The ATRX Syndrome Protein Forms a Chromatin-Remodeling Complex with Daxx and Localizes in Promyelocytic Leukemia Nuclear Bodies

The ATRX syndrome protein forms a chromatin-remodeling complex with Daxx and localizes in promyelocytic leukemia nuclear bodies

Yutong Xue*, Richard Gibbons†, Zhijiang Yan*, Dafeng Yang*, Tarra L. McDowell†, Salvatore Sechi‡§, Jun Qin¶, Sharleen Zhouʈ, Doug Higgs†, and Weidong Wang*,**

*Laboratory of Genetics and ‡Research Resources Branch, National Institute on Aging, National Institutes of Health, 333 Cassell Drive, TRIAD Center Room 4000, Baltimore, MD 21224; †Medical Research Council Molecular Haematology Unit, Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DU, United Kingdom; ¶Departments of Biochemistry and Cell Biology, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030; and ʈHoward Hughes Medical Institute, University of California, Berkeley, CA 94720

Edited by Gerald R. Crabtree, Stanford University School of Medicine, Stanford, CA, and approved July 23, 2003 (received for review December 16, 2002) ATRX syndrome is characterized by X-linked mental retardation result, the etiology of ATRX syndrome is poorly understood. associated with ␣-thalassemia. The gene mutated in this disease, Several lines of evidence hinted that ATRX protein may be part ATRX, encodes a plant homeodomain-like finger and a SWI2͞SNF2- of a chromatin-remodeling complex. First, ATRX protein not like ATPase motif, both of which are often found in chromatin- only has a SWI͞SNF2-type ATPase͞helicase motif but also has remodeling enzymes, but ATRX has not been characterized bio- a plant homeodomain-like zinc finger (10), both of which have chemically. By immunoprecipitation from HeLa extract, we found been found in molecules that modify chromatin structure. Sec- that ATRX is in a complex with transcription cofactor Daxx. The ond, ATRX in nuclear extracts fractionates as a complex be- following evidence supports that ATRX and Daxx are components tween 700 and 2,000 kDa (11), a size similar to that of other of an ATP-dependent chromatin-remodeling complex: (i) Daxx and SWI͞SNF complexes. Third, ATRX localizes at pericentromeric ATRX can be coimmunoisolated by antibodies specific for each heterochromatin (12) and has been identified in yeast two-hybrid protein; (ii) a proportion of Daxx cofractionates with ATRX as a screens by its interaction with the heterochromatin protein HP1 complex of 1 MDa by gel-filtration analysis; (iii) in extract from cells as well as a polycomb group protein EZH2 (12–14). Fourth, of a patient with ATRX syndrome, the level of the Daxx–ATRX ATRX mutations have been correlated with changes in DNA complex is correspondingly reduced; (iv) a proportion of ATRX and methylation patterns at several genomic loci (15). Here, we Daxx colocalize in promyelocytic leukemia nuclear bodies, with demonstrate that ATRX forms a complex with a transcription which Daxx had previously been located; and (v) the ATRX complex cofactor, Daxx, and this complex displays chromatin-remodeling displays ATP-dependent activities that resemble those of other activities. The results provide a step toward understanding the chromatin-remodeling complexes, including triple-helix DNA dis- physiological function of ATRX. placement and alteration of mononucleosome disruption patterns. But unlike the previously described SWI͞SNF or NURD complexes, Materials and Methods the ATRX complex does not randomize DNA phasing of the mono- Cell Culture. HeLa cells were purchased from the National Cell nucleosomes, suggesting that it may remodel chromatin differ- Culture Center (Minneapolis). The cell line derived from an ently. Taken together, the results suggest that ATRX functions in ATRX patient carries an R37X mutation (16). conjunction with Daxx in a novel chromatin-remodeling complex. The defects in ATRX syndrome may result from inappropriate Antibodies (Abs). An affinity-purified polyclonal Ab, FXNP5 expression of genes controlled by this complex. (NP5), and an ATRX mAb (23c) have been described (17). Two

ATRX Abs, D19 and C16, were obtained from Santa Cruz BIOCHEMISTRY SWI͞SNF Biotechnology. Two additional rabbit Abs were raised against fusion proteins also containing maltose-binding protein (New hromatin-remodeling complexes play major roles in regula- England Biolabs) and regions of ATRX (amino acids 365–473 Ction of gene expression in eukaryotes (1, 2). These com- and 2239–2492). Two different Daxx Abs, SC7152 and SC7001, plexes can modify chromatin structure through two distinct were obtained from Santa Cruz Biotechnology. mechanisms. One is through covalent modification, including methylation, phosphorylation, and acetylation. The other is Nuclear Extract Fractionation and Immunoprecipitation. HeLa nu- through noncovalent mechanisms, which include ATP- clear extract was prepared as described (4). ATRX-associated dependent chromatin remodeling. We and others have previ- complex was isolated from HeLa nuclear extract by using an ously purified and characterized several ATP-dependent chro- immunoprecipitation protocol (C. S. Lee, Y.X., X. Zhang, and matin-remodeling complexes of the human SWI͞SNF family W.W., unpublished data). This protocol and methods for mass (3–6). In attempts to characterize additional remodeling com- spectrometry (MS) analysis, have been described in another plexes, we reasoned that because all known ATP-dependent study (18). The Superose 6 gel-filtration analysis has been remodeling complexes contain a subunit with a SWI2͞SNF2-like described (19). After fractionation, the ATRX peak fractions ATPase motif, human proteins containing this motif might were collected and incubated with or without ethidium bro- reveal unique remodeling complexes. A previous study based on mide (100 ␮g͞ml) for 1 h before immunoprecipitation. this strategy indeed identified the NURD complex (7). Here, we characterize a complex containing ATRX, a protein initially identified by mutations in patients with ATRX syndrome (8). This paper was submitted directly (Track II) to the PNAS ofﬁce. Mutations in the ATRX gene are now known to cause several Abbreviation: PML, promyelocytic leukemia. X-linked mental retardation syndromes. The phenotypes include §Present address: National Institute of Diabetes and Digestive and Kidney Diseases, facial dysmorphism, urogenital defects, and ␣-thalassaemia (re- National Institutes of Health, Bethesda, MD 20892-5460. sulting from reduced ␣-globin expression) (9). However, the **To whom correspondence should be addressed. E-mail: [email protected]. ATRX protein has not been biochemically characterized. As a © 2003 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.1937626100 PNAS ͉ September 16, 2003 ͉ vol. 100 ͉ no. 19 ͉ 10635–10640 Downloaded by guest on October 1, 2021 Enzymatic Assays. The mononucleosome disruption assay (20, 21) and ATPase assay (22) followed the same procedure used for human SWI͞SNF complexes. One modification of the ATPase 32 assay is that the Pi is separated from [␥- P]ATP by TLC, and the radioactivity was quantitated by using a PhosphorImager. The triple-helix displacement assay used the same protocol and substrates used for RSC and STH1 (23). The double-helix displacement assay used assay conditions and substrates identical to those described (24).

Immunofluorescence. Immunofluorescence experiments followed a previous procedure (12), except that cells were fixed in 100% methanol and then air-dried before immunostaining. Primary Abs included: mouse monoclonal anti-ATRX (23c), rabbit polyclonal anti-Daxx (Ab-1, Oncogene Research Products), rabbit polyclonal anti-promyelocytic leukemia (PML; kindly provided by Paul Freemont, Imperial College, London). All cells were counterstained with 4Ј,6-diamidino-2-phenylindole (Sigma). Results Identification of an ATRX-Associated Complex. Our initial attempts to purify the ATRX-associated complexes used a combination of conventional and immunoaffinity chromatography, a strategy ͞ Fig. 1. Purification of an ATRX-associated complex from HeLa nuclear that successfully isolated the human SWI SNF and NURD extract. (a) Silver-stained SDS gels showing polypeptides immunoisolated by complexes. However, such approaches obtained ATRX as a three different ATRX Abs (NP5, D19, and C16) from nuclear extract. Mock single polypeptide with no associated proteins [data not shown; immunoprecipitation was done by using either protein A beads alone (lane 2) Yi Zhang’s group has independently obtained the same result or a preimmune (PI) serum (lane 5). MS has been used for identification of (Y. Zhang, personal communication)]. The ATRX-associated ATRX and its associated polypeptides in all three preparations. The number of complex was later found to be unstable in salt solutions of 0.5 M peptides that matched the indicated protein and the percentage of these or higher (see Fig. 4e). Fractionation of nuclear extract often peptides among total peptides obtained from matrix-assisted laser desorption resulted in partial or complete dissociation of the ATRX com- ionization–time-of-flight analysis are shown in parentheses for ATRX and Daxx. The other polypeptides (marked by lines) appeared to either loosely plex. For this reason, we adopted an optimized immunoprecipi- associate with ATRX or associate with ATRX through DNA by subsequent tation method (C. S. Lee, Y.X., X. Zhang, and W.W., unpub- analysis. (b) Immunoblot confirming the presence of Daxx in the polypeptides lished data) to directly isolate ATRX-associated complexes from immunoisolated by ATRX Ab. NE, nuclear extract; SN, supernatant; IP, immu- HeLa nuclear extract. This method has been used successfully noprecipitate. The lower Daxx band in nuclear extract could represent a to isolate complexes involved in Bloom syndrome and Fanconi posttranslationally modified version of Daxx that associates poorly with anemia (18). SDS. (c) Silver-stained gel showing polypeptides immunoisolated by an ATRX Six major polypeptides were immunoisolated by a polyclonal Ab from the ATRX peak fractions after Superose 6 fractionation of nuclear Ab against ATRX (Fig. 1a, lane 3). The 300-kDa polypeptide was extract (Fig. 2c). The presence or absence of ethidium bromide (EtBr) is identified as ATRX by both MS analysis and immunoblotting indicated. Note that p90 and p70 polypeptides are lost in IP in the presence of EtBr, hinting that they may associate with ATRX through DNA. Also, Daxx (Fig. 1b). A minor polypeptide of 180 kDa was also identified as becomes substoichiometric compared to ATRX, which could be caused by ATRX, which likely represents an alternatively spliced product partial dissociation of the ATRX–Daxx complex during Superose 6 fraction- (R.G. and D.H., unpublished data). The 110-kDa polypeptide ation. (d) Immunoblot showing that association between ATRX and Daxx is was identified as Daxx (Fig. 1 a and b), a protein previously not through DNA. shown to interact with multiple transcription factors (25). Daxx and other copurifying polypeptides were further ana- lyzed to determine their possible association with ATRX. All of supernatant after immunoprecipitation by the ATRX Ab (Fig. them can be isolated by four additional ATRX Abs against 1b). The results suggest that a significant proportion of Daxx in different regions of ATRX (Fig. 1a, lanes 6 and 7, and data not the nuclear extract forms a stable complex with ATRX. shown) but not by protein A beads alone (Fig. 1a, lane 2) or a Daxx was originally identified by a yeast two-hybrid screen preimmune serum (Fig. 1a, lane 5), suggesting that they are not designed to detect proteins interacting with the Fas receptor, isolated because of Ab crossreactivity. Immunoprecipitation by which localizes at the cytoplasmic membrane and mediates using ATRX peak fractions from Superose 6 fractionation of apoptosis (27). Immunofluorescence studies later showed that nuclear extract (see Fig. 2c) obtained fewer major polypeptides, the endogenous Daxx is exclusively a nuclear protein (28), and including ATRX, Daxx, p90, and p70 (Fig. 1c). To rule out the its function is likely to modulate transcription (29, 30). Consis- possibility that these polypeptides may associate with ATRX tent with a primary role in transcription, Daxx has been iden- through DNA, immunoprecipitation was performed in the pres- tified in two-hybrid screens for its interaction with multiple ence of ethidium bromide, a DNA-intercalating drug that can transcription factors, DNA methyltransferase I, histone deacety- dissociate proteins from DNA and has often been used to lases, and core histones (30–37). However, endogenous Daxx has identify DNA-independent protein associations (26). The im- not been purified by unbiased biochemical approaches. Conse- munoprecipitation yielded Daxx but not p90 and p70 (Fig. 1 c quently, basic questions regarding Daxx remain unanswered, and d), suggesting that Daxx, but not the latter two proteins, including the number and composition of Daxx complexes that forms a DNA-independent complex with ATRX. exist in a given cell type. To assess independently whether Daxx and ATRX form a ATRX Forms a Complex with Transcriptional Coactivator Daxx. Daxx complex, we used an unbiased approach to isolate Daxx- is present in nearly equimolar amounts with ATRX among associated polypeptides. Immunopurification by using a Daxx Ab polypeptides isolated by ATRX Ab directly from nuclear extract isolated four major polypeptides in near equimolar ratios (Fig. (Fig. 1a). The level of Daxx is also significantly reduced in the 2a). Immunopurification was highly efficient because little Daxx

10636 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.1937626100 Xue et al. Downloaded by guest on October 1, 2021 proportion of Daxx can form a complex independent of ATRX. The results are in agreement with the immunoprecipitation data above: not all Daxx was immunodepleted by the ATRX Ab (Fig. 1d), but almost all ATRX was immunodepleted by the Daxx Ab (Fig. 2b).

The Level of the ATRX–Daxx Complex Is Significantly Decreased in an ATRX Patient Cell Line. If two proteins are components of a complex, one would expect that when one protein is absent, the complex containing the other protein should become smaller when it is fractionated by gel-filtration chromatography. This experimental strategy has previously been used to demonstrate that five SWI͞SNF proteins are parts of one complex (38). Here, we used the same strategy to determine whether the ATRX– Daxx complex becomes smaller in an extract from a cell line derived from a patient with ATRX syndrome that contains ATRX at a very low level (Fig. 2d). So far, all ATRX patient cell lines have a residual level of full-length ATRX, even though nonsense or frameshift mutations were found near the N ter- minus of ATRX gene products. The patient cells apparently have some read-through of premature termination codons (R.G., unpublished data). The level of Daxx in the nuclear extract of the ATRX patient cell line is comparable to that from HeLa cells (Fig. 2d). After Superose 6 fractionation, the expected Daxx peak corresponding the ATRX–Daxx complex (Fig. 2c, fraction 20) is almost gone; only the second, ‘‘non-ATRX’’ peak of Daxx remains (around fraction 25). The residual ATRX in this cell line still peaked in Fig. 2. Daxx and ATRX form a complex, and the level of this complex is fraction 20. These results confirmed that the 1-MDa complex significantly decreased in a ATRX patient cell line. (a) Silver-stained SDS gel showing the polypeptides immunoisolated by a Daxx Ab compared to those by corresponds to that of ATRX–Daxx, whereas the 700-kDa Daxx the ATRX Ab. All polypeptides marked by an asterisk have been identified by complex is independent of ATRX. MS. The number of peptides that match the indicated protein and the percentage of these peptides among total peptides obtained from matrix- ATRX Complex Alters the DNase I Digestion of a Nucleosome in the assisted laser desorption ionization–time-of-flight analysis are shown in pa- Presence of ATP. A mononucleosome disruption assay (7, 20, 21) rentheses for ATRX and Daxx. (b) Immunoblot confirming the presence of was used to determine whether ATRX complex has chromatin- ATRX in the polypeptides isolated by the Daxx Ab. (c) Immunoblot showing remodeling activity. In this assay, a 176-bp fragment of the Superose 6 gel-filtration profile of ATRX and Daxx in nuclear extracts DNA containing a nucleosome positioning sequence from sea prepared from either HeLa cells (Upper) or a cell line derived from an ATRX urchin 5S rRNA gene was assembled into a rotationally phased patient (Lower). The peaks corresponding to proteins with known molecular masses are denoted at the bottom. Note that the peak of the ATRX–Daxx mononucleosome. DNase I digestion of this nucleosome pro- complex (fraction 20, indicated by an arrowhead) is significantly reduced in duces a pattern of distinctive 10-bp ladders on denaturing PAGE the ATRX patient cell line. (d) Immunoblot analysis showing the levels of ATRX (Fig. 3a, lane 2 vs. 1). The human SWI͞SNF complex disrupted (Upper) and Daxx (Lower) in the ATRX patient cell line in comparison to those the 10-bp ladders in the presence of ATP but not with non- in HeLa cells. hydrolyzable ATP-␥-S (Fig. 3a, lanes 5–7). This can be inferred because those nucleotides inaccessible to DNase I in the unre- BIOCHEMISTRY modeled nucleosome now become accessible, suggesting that remained in the supernatant after the purification (Fig. 2b), these nucleotides changed their phasing after chromatin remod- suggesting that most of the endogenous Daxx in the extract was eling. This finding is consistent with a model that the isolated by this approach. The 110-kDa polypeptide was identi- DNA phasing of these nucleosomes has been randomized by fied as Daxx by both MS and immunoblotting analysis (Fig. 2 a SWI͞SNF (39). and b). Similarly, the 300-kDa polypeptide was identified as ATRX complex had no obvious effect on the DNase I ATRX. The findings that ATRX and Daxx can be immunoiso- digestion pattern in the absence of ATP or in the presence of lated in near equimolar amounts by Abs specific for each protein ATP-␥-S (Fig. 3a, lanes 8, 9, and 11), but in the presence of ATP, strongly suggest that they are components of one complex. In the complex altered the digestion pattern (Fig. 3a, lanes 10 and addition, after immunopurification by the Daxx Ab, the amount 12). The region that shows the most alteration is between of ATRX in the supernatant was almost completely depleted, nucleotides 10 and 30 of the 5S DNA where nucleotides facing suggesting that most, if not all, ATRX is in a complex with Daxx inside toward the nucleosome showed an increased level of in HeLa nuclear extracts (Fig. 2b). digestion. This region is near the entry site of the nucleosome, because nucleotides before 10 are accessible to DNase I diges- ATRX and Daxx Cofractionate as a 1-MDa Complex. ATRX in HeLa tion. For regions beyond the first 30 nucleotides of 5S DNA, only nuclear extract fractionated with a single peak in Superose 6 certain nucleotides facing out from the nucleosome exhibited gel-filtration analysis, corresponding to a complex of 1 MDa enhanced digestion, whereas nucleotides facing in show no (Fig. 2c). This result is consistent with a previous report that enhancement. The results suggest that the ATRX complex ATRX fractionates in a complex of 0.7–2 MDa (11). Daxx mainly disrupts the DNA–histone interaction at the entry site of fractionated in two peaks: one of them overlaps with that of the nucleosome, and disruption does not alter nucleosome ATRX (at fraction 20) and could correspond to the ATRX–Daxx phasing. This is different from remodeling by SWI͞SNF (com- complex. The other one is Ϸ700 kDa (fraction 25) and should pare Fig. 3a, lanes 10 and 7). represent a complex without ATRX. Thus, most ATX in the The ATRX complex was found to have an ATPase activity extract is present in the ATRX–Daxx complex, but a significant that can be stimulated by DNA or nucleosomes (Fig. 3b), a

Xue et al. PNAS ͉ September 16, 2003 ͉ vol. 100 ͉ no. 19 ͉ 10637 Downloaded by guest on October 1, 2021 property that resembles SWI͞SNF. Different preparations of ATRX complex with ATPase activity 1- to 2-fold that of SWI͞SNF have been used for the mononucleosome disruption assay, with similar results obtained.

The ATRX–Daxx Complex Has an ATP-Dependent Triple-Helix Displace- ment Activity. The yeast RSC and SWI͞SNF chromatin- remodeling complexes, as well as the ATPase subunit of RSC, STH1p, have been shown to be ATP-dependent DNA translo- cases (23, 40). One way to show translocation is the triple-helix displacement assay. In this assay, a triple helix (H-DNA) is formed in which each nucleotide of the third strand forms Hoogsteen base pairs with Watson–Crick base pairs of the duplex (41). When the translocase proceeds through the triple helix, the third strand is displaced. Complexes isolated by different ATRX and Daxx Abs all displayed triple-helix displacement activity (Fig. 4a), suggesting that the ATRX–Daxx complex has a translocase property like yeast RSC and SWI͞SNF. One possible mechanism for the ATRX–Daxx complex to displace triple helix is through direct binding and dissociation. However, complexes isolated by either ATRX or Daxx do not displace the third strand in a blunt triplex (Fig. 4c). This result is consistent with the idea that the ATRX–Daxx complex needs to bind the flanking duplex DNA and translocate into the triple helix to displace the third strand. We also investigated whether ATRX–Daxx may function as a DNA helicase. Using a standard duplex DNA displacement assay, complexes isolated by either ATRX or Daxx Ab showed no detectable activity (Fig. 4d), consistent with that ATRX– Daxx is functioning as a chromatin-remodeling complex but not a helicase.

Daxx Is Dispensible for the Triple-Helix Displacement Activity of ATRX Complex. We examined whether Daxx is required for the triple- helix displacement activity of the ATRX–Daxx complex. Daxx was completely washed away from ATRX by using buffer containing salt of 0.5 M or higher (Fig. 4e; silver-staining gel not shown). The remaining ATRX protein displays triplex activity indistinguishable from that of the ATRX–Daxx complex (Fig. 4f), suggesting that Daxx is probably not needed for the remodeling activity of ATRX.

ATRX and Daxx Colocalize in PML Nuclear Bodies. ATRX is predom- inantly associated with heterochromatin both in interphase and metaphase (12). However, a significant proportion of ATRX is present in nuclear speckles in human cells (17). The distribution of ATRX between heterochromatin and nuclear speckles varies from one cell type to another. Consistent with previous obser- vations (12, 17), ATRX was found to localize in the nuclei of interphase human cells with a typical speckled pattern (Fig. 5). Daxx was found to display a similar speckled pattern in the nuclei, which is also in agreement with prior studies (28–30, 32, 33). Importantly, the majority of ATRX signals colocalize with those of Daxx, providing a further indication that these two proteins may work together in vivo. Daxx has been shown to associate with the PML gene product, PML, and localize in PML nuclear bodies (29, 30, 32, 33). PML

Fig. 3. The ATRX complex alters the DNase I digestion pattern of a nucleosome in the presence of ATP. (a) Autoradiograph showing the results of the positions of the nucleotides in 5S DNA. The 10-bp ladders can be observed mononucleosome disruption assay. Complexes isolated by protein A alone between nucleotides 10 and 130 of 5S DNA. (b) Graphic presentation showing (PnA) or Abs against ATRX and human SWI͞SNF (hSWI) are indicated at the that the ATRX complex has DNA- or nucleosome-stimulated ATPase activity top. The templates used and the presence of ATP (A) or ATP-␥-S (␥) are (shown in percentage of released inorganic phosphate from total ATP multi- indicated. The amounts of DNase used, 1 and 2, represent 0.2 and 0.4 units, plied by 100). The ATPase activity of mock immunoprecipitation (using pre- respectively. The solid arrows mark the nucleotides whose digestion was immune serum) is indistinguishable from that of the background (using enhanced by the ATRX–Daxx complex in the presence of ATP. The solid dots protein A beads), which was subtracted during calculation. The presence of denote the nucleotides between the 10-bp ladders whose digestion was nucleosomes (NЈsome), single-stranded DNA (ssDNA), and double-stranded stimulated by human SWI͞SNF but not ATRX–Daxx. The solid lines mark the DNA (dsDNA) is indicated.

10638 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.1937626100 Xue et al. Downloaded by guest on October 1, 2021 Fig. 5. ATRX and Daxx colocalize in PML nuclear bodies. (a) Immunofluo- rescence images showing colocalization of ATRX and Daxx in human fibro- blast cell lines. Images for each protein and the merged image are shown. (b) Immunofluorescence images showing colocalization of ATRX with PML in human fibroblasts. The nucleus was costained with 4Ј,6-diamidino-2- phenylindole (DAPI).

support the notion that ATRX functions as part of a chromatin- remodeling complex and may thereby participate in transcriptional regulation like other remodeling complexes. In addition, we found that the ATRX–Daxx complex lacks any detectable Fig. 4. The ATRX complex has an ATP-dependent triple-helix displacement DNA helicase activity, distinguishing it from proteins involved in activity. (a and b) Autoradiographs showing that the ATRX–Daxx complex DNA damage repair and genome instability diseases (e.g., displaces a triple helix in the presence of ATP. The positions of the labeled Bloom syndrome and Werner syndrome). This finding is con- triple-helix substrate and the displaced third strand in the gel are depicted at sistent with previously published clinical data that ATRX pa- the left. The complexes isolated by different Abs to ATRX, Daxx, and human SWI͞SNF are shown at the top. Polypeptides isolated by mock immunopre- tients have no increased incidence of chromosome abnormalities cipitation by using either protein A beads (PnA) or preimmune serum (mock or increased propensity to develop malignancy. IP) are also indicated. Human SWI͞SNF in our hands does not show signiﬁcant Interestingly, the ATRX–Daxx complex does not randomize activity in this assay. To serve as a positive control, recombinant STH1 (a DNA phasing of a nucleosome, which is in contrast to remod- generous gift of B. Cairns, Huntsman Cancer Institute, University of Utah eling by SWI͞SNF or NURD complexes. Its disruption of the School of Medicine, Salt Lake City) was used. (c) Autoradiograph showing that DNase I digestion pattern of a mononucleosome occurs mainly the ATRX–Daxx complex does not displace a blunt triple helix. The substrates near the entry site of the nucleosome, which is also different used are indicated at the bottom. The substrates used in the blunt triplex assay from remodeling by SWI͞SNF or NURD. The ATPase motif of contain two bands. The major band (with the faster mobility) represents the blunt triplex. The minor band (with slower mobility) represents a triplex ATRX is most similar to that in Rad54 (42). Rad54 displays an substrate with one blunt end and one overhang end of 17-bp double-stranded ATP-dependent chromatin-remodeling activity in the presence DNA. This overhang was designed to help annealing of the double-stranded of Rad51 (43). A recent study has shown that Rad54 has DNA and was supposed to be removed by restriction digestion, but it was not remodeling properties similar to ATRX, including triplex dis-

completely removed in the current experiment. (d) Autoradiograph showing placement activity and failure to randomize DNA phasing of BIOCHEMISTRY the results of a double-helix displacement assay. The duplex substrate and the nucleosomes (40). Speculatively, ATRX–Daxx and Rad54 may displaced oligo are illustrated at the left. The BLM (Bloom Syndrome gene be members of a group of complexes that remodel chromatin product) DNA helicase complex was used as a positive control (18). (e) Immu- differently from the SWI͞SNF and NURD families. noblot showing that Daxx was completely removed by washing with buffer containing 0.5 or 0.75 M salt, as shown at the top. (f) Autoradiograph showing How may ATRX and Daxx work together in a complex? that dissociation of Daxx does not affect triplex unwinding activity of ATRX. Chromatin-remodeling complexes are often recruited to their targets by sequence-specific transcription factors that bind to these promoters (44). Several lines of evidence hint that Daxx frequently is involved in chromosomal translocations in acute may be a targeting subunit for ATRX complex. First, Daxx PML. In normal cells, PML is concentrated within 10–20 nuclear contains two regions predicted to form paired amphipathic structures known as PML bodies, which are believed to play roles helices that resemble those of Sin3 (31). A paired amphipathic in transcriptional activation, DNA replication, apoptosis, and helix motif in Sin3 targets the Sin3 complex through interactions viral infection. The colocalization of ATRX and Daxx raised the with a sequence-specific transcription factor (45). The paired possibility that ATRX also localizes in PML bodies. Costaining amphipathic helix regions of Daxx may play a targeting role experiments showed that the majority of ATRX foci colocalize analogous to that of Sin3. Second, by yeast two-hybrid screens, with those of PML (Fig. 5c), suggesting that the ATRX–Daxx Daxx interacts with several sequence-specific transcription fac- complex may play a role in the function of PML bodies. tors (31, 34, 35). Third, in ongoing work, we found that the two other major Daxx-associated polypeptides, Daxx-associated Discussion polypeptide-50 and -60 (Fig. 2a), are both sequence-specific In this study, we show that ATRX, the protein involved in ATRX transcription factors (unpublished data). Our data are therefore syndrome, constitutes a novel complex with Daxx. This complex consistent with a model that the ATRX–Daxx complex partic- has several activities consistent with it being a chromatin- ipates in chromatin remodeling for genes that are controlled by remodeling complex. They include ATP-dependent activity that Daxx-interacting sequence-specific transcription factors. The alters the DNase I digestion pattern of a nucleosome and molecular defects in ATRX patients may result from inappro- ATP-dependent triple-helix displacement activity. These data priate regulation of these target genes.

Xue et al. PNAS ͉ September 16, 2003 ͉ vol. 100 ͉ no. 19 ͉ 10639 Downloaded by guest on October 1, 2021 We thank Drs. B. Cairns, C. Peterson, A. Saha, J. Cote, T. Owen-Hughes, lished results. We thank the National Cell Culture Center for providing and C. Wu for reagents and advice on different assays; Drs. H. Lu, D. cells. We also thank Dr. D. Schlessinger for critical reading of the Reinberg, M. Baron, and S. Lippard for Abs; Dr. N. Sherman for MS manuscript. W.W. is a scholar from The Ellison Medical Foundation and analysis; and Drs. C. Peterson and Y. Zhang for communicating unpub- has received a grant from the Rett Syndrome Foundation.

1. Krebs, J. E. & Peterson, C. L. (2000) Crit. Rev. Eukaryotic Gene Expression 10, 23. Saha, A., Wittmeyer, J. & Cairns, B. R. (2002) Genes Dev. 16, 2120–2134. 1–12. 24. Shen, J. C., Gray, M. D., Oshima, J., Kamath-Loeb, A. S., Fry, M. & Loeb, L. A. 2. Neely, K. & Workman, J. (2002) Biochim. Biophys. Acta 1603, 19–29. (1998) J. Biol. Chem. 273, 34139–34144. 3. Kwon, H., Imbalzano, A. N., Khavari, P. A., Kingston, R. E. & Green, M. R. 25. Michaelson, J. S. (2000) Apoptosis 5, 217–220. (1994) Nature 370, 477–481. 26. Lai, J. S. & Herr, W. (1992) Proc. Natl. Acad. Sci. USA 89, 6958–6692. 4. Wang, W., Cote, J., Xue, Y., Zhou, S., Khavari, P. A., Biggar, S. R., Muchardt, 27. Yang, X., Khosravi-Far, R., Chang, H. Y. & Baltimore, D. (1997) Cell 89, C., Kalpana, G. V., Goff, S. P., Yaniv, M., et al. (1996) EMBO J. 15, 5370–5382. 1067–1076. 5. Xue, Y., Canman, J. C., Lee, C. S., Nie, Z., Yang, D., Moreno, G. T., Young, 28. Pluta, A. F., Earnshaw, W. C. & Goldberg, I. G. (1998) J. Cell Sci. 111, 97, M. K., Salmon, E. D. & Wang, W. (2000) Proc. Natl. Acad. Sci. USA 2029–2041. 13015–13020. 29. Torii, S., Egan, D. A., Evans, R. A. & Reed, J. C. (1999) EMBO J. 18, 6. Lemon, B., Inouye, C., King, D. S. & Tjian, R. (2001) Nature 414, 924–938. 6037–6049. 7. Xue, Y., Wong, J., Moreno, G. T., Young, M. K., Cote, J. & Wang, W. (1998) 30. Zhong, S., Salomoni, P., Ronchetti, S., Guo, A., Ruggero, D. & Pandolfi, P. P. Mol. Cell 2, 851–861. J. Exp. Med. 191, 8. Gibbons, R. J., Picketts, D. J., Villard, L. & Higgs, D. R. (1995) Cell 80, (2000) 631–640. 837–845. 31. Hollenbach, A. D., Sublett, J. E., McPherson, C. J. & Grosveld, G. (1999) 9. Gibbons, R. J. & Higgs, D. R. (2000) Am. J. Med. Genet. 97, 204–212. EMBO J. 18, 3702–3711. 10. Gibbons, R. J., Bachoo, S., Picketts, D. J., Aftimos, S., Asenbauer, B., 32. Ishov, A. M., Sotnikov, A. G., Negorev, D., Vladimirova, O. V., Neff, N., Bergoffen, J., Berry, S. A., Dahl, N., Fryer, A., Keppler, K., et al. (1997) Nat. Kamitani, T., Yeh, E. T., Strauss, J. F., III, & Maul, G. G. (1999) J. Cell Biol. Genet. 17, 146–148. 147, 221–234. 11. Berube, N. G., Jagla, M., Smeenk, C., De Repentigny, Y., Kothary, R. & 33. Li, H., Leo, C., Zhu, J., Wu, X., O’Neil, J., Park, E. J. & Chen, J. D. (2000) Mol. Picketts, D. J. (2002) Hum. Mol. Genet. 11, 253–261. Cell. Biol. 20, 1784–1796. 12. McDowell, T. L., Gibbons, R. J., Sutherland, H., O’Rourke, D. M., Bickmore, 34. Li, R., Pei, H., Watson, D. K. & Papas, T. S. (2000) Oncogene 19, 745–753. W. A., Pombo, A., Turley, H., Gatter, K., Picketts, D. J., Buckle, V. J., et al. 35. Emelyanov, A. V., Kovac, C. R., Sepulveda, M. A. & Birshtein, B. K. (2002) (1999) Proc. Natl. Acad. Sci. USA 96, 13983–13988. J. Biol. Chem. 277, 11156–11164. 13. Cardoso, C., Timsit, S., Villard, L., Khrestchatisky, M., Fontes, M. & Colleaux, 36. Michaelson, J. S., Bader, D., Kuo, F., Kozak, C. & Leder, P. (1999) Genes Dev. L. (1998) Hum. Mol. Genet. 7, 679–684. 13, 1918–1923. 14. Le Douarin, B., Nielsen, A. L., Garnier, J. M., Ichinose, H., Jeanmougin, F., 37. Hollenbach, A. D., McPherson, C. J., Mientjes, E. J., Iyengar, R. & Grosveld, Losson, R. & Chambon, P. (1996) EMBO J. 15, 6701–6715. G. (2002) J. Cell Sci. 115, 3319–3330. 15. Gibbons, R. J., McDowell, T. L., Raman, S., O’Rourke, D. M., Garrick, D., 38. Peterson, C. L., Dingwall, A. & Scott, M. P. (1994) Proc. Natl. Acad. Sci. USA Ayyub, H. & Higgs, D. R. (2000) Nat. Genet. 24, 368–371. 91, 2905–2908. 16. Guerrini, R., Shanahan, J. L., Carrozzo, R., Bonanni, P., Higgs, D. R. & 39. Cote, J., Peterson, C. L. & Workman, J. L. (1998) Proc. Natl. Acad. Sci. USA Gibbons, R. J. (2000) Ann. Neurol. 47, 117–121. 95, 4947–4952. 17. Berube, N. G., Smeenk, C. A. & Picketts, D. J. (2000) Hum. Mol. Genet. 9, 40. Jaskelioff, M., Van Komen, S., Krebs, J. E., Sung, P. & Peterson, C. L. (2003) 539–547. J. Biol. Chem. 278, 9212–9218. 18. Meetei, A. R., Sechi, S., Wallisch, M., Yang, D., Young, M. K., Joenje, H., 41. van Dongen, M. J., Doreleijers, J. F., van der Marel, G. A., van Boom, J. H., Hoatlin, M. E. & Wang, W. (2003) Mol. Cell. Biol. 23, 3417–3426. 19. Wang, W., Xue, Y., Zhou, S., Kuo, A., Cairns, B. R. & Crabtree, G. R. (1996) Hilbers, C. W. & Wijmenga, S. S. (1999) Nat. Struct. Biol. 6, 854–859. Genes Dev. 10, 2117–2130. 42. Picketts, D. J., Tastan, A. O., Higgs, D. R. & Gibbons, R. J. (1998) Mamm. 20. Owen-Hughes, T., Utley, R. T., Cote, J., Peterson, C. L. & Workman, J. L. Genome 9, 400–403. (1996) Science 273, 513–516. 43. Alexiadis, V. & Kadonaga, J. T. (2002) Genes Dev. 16, 2767–2771. 21. Wang, W., Chi, T., Xue, Y., Zhou, S., Kuo, A. & Crabtree, G. R. (1998) Proc. 44. Hassan, A. H., Neely, K. E., Vignali, M., Reese, J. C. & Workman, J. L. (2001) Natl. Acad. Sci. USA 95, 492–498. Front. Biosci. 6, D1054–D1064. 22. Zhao, K., Wang, W., Rando, O. J., Xue, Y., Swiderek, K., Kuo, A. & Crabtree, 45. Brubaker, K., Cowley, S. M., Huang, K., Loo, L., Yochum, G. S., Ayer, D. E., G. R. (1998) Cell 95, 625–636. Eisenman, R. N. & Radhakrishnan, I. (2000) Cell 103, 655–665.

10640 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.1937626100 Xue et al. Downloaded by guest on October 1, 2021