Bivalent recognition of nucleosomes by the tandem PHD fingers of the CHD4 ATPase is required for CHD4-mediated repression

Catherine A. Musselmana, Julita Ramírezb, Jennifer K. Simsc, Robyn E. Mansfieldd, Samuel S. Olivere, John M. Denue, Joel P. Mackayd, Paul A. Wadec, James Hagmanb, and Tatiana G. Kutateladzea,1

aDepartment of Pharmacology, University of Colorado Denver School of Medicine, Aurora, CO 80045; bIntegrated Department of Immunology, National Jewish Health, Denver, CO 80206; cLaboratory of Molecular Carcinogenesis, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709; dSchool of Molecular Biosciences, University of Sydney, Sydney, New South Wales 2006, Australia; and eDepartment of Biomolecular Chemistry, University of Wisconsin, Madison, WI 53706

Edited by C. David Allis, The Rockefeller University, New York, NY, and approved November 22, 2011 (received for review August 19, 2011)

CHD4 is a catalytic subunit of the NuRD (nucleosome remodeling ious nuclear complexes to (reviewed in refs. 17 and and deacetylase) complex essential in transcriptional regulation, 18). PHD fingers bind primarily to H3 trimethylated at chromatin assembly and DNA damage repair. CHD4 contains tan- Lys4 (H3K4me3) or unmodified H3K4. The biological outcome dem plant homeodomain (PHD) fingers connected by a short linker, of these interactions is highly context-dependent and is deter- the biological function of which remains unclear. Here we explore mined by a number of factors including the composition of the the combinatorial action of the CHD4 PHD1/2 fingers and detail the complex, adjacent effector domains present in the same , molecular basis for their association with chromatin. We found and neighboring posttranslational histone modifications. An that PHD1/2 targets nucleosomes in a multivalent manner, conco- adjacent effector domain can be another PHD finger, a distinct mitantly engaging two histone H3 tails. This robust synergistic histone-binding domain such as , chromodomain, interaction displaces HP1γ from pericentric sites, inducing changes Tudor, or a catalytic histone-associating module. The combina- in chromatin structure and leading to the dispersion of the hetero- torial action of linked effectors often gives rise to binding me- chromatic mark H3K9me3. We demonstrate that recognition of the chanisms that are fundamentally different from those seen for histone H3 tails by the PHD fingers is required for repressive activ- the individual modules. Simultaneous interactions of adjacent do- ity of the CHD4/NuRD complex. Together, our data elucidate the mains modulate the affinity and specificity of their host molecular mechanism of multivalent association of the PHD fingers or protein complexes for specific chromatin states and thus play with chromatin and reveal their critical role in the regulation of a significant role in epigenetic mechanisms. Despite the wide- CHD4 functions. spread phenomena of the multivalent engagement of epigenetic readers, characterization of the structure and binding properties repression ∣ epigenetics ∣ histone ∣ posttranslational modifications of multidomain constructs poses unique challenges. To date few examples have been reported (19, 20), and we have only begun hromodomain helicase DNA-binding protein 4 (CHD4) is an understanding the complexity of crosstalk within polyeffector CATP dependent chromatin remodeler and a major subunit of systems that defines a specific biological outcome. the repressive NuRD complex (1–4). The CHD4/NuRD complex In this study, we elucidate the biological function and binding plays pivotal roles in transcriptional regulation and reorganiza- mechanism of the tandem CHD4 PHD1/2 fingers connected by tion and maintenance of chromatin structure (5, 6). It targets the native linker. Our data reveal that each PHD module of pericentromeric and has recently been impli- PHD1/2 possesses individual histone-binding activity. The do- cated in DNA damage repair, being rapidly recruited to DNA main organization positions PHD1/2 to concomitantly bind two double strand breaks (7, 8). Other components of the complex histone H3 tails, directing CHD4 to chromatin in a multivalent include a second catalytic subunit HDAC1/2 (histone deacety- manner. We employ a combination of NMR, mutagenesis, tryp- lase) and the nonenzymatic proteins MBD2/3 (methyl-CpG-bind- tophan fluorescence, and gel shift assays to define this unique ing domain), RbAp46/48 (retinoblastoma-associated), MTA1/2/3 mechanism and in vivo immunofluorescence, immunoprecipita- α β (metastasis-associated) and p66 / (Fig. 1A). The exact composi- tion, shRNA knockdown, and flow cytometry analyses to estab- tion of NuRD can vary, reflecting alterations in the activity and lish the significance of the PHD1/2 fingers for chromatin localization of the complex. localization and transcriptional regulation by CHD4. BIOCHEMISTRY CHD4 (also known as Mi-2β) is a large multimodular protein. It belongs to the family of nine CHD proteins and, similar to its Results and Discussion α closest family members CHD3 (or Mi-2 ) and CHD5, contains The Tandem PHD Fingers of CHD4 Possess Individual Histone-Binding tandem plant homeodomain (PHD) fingers, two chromodomains, Activities. The CHD4 ATPase contains an N-terminal tandem and a catalytic ATPase module (Fig. 1 B and C). The SNF2-like of PHD fingers (PHD1/2) linked by a stretch of 30 amino acids. ATPase domain mediates nucleosome mobility, providing the energy necessary for histone displacement and sliding (9). Unlike other chromodomains that bind methylated histone marks en- Author contributions: C.A.M., J.H., and T.G.K. designed research; C.A.M., J.R., J.K.S., and riched in heterochromatic regions (10–13), the CHD4 chromodo- R.E.M. performed research; S.S.O. and J.M.D. contributed new reagents/analytic tools; C.A.M., J.R., J.K.S., R.E.M., J.P.M., P.A.W., J.H., and T.G.K. analyzed data; C.A.M., J.R., mains exhibit a preference for DNA (14). The individual PHD J.K.S., R.E.M., J.P.M., P.A.W., J.H., and T.G.K. summarized results; and C.A.M. and T.G.K. fingers of CHD4 (PHD1 and PHD2) have been shown to possess wrote the paper. histone-binding activity (15, 16); however, the target of the tandem The authors declare no conflict of interest. PHD1/2 module and the functional consequences of its association This article is a PNAS Direct Submission. with chromatin remain unclear. 1To whom correspondence may be addressed. E-mail: Tatiana.Kutateladze@UCDenver. PHD fingers comprise one of the largest families of epigenetic edu. “ ” effectors capable of recognizing or reading posttranslationally This article contains supporting information online at www.pnas.org/lookup/suppl/ modified or unmodified histone tails, consequently recruiting var- doi:10.1073/pnas.1113655109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1113655109 PNAS ∣ January 17, 2012 ∣ vol. 109 ∣ no. 3 ∣ 787–792 Downloaded by guest on September 29, 2021 added stepwise confirmed that the two domains do not physically interact (Fig. S2A). Thus, in the absence of ligands, the two do- mains of PHD1/2 are flexibly linked and structurally independent. Histone-binding analysis reveals that each domain in PHD1/2 has distinct binding activity toward the N-terminal tail of H3, with PHD2 exhibiting a higher affinity than PHD1. We charac- terized the binding of PHD1/2 to histone tail peptides by NMR titration experiments (Fig. 1 E and F). 1H, 15N HSQC spec- tra of uniformly 15N-labeled PHD1/2 were collected while titrat- ing in a peptide corresponding to the first 12 residues of the H3 tail. Addition of increasing amounts of the peptide induced sub- stantial chemical shift changes in both of the linked modules. The pattern of chemical shift changes in PHD1/2 was similar to that observed in the separated PHD1 and PHD2 fingers mixed to- gether upon addition of the H3 peptide (Fig. 1E). This suggests that each PHD finger in the tandem PHD1/2 module binds an individual histone tail. The pattern of resonance perturbations also indicated two distinct binding events in PHD1/2, with one set of peaks shifting significantly upon the first addition of pep- tide and reaching saturation at a 1∶5 protein:peptide molar ratio and another set only beginning to shift at 1∶2. Careful examina- tion reveals that cross peaks associated with the first pattern of changes belong to the PHD2 finger and peaks associated with the second pattern belong to PHD1 (Fig. 1E). Thus, when linked, the PHD2 finger has a higher affinity for the H3 peptide than does the PHD1 finger. The recognition of individual histone tails by each domain in PHD1/2 was supported by mutational analysis. Point mutants of PHD1/2 were generated in which histone binding of either PHD1 or PHD2 was compromised. Titration experiments using PHD1/ 2-L381W (a PHD1 mutant) and the unmodified H3 peptide de- monstrated that the interaction of PHD1 with the peptide was abolished, whereas the interaction of PHD2 was unchanged (Fig. 1F). Similarly, mutation of the PHD2 finger (PHD1/2- C460W) led to significantly decreased binding of PHD2 to the H3 tail, whereas the interaction of PHD1 was not affected (Fig. 1F). Fig. 1. The tandem PHD fingers of CHD4 possess individual histone-binding To determine how the H3 tail peptide is recognized, we per- activities. (A) The CHD4/NuRD complex. (B) CHD4 contains two N-terminal formed NMR titration experiments using the unmodified H3 PHD fingers (orange), two chromodomains (blue) and a C-terminal ATPase peptide in which Ala1 is acetylated (H3A1ac) (Fig. S2B). The ab- module (purple). (C) Alignment of the CHD3-CHD5 PHD1/2 sequences. Resi- sence of chemical shift changes in PHD1/2 upon addition of dues that are strongly, moderately, and weakly conserved in PHD1 and PHD2 H3A1ac implied that binding of both PHD fingers is abrogated. fingers are highlighted in orange, blue, and yellow, respectively. (D) An over- We concluded that each PHD finger in the tandem module lay of the 1H,15N HSQC spectra of unbound PHD1/2 (black) and the individual 1 15 requires the N-terminal Ala1 residue for binding. PHD1 (red) and PHD2 (blue) fingers. (E) An overlay of H, N HSQC spectra of Together these results reveal a unique mode of histone binding the tandem PHD1/2 construct (Left) and PHD1 and PHD2 separately (Right)as increasing amounts of the histone H3 tail peptide (1–12) is titrated in. Spectra in which the two fingers of the tandem module are structurally are color coded according to the molar ratio of protein:peptide. Resonances independent and exhibit distinct binding activity, each recogniz- of the PHD1 domain are labeled in brown and those for the PHD2 domain are ing an individual histone H3 tail with PHD2 playing a dominant labeled in blue. (F) Overlays of 1H,15N HSQC spectra of PHD1/2 (Top Left) and role. This mechanism contrasts with the mechanisms for associa- mutants PHD1/2(L381W) (Top Middle), PHD1/2(C460W) (Top Right), PHD1/2 tion of the double PHD finger of DPF3 and the double chromo- (W481A) (Bottom Left), and PHD1/2 (W481A/E439A/E440A/E441A) (Bottom domain of CHD1 with a single histone tail (21, 22). The PHD Right) as the histone H3 tail peptide (1–12) is titrated in. fingers of DPF3 bind sequentially to a long stretch of the tail, with one finger recognizing the N terminus of H3 and another To establish whether the linked PHD fingers function together, interacting with the central region of the tail acetylated at we investigated their structures and abilities to bind 14, whereas the chromodomains of CHD1 form a single binding by NMR (Fig. 1 and Figs. S1 and S2). To examine the structural 1 15 pocket for the H3 tail methylated at lysine 4. Thus, CHD4 PHD1/ interrelation of the domains, the H, N HSQC spectra of 2 is an example in which two effector domains within the same PHD1/2 and the individual PHD1 and PHD2 fingers in the free protein are seen to have duplicate histone-binding function. state were collected and compared. An overlay of the spectra showed that the majority of cross peaks of the globular PHD CHD4 PHD1/2 Associates Bivalently with a Single Nucleosome. Our modules have similar positions, indicating no significant change data demonstrate that PHD1/2 binds two histone H3 tails, suggest- in the fold of these domains in the tandem construct (Fig. 1D). ing that PHD1/2 associates with nucleosomes in a multivalent Additional cross peaks, which belong to the linker region, were manner. Using NMR data and modeling we show that the inter- clustered primarily in the middle of the PHD1/2 spectrum, imply- domain orientation of PHD1/2 positions it to bivalently target a ing that the linker is largely unstructured. A comparable linewidth single nucleosome, concomitantly recognizing both of the histone (at half height) measured in the 1H,15N HSQC spectra suggested H3 tails. The histone-binding sites of PHD1/2 were determined by that the PHD fingers in the tandem module are rotating nearly analysis of chemical shift changes observed in the 1H,15NHSQC independently of each other. The lack of resonance perturbations spectra of PHD1/2 upon titration of the H3 peptide. Those resi- in the spectrum of 15N-labeled PHD1 when unlabeled PHD2 was dues that were perturbed the most (greater than the average shift

788 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1113655109 Musselman et al. Downloaded by guest on September 29, 2021 H3 tails of a single nucleosome, which protrude from the core at a distance of 70 Å in a parallel orientation (Fig. 2B). In agreement, few chemical shift perturbations were observed in the linker re- gion, suggesting that it is not involved in the interaction. Further- more, mutation of the conserved Trp residue and the conserved acidic (EEE) patch in the linker did not affect the bind- ing (Fig. 1F). We next examined the ability of CHD4 PHD1/2 to associate with nucleosomes using reconstituted tetrasomes (the H3/H4 tet- ramer wrapped by 80 bp of DNA). Our data reveal that PHD1/2 can bind to a single tetrasome in a bivalent manner (Fig. 3). To evaluate the strength of the association with the tetrasomal par- ticle, we measured binding affinities by tryptophan fluorescence. Analysis of the binding curves of PHD1/2 required a two site binding model, indicating that both PHD fingers are involved in the interaction with the tetrasome (Fig. 3A). This fitting yielded two Kd values of 0.33 μM and 0.96 μM, which in keeping with the NMR data correspond to PHD2 and PHD1, respectively. As expected, binding of the separate PHD fingers was well described by a single site model (Kds ¼ 3–4 μM, Fig. 3A). The three- and 11-fold increase in affinity of the linked PHD fingers as compared to binding affinities of the individual domains is likely entropi- cally driven, with both the PHD fingers and histone tails being prealigned for the interaction. Binding of the CHD4 PHD1/2 fingers to a single tetrasome was substantiated by gel shift assays. PHD1/2 was first incubated with the tetrasome in 0.2∶1, 1∶1 and 5∶1 molar ratios, and the reaction Fig. 2. The interdomain organization of PHD1/2 results in a perfect fit for mixtures were run on a nondenaturing acrylamide gel and stained binding to two histone H3 tails in a single nucleosome. (A) The histograms with ethidium bromide for detection. A shift of the tetrasome show normalized 1H, 15N chemical shift changes in backbone amides of band (approximately 102 kDa) upon formation of the complex the PHD1 (Left) and PHD2 (Right) domains of PHD1/2 upon addition of the with PHD1/2 (approximately 119 kDa) indicated that PHD1/2 H3 peptide as a function of residue. Shifts greater than the average plus one- binds to a single tetrasome but does not crosslink a pair of tetra- quarter (yellow), one-half (orange) and one (red) standard deviation are somes of the overall size of 221 kDa (Fig. 3B). No shift was highlighted and the residues labeled. These changes are mapped onto the surface of the unbound PHD1 and H3K9me3-bound PHD2 structures in B. observed in the case of free DNA, suggesting that PHD1/2 does (B) A model of the association of CHD4 PHD1/2 with the nucleosome. The not interact with the tetrasomal DNA. Together, the NMR, tryp- PHD1-H3 peptide complex, the linker and orientation with respect to the tophan fluorescence, and gel shift assays demonstrate that PHD1/ nucleosome were modeled. 2 can associate bivalently with a single nucleosome.

plus one-quarter standard deviation) were assumed to be PHD1/2 Association with H3 Is Modulated by Histone PTMs. Histones directly or indirectly involved in binding to H3 (Fig. 2A). Mapping undergo a number of posttranslational modifications (PTMs), these changes onto the structures of the individual PHD fingers particularly in their N-terminal tails. We found that (2L5U.pdb and 2L75.pdb) reveals well-defined binding pockets or acetylation of Lys9 strengthens binding of both domains in that are similar to those previously reported for the individual PHD1/2 to the histone H3 tail, whereas methylation of Lys4 PHD fingers (Fig. 2B). Association of PHD1/2 with the two pep- significantly weakens it. The association of the tandem construct tides was modeled using NMR data and the structures of unbound with several modified histone H3 tail peptides was tested by PHD1 and of PHD2 in complex with a histone peptide (16, 23). NMR and pull-down assays (Fig. 4 A and B). 1H, 15N HSQC spec- PHD1 and PHD2 were oriented in order for the C-terminal end of tra of PHD1/2 were recorded in the presence of increasing PHD1 to align with the N-terminal end of PHD2. Further, NMR amounts of H3K9ac, H3K9me3, H3K4me3, or H4K20me3 and data suggest that the 30-residue linker between the two domains is recognition of methylation marks on H3K4, H3K9, H3K27, largely unstructured, therefore the distance between the PHD fin- and H3K36 was examined by Western blot analysis. As compared 70 gers would be just over Å when the linker is fully extended. This to the unmodified H3 peptide, titration of H3K9ac or H3K9me3 BIOCHEMISTRY interdomain configuration could allow for binding to two histone into 15N-labeled PHD1/2 showed that both PHD fingers reach

Fig. 3. PHD1/2 associates bivalently with a single nu- cleosome. (A) A representative tryptophan fluores- cence curve and the two-site fit for PHD1/2 as the tetrasome particle is titrated in. The table shows Kd values of the linked PHD1/2 fingers and individual PHD1 and PHD2 for the tetrasome and the unmodi- fied H3 tail peptides, residues 1–12 (averaged over three experiments). (*) Taken from ref. 15. (B) The tet- rasome particle free and in the presence of increasing amounts of PHD1/2 was run on a nondenaturing ac- rylamide gel and detected by ethidium bromide staining. The tetrasome bands (Top) and the free DNA bands (Bottom) are shown. (C) A model of the concomitant interaction of both PHD modules with the two histone H3 tails.

Musselman et al. PNAS ∣ January 17, 2012 ∣ vol. 109 ∣ no. 3 ∣ 789 Downloaded by guest on September 29, 2021 saturation at lower concentration of the peptide, indicating that specifically disrupt pericentric heterochromatin structure, possi- binding is facilitated by an increase in the hydrophobicity of Lys9 bly through displacement of HP1γ. (Fig. 4A). However, the linked PHD fingers retained distinct Schizosaccharomyces pombe HP1 displays a 0.12 μM affinity binding modes, with PHD2 prevailing over PHD1. In contrast, for the H3K9me3-containing nucleosomes (24). We found that titration of H3K4me3 diminished binding of both modules. the linked PHD fingers of CHD4 bind to the unmodified H3 Moreover, the more methyl groups Lys4 contained, the weaker tails of the tetrasome with the Kds of 0.33 and 0.96 μM, and these this interaction became (Fig. 4B). A lack of chemical shift per- affinities are expected to increase significantly upon methylation turbations in either domain implied that PHD1/2 does not recog- of Lys9 (a 10-fold increase is seen with H3 peptides alone). These nize H4K20me3 and little to no association with methylated data support a model wherein the tandem PHD1/2 fingers of γ Lys27 or Lys36 was observed in pull-down assays. CHD4 compete with chromodomain (CD) of HP1 for H3K9me3- enriched nucleosomes, resulting in displacement of HP1γ γ CHD4 PHD1/2 Disrupts Pericentric Heterochromatin. NuRD has been (Fig. 4D). In contrast to CHD4 PHD1/2, HP1 associates with two shown to localize to heterochromatin and display a CHD4- spatially proximal nucleosomes through binding of its CD to dependent activity in heterochromatin maintenance and assem- H3K9me3 and dimerization via its chromoshadow domain (CSD) bly. To determine if the PHD fingers of CHD4 are necessary for (24), providing a mechanism for the formation and spreading of localization of CHD4 to pericentric heterochromatin, we gener- heterochromatin. The antagonistic action of CHD4 PHD1/2 would ated a GFP-fusion of PHD1/2 (GFP-CHD4-PHD1/2) (Fig. 4). lead to the disruption of pericentric heterochromatin assembly and the subsequent dispersion of H3K9me3 and HP1γ observed by HEK293T cells were transfected to express the fusion protein or þ light microscopy (Fig. 4C). Totest this idea, we generated the GFP- the GFP tag alone. GFP cells were sorted 48 h post transfection fusion mutants of CHD4 PHD1/2, in which histone binding of and the subcellular localization of GFP was examined (Fig. 4C). either PHD1 or PHD2 was compromised, and examined the im- As a control for the integrity of pericentric heterochromatin re- pact of expression of these mutants on pericentric heterochroma- gions, we assessed the localization pattern of H3K9me3 and tin. As shown in Fig. 4C, the L381W mutant (PHD1 is impaired) or HP1γ. As expected, in cells transfected with GFP alone we ob- γ the E447A and C460W mutants (PHD2 is impaired) were unable served the normal focal accumulation of H3K9me3 and HP1 . to disrupt the normal focal accumulation of H3K9me3 and HP1γ, However, in cells transfected with GFP-CHD4-PHD1/2, we ob- therefore supporting the notion that the concomitant engagement served diffuse nuclear localization of the fusion protein and loss of both PHD fingers of CHD4 is required to elicit this effect. of the normal focal accumulation of the pericentric heterochro- Lastly, we note that further studies are necessary to more fully ex- matin markers. Such a disruption of heterochromatin markers plore the interplay of PHD1/2 and HP1γ in the context of full by the PHD1/2 fingers of CHD4 could result from decreased length CHD4 and in the context of intact NuRD complex. levels of these markers or their redistribution within the nucleus. To differentiate between these possibilities, we performed Wes- PHD1/2 Histone-Binding Activity Is Crucial for Repressive Functions of tern blot analysis and examined global levels of H3K9me3 and CHD4/NuRD CRCs in Vivo. The role of the PHD1/2-histone interac- HP1γ (Fig. 4E). H3 acetylation and H4K20me3 were measured tions in transcriptional regulation by CHD4/NuRD was examined as controls. The absence of any changes in these markers 48 h by assessing the ability of wild-type and mutated CHD4 to repress after transfection suggested that the PHD1/2 fingers of CHD4 transcription of the mb-1 (Cd79a) gene in B cells. Display of

Fig. 4. Histone modifications modulate the interac- tion of PHD1/2. (A) Overlays of 1H, 15N HSQC spectra of PHD1/2 as peptides corresponding to unmodified H3, H3K9ac, H3K9me3, H3K4me3 or H4K20me3 (from left to right) are titrated in. Spectra are color coded according to the protein:peptide molar ratio. (B) Pull-down assays of GST-PHD1/2 with different singly modified histone tail peptides. (C) CHD4 PHD1/2 disrupts pericentric heterochromatin. Repre- sentative images of immunofluorescence performed in sorted GFP and GFP-CHD4 PHD1/2 (wild type and the indicated mutants) cells. Cells were spotted onto slides and stained for H3K9me3 and HP1γ. Scale bar, 10 μm. (D) A model of how intranucleosomal binding of CHD4 PHD1/2 can modulate the internucleosomal interaction of HP1γ.(E) Western blot analysis of H3K9me3, H4K20me3, HP1γ, and total H3 acetyla- tion 48 hr after transfection with either GFP or GFP- CHD4 PHD1/2. Total histone H3 (H3 general) was used as a loading control.

790 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1113655109 Musselman et al. Downloaded by guest on September 29, 2021 membrane-bound IgM (mIgM) on B cells requires transcription of the mb-1 gene, which encodes the Ig-α subunit of the B cell receptor (25). The display of mIgM on μM.2 murine plasmacy- toma cells is proportional to mb-1 transcript abundance (26). In μM.2 cells, mb-1 transcription is activated approximately 80-fold by the Early B cell factor 1 (EBF1) and Paired box protein 5 (Pax5) transcription factors but is restrained by CHD4/NuRD CRCs (27). In this context, depletion of endogenous CHD4 by shRNA increases activation of mb-1 promoters significantly by enhancing chromatin accessibility and demethylation of mb-1 promoter CpGs (27). In μM.2 cells expressing estrogen-dependent versions of EBF1 and Pax5 (EBF1:ER, Pax5:ER) and 3′ UTR-specific Chd4 shRNAs, highly efficient display of mIgM was detected due to activation of endogenous mb-1 transcription (Fig. 5A, Right, and Fig. 5B, Top, Second from Left). In this context, expression of exo- genous wild-type CHD4 attenuated mb-1 transcription as mea- sured by a significant reduction in mIgM (Fig. 5B, Top, Third from Left). To demonstrate that enzymatically active CHD4 is re- quired for this activity, we generated a point mutation (K757R) that inactivates the ATPase domain of CHD4 (28). This mutation is predicted to block nucleosome mobilization by CHD4. We also generated a mutation (R1034A) with the potential to interfere with DNA-binding function as predicted based on the crystal structure of SWI2/SNF2 (29). As expected, the K757R mutation abrogated reduction of IgM display by CHD4 in the presence of activated transcription factors (Fig. 5B, Top, Fourth from Left). However, the R1034A mutant retained wild type activity. This mutant demonstrates that mutation of CHD4 does not intrinsi- cally result in loss of repressive activity (Fig. 5B, Top Right). The add-back assay was used to assess the importance of recognition of histone H3 tails by the PHD1/2 fingers for CHD4 transcriptional repressive function in vivo. We introduced muta- tions in the PHD1 and PHD2 fingers of the full length protein that have been shown to significantly reduce binding of the indi- vidual PHD modules to histone H3 peptides (15, 16). Addition- ally, PHD finger deletion mutants (ΔD365-C411, ΔPHD1 and ΔD443-C490, ΔPHD2) of CHD4 were generated. Of the five PHD1 mutants tested, D368A, G377A and ΔPHD1 attenuated re- pressive activity of CHD4, while the L381W and W406A mutants did not result in significant effects (Fig. 5B, Second Row). How- ever, expression of the PHD2 finger mutants L458A, C460W, and ΔPHD2 completely abrogated CHD4 repressive function, as evi- denced by increased IgM display (Fig. 5B, Third Row). Statistical Fig. 5. PHD fingers of CHD4 are necessary for transcriptional repressive func- analysis of both mIgM mean fluorescence intensities (MFI) and tions in vivo. (A) Increased display of membrane-bound IgM (mIgM) in re- mb-1 transcript abundance confirmed that PHD1 finger mutants sponse to activated transcription factors (EBF1:ER, Pax5:ER) and depletion D368A, G377A, and ΔPHD1 and PHD2 finger mutants L458A, of CHD4 · μM:2 plasmacytoma cells do not express mIgM in the absence of C460W, and ΔPHD2 exhibit significantly reduced activity com- transcriptional activators (Left). When activated by 4-hydroxytamoxifen (4- pared to wild-type CHD4 (Fig. 5C and Fig. S4). As controls, Wes- OHT), the transcription factors EBF1 (as EBF1:ER) and Pax5 (as Pax5:ER) syner- tern blotting detected the expression of (3X)T7-tagged wild-type gistically activate endogenous Cd79a , which enables the assembly of μ mIgM on the B cell surface at low levels (Middle). shRNA-mediated knock-

and mutant CHD4 in transduced M.2 cells (Fig. 5D). These BIOCHEMISTRY results support two significant conclusions: (i) histone tail binding down of CHD4 greatly increases Cd79a transcription (shown in C) and the display of mIgM in response to EBF1 and Pax5 (Right). Empty retroviruses by CHD4 PHD fingers is necessary for transcriptional repression and luciferase shRNA were used as controls. (B) Effect of mutations in by the CHD4/NuRD complex in B cells, and (ii)althoughPHD1 PHD1 and PHD2 in CHD4 on mIgM expression in μM.2 cells. Expression of and PHD2 each contribute to the activity of CHD4, PHD2 plays a wild-type CHD4 restores repressive functions when endogenous CHD4 is de- more important role than PHD1, which is in agreement with mea- pleted by 3′UTR-specific shRNA. (Top Row, First Panel) Background mIgM in sured binding affinities. the presence of EBF1:ER and Pax5:ER. Depletion of CHD4 greatly enhances mIgM (Top Row, Second Panel). Add-back of wild type CHD4 (Top Row, Third Concluding Remarks Panel) or mutant CHD4 (Top Row, Fourth and Fifth Panels and Middle and In this study, we determined the biological function and a unique Bottom Rows). (C) Relative abundance of mb-1 transcripts from sorted cells binding mechanism of the tandem PHD1/2 fingers of the CHD4 in B. CHD4 R1034A (positive control) did not reduce CHD4 activity, while ATPase. We found that the PHD1/2 fingers target nucleosomes in CHD4 K757R (negative control) is inactive. All conditions have activated EBF1:ER and Pax5:ER except control (no activators). Asterisks indicate statis- a bivalent manner, concomitantly recognizing two histone H3 tical significance relative to wild type: **¼p < 0.001, ***¼p < 0.0001.(D) tails with high affinity. This interaction is further modulated by Western blot of CHD4 mutants. Whole cell extracts were prepared from histone modifications, with methylation or acetylation of Lys9 en- μM.2 cells transduced with CHD4 or control (empty vector) retroviruses hancing binding, and methylation of Lys4 or acetylation of Ala1 and analyzed by blotting with anti-T7 epitope antibodies. HDAC2 served abolishing it. Our functional data reveal that histone binding by as a nuclear protein loading control.

Musselman et al. PNAS ∣ January 17, 2012 ∣ vol. 109 ∣ no. 3 ∣ 791 Downloaded by guest on September 29, 2021 the CHD4 PHD1/2 fingers is required for transcriptional repres- salt-gradient dialysis method, starting in 300 mL of 10 mM Tris pH 7.5, 0.5 mM sive activity of the CHD4/NuRD complex and support the me- TCEP and 2M NaCl and adding 10 mM Tris pH 7.5 and 0.5 mM TCEP at chanism of multivalent engagement. 1.5 mL∕ min to reach a final NaCl concentration of 150 mM. The interdomain organization of CHD4 PHD1/2 allows for in- Cell Lines, Transfection, Retroviral Infection, and Flow Cytometry. The μM.2 and tranucleosomal interactions. Although the biochemical data do μ not preclude the ability of PHD1/2 to engage adjacent nucleo- M.2 EBF:ER stably transfected cell lines were cultured as described previously (27). Generation of retroviruses and infection of cells were performed as de- somes in trans, disruption of heterochromatin structure upon dis- scribed (26) with the following exceptions: For generation of CHD4 wild type placement of HP1γ suggests that the CHD4 PHD fingers may not μ μ γ and mutant retroviruses, 67.2 g retroviral plasmid DNA and 56 L Lipofec- be as effective as HP1 at internucleosomal bridging. Thus, the tamine™ 2000 (Invitrogen) were used to transfect 60–80% confluent 100- crosstalk between linked effector domains may be essential not mm dishes of ΦNX cells. Resulting retroviral supernatants were concentrated only for recognition of a set of posttranslational histone modifi- using Retro-Concentin™ (System Biosciences). Precipitated viral pellets were cations but also for distinguishing a particular chromatin land- resuspended in μM.2 growth media to 1∕100 of the original volume of super- scape as defined by internucleosomal spacing. It is also of note natant. All viral transductions were performed in 100-mm dishes. Eight ml of that rather than defining the binding interaction, histone modi- retroviral supernatant was used for transductions of shRNA and control ret- fications appear to fine-tune the affinity of CHD4 for nucleo- roviruses, while the equivalent of 24 mL retroviral supernatant (concentrated retrovirus) was used for transductions of CHD4 wild type and mutant retro- somes in a functionally relevant way. The ability to recognize 5 unmodified histone H3 tails is critical in CHD4 function in tran- viruses. Eight mL of cells (6.25 × 10 cells∕mL) and 18.8 μg∕mL polybrene (Sig- ma-Aldrich) was used for all transductions. ER fusion proteins (EBF1:ER and scriptional regulation, whereas the additional affinity imparted by μ Lys9 methylation is important in heterochromatin targeting. Pax5:ER) were induced using 0.5 M 4-OHT (Sigma-Aldrich) 48 hrs after retro- viral transduction. Flow cytometry was performed 5 d after retroviral trans- Materials and Methods duction. For flow cytometry, APC-eFluor® 780- or Pacific Blue-conjugated anti- Protein Purification, PCR Mutagenesis, NMR Spectroscopy, Fluorescence Spec- CD90.1 was obtained from eBioscience and Cy5-conjugated anti-IgM from troscopy, Gel Shift Assays, Cell Culture and Western Analysis, Immunofluores- Jackson Research Laboratories. Staining of cells was performed as previously cence, Pull-Down Assays, RNA Isolation, and qPCR. See SI Materials and described and detected using a CyAn™ flow cytometer (Dako). Live cells were Methods. gated mCherry fluorescent proteinþ (Chd4 shRNA), CD90.1-PBþ (exogenous Chd4) and YFPþ (Pax5:ER). Data was analyzed using FloJo™ software. Tetrasome Reconstitution. The tetrasome was reconstituted as previously re- ported (30). Briefly, histones H3 and H4 (with the modifications C110A and T71C) were expressed in Escherichia. coli Rosetta2(DE3) pLysS cells grown in ACKNOWLEDGMENTS. We thank F. Davrazou, J. Scorgie, and B. Hirsch for help 2XTY media. The histones were solubilized, purified over ion exchange col- with the experiments and discussions; G. Blobel for the CHD4 cDNA plasmid; umns, and lyophilized. An 80-bp segment corresponding to a portion of the and K. Luger for the histone H3 and H4 plasmids. This research is supported by grants from the National Institutes of Health (NIH) (GM096863 and Widom 601 sequence was ordered from Integrated DNA Technologies. The CA113472) and the Cancer League of Colorado to T.G.K. and NIH Grants single-stranded DNA was purified by ethanol precipitation before annealing. R01 AI054661 and AI081878 to J.H. This work is also supported in part by The histones were refolded into tetramer form into 10 mM Tris pH 7.5, 1 mM the Intramural Research Program of the National Institute of Environmental β EDTA, 5 mM -mercaptoethanol and 2M NaCl following the protocol in ref. 31 Health Sciences, NIH (Z01ES101765 to P.W.). J.R. was supported by a generous and further purified over a sephacryl S-100 column (without EDTA). The pur- grant from the Rocky Mountain Chapter of the Arthritis Foundation and NIH ified tetramer was then mixed at a 1∶1 molar ratio with the DNA at approxi- Postdoctoral Training Grant T32 AI07405. C.A.M. is an NIH National Research mately 0.1 mg∕mL and reconstituted into tetrasome form using a continuous Service Award postdoctoral fellow (NHLBI, F32HL096399).

1. Tong JK, Hassig CA, Schnitzler GR, Kingston RE, Schreiber SL (1998) Chromatin deace- 17. Musselman CA, Kutateladze TG (2009) PHD fingers: Epigenetic effectors and potential tylation by an ATP-dependent nucleosome remodelling complex. Nature 395:917–921. drug targets. Mol Interv 9:314–323. 2. Wade PA, Jones PL, Vermaak D, Wolffe AP (1998) A multiple subunit Mi-2 histone dea- 18. Musselman CA, Kutateladze TG (2011) Handpicking epigenetic marks with PHD fin- cetylase from Xenopus laevis cofractionates with an associated Snf2 superfamily AT- gers. Nucleic Acids Res, in press. – Pase. Curr Biol 8:843 846. 19. Wang Z, Patel DJ (2011) Combinatorial readout of dual histone modifications by 3. Xue Y, et al. (1998) NURD, a novel complex with both ATP-dependent chromatin-re- paired chromatin-associated modules. J Biol Chem 286:18363–18368. – modeling and histone deacetylase activities. Mol Cell 2:851 861. 20. Ruthenburg AJ, et al. (2011) Recognition of a mononucleosomal histone modification 4. Zhang Y, LeRoy G, Seelig HP, Lane WS, Reinberg D (1998) The dermatomyositis-specific pattern by BPTF via multivalent interactions. Cell 145:692–706. autoantigen Mi2 is a component of a complex containing histone deacetylase and 21. Zeng L, et al. (2010) Mechanism and regulation of acetylated histone binding by the nucleosome remodeling activities. Cell 95:279–289. tandem PHD finger of DPF3b. Nature 466:258–262. 5. Denslow SA, Wade PA (2007) The human Mi-2/NuRD complex and gene regulation. 22. Flanagan JF, et al. (2005) Double chromodomains cooperate to recognize the methy- Oncogene 26:5433–5438. – 6. Wang HB, Zhang Y (2001) Mi2, an auto-antigen for dermatomyositis, is an ATP-depen- lated histone H3 tail. Nature 438:1181 1185. dent nucleosome remodeling factor. Nucleic Acids Res 29:2517–2521. 23. Kwan AHY, et al. (2003) Engineering a protein scaffold from a PHD finger. Structure – 7. Smeenk G, et al. (2010) The NuRD chromatin-remodeling complex regulates signaling 11:803 813. and repair of DNA damage. J Cell Biol 190:741–749. 24. Canzio D, et al. (2011) Chromodomain-mediated oligomerization of HP1 suggests a 8. Larsen DH, et al. (2010) The chromatin-remodeling factor CHD4 coordinates signaling nucleosome-bridging mechanism for heterochromatin assembly. Mol Cell 41:67–81. and repair after DNA damage. J Cell Biol 190:731–740. 25. Hombach J, Tsubata T, Leclercq L, Stappert H, Reth M (1990) Molecular components of 9. Hall JA, Georgel PT (2007) CHD proteins: A diverse family with strong ties. Biochem Cell the B-cell antigen receptor complex of the IgM class. Nature 343:760–762. Biol 85:463–476. 26. Maier H, et al. (2003) Requirements for selective recruitment of Ets proteins and acti- 10. Jacobs SA, Khorasanizadeh S (2002) Structure of HP1 chromodomain bound to a lysine vation of mb-1/Ig-alpha gene transcription by Pax-5 (BSAP). Nucleic Acids Res 9-methylated histone H3 tail. Science 295:2080–2083. 31:5483–5489. 11. Nielsen PR, et al. (2002) Structure of the HP1 chromodomain bound to histone H3 27. Gao H, et al. (2009) Opposing effects of SWI/SNF and Mi-2/NuRD chromatin remodel- – methylated at lysine 9. Nature 416:103 107. ing complexes on epigenetic reprogramming by EBF and Pax5. Proc Natl Acad Sci USA 12. Fischle W, et al. (2003) Molecular basis for the discrimination of repressive methyl-ly- 106:11258–11263. sine marks in histone H3 by Polycomb and HP1 chromodomains. Genes Dev 28. Linder B, et al. (2007) CHD4/Mi-2beta activity is required for the positioning of the 17:1870–1881. mesoderm/neuroectoderm boundary in Xenopus. Genes Dev 21:973–983. 13. Min J, Zhang Y, Xu RM (2003) Structural basis for specific binding of Polycomb chro- 29. Durr H, Korner C, Muller M, Hickmann V, Hopfner KP (2005) X-ray structures of the modomain to histone H3 methylated at Lys 27. Genes Dev 17:1823–1828. 14. Bouazoune K, et al. (2002) The dMi-2 chromodomains are DNA binding modules im- Sulfolobus solfataricus SWI2/SNF2 ATPase core and its complex with DNA. Cell – portant for ATP-dependent nucleosome mobilization. EMBO J 21:2430–2440. 121:363 373. 15. Musselman CA, et al. (2009) Binding of the CHD4 PHD2 finger to histone H3 is modu- 30. Donham DC, 2nd, Scorgie JK, Churchill ME (2011) The activity of the histone chaperone lated by covalent modifications. Biochem J 423:179–187. yeast Asf1 in the assembly and disassembly of histone H3/H4-DNA complexes. Nucleic 16. Mansfield RE, et al. (2011) Plant homeodomain (PHD) fingers of CHD4 are histone H3- Acids Res 39:5449–5458. binding modules with preference for unmodified H3K4 and methylated H3K9. J Biol 31. Dyer PN, et al. (2004) Reconstitution of nucleosome core particles from recombinant Chem 286:11779–11791. histones and DNA. Methods Enzymol 375:23–44.

792 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1113655109 Musselman et al. Downloaded by guest on September 29, 2021