumor suppressors, lying at the heart of many different but interconnected stress response pathways, whose action is required to prevent genomic instability (1). Once activated, p53 coordinates a complex cellular response ending in reversible cell-cycle arrest, irreversible senescent-like state, or apoptosis (2).
Multiple mechanisms are responsible for controlling p53 activation within cells, and most of them involve post-translational modifications of the protein, such as phosphorylation, acetylation, and sumoylation (2,3). Stress-induced post-translational modifications may affect the stability and the conformation of p53 (4), its ability to interact with positive and negative regulators and its subcellular localization (3).
Much less understood are the mechanisms that control the specificity of p53 response and that allow, depending on the stress and cell type, discrimination between growth arrest and cell death. p53 exerts its functions mostly as a transcription factor; although induction of cell cycle arrest is mainly mediated by the CDK inhibitor p21, a number of effectors are required to coordinate the apoptotic response (2). Experimental evidence indicates that promoter specificity may be determined by the sequence of the p53 binding site on DNA, for which the protein can have different affinities (5), and also by the interaction of p53 with co-factors responsible for directing it toward specific gene subsets (6,7). p53 belongs to a family of proteins comprising two additional members, p63 and p73, which share extensive structural and functional homologies (7). p63 and p73, despite being able to bind to p53 consensus sites and to activate transcription of several p53-responsive genes, do not behave as classic tumor suppressors and have been implicated in control of differentiation and developmental pathways (8,9). The mechanisms governing the activation of p53 homologues as well as their functional specificity remain poorly understood.
To gain insights on the mechanism of regulation of p53 family members, we sought to analyze their interaction profile and to isolate proteins able to bind all three proteins or specifically to only one of them. Among the common interactors for all p53 family members, we identified hDaxx, a highly conserved protein, initially isolated in mouse as a Fas interactor (10). Daxx contains in its primary sequence two nuclear localization signals that are conserved between murine and human proteins. Therefore, it is predominantly nuclear and has been shown to associate with PML 1 and to localize within PML nuclear bodies (NBs) (11)(12)(13).
The ability of hDaxx to interact with multiple cellular factors has resulted in its assignment to several putative functions (14). Many reports have implicated Daxx in apoptosis; however, whether it functions as a pro-or antiapoptotic molecule has not yet been clarified. Despite the lack of evidence of interaction between Daxx and Fas in human cells, hDaxx overexpression is able to potentiate Fas-induced apoptosis and a direct role of hDaxx in activation of the apoptosis signal-regulating kinase 1/Jun N-terminal kinase pathway has been demonstrated, at least in some cell types and under some experimental conditions (10,(15)(16)(17). Tumorigenic mutant p53 has recently been shown to interact with hDaxx and to counteract its ability to activate the Jun N-terminal kinase pathway and to induce cell death (18). Moreover, hDaxx has been involved also in transforming growth factor ␤-induced (19) as well as in nuclear, PML-dependent apoptotic pathways (13,20). Depletion of hDaxx by antisense RNA showed a protective effect toward transforming growth factor ␤-induced apoptosis (19).
In contrast with these observations, Daxx knock-out embryos showed early embryonic lethality, and studies with Daxxnull murine embryonic stem cells revealed an antiapoptotic role for this protein (21). It has recently been shown that ablation of Daxx expression by RNA interference can increase apoptosis in different cell types and that this effect is rescued by Bcl-2 overexpression (22,23).
The proapoptotic function of Daxx within the nucleus has been linked with its ability to function as a transcriptional repressor, possibly as a result of interaction with histone deacetylases and core histones (12,24). Daxx can inhibit basal transcription when fused to a Gal4-DNA binding domain, but under more physiological conditions, this effect is not generalized and has been observed only for some promoters, where hDaxx can be recruited by binding to specific transcription factors, such as Pax3 and ETS1 (25,26). In this respect, it is conceivable that the dual role played by Daxx in apoptosis may be explained, at least in part, by subtle differences in the transcriptional status of a given cell.
In this study, we demonstrated that hDaxx interacts with wild-type (wt) p53 and its homologues p73 and p63 both in vitro and in vivo. Binding to hDaxx caused repression of the transcriptional activity of p53 family members toward the p21 promoter. In the case of p53, the effect of hDaxx seemed to be promoter-dependent, repressing preferentially genes involved in cell-cycle arrest rather than proapoptotic genes. Therefore, hDaxx overexpression sensitizes cells to drug-induced apoptosis, whereas RNAi knock-down of hDaxx is protective toward p53-dependent cell death.

EXPERIMENTAL PROCEDURES
Plasmids-pLexA-p53wt and deletions pcDNA 3 p53wt and pGEX-p53wt have been described previously (27). To generate pLexA-p73␣, the human p73␣ cDNA was PCR-amplified from amino acid 112 to the stop codon. pLexA-p63␣ contains the human p63␣ cDNA from amino acid 132 to the stop codon and was a kind gift of S. Piccolo. pcDNA 3 -HAp73␣ and pcDNA 3 -mycp63␣ were kindly provided by G. Melino and F. McKeon, respectively. pGEX-p73␣ and pGEX-p63␣ contain the whole human p73␣ and p63␣ open reading frame fused to the glutathione S-transferase (GST) coding sequence and were provided by G. Blandino. cDNA encoding hDaxx was obtained from I.M.A.G.E. consortium (RZPD, Deutsches HumanGenomeProjekt) and subcloned into pcDNA 3 and into the retroviral vector pLPC. To generate the HA-hDaxx-pIND construct for inducible expression of HA-tagged hDaxx, the hDaxx cDNA was first introduced into pcDNA 3 -HA and then subcloned into pIND vector (Invitrogen). p53-responsive reporter constructs have been described previously and were provided by M. Oren, B. Vogelstein, M. Levrero, and X. Lu. Constructs encoding hDaxx deletions fused to GST were kindly provided by H. Will. pcDNA 3 -HATNV contains the Escherichia coli thioredoxin cDNA cloned upstream of an HA tag and before a nuclear localization signal and vesicular stomatitis virus tag. Oligonucleotides encoding for p53 peptides were cloned within the RsrII site of the thioredoxin cDNA.
Cell Lines-All cell lines were cultured at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, penicillin (100 units/ml), and streptomycin (100 mg/ml). U2OS and SaOS-2 are human osteosarcoma cells wt and null, respectively, for p53. 293 is a human embryonic kidney cell line.
hDaxx-inducible Cell Line-To generate the ecdysone-inducible hDaxx cell line, U2OS cells were co-transfected with pVgRXR (Invitrogen), expressing the two subunits of the Drosophila melanogaster ecdysone receptor, and with pIND-hDaxx, expressing HA-tagged hDaxx under the control of the ecdysone-inducible promoter. After selection of cells with Zeocin (150 g/ml) and G418 (150 g/ml) for 2 weeks, clones were isolated and analyzed for HA-hDaxx expression by Western blot. Selected positive clones were further expanded and processed.
Western blot was performed with standard procedures, using horseradish peroxidase-conjugated secondary antibodies (Sigma). Bound antibodies were visualized by enhanced chemiluminescence (Pierce).
For immunofluorescence analysis, cells seeded on coverslips were fixed with 3% paraformaldehyde, permeabilized by treatment with 0.1% Triton X-100 in PBS, and then incubated for 1 h at 37°C with the indicated primary antibodies. Primary antibodies were revealed by a 30-min incubation with goat anti-mouse tetramethylrhodamine B isothiocyanate-and anti-rabbit fluorescein isothiocyanate-conjugated secondary antibodies (Sigma). Images were obtained with the use of a Leica DMLB microscope and Photometrics Coolsnap camera.
Yeast Two-hybrid Screening-The yeast two-hybrid screening with LexA-p53wt and LexA-p73␣ strains was performed as described previously (27) using a human fetal brain cDNA library cloned into the galactose-inducible expression vector pJG4 -5. At least 2 million primary clones were analyzed in each screening. Positive interaction between bait and fish protein resulted in the transcription of two different reporters, lacZ and Leu2, and was evaluated by blue staining on X-Galcontaining plates and growth on medium lacking leucine. For Western blot analysis, cells grown in medium containing either glucose or galactose and raffinose were subjected to mechanical lysis with glass beads in SDS-containing sample buffer.
In Vitro Binding and Immunoprecipitation Assays-In all cases, cells were seeded in 10-cm plates, transfected with the indicated vectors by standard calcium phosphate method, and further processed 36 h later. GST pull-down assays were performed by lysing cells in 300 mM NaClcontaining buffer (300 mM NaCl, 50 mM Tris, pH 7.5, 0.5% Nonidet P-40, and 10% glycerol) supplemented with protease inhibitors (1 mM phenylmethylsulfonyl fluoride and 10 g/ml each chymostatin, leupeptin, antipain, and pepstatin). Lysates were then diluted 1:2 in the same buffer without NaCl and incubated for 2 h at 4°C with 4 g of Sepharose-GSH-bound GST proteins. For co-immunoprecipitation analysis, cells were lysed in 150 mM NaCl-containing buffer (150 mM NaCl, 50 mM HEPES, pH 7.5, 0.1% Tween 20, and 10% glycerol) and incubated for 1.5 h at 4°C with 1 g of anti-p53 (FL-393) or anti-Daxx (M-112) polyclonal antibodies. 20 l of Protein A-Sepharose CL-4B (Amersham Biosciences) were then added to each sample, and incubation at 4°C was carried on for an additional 1.5 h. After three washes in lysis buffer, immunoprecipitated proteins were separated by SDS-PAGE and analyzed by Western blot with the indicated antibodies.
Retroviral Infection-Phoenix packaging cells were transfected with empty pLPC or with pLPC-hDaxx by standard calcium phosphate method. After 48 h of incubation at 32°C, the supernatants containing viral particles were collected, diluted 1:2 with fresh medium, and used to infect sub-confluent U2OS or SaOS-2 cells. After overnight incubation at 32°C, cells were split and kept under selection with 1 g/ml puromycin for 1 week. The polyclonal populations of infected cells were then analyzed for hDaxx expression, expanded, and further processed. Four U2OS and three SaOS-2 independent polyclonal pLPC and pLPC-hDaxx lines were generated and analyzed.
Reporter Assays-SaOS-2 or U2OS cells were seeded in 24-well plates and transfected with LipofectAMINE 2000 (Invitrogen) with 200 ng of Luc reporters, 20 ng of p53, p73␣, or p63␣ expression plasmids (only in the case of SaOS-2), and 400 ng of pcDNA 3 -hDaxx or empty pcDNA 3 . In all the samples, 20 ng of the reporter pRL-CMV (Promega) was included for normalization of the transfection efficiency. 24 h after transfections, cells were lysed and assayed for luciferase activity using the Dual Luciferase kit (Promega). For reporter assays upon silencing of Daxx expression in SaOS-2 cells, 1 pmol of Daxx-specific small interfering RNA or control oligonucleotides were included (see below).
Apoptosis Assays and RNA Interference-U2OS and SaOS-2 stable pLPC and pLPC-hDaxx lines were seeded in 6-cm plates; 24 h later, cisplatin (cis-diamminedichloroplatinum II (CDDP) prepared in PBS at a concentration of 500 g/ml) was added to the culture medium at the final concentration of 2.5 g/ml. After 48 or 72 h, cells were collected by trypsinization, recovering also the supernatants, washed in PBS, and fixed with 70% ethanol at Ϫ20°C. After several washes in PBS, cells were resuspended in 50 l of PBS/0.1% Nonidet P-40 containing 2 g/ml RnaseA; 5 min later, 200 l of 50 g/ml propidium iodide in PBS were added. Cells were analyzed by cytofluorimeter (Bio-Rad Bryte HS) after an additional 20 min of incubation. Apoptosis was evaluated by scoring the percentage of cells with sub-G 1 DNA content.
For RNA interference of p53, cells seeded in 6-cm plates were transfected with Oligofectamine (Invitrogen) according to the instructions of the manufacturer, with 6 pmol of dsRNA oligonucleotides specific for human p53 (GACUCCAGUGGUAAUCUACdTdT) or with control scrambled oligonucleotides (CCUUUUUUUUUGGGGAAAAdTdT). Cells were further processed 24 h after transfection. For apoptosis experiments upon silencing of hDaxx, SaOS-2 cells seeded on glass coverslips in 2-cm dishes were transfected by LipofectAMINE 2000 (Invitrogen) with 2 pmol of hDaxx-specific double-stranded RNA oligonucleotides (GGAGUUGGACCUGUCAGAGCdTdT) or with control scrambled oligonucleotides, together with 1 g of p53 expression vector or pEGFP. 24 h after transfection, cells were fixed with 3% paraformaldehyde and stained for the expression of p53 (DO-1 monoclonal antibody; Santa Cruz Biotechnology). DNA was stained with Hoechst 33342 and, upon observation through a UV-light microscope, nuclei exhibiting condensed chromatin were scored as apoptotic. At least 200 cells that were p53-positive (or GFP-positive in control experiments) for each of four independent experiments were counted in randomly selected fields from each plate.

RESULTS
hDaxx Interacts with p53 Family Members-With the aim to gain information about the interaction profile of p53 and its homologues p73 and p63, we performed a series of yeast twohybrid screenings to identify proteins able to bind to each family member. These three proteins share a similar protein domain organization (see Fig. 1A), with the highest homology in the DNA binding and oligomerization domains. In contrast to p53, however, p73 and p63 are expressed as a series of alternative splice variants, differing mostly at their C termini; the longest isoform (␣) contains the sterile ␣ motif (8).
Yeast strains expressing transactivation domain-deleted human p53 and p73␣ fused to the LexA DNA binding domain were transformed with a fetal brain cDNA library, and screenings were performed as described previously (27). Putative p53 and p73 interactors were isolated and sequenced. The encoding cDNAs were subsequently assayed for interaction with LexA-p63␣. One of the most representative clones isolated in both screenings (accounting for more than 25% of the p53-interact-FIG. 1. hDaxx interacts with p53 family members. A, interaction with hDaxx was analyzed by yeast two-hybrid assay using p53, p73␣, or p63␣ deleted in their transactivation domain and fused with LexA, as indicated in the upper scheme. The various domains are indicated. "ϩ" indicates positive interaction, as judged by ␤-galactosidase activity and by the ability to grow in the absence of leucine. The same strains used for the interaction assays were analyzed for the expression of the bait (anti-LexA, top) and of the fish (anti-HA, bottom) proteins, either in the presence of glucose (G) or galactose/raffinose (G/R) as carbon source. B, in vitro binding assay with GST-p53, GST-p73␣, GST-p63␣ or GST, as indicated, was performed on lysates from p53-null SaOS-2 cells overexpressing hDaxx. Western blot (WB) was then performed with anti-Daxx polyclonal antibody. The bands corresponding to hDaxx protein and running positions of molecular mass markers are indicated. C, co-immunoprecipitation on lysates from 293 cells overexpressing hDaxx was performed with anti-p53 polyclonal antibody or with normal rabbit serum as a negative control. Immunoprecipitates were analyzed by Western blotting with anti-Daxx antibody (top). An aliquot of each total lysate was then checked for the expression of hDaxx and p53 proteins (bottom). D, co-immunoprecipitation on lysates from 293 cells overexpressing HA-tagged p73␣ alone or together with hDaxx was performed with anti-Daxx polyclonal antibody, and immunoprecipitates were analyzed by Western blotting with anti-HA monoclonal antibody (top). An aliquot of each total lysate was then checked for the expression of the hDaxx and p73 proteins (bottom).
ing and 40% of the p73-interacting clones) contained cDNAs of different lengths, all encoding the C-terminal region of hDaxx. As represented schematically in Fig. 1A, hDaxx showed a strong and specific interaction with wt p53 and p73␣, as well as with p63␣.
Because the binding between hDaxx and p53 family members in yeast was strong and specific, we first sought to confirm the interaction by performing a GST in vitro binding assay. Lysates prepared from p53-null SaOS-2 cells transfected with vector expressing hDaxx were subjected to pull-down assay with beads loaded with GST-p53, GST-p73␣, GST-p63␣ or with GST alone as negative control. Subsequent Western blot analysis of the resin-bound proteins (Fig. 1B) revealed that hDaxx interacted specifically with GST-fused p53 family members.
To verify the binding in human cells, we next performed a co-immunoprecipitation assay on lysates from 293 cells transfected with hDaxx expression vector, using anti-p53 polyclonal antibody or normal rabbit serum as a control. As shown in Fig.  1C, hDaxx was clearly detectable in the p53-bound fraction, whereas no specific signal was observed in the negative control.
Likewise, we immunoprecipitated with anti-Daxx antibody lysates of 293 cells transfected with vector expressing HAtagged p73␣ together with hDaxx expression plasmid or with empty plasmid as a control. The immunoprecipitated proteins were then analyzed by Western blot with anti-HA antibody (Fig. 1D), demonstrating also in this case a specific interaction between hDaxx and p73. These results therefore demonstrated the ability of hDaxx to bind to all p53 family members hDaxx Acts As a Repressor of the Transcriptional Activity of p53 Family Members-Because hDaxx has been shown to interact with several transcription factors and to modulate their activity, we sought to verify whether this binding could affect the transcriptional activity of p53 family members. We therefore performed reporter assays with a construct containing the p53-responsive p21 promoter cloned upstream of the firefly luciferase gene (p21-Luc). p53-null SaOS-2 cells were transfected with p21-Luc and either p53, p73␣, or p63␣ expression vectors, together with hDaxx expression vector or empty plas-mid as a control. In each experiment, transfection efficiency was monitored by co-expressing a fixed amount of a second reporter construct containing the CMV promoter upstream of the Renilla reniformis luciferase gene (pRL-CMV). As shown in Fig. 2, hDaxx overexpression was able to reduce the transcriptional activity of all p53 family members, even though the effect was more pronounced in the case of p53. It is noteworthy that hDaxx overexpression did not affect basal p21 transcription in the absence of overexpressed p53, p73, or p63, demonstrating the specificity of the effect observed. Western blot analysis on the same lysates used for reporter assays confirmed similar levels of expression of p53, p73␣, and p63␣, in either the absence or the presence of overexpressed hDaxx (Fig. 2, bottom).
These results suggest that binding with hDaxx may influence the ability of all p53 family members to activate transcription, as previously reported for other transcription factors, with no major consequences on their expression levels and/or stability.
Distinct Domains of hDaxx Are Responsible for the Binding to p53-We decided to further characterize the interaction between hDaxx and the founder member of the family, p53. hDaxx is a protein of 740 amino acids and the shortest cDNA clone isolated from the yeast two-hybrid screening encodes its C-terminal portion, from amino acid 605 to 740. The same region has also been shown to be involved in the interaction with several other factors, such as Fas (10) and PML (11).
To verify that the C terminus of hDaxx mediates the interaction with p53 family members also in other experimental systems, we took advantage of different constructs bearing hDaxx deletions (represented in Fig. 3A) fused to GST, corresponding to amino acids 1-188 (GST-hDaxx A), 189 -400 (GST-hDaxx B), 410 -600 (GST-hDaxx C), and 600 -740 (GST-hDaxx D). In vitro pull-down assays performed with lysates of wt p53-containing 293 cells (Fig. 3A, bottom) clearly indicated that the C-terminal 140 amino acids of hDaxx were sufficient to mediate the interaction with p53 (Western blot, lane 6). Interestingly however, binding was also detected with GST-hDaxx A and, to a lesser extent, with B (lanes 3 and 4), indicating that N-terminal regions could also interact with p53. Similar results FIG. 2. hDaxx represses transcriptional activity of p53 family members. Luciferase activity assays were performed on lysates from p53-null SaOS-2 cells, transiently transfected with the p21-Luc reporter together with wt p53, p73␣, or p63␣ and empty pcDNA3 or pcDNA3-hDaxx, as indicated. To normalize the transfection efficiency, a fixed amount of pRL-CMV reporter, constitutively expressing the R. reniformis luciferase gene, was included in each sample. were obtained using in vitro-translated p53 and p73 (data not shown), suggesting that multiple domains of hDaxx might be responsible for contacting p53 and its homologue p73.
The Oligomerization Domain of p53 Mediates the Interaction with hDaxx-To identify the region of p53 responsible for the binding, we first tested by yeast two-hybrid assay the C-terminal part of hDaxx isolated from the screening and several p53 deletion mutants. As summarized in Fig. 3B, deletion of p53 C terminus to amino acid 355 did not affect the binding, whereas removal of 17 more residues almost abolished the interaction. This indicates that amino acids between 338 and 355 of wildtype p53, within the oligomerization domain, are required for binding to hDaxx. Therefore, a construct encoding only the C-terminal region of p53 (amino acids 294 -393) showed strong and specific interaction with hDaxx.
To confirm these findings in mammalian cells, we used constructs encoding different peptides derived from the p53 Cterminal region, inserted within the scaffold of the thioredoxin protein. These plasmids contain a cassette in which the thioredoxin cDNA is fused with an N-terminal HA tag and with C-terminal nuclear localization signal and vesicular stomatitis virus tag (HA-TNV, schematically represented in Fig. 3C). SaOS-2 cells were transfected with hDaxx expression vector together with a plasmid expressing the empty HA-TNV cassette or p53 peptides corresponding to regions 322-355, 355-363, and 363-384. Lysates were then immunoprecipitated with anti-Daxx polyclonal antibody and analyzed by Western blot with anti-HA antibody. As presented in Fig. 3C, the peptide corresponding to p53 oligomerization domain (amino acids 322-355) showed a specific interaction with hDaxx, whereas no binding was detectable with any of the other peptides or with the empty HA-TNV cassette.
hDaxx Sensitized Cells to p53-dependent Apoptosis-The above results, together with evidence indicating a complex role for hDaxx in regulation of cell death, prompted us to test whether it could be also involved in modulating p53-dependent apoptosis. We generated cell lines stably expressing hDaxx in the background of wt p53-containing U2OS cells or in p53-null SaOS-2 cells as a control.
Cells were infected with retroviruses expressing hDaxx (pLPC-hDaxx) or with the empty pLPC vector and, after 4 days of selection, the polyclonal infected cell populations were expanded and further analyzed. As shown in Fig. 4B, a clear increase in hDaxx expression was detectable in the U2OS stable cell lines (U2OS/hDaxx) compared with control cells (U2OS/ pLPC). Similar results were also obtained in SaOS-2 cell lines Bottom, in vitro binding assays were performed with lysates of 293 cells and different hDaxx deletions fused to GST. Samples were then separated by SDS-PAGE and analyzed by Western blot with anti-p53 monoclonal antibody, and ponceau red staining of the membrane was used to verify the amounts of the different GST-fusions loaded in each sample. B, interaction with hDaxx was analyzed by yeast two hybrid assay using different p53 deletions fused with LexA. ϩ, positive interaction; Ϫ, no detectable interaction. C, p53-null SaOS-2 cells were transiently transfected with hDaxx together with vectors expressing peptides corresponding to different p53 regions within the thioredoxin scaffold (HA-TNV), as indicated. Immunoprecipitation was performed with anti-Daxx polyclonal antibody; subsequently, Western blot (WB) was performed with anti-HA monoclonal antibody. (Fig. 4D). hDaxx localization in the stable cell lines was comparable with the endogenous protein, as detected by immunofluorescence analysis with a polyclonal anti-Daxx antibody (Fig. 4C) with clear nuclear diffuse and punctate staining. No effect of hDaxx overexpression on growth rate and doubling capacity of the cells was detected, in agreement with previously reported data indicating that simple overexpression of hDaxx is not sufficient to trigger cell death (10,13).
To analyze p53-dependent apoptosis in hDaxx-overexpressing stable lines, cells were treated for 48 h with the chemotherapeutic drug CDDP and then analyzed for induction of apoptosis by scoring the sub-G 1 DNA content by FACS analysis. As reported in Fig. 4A, results obtained from several experiments performed with polyclonal populations of hDaxx-overexpressing U2OS cells derived from four independent infections clearly indicated that hDaxx-overexpressing cells are sensitized to CDDP-induced cell death. Increased apoptosis was confirmed by monitoring the levels of cleaved poly(ADP-ribose) polymerase protein, a well established caspase target, using an antibody specific for the processed p85 fragment (Fig. 4B, second  panel).
In search for a mechanism underlying the observed effect of hDaxx overexpression on apoptosis, we initially verified p53 protein levels in the stable cell lines, but we could observe no significant change in p53 expression and subcellular localization in hDaxx-overexpressing cells compared with control pLPC cells, either under untreated conditions or after CDDP treatment (see Fig. 4, B, third panel, and C). In addition, the levels of p53 phosphorylated on Ser-15 were comparable in the two cell lines (Fig. 4B, fourth panel). We next performed the same experiments in p53-null SaOS-2 cells. Although CDDP treatment for up to 72 h was able to efficiently induce cell death in these cells, no significant differences in apoptosis rates were observed in three independent polyclonal populations of hDaxx-infected SaOS-2 cells compared with pLPC-infected (Fig. 4D). These results strongly imply that the ability of hDaxx to increase CDDP-induced cell death is dependent on p53. To further confirm the involvement of p53 in the effect observed, we repeated the experiments after ablation of p53 expression by RNA interference in the U2OS/ hDaxx stable lines. U2OS/hDaxx or control U2OS/pLPC were transfected with small interfering RNAs specific for p53 (sip53) or with control scrambled oligonucleotides (siC). 24 h after transfection, cells were split in two plates and one of them was subjected to CDDP treatment for additional 48 h. Efficient reduction of p53 expression levels following small interfering RNA transfection was verified by Western blotting (Fig. 4E, lower panels). FACS analysis of the sub-G 1 DNA content revealed that, in the absence of p53, the sensitivity of hDaxxoverexpressing cells to CDDP-induced apoptosis was similar to that of control pLPC-infected cells (Fig. 4E). This was also confirmed by analyzing the levels of cleaved poly(ADP-ribose) polymerase in the different cell lysates (Fig. 4E, bottom).
Taken together, these results demonstrate that hDaxx sensitizes cells to p53-dependent cell death, therefore identifying a novel link between this multifunctional protein and another apoptotic pathway.
hDaxx Differentially Modulates p53-responsive Genes-The results obtained with hDaxx-overexpressing cells are in apparent contrast with the data from reporter assays, where we observed a negative effect of hDaxx on p53 transactivation ability. In these experiments, however, we used only a reporter construct containing the promoter of the p53-responsive p21 gene, which is mainly involved in mediating growth arrest rather than apoptosis. We then decided to analyze in more detail the effect of hDaxx on p53 transcriptional response, employing reporter constructs containing different p53-responsive promoters. wt p53-containing U2OS cells were transiently transfected with each reporter construct together with a vector expressing hDaxx or with empty pcDNA 3 as a control. Transfection efficiency was monitored by co-expressing a fixed amount of pRL-CMV. As presented in Fig. 5A, and in agreement with what we observed in SaOS-2 cells (Fig. 2), hDaxx overexpression caused a marked reduction of the activity of p21 and Mdm2 promoters. No significant effect was observed, how-ever, on several promoters of proapoptotic genes, such as Bax, PIG3, and AIP1, whereas transactivation of the PUMA promoter was increased. Immunoblotting of the same lysates confirmed that the difference in transcriptional activity was not reflecting changes in p53 protein levels (Fig. 5A, bottom).
To prove that the observed effect of hDaxx on transcription was dependent on p53, we performed similar reporter assays in p53-null SaOS-2 cells (see Fig. 5B). In the absence of p53, when hDaxx was overexpressed along with the various reporter constructs, no significant changes in transcriptional activity were observed. On the contrary, when a p53 expression vector was co-transfected, hDaxx repressed transcription of the Mdm2 promoter, whereas it did not significantly affect PIG3 promoter and activated transcription of the PUMA reporter.
We then sought to verify the effects of hDaxx overexpression on endogenous p53 target genes. Lysates were prepared from U2OS/pLPC and U2OS/hDaxx cell lines, either untreated or treated with cisplatin for 48 h. Western blot analysis clearly indicated that CDDP-induced up-regulation of p21 was impaired in cells overexpressing hDaxx (Fig. 5C), confirming the result obtained with reporter assays.
To further confirm and expand these findings, we generated a U2OS cell line in which hDaxx expression could be induced by treatment with Ponasterone A, a synthetic analogue of ecdysone. Upon induction of hDaxx expression and treatment with cisplatin for 48 h, cell lysates were analyzed by Western blot for the expression of several p53-induced genes. As shown in Fig.  5D, DNA damage-mediated up-regulation of both p21 and Mdm2 was reduced upon induction of hDaxx overexpression compared with control, uninduced cells, whereas the levels of two p53 proapoptotic targets, PIG3 and Bax, were slightly  5. hDaxx differentially modulates transcription from different p53-responsive promoters. A, U2OS cells expressing endogenous wt p53 were transfected with either empty pcDNA 3 or pcDNA 3 -hDaxx expression vector together with reporter constructs bearing different p53-responsive promoters upstream of the luciferase gene, as indicated. 36 h after transfection, luciferase activity assays were performed. To normalize the transfection efficiency, a fixed amount of pRL-CMV reporter was included in each sample. Graphs represent the mean of at least four independent experiments. Standard deviations are indicated. An aliquot of each sample was analyzed by Western blot (WB) to verify the expression Role of hDaxx in p53-dependent Apoptosis 48020 increased by hDaxx overexpression. All together, the above results clearly indicate that hDaxx was able to modulate p53 transcriptional activity. This effect is promoter-dependent, resulting in repression of p53-mediated induction of genes involved in cell cycle arrest (p21), whereas genes involved in apoptosis (Bax, PIG3, AIP1) are either stimulated or remain unaffected.
Endogenous hDaxx Required for p53 Apoptotic Function-Having demonstrated that overexpression of hDaxx potentiated p53-dependent apoptosis, we then sought to investigate whether the endogenous hDaxx protein could exert the same effect on p53 activity. Therefore, we knocked down hDaxx expression by transfecting p53-null SaOS-2 cells with small interfering RNA specific for hDaxx (sihDaxx) or control scrambled oligonucleotides (siC), together with either wt p53 or control plasmid, to compare the effects of hDaxx depletion in cells expressing or not p53. Twenty-four hours later, cells were fixed, stained with Hoechst dye 33342, and nuclear morphology was analyzed by epifluorescence to identify apoptotic cells. As can be seen in Fig. 6A, transfection with sihDaxx oligonucleotides effectively reduced endogenous hDaxx expression, and this resulted in a moderate increase in apoptosis in the absence of p53, in agreement with previous observations (22,23). However, in cells that expressed p53, ablation of hDaxx expression significantly reduced p53-dependent apoptosis (Fig. 6A). To further extend this analysis, we performed reporter assays in SaOS-2 cells transfected with p21-Luc or Bax-Luc reporter constructs together with wtp53 and either sihDaxx or siC oligonucleotides. As shown in Fig. 6B, and consistent with results from overexpression experiments (Figs. 2 and 5), reduction of hDaxx levels caused a 2-fold increase in p53-dependent transactivation of the p21 promoter, whereas no significant effect was observed on Bax promoter. These results therefore confirmed the evidences obtained after hDaxx ectopic expression and demonstrated an important role of this multifunctional protein in regulation of p53 functions.

DISCUSSION
In this study, we demonstrated that the multifunctional protein hDaxx bound to p53 and its homologues p73 and p63. The interaction took place both in vitro and in vivo and requires the oligomerization domain of p53. This region is conserved also in p73 and p63 (8), allowing us to predict, therefore, a common hDaxx interaction domain in all three proteins of the p53 family.
The exact biological function of hDaxx awaits elucidation; despite some evidence implicating this protein in control of apoptosis, its exact role in this process remains controversial. Although Daxx overexpression potentiates apoptosis induced by stimuli such as treatment with Fas-ligand and transforming growth factor ␤ (13,15,19), an antiapoptotic effect of this protein has been proposed based on knock-out mice models and RNA interference studies (21,22). It seems, therefore, that Daxx might have a dual function in regulation of apoptosis, probably depending also on the cellular context, as underlined by the fact that no increase in Fas-induced apoptosis was observed after overexpression of hDaxx in lymphoid cells (28) and that Daxx exerted an antiapoptotic role in myeloid precursors (29).
Our results clearly indicated that physiological levels of hDaxx protein are important in setting the threshold for p53dependent apoptosis. In fact, we showed that whereas overexpression of hDaxx sensitized wt p53-containing cells to CDDPinduced cell death, ablation of endogenous hDaxx expression resulted in a protective effect toward p53-dependent apoptosis. Such an effect requires p53, because we observed in p53-null cells a moderate increase in cell death after down-regulation of hDaxx protein levels, as reported previously (22,23). These results represent one of the few demonstrations of a proapoptotic role exerted by hDaxx at physiological expression levels and further underline the complex role of this multifunctional protein in controlling the apoptotic response.
In agreement with previous observations that hDaxx binds to several transcription factors and modulates their activity (5,25,26,30), we were able to show that hDaxx represses the transcriptional ability of p53 and its homologues p73 and p63 toward the p21 promoter. However, at least in the case of p53, this negative effect was significant with respect to promoters derived from the p21 or Mdm2 genes, involved in cell cycle arrest or in p53 regulation, whereas no differences were observed in the induction of promoters of proapoptotic genes such as bax, PIG3, and AIP1. It is noteworthy that hDaxx overexpression could also modulate the levels of endogenous p53 target genes, with effects mirroring those observed in reporter assays. Although the exact mechanism mediating p53 promoter specificity is not clear, it is now well established that at least two classes of p53-binding sites exist within p53-responsive promoters, with p21 and Mdm2 belonging to a "high affinity" class of sites, whereas bax and PIG3 to a "low affinity" one (5). Moreover, specific post-translational modifications of p53 as well as interaction with cellular cofactors might differentially regulate its ability to interact with a defined subset of DNA targets. Therefore, it has recently been shown that the p53-binding protein ASPP1 specifically stimulates the apoptotic function of p53, by promoting its binding to apoptosisrelated promoters (6). In addition, the p53 family members p73 and p63 are required to modulate the capability of p53 to efficiently bind and transactivate promoters of proapoptotic genes (31). Our data are in line with these observations and suggest that hDaxx may be another factor involved in such a complex regulation. However, the exact mechanism by which the repressive function of hDaxx operates specifically toward some p53-induced genes requires further elucidation. It should be noted that, in the case of the PUMA promoter, hDaxx overexpression causes an increase in transcriptional activity in reporter assays. Even though hDaxx was shown to behave as repressor in most cases, it has also been demonstrated that, in the case of Pax5, interaction with hDaxx may result either in transcriptional repression or activation, depending on the cell type (30). The repressive function of hDaxx has been linked to its ability to recruit histone deacetylases to target promoters; indeed, a specific interaction between hDaxx and several histone deacetylases has been reported (12,24). However, hDaxx can bind also to core histones, particularly to their acetylated forms (24); therefore, its activity in a particular promoter context might be differentially regulated by association with tranlevels of endogenous p53 and overexpressed hDaxx proteins (bottom). B, luciferase activity assays were performed on p53-null SaOS-2 cells transfected with the indicated p53-responsive reporters together with empty pcDNA 3 or pcDNA 3 -hDaxx, either in the presence or in the absence of wt p53 expression vector. Graphs represent the mean of three independent experiments; standard deviations are indicated. C, U2OS/pLPC and U2OS/hDaxx cell lines were treated with CDDP or left untreated, as indicated. 48 h later, lysates were analyzed by Western blotting with anti-Daxx and anti-p21 primary antibodies. Stabilization of p53 after DNA damage was verified by anti-p53 staining, whereas anti-actin staining was used as loading control. D, U2OS cells expressing inducible HA-tagged hDaxx were treated with CDDP for 48 h, either inducing Daxx expression with Ponasterone A (PonA) or not, as indicated. Lysates were then analyzed by Western blotting with the indicated primary antibodies. Expression of HA-hDaxx as well as stabilization of p53 after DNA damage were verified by anti-Daxx and anti-p53 staining, respectively. Anti-actin staining was used as loading control. scription factors and additional co-repressors/co-activators. In the case of Pax5, for example, hDaxx was also found in complex with the transcriptional co-activator CBP (30).
Although we observed no major changes in hDaxx protein levels after DNA-damage treatment, it is possible that the interaction between hDaxx and p53 family members is regulated by other mechanisms. Several factors seem to modulate hDaxx subcellular distribution by recruiting it to specific compartments, such as the cytoplasm in the case of interaction with apoptosis signal-regulating kinase 1 (17,32) or the nucleolus in the case of the interaction with the 58-kDa microspherule protein (33). The interaction with PML is particularly interesting in this context because hDaxx recruitment to PML nuclear bodies has been reported not only to be essential for its proapoptotic effect but also to relieve its transcriptional repressive activity (11,12,20). PML has been implicated in transcriptional regulation and has been shown to modulate responses on different promoters via interaction and sequestration of coactivators and repressors to the NBs (12,34,35). PML also regulates the function of p53 (36 -38) and recruitment of p53 to NBs has been related, at least in some experimental settings, with increased transactivation of proapoptotic targets and reduction of cell survival (36 -38). The picture has been further complicated by the recent demonstration that PML is a direct target of p53, thus contributing to its antiproliferative effects (39). How the observed interaction between p53 and hDaxx fits in this scenario still requires elucidation, but it should be noted that a recent report demonstrated that, similarly to what we observed, hDaxx can repress p53 transcriptional activity, and PML is able to counteract this effect (40). Although all the PML splice variants are able to interact with hDaxx, binding to p53 is restricted to a specific isoform, PMLIV (36). Therefore, the pattern of expression of PML isoforms as well as the availability of other proteins to interact with them within the NBs may account for cell-type specific regulation of both hDaxx and p53 functions.
A role can also be postulated for protein modifications. Not only is p53 subjected to a complex series of post-translational modifications, but also hDaxx is phosphorylated on several sites (25) as well as SUMO-1 modified (41). Different kinases have been shown to interact with hDaxx (such as HIPK1 (42), HIPK2 (43), HIPK3 (44), and the ZIP kinase (45)) even though FIG. 6. Endogenous hDaxx protein is required for p53 dependent apoptosis. A, p53-null SaOS-2 cells were transfected with p53 expression vector or with control plasmid (pEGFP to identify transfected cells), together with either hDaxx-specific double-stranded RNA oligonucleotides (sihDaxx) or control scrambled oligonucleotides (siC) as indicated. After 24 h, cells were fixed and subjected to immunofluorescence to identify cells expressing p53. Thereafter, cells were stained with Hoechst dye 33342 to identify apoptotic nuclei. The histogram represents the mean results of four independent experiments, in which p53-positive and control-transfected apoptotic nuclei were scored. Standard deviations are indicated. Aliquots of each sample were analyzed by Western blot to verify the expression levels of endogenous hDaxx and overexpressed p53 proteins (bottom). B, p53-null SaOS-2 cells were transfected with reporter constructs bearing p53-responsive promoters Bax and p21 upstream of the luciferase gene, in either the presence or the absence of wt p53 expression vector, and together with either hDaxx-specific double-stranded RNA oligonucleotides (sihDaxx) or control scrambled oligonucleotides (siC) as indicated. 24 h after transfection, luciferase activity assays were performed. To normalize for transfection efficiency, a fixed amount of pRL-CMV reporter was included in each sample. Graphs represent the mean of at least three independent experiments. Standard deviations are indicated. their exact role in hDaxx phosphorylation is not yet clear. In the case of HIPK1, it has been shown to phosphorylate hDaxx, to relocalize it from NBs to chromatin, and to modulate its transcriptional repressive functions. HIPK2 may as well modulate hDaxx localization and function, because it acts by disgregating NBs and releasing NB-associated factors (43). It is interesting that HIPK2 can also bind to and phosphorylate p53, in that it is involved in the induction of p53-dependent apoptotic response (46,47).
The ability of hDaxx to interact with different cellular factors may depend on its phosphorylation status, as observed in the case of PML, which interacts with hyperphosphorylated hDaxx, or of Pax3, which instead binds to the unphosphorylated form (25). In particular, the C-terminal region of hDaxx, which is responsible for binding to several factors and, as we observed, also to p53 family members, contains a Ser/Thr/Pro rich domain that is a potential target for phosphorylation events. It is interesting that we observed an increased interaction between hDaxx and p53 family members upon phosphatase treatment. 2 Further analysis of the post-translational modification pattern of hDaxx will surely help clarify how its affinity for specific cellular factors and its functions are modulated.
The evidence that hDaxx specifically represses the p21 promoter, but not proapoptotic p53 targets, may also provide an explanation for the observed effect on p53-dependent apoptosis, because several reports demonstrated a protective role of p21 against apoptosis (48 -50). Recently, the ability of c-jun to inhibit p53-dependent induction of p21 has been shown to mediate its death-promoting effect after UV irradiation (51). In this light, hDaxx may be envisioned as a factor that influences the balance of transcription between genes that induce cell cycle arrest or apoptosis. Although further experiments are required to confirm this hypothesis, our results indicated that hDaxx could modulate the threshold of induction of the apoptotic response to p53 and clearly implicate this protein, not only in Fas-and transforming growth factor ␤-induced apoptosis but also in DNA damage-induced apoptosis. As a further indication of the complex role of hDaxx in such processes, expression of mutant p53 in tumor cells has been shown to abrogate the ability of hDaxx to promote Fas-dependent apoptosis (18). Mutant p53 binds to hDaxx and inhibits hDaxx-mediated activation of apoptosis signal-regulating kinase 1/Jun N-terminal kinase pathway. Furthermore, mutant p53 may also downregulate Fas expression, contributing to protection against Fas-induced cell death (52).
It has been demonstrated that mouse embryo fibroblasts lacking p73 and p63 are resistant to p53-dependent apoptosis and this correlates with the inability of p53 to efficiently transactivate proapoptotic targets (31). In light of these findings, the evidence that hDaxx binds to and regulates the activity of all p53 family members allows us to speculate that it might represent an integrating factor to coordinate their physiological response.