Structural Characterization of the DAXX N-Terminal Helical Bundle Domain and Its Complex with Rassf1c

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Structure Article

Structural Characterization of the DAXX N-Terminal Helical Bundle Domain and Its Complex with Rassf1C

Eric Escobar-Cabrera,1,2,3 Desmond K.W. Lau,1,2,3 Serena Giovinazzi,4 Alexander M. Ishov,4,5 and Lawrence P. McIntosh1,2,3,* 1Department of Biochemistry and Molecular Biology 2Department of Chemistry 3Michael Smith Laboratories, University of British Columbia, Vancouver, BC V6T 1Z3, Canada 4Department of Anatomy and Cell Biology 5Shands Cancer Center University of Florida, 1376 Mowry Road, Gainesville, FL 32610, USA *Correspondence: [email protected] DOI 10.1016/j.str.2010.09.016

SUMMARY DAXX localization into PML-NBs has been confirmed in a number of reports (Lindsay et al., 2008). The ability of DAXX DAXX is a scaffold protein with diverse roles to bind sumoylated PML via one of its two SUMO interaction including transcription and cell cycle regulation. motifs (SIMs) establishes its subnuclear localization (Lin et al., Using NMR spectroscopy, we demonstrate that the 2006; Santiago et al., 2009). Moreover, DAXX can modulate C-terminal half of DAXX is intrinsically disordered, the function of other proteins by sequestering them into whereas a folded domain is present near its N PML-NBs. For example, Kitagawa et al. (2006) proposed that terminus. This domain forms a left-handed four-helix DAXX recruits the Ras-association domain family 1C (Rassf1C) into PML-NBs and only releases it when DAXX is degraded bundle (H1, H2, H4, H5). However, due to a crossover upon DNA damage. helix (H3), this topology differs from that of the Sin3 The functions of DAXX in the nucleus are diverse and often PAH domain, which to date has been used as a model controversial. Best characterized is that of a transcription core- for DAXX. The N-terminal residues of the tumor pressor (for instance, see Morozov et al., 2008). This role suppressor Rassf1C fold into an amphipathic a helix appears to depends on the ability of DAXX to bind sumoylated upon binding this DAXX domain via a shallow cleft transcription factors (Chang et al., 2005; Shih et al., 2007), as along the flexible helices H2 and H5 (KD 60 mM). well as histone deacetylases (Hollenbach et al., 2002) and DNA Based on a proposed DAXX recognition motif as methyltransferases (Puto and Reed, 2008). A second function hydrophobic residues preceded by negatively of DAXX lies with regulating Mdm2 and hence p53-mediated charged groups, we found that peptide models of apoptosis. In one scenario, DAXX prevents Mdm2 self-ubiquiti- p53 and Mdm2 also bound the helical bundle. These nation, enabling this E3 ligase to target p53 for proteolytic degradation. However, upon DNA damage, DAXX dissociates from data provide a structural foundation for under- Mdm2 by an unknown mechanism, and the subsequent loss of standing the diverse functions of DAXX. Mdm2 leads to p53-activated apoptosis (Song et al., 2008; Tang et al., 2006). Most recently, DAXX has also been identified as a specific histone 3.3 chaperone (Drane et al., 2010; Goldberg INTRODUCTION et al., 2010). Compounding the enigmatic functions of DAXX, this essential DAXX was first discovered and characterized in 1997 as a Fas protein has been poorly characterized at the structural level. death-domain associating protein. This association was thought Based on early sequence analyses, it was proposed that resi- to initiate FADD-independent activation of the Jun N-terminal dues 180–212 and 356–388 form coiled-coils (Pluta et al., kinase pathway, which ultimately led to apoptosis (Yang et al., 1998), whereas residues 64–108 and 192–240 fold as two ‘‘pair 1997). However, conflicting reports on the role of DAXX ap- of amphipathic helices’’ (PAH) domains (Hollenbach et al., peared shortly after its discovery (Salomoni and Khelifi, 2006). 1999). These ill-defined boundaries have been used to design In particular, mouse embryo fibroblasts deficient in FADD or cas- deletion constructs and coarsely map direct or indirect interact- pase-8 were unable to produce apoptosis in response to Fas ing regions of DAXX with over 50 putative partner proteins (Lind- activation, suggesting that DAXX did not dictate a parallel say et al., 2008). With the aim of better defining its modular signaling pathway (Juo et al., 1998; Zhang et al., 1998). Further- architecture, we undertook an NMR spectroscopic characteriza- more, Torii et al. failed to detect a direct DAXX-Fas interaction tion of DAXX. Most importantly, we show that the C-terminal half and instead found that DAXX is primarily located in the nucleus, of DAXX is intrinsically disordered, whereas its N-terminal region often in PML-nuclear bodies (PML-NBs) (Torii et al., 1999; Ishov contains a well-folded helical bundle that is distinct in structure et al., 1999). and binding mechanism from the PAH domains of Sin3.

SIM SIM DAXX DHB helical acidic SPE SPT 1 17 55 144 180 400 434 485 495596 665724 740

Sequence conservation

Disorder consensus

SS consensus

100 200 300 400 500 600 700 Residue

Figure 1. Sequence-Based Predicted Modular Organization of DAXX DAXX has six regions of sequence conservation. The ﬁrst and last correspond to SUMO-interaction motifs (SIMs). A consensus of secondary structure prediction algorithms (SS: a helices black, b strands gray) suggested the presence of several clustered helical segments, including the DHB domain, characterized herein. Additional helical domains may occur within the third conserved region. In contrast, disorder prediction algorithms indicated that the C-terminal half of DAXX is intrinsically unstructured. This includes the ‘‘acidic’’ region, which contains 80% Glu/Asp residues, and segments rich in Ser/Pro/Glu residues (SPE) and in Ser/Pro/Thr (SPT) residues. See Figure S1 for details.

Additionally, the N terminus of Rassf1C forms a short amphi- respectively. The C-terminal fragment mDAXX566-739 had pathic a helix upon binding this DAXX helix bundle (DHB) random coil 1HN shifts, indicating that indeed, as predicted, it domain. The structure of the DAXX/Rassf1C complex also is predominantly disordered. allowed us to identify similar DHB interacting motifs in Mdm2 and p53. Based on these results, we propose a revised model Refinement of Domain Boundaries for the modular organization of DAXX. To refine the boundaries of the structured domain within DAXX46-160, the NMR signals from the main chain nuclei in this RESULTS fragment were assigned. Based on 13Ca chemical shift patterns and heteronuclear 15N-NOE measurements, residues 60 DAXX Sequence Analysis through 136 were clearly ordered, folding as at least four To identify possible folded domains in DAXX, the protein a helices (Figures S2C and S2D). This agreed remarkably well sequence was analyzed for conservation and predicted with the results of secondary structure prediction algorithms. secondary structure or disorder. As summarized in Figure 1 Four additional constructs were tested, and of these, and Figure S1 (available online), the N-terminal 1/2 of DAXX DAXX55-144 showed the best soluble expression and NMR spec- is expected to contain many a helices that cluster into three tral quality (Figure S2E). A superimposition of 15N-HSQC spectra regions of high sequence conservation. In contrast, the confirmed that this optimized fragment indeed retained the C-terminal 1/2 of the protein, which is abundant in polar Ser/ structured elements present in the initially indentified DAXX46-160 Thr/Pro/Glu residues, is likely to be predominantly ‘‘intrinsically (Figure S2F). The fully assigned spectrum of DAXX55-144 is disordered.’’ Thus, secondary structure prediction algorithms presented in Figure S2G. and sequence conservation analyses suggested the presence of folded domains within the N-terminal portion of DAXX. DAXX55-144 Tertiary Structure: The DHB Domain The three-dimensional structure of DAXX55-144 was determined Characterization of DAXX Deletion Fragments using extensive NMR-derived distance, dihedral angle, and With the aim of using NMR spectroscopy to characterize poten- orientational restraints (Figure 2 and Table 1). A final ensemble tial structural domains of DAXX, an extensive set of gene of 25 water-refined structures was obtained with an rmsd value constructs encoding various truncation fragments correspond- of 0.22 ± 0.04 A˚ for backbone atoms, excluding the flexible ing to regions of high sequence conservation were generated. termini. DAXX55-144 folds as a left-handed four-helix bundle The constructs were tested first for protein expression in Escher- (H1, residues 60–77; H2, 84–93; H4, 103–118; H5, 123–136) ichia coli. In the cases where protein of the expected size was with a fifth small crossover helix (H3, 97–100). Due to its helical produced, cells were fractionated to differentiate soluble DAXX content, we denote this structure as the DAXX Helix Bundle fragments from those forming insoluble inclusion bodies. Of 23 (DHB) domain. A comparison of the DHB domain architecture constructs tested, 4 expressed soluble proteins (Figure S2A): to that of other proteins is presented below. human DAXX46-160, DAXX161-243, and DAXX347-420, and murine An examination of the DAXX55-144 structural surface revealed mDAXX566-739. These truncation fragments were isotopically a number of physicochemical properties that provided clues to labeled and screened by NMR spectroscopy (Figure S2B). By the function of the DHB domain (Figure 2C). Overall, the molecule far, the best construct was DAXX46-160, yielding a 15N-HSQC is predominantly positively charged, as expected from its theo- spectrum with well-dispersed, sharp signals indicative of a struc- retical pI value of 9.45. However, several positively charged tured, monomeric domain. In contrast, DAXX161-243 and (Arg91, Arg94, Lys135, His137, Lys140, Lys141, and Lys142) DAXX347-420 appeared partially aggregated or mostly unfolded, and negatively charged (Glu62, Glu64, Glu68, Glu69, and

AB Figure 2. The DHB Domain Is a Four-Helix Bundle with a Short Crossover Helix (A) Superimposed backbone trace of the 25 member ensemble. (B and C) The lowest-energy NMR-derived structure of the DHB domain in DAXX55-144 is shown in (B) ribbon diagram and (C) surface representa- H1 tion formats. In the latter, residues are colored as negatively charged (red), positively charged (blue), hydrophobic (green), and neutral polar (gray). H5 H2 H4 See also Figure S2. N144

H3 Figure S3). This order is reflected by the G55 low RMS deviations of both the helices and loops in the structural ensemble of 900 900 DAXX55-144 (Figure 2A). Amide HX rates were also measured to C probe the local stability of DAXX55-144 (Figure 3B). The exchange rate of A121 Y124 D80 L73 H81 a specific amide in a protein is dependent P82 P120 R119 A79 R117 R119 upon its structural environment, as well V84 S118 as local and global fluctuations leading V125 E83 M76 K75 K122 to hydrogen bond breakage and water K122 P86 L113 E72 N128 0 S114 contact. Under EX2 conditions, HX I127 180 E69 L66 R115 protection factors can be interpreted as F87 Y89 R115 E129 C131 E68 S110 the inverse of an equilibrium constant K65 E62 R91 for fluctuations between a closed nonex- R111 Q93 R111 T132 R94 L61 N107 changeable state and a transiently open I108 K60 A103 exchange-competent state (Englander E104 N144 Y59 and Kallenbach, 1983). Based on the A136 L143 K56 H137 C58 K57 largest protection factor measured K142 3 5 K135 L143 K140 (Leu71, 1.1 10 ), the minimum estimate for the global unfolding free energy L98 G55 change of DAXX55-144 under benign conditions is 6.8 kcal/mol. More interestingly, the measured exchange life-

Glu72) residues cluster on opposite sides of the molecule, giving times (1/kex) of the DHB domain reveal a striking difference in the DAXX DHB domain an asymmetric electrostatic surface. In the protection of each helix. Residues in helix 1 and helix 4 had addition, a patch of exposed hydrophobic residues (Pro86, the largest protection factors, residues in helix 3 and helix 5 Phe87, Val84, Tyr124, Val125, and Ile127) flanked by positively had intermediate protection, and residues in helix 2 underwent charged groups is present on the surface of DAXX55-144.As facile exchange. Thus, in contrast to their relatively uniform fast shown below, these features are key to the function of the timescale order, the helices of the DHB domain differ signifi- DHB domain as a protein-protein interaction module. cantly in their local stability against fluctuations leading to HX.

Dynamic Properties of DAXX55-144 Residues 28-38 of Rassf1C Bind the DHB Domain In parallel with structure calculations, the dynamic properties of Kitagawa et al. (2006) recently identified Rassf1C as a binding DAXX55-144 were examined by NMR relaxation and amide partner of DAXX, and found that the interaction mediated the hydrogen exchange (HX) approaches (Figure 3). Based on cellular localization of Rassf1C. The regions responsible for the 15 a model-free analysis of NT1,T2, and heteronuclear NOE interaction were mapped coarsely to residues 1–220 of Rassf1C data (Dosset et al., 2000), the protein fragment has a correlation and 1–260 of DAXX. Using a yeast two-hybrid screen, we also time tc of 7.5 ± 0.1 nsec for global tumbling (Figure S3). This identified Rassf1C as a DAXX interactor (Lindsay, 2008). With value is within the range expected for a monomeric globular an in vitro binding assay, we further refined the minimal interact- protein with the molecular mass of DAXX55-144 (Bernado et al., ing regions to residues 1–50 of Rassf1C and 1–142 of DAXX 2002). Inspection of the site-specific relaxation properties of (S.G., A.M.I., unpublished data). We now know that the latter DAXX55-144 demonstrates that the terminal residues 55–58 and contains the structured DHB domain. 141–144 are flexible, whereas the backbone of the DHB domain To better map the interface between DAXX and Rassf1C, we is uniformly well ordered on the nsec-psec timescale (Figure 3A; undertook NMR-monitored titrations of DAXX55-144 and

1-50 Table 1. NMR Restraints and Statistics for apo DAXX55-144 and Rassf1C . The NMR spectra of bacterially expressed 1-50 the DAXX55-144/Rassf1C23-38w Complex Structural Ensembles Rassf1C revealed that the latter is predominantly disordered Summary of Restraints Apo Complex in isolation (Figure S4A). Although at low ionic strength the two species aggregated when mixed, residues 28–38 of 15N-labeled NOEsa Rassf1C exhibited the largest spectral perturbations upon addi- Intraresidue 1025 (137) 838 (115) tion of substoichiometric amounts of unlabeled DAXX55-144 (not Sequential 562 (151) 384 (95) shown). A titration with the shorter construct, 15N-labeled Medium range (i–j = 2–4) 664 (289) 246 (164) Rassf1C8-38 at higher ionic strength (100 mM KCl, 10 mM potas- Long range (i–j R 5) 454 (301) 216 (123) sium phosphate) conﬁrmed that residues 31–36 of Rassf1C were

Intramolecular 35 (70) most affected upon binding (Figure S4B). The KD for this interac- m Total 2705 (878) 1719 (567) tion was 210 M based on least-squares fitting of fast- Dihedral angles exchange chemical shift changes for six Rassf1C amides as a function of added DAXX55-144 (Table S1). Truncating further, c, f 69, 69 81, 81 reciprocal titrations confirmed that Rassf1C23-38w and 1HN-15N Residual dipolar 87 DAXX55-144 bound with K 60 mM in low ionic strength buffer couplings (RDC) D (Figures 4A and 4B and Table 2; Table S1). Rassf1C23-38w Deviation from restraints contains a nonnative C-terminal tryptophan (w) added to facilitate ˚ NOE(A) 0.05 ± 0.00 0.05 ± 0.00 quantification. Dihedral restraints 0.57 ± 0.15 0.65 ± 0.11 (degrees) Coil-to-Helix Transition of Rassf1C23-38w upon DHB RDC (Hz) 0.88 ± 0.02 Domain Binding Deviation from idealized geometry Isolated Rassf1C23-38w is predominantly unstructured as evi- 15 Bonds (A˚ ) 0.007 ± 0.000 0.005 ± 0.000 denced by its random coil chemical shifts and N relaxation 55-144 Angles (degrees) 0.71 ± 0.02 0.63 ± 0.02 properties (Figure 4C). In contrast, upon binding DAXX , residues 28–38 show both increased heteronuclear 15N{1H} Improper angles (degrees) 1.54 ± 0.06 1.4 ± 0.06 NOE values and upfield/downfield perturbations of 1Ha/13Ca Residues in generously 99.3% 99.5% chemical shifts (Figures 4C–4E). This behavior is diagnostic of allowed regions of the Ramachandran plot a coil-to-helix transition induced upon complex formation. In contrast, upon binding unlabeled Rassf1C1-50 and Mean energiesb (kcal mol-1) Rassf1C23-38w, only a limited number of residues in 15N-labeled À À Evdw 165 ± 20 306 ± 22 DAXX55-144 displayed chemical shift perturbations (Figure 4F). Ebonds 70 ± 5 53 ± 3 This indicated that the secondary and tertiary structure of the Eangles 220 ± 9 202 ± 11 protein was not significantly altered by the bound Rassf1C frag-

Eimproper 94 ± 6 81 ± 6 ments. Importantly, these residues (84–85 and 124–130) form

ENOE 408 ± 29 287 ± 24 the above mentioned exposed hydrophobic patch between the flexible helices H2 and H5 of DAXX55-144 (Figure 4G). These shift Ecdih 3±2 4±2 perturbation data clearly identified the Rassf1C-binding inter- Esani 68 ± 4 face on the DHB domain. Residual dipolar couplings c Q value 0.048 ± 0.002 Tertiary Structure of the DAXX55-144/Rassf1C23-38w ˚ Rmsd from average structure (A) Complex Residues in a-helicesd The structural ensemble of the DAXX55-144/Rassf1C23-38w Backbone atoms 0.14 ± 0.02 0.23 ± 0.04 complex was determined from an extensive set of NMR-derived Heavy atoms 0.53 ± 0.07 0.59 ± 0.09 inter- and intramolecular restraints (Figure 5 and Table 1). As ex- All residuese pected from chemical shift comparisons, peptide-bound 55-144 Backbone atoms 0.16 ± 0.03 0.27 ± 0.07 DAXX retained the secondary and tertiary structure of the free DHB domain. This is reflected quantitatively by the low Heavy atoms 0.54 ± 0.06 0.68 ± 0.09 RMS deviations between the helices in the structural ensembles The 25 lowest energy structures after the water refinement iteration of the of unbound versus bound DAXX55-144 (0.29 ± 0.04 and 0.80 ± ARIA protocol (Habeck et al., 2004). a 0.09 A˚ for main chain and all heavy atoms, respectively). Number of unambiguous and, in parentheses, ambiguous restraints. 23-38w b Final ARIA/CNS energies for van der Waals (vdw), bonds, angles, NOE Rassf1C lies ‘‘side-on’’ along the shallow cleft between restraints (NOE), dihedral angle restraints (cidh) and residual dipolar helices 2 and 5 of the DHB. Consistent with chemical shift and coupling restraints (sani). relaxation changes, the peptide forms a short amphipathic helix c Defined as Q = RMS(Dobs-Dcalc)/RMS(Dobs)(Bax et al., 2001). (residues 29–34) with Leu31, Tyr34, Phe35, and Thr36 making d Residues in a helices in the apo-protein (60–77, 84–93, 97–100, most of the intermolecular contacts with Val84, Phe87, Tyr124, 103–118, 123–136) and in the complex-protein (DAXX: 60–77, 84–93, and Ile127 of the DHB domain (Figures 5C–5F). On the outer 98–100, 103–118, 123–137; Rassf1C: 29-34). side of the amphipathic helix, the hydrophilic Ser29, Glu30, e All residues excluding disordered termini in the apo protein (60-136) and Glu32, and Gln33 face the solvent. Preceding this helix, a cluster the complex (DAXX: 60-137; Rassf1C: 27-39). of negatively-charged residues in Rassf1C23-38w (Glu25, Asp26,

H1 H2 H3 H4 H5 Figure 3. The Global and Local Backbone 1.0 Dynamics of DAXX55-144 A (A) Heteronuclear 15N{1H}NOE of DAXX55-144 measured at 25C, pH 6.5 using a 500 MHz spec- 0.5 trometer. These data show that, with the exception of the terminal residues, the DHB domain is well ordered on the nsec-psec timescale. Lower NOE values indicate increasing mobility on a sub-nsec OE

N 0 timescale. Missing data points corresponded to prolines or residues with overlapping signals. (B) Amide HX lifetimes for DAXX55-144 domain at an -0.5 uncorrected pH 6.5 and 25C. Data points are only -1.0 shown for amides with resolved 1HN-15N signals 15 6 present at least in the first N-HSQC spectrum 10 B recorded after transfer to D2O buffer (i.e., 1/kex >300 s; dashed line). Helix H1 forms the 5 stable core of the DHB domain, whereas helix H2 10 under goes relatively facile HX. 15 See Figure S3 for NT1 and T2 relaxation data. 4 ex 10 1/k (s) 3 10 DISCUSSION 2 10 Structural Comparisons 60 70 80 90 100 110 120 130 140 with the PAH Domain Residue Prior to this study, the region of DAXX encompassing the DHB domain was hypothesized to contain a PAH domain (Hollenbach et al., 1999). Historically, and Asp28) is juxtaposed to a cluster of positively charged resi- this name comes from an early sequence-based prediction dues in DAXX55-144 (Arg91, Arg94, Lys135, His137, Lys140, that the yeast Sin3 histone deacetylase-associated corepres- Lys141, and Lys142) (Figure 5D). The importance of these elec- sors contain two adjacent amphipathic helices (Wang et al., trostatic interactions is demonstrated by the observed ionic 1990). The subsequence structural analyses of Sin3A and strength dependence of binding (Table S1 and Figure S5). Sin3B fragments demonstrated that the so-called PAH domain Although the structural ensemble of the DHB domain does not is in fact a four-helix bundle that binds helical segments of change significantly upon Rassf1C23-38w binding, it is interesting partner proteins ‘‘end-on’’ via a hydrophobic interface wedged that the interfacial residues lie along helices H2 and H5, which by between these helices (Figure 6)(Brubaker et al., 2000; Spronk HX measurements are locally dynamic (Figure 3B). Their flexi- et al., 2000). Based on these observations, Sponk et al. sug- bility may facilitate subtle changes to accommodate a bound gested the descriptive name of ‘‘wedged helical bundle’’ for partner protein. the domain. However, the somewhat erroneous PAH nomencla- ture has prevailed in the literature. A DHB Domain Interacting Motif: Peptides from p53 Although at first sight the DHB domain indeed looks similar to and Mdm2 bind DAXX55-144 the PAH domain, closer inspection reveals a number of very signif- The structure of the DAXX/Rassf1C complex suggested that icant differences (Figure 6; Figure S7). Most importantly, the DHB a DHB domain interaction motif is characterized by the presence domain contains a fifth helix (helix H3). Although short, this helix is of four negatively charged residues (Glu25, Asp26, Asp28, well defined and traverses diagonally from helix H2 to H4. As Glu30), followed by an aliphatic (Leu31), two polar (Glu32, a result of this crossover, the Sin3 PAH domains and the DHB Gln33), two aromatic (Tyr34, Phe35), and a final polar (Thr36) domain have distinct arrangements of their core four-helix residue.As summarized in Table 2, we identified suchapproximate bundles. By this criteria, the DHB domain can be classified in the motifs in four reported DAXX interacting partners. In vitro binding same topological family as domain I of the transcription elongation was tested by NMR-monitored titrations of 15N-DAXX55-144 with factor TFIIS (Booth et al., 2000). The polypeptide connectivity of unlabeled peptides corresponding to these sequences. the core four-helix bundles of TFIIS, DAXX, and Sin3 are all up- STAT3443-458 and NHE1565-580, which only contain two negatively down-up-down and left-handed. However, the direction vectors charged residues and a positively charged histidine, did not of nearest neighbor helices in the DAXX DHB and TFIIS domains show any significant binding. On the other hand, the peptides are not all antiparallel to each other, whereas they are in most Mdm2296-313 and p5339-57, which more closely resemble the helical bundles including the Sin3 PAH domains (Presnell and proposed interaction motif, bound DAXX55-144 withaffinities similar Cohen, 1989). This is due to the crossover of helix H3, making to that of Rassf1C23-38w. Chemical shift perturbation mapping helices H1 and H4 and helices H2 and H5 parallel to one another. confirmed that these peptides also bound the interface of The structure of the DHB domain/Rassf1C peptide complex DAXX55-144 formed by helices H2 and H5 (Figure S6). also confirms that its binding mode is completely different to

F35 A B T36 A37 Q33 Q33 Y34 0.4 120 E32 L31 F35 0.3

121 R38 (ppm)

N (ppm) 0.2

15 E32

122 D26 0.1 L31 ± KD= 55 5 M

8.2 8.1 8.0 7.9 7.8 1234 1H (ppm) DAXX:Rassf1C ratio

1.0 0.4 C DE 0.05 0.8 0 0 0.6 -0.4 -0.05

0.4 (ppm)

NOE -0.10

-0.8 C (ppm) 0.2 -0.15 -1.2 0 -0.20 -1.6 -0.2 26 30 34 38 25 30 35 25 30 35 Rassf1C residue Residue Residue

G F Y124 0.30 H1 H5 V84 0.20 V125 (ppm) H4 E129 I127 H2 0.10 N128

K60 H3 60 70 80 90 100 110 120 130 140 C DAXX residue N

Figure 4. Identifying the Binding Interfaces between Rassf1C and DAXX by Spectral Perturbation Mapping (A) A section of the 15N-HSQC spectrum of 15N-Rassf1C23-38w (125 mM), showing progressive amide chemical shift changes upon addition of aliquots of 2 mM unlabeled DAXX55-144 to a final 4.7 molar excess. 2 2 23-38w 55-144 (B) Fitting the chemical shift changes ([(Dd1H) + [(0.2Dd15N) ]) of residues 31–37 of Rassf1C as a function of added DAXX yielded an average KD value of 55 ± 5 mM for dissociation of the complex. The proteins were in low ionic strength NMR sample buffer (10 mM potassium phosphate [pH 6.5], 25C) without added KCl. (C) Heteronuclear 15N{1H}NOE values of 15N-Rassf1C23-38w free (gray) and bound to DAXX55-144 (black) demonstrate that residues 28-38 are unstructured in isolation and become ordered upon complex formation. Decreasing NOE values result from increasing mobility on a sub-nsec timescale. (D and E) Downfield and upfield 13Ca and 1Ha chemical shift changes, respectively, of the DAXX55-144-bound Rassf1C23-38w (95% saturated) relative to the free peptide demonstrate that residues 30–34 of Rassf1C23-38w undergo a coil-to-helix transition upon binding DAXX55-144. (F) Amide chemical shift perturbations of 15N-labeled DAXX55-144 resulting from addition of 7.5:1 unlabeled Rassf1C23-38w identify the peptide-binding interface. (G) Ribbon and surface representations of DAXX55-144, with residues showing the largest spectral perturbations upon titration with Rassf1C23-38w highlighted in red (i.e., dotted line in F; Dd > 0.085 ppm). Also in red are Cys58 and Lys60, which experienced spectral changes independent of the binding of Rassf1C23-38w. The 1HN-15N signal of Lys60 is exquisitely sensitive to the sample salt concentration, possibly due to an ionic strength-dependent hydrogen bond with Leu100. See also Table 2, and Figure S4 and Table S1. those of the Sin3 PAH domains (Figure 6). For example, the DHB domain blocks this cleft, thus sterically preventing such Sin3A and Sin3B PAH2 domains binds a Mad1 peptide ‘‘end- ‘‘end-on’’ binding. The H1/H2 interhelical angle of the DHB on’’ via a cleft between all four of their helices (Brubaker et al., domain are also 165, creating a more cylindrical than wedge 2000; Spronk et al., 2000). This cleft results from the 140 inter- shape. Instead, the Rassf1C peptide is bound ‘‘side-on’’ via helical angle between helices H1/H2 and between H3/H4, a shallow 570 A˚ 2 hydrophobic groove along helices H2 and which creates a wedge shape to expose a large 2100 A˚ 2 hydro- H5. In addition, the Mad1 peptide forms a longer amphipathic phobic interface. In contrast, the crossover helix H3 of the DAXX helix (11 amino acids) when bound to the Sin3B PAH2 domain

Table 2. Identification of a DHB Domain Interacting Motif in Reported DAXX Partners a b c Partner (Accession No.) Technique Reference Peptide NMR-measured KD (mM) Rassf1C Y2H, Co-IP (Kitagawa et al., 2006)23SQEDSDSELEQYFTAR 38 65 ± 16d 55 ± 5d Mdm2 (NP_002383) Co-IP (Tang et al., 2006) 296 FEEDPEISLADYWKCT 313 70 ± 13 p53 (BAC16799) GST, Y2H (Gostissa et al., 2004)39AMDDLMLSPDDIEQWFTED 57 15 ± 13 STAT3 (NP_644805) GST, Co-IP (Muromoto et al., 2006) 458 SHTELDIKLGQHYVET 443e N.D.f NHE1 (NP_003038) Y2H, Co-IP (Jung et al., 2008) 565 GERSKEPQLIAFYHKM 580 N.D.f a Binding detected by coimmunoprecipitation or GST pull-down of exogenous protein or a yeast two-hybrid screen. b Negatively charged residues (bold) preceding the amphiphatic core (bold and underlined) are highlighted. c 55-144 2 2 In vitro binding of the indicated peptide to DAXX .KD values derived by fitting the amide chemical shift changes ([(Dd1H) + [(0.2Dd15N) ]) of residues 85 and 125–128 (for Rassf1C), 78 and 124–125 (for p53) and 85, 125–126, and 128–129 (for Mdm2) of 15N-DAXX55-144 (initially 100 mM) as a function of added unlabelled peptide to a final peptide/protein ratio of 6:1 (90% saturation). The samples were at 25C in low ionic strength NMR buffer without added KCl. d Dissociation constants measured from the titration of 15N-DAXX55-144 with unlabelled Rassf1C23-38w or 15N-Rassf1C23-38w with unlabelled DAXX55-144. e Inverted sequence. f No binding detected by NMR titrations (hence KD > > 1 mM). than does the Rassf1C peptide (5 amino acids) in the resulting whereas Rassf1C is strictly cytoplasmic even in unstressed DHB domain complex. In both cases, the hydrophilic residues conditions (S.G., A.M.I., unpublished data). However, during of the bound peptides are negatively charged and complement prometaphase, DAXX and Rassf1C interact via the DHB domain. the positively charged residues the DHB and PAH domains. This interaction defines a DAXX-dependent mitotic check point Finally, the isolated DHB domain is well ordered on the sub- that is critical for the response to taxol (Lindsay, 2008). We nsec timescale, whereas the ligand-free Sin3 PAH domains also discovered that depletion of either Rassf1 or DAXX show substantial conformational dynamics involving partially increases taxol resistance in cellular and animal models via unfolded states (van Ingen et al., 2006; Sahu et al., 2008). Collec- elevating cyclin B stability (S.G., A.M.I., unpublished data). This tively, these differences unequivocally demonstrate that DAXX suggests that DAXX/Rassf1 establish a mitotic stress check- does not contain a PAH domain as is often cited in the literature. point, enabling cells to die efficiently when encountering specific mitotic stress stimuli, including taxol. Therefore, DAXX and Proteins with Similar Helical Bundle Architecture Rassf1 may become useful predictive markers for the proper Additional structural relatives of the DHB domain/Rassf1C23-38w selection of patients for taxane chemotherapy. Furthermore, complex were searched for using the DALI server. The best a detailed understanding of their interactions as described herein match was the five-helix bundle THATCH domain from the may help to overcome chemotherapy resistance. protein-trafficking control Huntingtin-interacting protein-1 related (HIP1R) protein (Z-score 5.5, rmsd 3.0 A˚ ). Although not Interactions of DAXX with Mdm2 and p53 a complex, the first four helices of the helix bundle follow the The DHB domain complex with Rassf1C also provides clues for same topology as the DHB domain. The fifth helix packs on dissecting the interaction of DAXX with potential binding part- the groove of helices 2 and 4, in a very similar manner to the ners. We hypothesized that such partners contain a similar nega- packing of the Rassf1C helix along the equivalent groove in the tively charged region juxtaposed with a short sequence that DHB domain (Figure S7). A more distant structural relative is folds as an amphiphatic helix. Of four such potential sequences the Focal Adhesion Targeting (FAT) domain in complex with Pax- tested, those of p5339-57 and Mdm2296-313 indeed bound the illin LD domains. The FAT domain forms an antiparallel right- same interface of the DHB domain as did Rassf1C and with handed four-helix bundle that binds Paxillin LD peptides. In the comparable affinity. Given the small size of this interface, it is structure of the complex, two LD peptides form amphipathic reasonable to predict that these peptides would bind DAXX in helices that pack ‘‘side-on’’ against helices of the FAT domain a mutually exclusive, competitive manner. (Figure 6; Figure S7). Therefore, the packing of a helical ligand The interactions of DAXX with p53 and Mdm2 provide another along a groove between two helices of a four-helix bundle is link for understanding the regulation of apoptosis. In unstressed not uncommon. cells, DAXX stabilizes the E3 ligase Mdm2 by also recruiting the deubiquitinase HAUSP (Tang et al., 2006). Stabilized Mdm2 can Biological Implications of the Rassf1C-DAXX Interaction ubiquitinate p53, targeting it for proteosomal degradation. Upon The structural characterization of the DHB domain and its cellular stress, DAXX dissociates from Mdm2, Mdm2 undergoes complex with Rassf1C23-38w provides a critically needed molec- self-ubiquitination and degradation, and the resulting increased ular framework for understanding the interaction of DAXX and levels of p53 induce cell-cycle arrest or apoptosis. Although the this tumor suppressor protein. Kitagawa et al. (2006) proposed molecular mechanism for the interactions of DAXX, HAUSP, that DAXX sequesters Rassf1C into PML-NBs. Upon ubiquitin- Mdm2, and p53 underlying this regulatory pathway are largely mediated degradation of DAXX in response to DNA damage, unknown, an important clue is provided by our observation Rassf1C is released to the cytoplasm activate the SAPK/JNK that, at least in vitro, both p5339-57 and Mdm2296-313 bind the signaling pathway. In contrast, we have found that DAXX is DHB domain. However, we stress that these regions of p53 a nuclear protein during interphase (Lindsay et al., 2009), and Mdm2 were chosen based only on a proposed DHB domain

R38 A B H1 H5

Rassf1C 0 45 H4

H2 S23 H3

N144 G55

V84 E83 P82 D80 W39 C A121 Y124 D R38 K122 H81 A37 R119 I127 T36 F35 V125 R115 0 Y34 0 E32 C131 180 180 L31 E129 E30 V133 D28 R91 D26 R111 I108 E104 E25 P86 H137 R94 A136 F87 K141 K135 L143 K56 K140 K142 N144 DHB Rassf1C DHB Rassf1C

I127 55-144 DAXX Helix 5 T132 -- N128 I127 - V125 Y124 - K122 A121 P120 R119 V84 23-38w Rassf1C D26 S27 D28 E30 L31 E32 Q33 Y34 F35 T36 A37 R38 W39 N128 Y124 F87 L31 55-144 A37 F35 DAXX Helix 2 L88 F87 P86 V85 V84 E83 Y34 T36

Figure 5. Structural Ensemble of the DAXX55-144/Rassf1C23-38w complex (A and B) (A) Superimposed backbone atoms from the 25 member DAXX55-144/Rassf1C23-38w complex ensemble and (B) a ribbon diagram of the lowest energy structure. Rassf1C23-38w is gold and DAXX55-144 colored as in Figure 1. (C and D) Interaction surfaces of the complex. The two molecules are separated and colored by surface physicochemical properties. (C) highlights the interfacial hydrophobic residues (green) and D shows negative (red), positive (blue), and neutral polar (white) side chains. The default values from Molmol (Koradi et al., 1996) were used for calculating the electrostatic potential at zero ionic strength. (E) Expanded view of the side-chain atoms forming the intermolecular interface. (F) A summary of the observed intermolecular NOE connections between residues in helices H2 and H5 of the DHB domain and the bound Rassf1C23-38w peptide. Note that the C-terminal nonnative Trp39 of Rassf1C23-38w also contacted DAXX55-144. Attempts to extend the C terminus with native residues did not result in any improvement in complex formation, and in some cases led to aggregation. This suggests that Trp39 does not contribute signiﬁcantly to the afﬁnity of the DAXX55-144/Rassf1C23-38w complex. See also Figure S5.

Rassf1C Paxillin

N Mad1 N N C C C N C TFIIS DHB FAT SIN3B PAH2

21 21 2 1 1 2

3 4 4 5 43 34

N N 3 N N Left hand Left hand Right hand Left hand

C C C C

Figure 6. Structural Comparison of the DHB Domain with Similar Helix Bundles The DHB domain shares the same left-handed helical bundle topology as the TFIIS domain I (1EO0.pdb), except for the presence of the short crossover helix H3 (orange). The right-handed FAT domain bundle binds a Paxillin helical peptide (1OW8.pdb) in a similar ‘‘side-on’’ conformation as seen in the DHB domain/ Rassf1C complex (asterisk). In contrast, although also left-handed, the Sin3B PAH2 domain displays both different relative arrangements and orientations of constituent helices and binds the helical Mad1 peptide in an ‘‘end-on’’ manner (1E91.pdb). See also Figure S7. binding motif defined by Rassf1C23-38. Further studies are kinases. Further in vivo studies, directed by structural knowledge required to establish the biological significance of these of the DAXX/Rassf1C complex, are needed to test this hypothesis. interactions. Interestingly, p5339-57 corresponds to part of the transactivation Modular Structure of DAXX domain of p53 (Candau et al., 1997). This domain has been shown Based on NMR analyses of a series of DAXX truncation frag- to be intrinsically disordered (Dawson et al., 2003), to interact with ments, combined with patterns of sequence conservation and a number of proteins including Mdm2 and CBP/p300 (Ferreon predicted secondary structure and intrinsic disorder, we et al., 2009), to be carefully regulated by phosphorylation and propose a revised model for the modular organization of DAXX protein-protein interactions (Kaustov et al., 2006), and to form (Figure 7). The N-terminal half of DAXX contains at least two an amphipathic helix upon binding target proteins (Kussie et al., helical domains. The first is the DHB domain, whose structure 1996). DAXX association with the highly regulated transactivation and function in mediating protein interactions is presented domain of p53 could thus form part of a competitive binding herein. The second, a putative helical domain that we could mechanism for controlling apoptosis. not express in soluble form in E. coli, awaits characterization. The potential interaction of residues 296-313 of Mdm2 with the In addition to the linker between these domains, the 50 DHB domain is also intriguing. The latter half of these residues N-terminal and 340 C-terminal residues of DAXX are predicted form an N-terminal 310 helix and first b strand of the C4 zinc finger to be unstructured. The intrinsic disorder of these regions was domain of Mdm2 (Yu et al., 2006). The functions of this domain, confirmed by NMR analysis of DAXX1-144 (not shown) and located within the otherwise disordered acidic region of Mdm2, mDAXX566-739. Interestingly, each of these flexible termini are not well established. However, a recent report has implicated contains a SIM (Lin et al., 2006; Santiago et al., 2009). the zinc finger in mediating binding to ribosomal proteins (Lind- Despite the lack of structural elements, numerous proteins strom et al., 2007). It will be interesting to investigate whether have also been reported to interact with the C-terminal portion zinc binding could regulate the interaction between DAXX and of DAXX. Indeed, intrinsically disordered regions often serve as native Mdm2. linear peptide recognition motifs and are the sites of regulatory Finally, the electrostatic nature of the DHB domain/ posttranslational modifications. For example, DAXX contains Rassf1C23-38w interactions suggests a possible mechanism for a nuclear localization signal in this region (Yeung et al., 2008), regulating DAXX function via phosphorylation of partner proteins at least two phosphorylation sites (Ecsedy et al., 2003; Song to increase the negative-charge flanking an amphipathic helix. and Lee, 2003), and multiple sumoylation sites (Lin et al., The peptides tested in this report either contain residues known 2006). Thus, the flexibility of the C-terminal portion of DAXX to be phosphorylated or to form consensus target sequences for could play a role in facilitating access to modifying enzymes

DHB Helical acidic SPE SPT SIM SIM 1 17 55 144 180 400 434 485 495596 665724 740

Rassf1C

SUMO I- SMC SUMOSUMO Helical TF uns ructuredt ac dic unstructured SIM-N i E SPT SP DHB

Figure 7. Model of the Structural Organization of DAXX DAXX contains the characterized DHB domain, along with another predicted helical region (shown as a Robetta model) (Kim et al., 2004). The DHB domain can recognize intrinsically disordered regions with specific motifs that form an amphipathic helix upon binding, such as that found at the N terminus of Rassf1C. The Ras association domain of Rassf5A (3DDC.pdb) is used as a model for Rassf1C, and other Rassf1C domains omitted for clarity. The intrinsically disordered regions (acidic, SPE, SPT, and SIMs) of DAXX are to scale in an arbitrary extended conformation. In reality, they would sample a much smaller average radius of gyration. Also shown are SUMO molecules noncovalently binding each SIM. SUMO is often covalently attached to an unstructured region of a transcription factor (TF), such as Ets-1, via a flexible isopeptide bond (Macauley et al., 2006). and interacting partners. We also speculate that, as with the Structure Determination case of Ets-1 (Macauley et al., 2006), many of these reported NOE-derived distance restraints were obtained from 3D simultaneous regular 13 15 interactions are not direct, but rather are mediated by the binding and constant time methyl C- and N-NOESY-HSQC spectra, all with 100 ms mixing times (Zwahlen et al., 1998). In the case of the DAXX55-144/ of a DAXX SIM to an unrecognized intervening SUMO-tag cova- 23-38w 13 15 Rassf1C complex, a C/ N filtered-edited NOESY-HSQC spectrum lently attached to the partner protein. (150 msec) was also recorded to obtain intermolecular distance restraints In closing, the structure of the DHB domain and the proposed (Zwahlen et al., 1997). Backbone dihedral angle restraints were derived using modular organization of DAXX provide a testable framework for TALOS (Cornilescu et al., 1999). 1HN-15N residual dipolar couplings (RDC’s) understanding the role of this essential scaffold protein in medi- were measured for unbound DAXX55-144 partially aligned in stretched poly- ating a diverse array of cellular processes. acrylamide gels (Chou et al., 2001). Structural calculations were carried out using ARIA 2.2 with CNS 1.2 (Habeck et al., 2004). The majority of long range intramolecular NOEs were assigned automatically, whereas all of the intermo- EXPERIMENTAL PROCEDURES lecular NOEs for the complex were assigned manually. Molecular graphics were rendered using PyMol (DeLano, 2004). Protein Production

The genes encoding the His6-tagged DAXX constructs were cloned by PCR methods from a human cDNA library (accession CAG33366; Invitrogen) into ACCESSION NUMBERS pET28a vector (Novagen). To produce GST-tagged fusion proteins, Rassf1C (obtained from Dr. G. Pfeifer, Beckman Research Institute, CA) was cloned The coordinates of DAXX55-144 and the DAXX55-144/Rassf1C23-38w complex into pGEX-2T expression vector, whereas E. coli codon optimized synthetic have been deposited in the Protein Data Bank under accession codes 2KZS oligonucleotides encoding p53, Mdm2, STAT3, and NHE1 constructs were and 2KZU, respectively. cloned into a pGEX-4T expression vector (GE Healthcare). Rassf1C23-38w contains an added C-terminal tryptophan for quantification by absorbance SUPPLEMENTAL INFORMATION spectroscopy. Proteins were expressed using E. coli BL21 (lDE3) cells and purified by standard protocols, as detailed in the Supplemental Experimental Supplemental Information includes Supplemental Experimental Procedures, Procedures. seven figures, and one table and can be found online at doi:10.1016/j.str. 2010.09.016. NMR Spectroscopy Spectra were recorded at 25C using 500 MHz Varian Unity and 600 MHz Var- ian Inova NMR spectrometers. Unless stated otherwise, proteins were in ACKNOWLEDGMENTS a sample buffer of 100 mM KCl, 10 mM potassium phosphate (pH 6.5),

0.1 mM EDTA, and 10 mM DTT with 5% D2O. The chemical shifts of We thank Lewis Kay, Cameron Mackereth, and Mark Okon for advice and help. DAXX55-144 and the DAXX55-144/Rassf1C23-38w complex have been deposited This research was funded by the Canadian Cancer Society Research Institute in the BioMagResBank under accession codes 17018 and 17019, respectively. (017308 to L.P.M.) and by NIH/NCI R01 CA127378-01A1 (to S.G. and A.M.I.). Amide 15N relaxation data (Farrow et al., 1994) were acquired on a 500 MHz Instrument support was provided by the Canadian Institutes for Health spectrometer at 25C and analyzed using Tensor2 (Dosset et al., 2000). Amide Research, the Canadian Foundation for Innovation, the British Columbia hydrogen exchange (HX) rates were measured from a series of 15N-HSQC Knowledge Development Fund, the UBC Blusson Fund, and the Michael Smith 55-144 spectra recorded at 25 C after dissolving lyophilized DAXX into a D2O Foundation for Health Research. sample buffer. NMR-monitored titrations were carried out by recording 15N-HSQC spectra of 15N-labeled protein to which the unlabeled protein partner Received: July 2, 2010 was added in small aliquots (Table S1). Equilibrium dissociation constants were Revised: September 6, 2010 determined by non-linear least-squares ﬁtting to the equation describing forma- Accepted: September 8, 2010 tion of a 1:1 complex in the fast exchange limit (Johnson et al., 1996). Published: December 7, 2010

REFERENCES Hollenbach, A.D., Sublett, J.E., McPherson, C.J., and Grosveld, G. (1999). The Pax3-FKHR oncoprotein is unresponsive to the Pax3-associated repressor Bax, A., Kontaxis, G., and Tjandra, N. (2001). Dipolar couplings in macromo- hDaxx. EMBO J. 18, 3702–3711. 339 lecular structure determination. Methods Enzymol , 127–174. Hollenbach, A.D., McPherson, C.J., Mientjes, E.J., Iyengar, R., and Grosveld, Bernado, P., Garcia de la Torre, J., and Pons, M. (2002). Interpretation of 15N G. (2002). Daxx and histone deacetylase II associate with chromatin through NMR relaxation data of globular proteins using hydrodynamic calculations an interaction with core histones and the chromatin-associated protein Dek. with HYDRONMR. J. Biomol. NMR 23, 139–150. J. Cell Sci. 115, 3319–3330. Booth, V., Koth, C.M., Edwards, A.M., and Arrowsmith, C.H. (2000). Structure Ishov, A.M., Sotnikov, A.G., Negorev, D., Vladimirova, O.V., Neff, N., Kamitani, of a conserved domain common to the transcription factors TFIIS, elongin A, T., Yeh, E.T.H., Strauss, J.F., and Maul, G.G. (1999). PML is critical for ND10 and CRSP70. J. Biol. Chem. 275, 31266–31268. formation and recruits the PML-interacting protein Daxx to this nuclear structure when modified by SUMO-1. J. Cell Biol. 147, 221–233. Brubaker, K., Cowley, S.M., Huang, K., Loo, L., Yochum, G.S., Ayer, D.E., Eisenman, R.N., and Radhakrishnan, I. (2000). Solution structure of the inter- Johnson, P.E., Tomme, P., Joshi, M.D., and McIntosh, L.P. (1996). Interaction acting domains of the Mad-Sin3 complex: implications for recruitment of of soluble cellooligosaccharides with the N-terminal cellulose-binding domain a chromatin-modifying complex. Cell 103, 655–665. of Cellulomonas fimi CenC 2. NMR and ultraviolet absorption spectroscopy. Biochemistry 35, 13895–13906. Candau, R., Scolnick, D.M., Darpino, P., Ying, C.Y., Halazonetis, T.D., and Berger, S.L. (1997). Two tandem and independent sub-activation domains in Jung, Y.S., Kim, H.Y., Kim, J., Lee, M.G., Pouyssegur, J., and Kim, E. (2008). the amino terminus of p53 require the adaptor complex for activity. Physical interactions and functional coupling between Daxx and sodium 283 Oncogene 15, 807–816. hydrogen exchanger 1 in ischemic cell death. J. Biol. Chem. , 1018–1025. Chang, C.C., Lin, D.Y., Fang, H.I., Chen, R.H., and Shih, H.M. (2005). Daxx Juo, P., Kuo, C.J., Yuan, J., and Blenis, J. (1998). Essential requirement for mediates the small ubiquitin-like modifier-dependent transcriptional repres- caspase-8/FLICE in the initiation of the Fas-induced apoptotic cascade. 8 sion of Smad4. J. Biol. Chem. 280, 10164–10173. Curr. Biol. , 1001–1008. Kaustov, L., Yi, G.S., Ayed, A., Bochkareva, E., Bochkarev, A., and Chou, J.J., Gaemers, S., Howder, B., Louis, J.M., and Bax, A. (2001). A simple Arrowsmith, C.H. (2006). p53 transcriptional activation domain: a molecular apparatus for generating stretched polyacrylamide gels, yielding uniform chameleon? Cell Cycle 5, 489–494. alignment of proteins and detergent micelles. J. Biomol. NMR 21, 377–382. Kim, D.E., Chivian, D., and Baker, D. (2004). Protein structure prediction and Cornilescu, G., Delaglio, F., and Bax, A. (1999). Protein backbone angle analysis using the Robetta server. Nucleic Acids Res. 32, W526–W531. restraints from searching a database for chemical shift and sequence homology. J. Biomol. NMR 13, 289–302. Kitagawa, D., Kajiho, H., Negishi, T., Ura, S., Watanabe, T., Wada, T., Ichijo, H., Katada, T., and Nishina, H. (2006). Release of RASSF1C from the nucleus by Dawson, R., Muller, L., Dehner, A., Klein, C., Kessler, H., and Buchner, J. Daxx degradation links DNA damage and SAPK/JNK activation. EMBO J. (2003). The N-terminal domain of p53 is natively unfolded. J. Mol. Biol. 332, 25, 3286–3297. 1131–1141. Koradi, R., Billeter, M., and Wu¨ thrich, K. (1996). MOLMOL: a program for DeLano, W.L. (2004). Use of PYMOL as a communications tool for molecular display and analysis of macromolecular structures. J. Mol. Graph. 14, 228 science. Abstracts of Papers of the American Chemical Society , U313– 51–55, 29–32. U314. Kussie, P.H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J., Levine, A.J., Dosset, P., Hus, J.C., Blackledge, M., and Marion, D. (2000). Efficient analysis and Pavletich, N.P. (1996). Structure of the MDM2 oncoprotein bound to the of macromolecular rotational diffusion from heteronuclear relaxation data. p53 tumor suppressor transactivation domain. Science 274, 948–953. J. Biomol. NMR 16, 23–28. Lin, D.Y., Huang, Y.S., Jeng, J.C., Kuo, H.Y., Chang, C.C., Chao, T.T., Ho, Drane, P., Ouararhni, K., Depaux, A., Shuaib, M., and Hamiche, A. (2010). The C.C., Chen, Y.C., Lin, T.P., Fang, H.I., et al. (2006). Role of SUMO-interacting death-associated protein DAXX is a novel histone chaperone involved in the motif in Daxx SUMO modification, subnuclear localization, and repression of 24 replication-independent deposition of H3.3. Genes Dev. , 1253–1265. sumoylated transcription factors. Mol. Cell 24, 341–354. Ecsedy, J.A., Michaelson, J.S., and Leder, P. (2003). Homeodomain-interact- Lindsay, C.R. (2008). Daxx and Rassf1 define a novel mitotic stress checkpoint ing protein kinase 1 modulates Daxx localization, phosphorylation, and tran- that is critical for cellular taxol response. Ph.D. Thesis, University of Florida. scriptional activity. Mol. Cell. Biol. 23, 950–960. Lindsay, C.R., Morozov, V.M., and Ishov, A.M. (2008). PML NBs (ND10) and Englander, S.W., and Kallenbach, N.R. (1983). Hydrogen-exchange and struc- Daxx: from nuclear structure to protein function. Front. Biosci. 13, 7132–7142. 16 tural dynamics of proteins and nucleic-acids. Q. Rev. Biophys. , 521–655. Lindsay, C.R., Giovinazzi, S., and Ishov, A.M. (2009). Daxx is a predominately Farrow, N.A., Muhandiram, R., Singer, A.U., Pascal, S.M., Kay, C.M., Gish, G., nuclear protein that does not translocate to the cytoplasm in response to cell Shoelson, S.E., Pawson, T., Formankay, J.D., and Kay, L.E. (1994). Backbone stress. Cell Cycle 8, 1544–1551. dynamics of a free and a phosphopeptide-complexed Src Homology-2 Lindstrom, M.S., Jin, A., Deisenroth, C., White Wolf, G., and Zhang, Y. (2007). 33 domain studied by N-15 NMR relaxation. Biochemistry , 5984–6003. Cancer-associated mutations in the MDM2 zinc finger domain disrupt ribo- Ferreon, J.C., Lee, C.W., Arai, M., Martinez-Yamout, M.A., Dyson, H.J., and somal protein interaction and attenuate MDM2-induced p53 degradation. Wright, P.E. (2009). Cooperative regulation of p53 by modulation of ternary Mol. Cell. Biol. 27, 1056–1068. complex formation with CBP/p300 and HDM2. Proc. Natl. Acad. Sci. USA Macauley, M.S., Errington, W.J., Scharpf, M., Mackereth, C.D., Blaszczak, 106, 6591–6596. A.G., Graves, B.J., and McIntosh, L.P. (2006). Beads-on-a-string, character- Goldberg, A.D., Banaszynski, L.A., Noh, K.M., Lewis, P.W., Elsaesser, S.J., ization of ETS-1 sumoylated within its flexible N-terminal sequence. J. Biol. Stadler, S., Dewell, S., Law, M., Guo, X., Li, X., et al. (2010). Distinct factors Chem. 281, 4164–4172. control histone variant H3.3 localization at specific genomic regions. Cell Morozov, V.M., Massoll, N.A., Vladimirova, O.V., Maul, G.G., and Ishov, A.M. 140, 678–691. (2008). Regulation of c-met expression by transcription repressor Daxx. Gostissa, M., Morelli, M., Mantovani, F., Guida, E., Piazza, S., Collavin, L., Oncogene 27, 2177–2186. Brancolini, C., Schneider, C., and Del Sal, G. (2004). The transcriptional Muromoto, R., Nakao, K., Watanabe, T., Sato, N., Sekine, Y., Sugiyama, K., repressor hDaxx potentiates p53-dependent apoptosis. J. Biol. Chem. 279, Oritani, K., Shimoda, K., and Matsuda, T. (2006). Physical and functional inter- 48013–48023. actions between Daxx and STAT3. Oncogene 25, 2131–2136. Habeck, M., Rieping, W., Linge, J.P., and Nilges, M. (2004). NOE assignment Pluta, A.F., Earnshaw, W.C., and Goldberg, I.G. (1998). Interphase-specific with ARIA 2.0: the nuts and bolts. Methods Mol. Biol. 278, 379–402. association of intrinsic centromere protein CENP-C with hDaxx, a death

domain-binding protein implicated in Fas-mediated cell death. J. Cell Sci. 111, Tang, J., Qu, L.K., Zhang, J., Wang, W., Michaelson, J.S., Degenhardt, Y.Y., 2029–2041. El-Deiry, W.S., and Yang, X. (2006). Critical role for Daxx in regulating Mdm2. Nat. Cell Biol. 8, 855–862. Presnell, S.R., and Cohen, F.E. (1989). Topological distribution of four-alpha- helix bundles. Proc. Natl. Acad. Sci. USA 86, 6592–6596. Torii, S., Egan, D.A., Evans, R.A., and Reed, J.C. (1999). Human Daxx regu- lates Fas-induced apoptosis from nuclear PML oncogenic domains (PODs). Puto, L.A., and Reed, J.C. (2008). Daxx represses RelB target promoters via EMBO J. 18, 6037–6049. DNA methyltransferase recruitment and DNA hypermethylation. Genes Dev. van Ingen, H., Baltussen, M.A., Aelen, J., and Vuister, G.W. (2006). Role of 22, 998–1010. structural and dynamical plasticity in Sin3: the free PAH2 domain is a folded Sahu, S.C., Swanson, K.A., Kang, R.S., Huang, K., Brubaker, K., Ratcliff, K., module in mSin3B. J. Mol. Biol. 358, 485–497. and Radhakrishnan, I. (2008). Conserved themes in target recognition by the Wang, H., Clark, I., Nicholson, P.R., Herskowitz, I., and Stillman, D.J. (1990). PAH1 and PAH2 domains of the Sin3 transcriptional corepressor. J. Mol. The Saccharomyces cerevisiae SIN3 gene, a negative regulator of HO, 375 Biol. , 1444–1456. contains four paired amphipathic helix motifs. Mol. Cell. Biol. 10, 5927–5936. Salomoni, P., and Khelifi, A.F. (2006). Daxx: death or survival protein? Trends Yang, X., Khosravi-Far, R., Chang, H.Y., and Baltimore, D. (1997). Daxx, a novel Cell Biol. 16, 97–104. Fas-binding protein that activates JNK and apoptosis. Cell 89, 1067–1076. Santiago, A., Godsey, A.C., Hossain, J., Zhao, L.Y., and Liao, D. (2009). Yeung, P.L., Chen, L.Y., Tsai, S.C., Zhang, A., and Chen, J.D. (2008). Daxx Identification of two independent SUMO-interacting motifs in Daxx: evolu- contains two nuclear localization signals and interacts with importin alpha 3. tionary conservation from Drosophila to humans and their biochemical func- J. Cell. Biochem. 103, 456–470. tions. Cell Cycle 8, 76–87. Yu, G.W., Allen, M.D., Andreeva, A., Fersht, A.R., and Bycroft, M. (2006). 15 Shih, H.M., Chang, C.C., Kuo, H.Y., and Lin, D.Y. (2007). Daxx mediates Solution structure of the C4 zinc finger domain of HDM2. Protein Sci. , SUMO-dependent transcriptional control and subnuclear compartmentaliza- 384–389. tion. Biochem. Soc. Trans. 35, 1397–1400. Zhang, J., Cado, D., Chen, A., Kabra, N.H., and Winoto, A. (1998). Fas-mediated apoptosis and activation-induced T-cell proliferation are defective in mice Song, J.J., and Lee, Y.J. (2003). Role of the ASK1-SEK1-JNK1-HIPK1 signal in lacking FADD/Mort1. Nature 392, 296–300. Daxx trafficking and ASK1 oligomerization. J. Biol. Chem. 278, 47245–47252. Zwahlen, C., Legault, P., Vincent, S.J.F., Greenblatt, J., Konrat, R., and Kay, Song, M.S., Song, S.J., Kim, S.Y., Oh, H.J., and Lim, D.S. (2008). The tumour L.E. (1997). Methods for measurement of intermolecular NOEs by multinuclear suppressor RASSF1A promotes MDM2 self-ubiquitination by disrupting the NMR spectroscopy: Application to a bacteriophage lambda N-peptide/boxB 27 MDM2-DAXX-HAUSP complex. EMBO J. , 1863–1874. RNA complex. J. Am. Chem. Soc. 119, 6711–6721. Spronk, C.A., Tessari, M., Kaan, A.M., Jansen, J.F., Vermeulen, M., Zwahlen, C., Gardner, K.H., Sarma, S.P., Horita, D.A., Byrd, R.A., and Kay, L.E. Stunnenberg, H.G., and Vuister, G.W. (2000). The Mad1-Sin3B interaction (1998). An NMR experiment for measuring methyl-methyl NOEs in C-13- involves a novel helical fold. Nat. Struct. Biol. 7, 1100–1104. labeled proteins with high resolution. J. Am. Chem. Soc. 120, 7617–7625.