|Home | About | Journals | Submit | Contact Us | Français|
ATP-dependent chromatin remodeling by the CHD family of proteins plays an important role in the regulation of gene transcription. Here we report that full-length CHD8 interacts directly with β-catenin and that CHD8 is also recruited specifically to the promoter regions of several β-catenin-responsive genes. Our results indicate that CHD8 negatively regulates β-catenin-targeted gene expression, since short hairpin RNA against CHD8 results in the activation of several β-catenin target genes. This regulation is also conserved through evolution; RNA interference against kismet, the apparent Drosophila ortholog of CHD8, results in a similar activation of β-catenin target genes. We also report the first demonstration of chromatin remodeling activity for a member of the CHD6-9 family of proteins, suggesting that CHD8 functions in transcription through the ATP-dependent modulation of chromatin structure.
The alteration of chromatin structure provides a key regulatory step for all processes that act upon DNA (41). The factors that regulate this structure, commonly referred to as chromatin remodeling enzymes, can be grouped into two broad categories: complexes that alter chromatin structure via the covalent modification of histones (25, 42, 83) and complexes that use the energy of ATP hydrolysis to alter the structure or position of the nucleosome (6, 47, 57, 67).
ATP-dependent chromatin remodeling enzymes modulate the contacts between histones and DNA. In vitro, these enzymes catalyze structural changes that allow factors to access nucleosomal DNA, reposition nucleosomes on a template, transfer histone octamers to donor DNA, and replace histones with histone variants (27, 43, 80). In vivo, these activities are crucial for transcription, replication, repair, and recombination of the eukaryotic genome (2, 21, 59, 65). These remodeling enzymes can be divided into numerous families based on domain architecture. One such family is the CHD (chromodomain, helicase, DNA binding) group of proteins, which are critical regulators of chromatin structure (23, 26, 29, 49). These enzymes are characterized by tandem chromodomains N-terminal to their catalytic Snf2 helicase domain.
The CHD family can be further subdivided into three subfamilies: CHD1-2, CHD3-5, and CHD6-9. While the first two subfamilies have been extensively studied, very little is known about the CHD6-9 family (29, 49). Previous studies have indicated that CHD8 may regulate the Wnt signaling pathway, since an N-terminal fragment of CHD8 was previously identified as a protein in Rattus norvegicus that binds β-catenin both in vivo and in vitro (61). This N-terminal fragment, termed Duplin, contains only the chromodomains and lacks the Snf2 helicase domain and C-terminal sequences. Overexpression of this N-terminal fragment results in inhibition of Tcf4-dependent transcription, and studies of Xenopus embryos demonstrated that this fragment inhibited axis formation and β-catenin-mediated axis duplication (61).
The “canonical” Wnt signaling pathway functions by controlling the soluble pool of β-catenin (11). In the absence of Wnt ligand, nonanchored β-catenin is bound by the APC complex. Within this complex, phosphorylation by glycogen synthase kinase 3β targets β-catenin for degradation by the proteasome (33, 35, 50). Wnt signaling results in the inhibition of glycogen synthase kinase 3β and allows for nonphosphorylated β-catenin to accumulate and enter the nucleus. Once in the nucleus, β-catenin binds to Tcf transcriptional enhancers and activates transcription (7, 53).
While the mechanism of transcriptional activation by β-catenin is not completely understood, it is evident that its regulation involves reconfiguring the chromatin structure (28, 78). β-Catenin has been shown to interact with several proteins that can serve to “open” the chromatin structure. These include p300/CBP (31, 52, 71), BRG1 (3), CARM1 (39), GRIP1 (44), pontin52/TIP49 (5, 24, 64), MLL1/2, p400, Snf2H, and TRRAP (64). BRG1, Snf2H, and p400 are of particular interest in that they are all members of the Snf2 family of ATP-dependent chromatin remodeling enzymes, suggesting that ATP-dependent chromatin remodeling plays a fundamental role in the regulation of β-catenin-mediated transcription. The association of Duplin (the N-terminal fragment of CHD8) with β-catenin suggests that additional chromatin remodeling factors may be required. The identification and characterization of these factors will be necessary in order to understand the role of ATP-dependent remodeling during transcriptional activation by β-catenin.
Here we report that a member of the CHD family of chromatin remodelers, CHD8, interacts directly with β-catenin. Using chromatin immunoprecipitation (ChIP) techniques, we demonstrate that CHD8 is also recruited specifically to the promoter regions of several β-catenin-responsive genes. To gain further insight into the importance of this association in the regulation of β-catenin-targeted genes, short hairpin RNA (shRNA) against CHD8 was utilized. Our results demonstrate that CHD8 can negatively regulate β-catenin-targeted gene expression. RNA interference (RNAi) against kismet, the Drosophila melanogaster ortholog of CHD8, similarly results in activation of β-catenin target genes, demonstrating that this regulation is conserved through metazoan evolution. Although CHD8 is predicted to be an ATP-dependent chromatin remodeler, evidence demonstrating chromatin remodeling activity for CHD8, or any members of the CHD6-9 family of proteins, is lacking. Here we report the first demonstration of chromatin remodeling activity for CHD8. Taken together, these results suggest that CHD8 functions in transcriptional regulation through the modulation of chromatin structure.
HeLa and HCT116 cells were grown in Dulbecco's modified Eagle medium (Invitrogen) supplemented with 10% fetal bovine serum (HyClone) and 1× penicillin-streptomycin-glutamine (Invitrogen) at 37°C under 5% CO2. HeLa Flag-βcat cells were grown as described above with 5 μg/ml puromycin. Drosophila S2 cells were grown in Schneider's Drosophila medium (Invitrogen) supplemented with 10% fetal bovine serum and 1× penicillin-streptomycin-glutamine at 24°C. SF9 cells were grown in 1× Grace's insect medium (Invitrogen) supplemented with 10% fetal bovine serum and 1× penicillin-streptomycin-glutamine at 24°C. Anti-acetyl-histone H4 antibodies (06-866) and anti-trimethyl-histone H3 Lys4 antibodies (07-473) were purchased from Upstate (Millipore). Rabbit polyclonal antibodies against CHD8 were generated against a 20-amino-acid peptide (HTETVFNRVLPGPIAPESSK) conjugated to keyhole limpet hemocyanin (Open Biosystems). Antibodies were affinity purified by using this peptide conjugated to Affi-Gel 10 (Bio-Rad). Oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA). Primer sequences are available at http://bochar.biochem.med.umich.edu/supdata.html. HeLa cells for the preparation of nuclear extracts were obtained from the National Cell Culture Center (Minneapolis, MN).
Constructs for the expression of glutathione S-transferase (GST)-β-catenin fusion proteins were a kind gift from K. A. Jones (75) or were cloned by PCR using full-length GST-β-catenin as a template. Constructs for the expression of GST-WDR5 were a kind gift from R. C. Trievel (14). GST and GST fusion proteins were expressed in Escherichia coli BL21. Cells were harvested, suspended in 150 mM KCl in buffer A (20 mM Tris-HCl [pH 7.9], 0.2 mM EDTA, 10 mM β-mercaptoethanol, 10% glycerol, 0.2 mM phenylmethylsulfonyl fluoride [PMSF]), and lysed by two passages through a French pressure cell. The cell lysates were cleared by centrifugation (105,000 × g for 60 min at 4°C) prior to use. The GST-β-catenin N-terminal fragment required additional chromatography on DEAE and butyl Sepharose (GE Healthcare) to remove partial fusion products. The concentration of GST or GST fusion proteins in the cell lysates was determined by Coomassie staining of sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) samples purified by affinity chromatography on glutathione-Sepharose (GE Healthcare).
Recombinant baculoviruses containing Flag-tagged human CHD8 or human Snf2H were created using the Bac-N-Blue baculovirus expression system (Invitrogen). For protein expression, SF9 cells (1 × 106/ml) were infected with the various recombinant viruses (multiplicity of infection, 2) and grown for 4 days. Cells were harvested, washed with phosphate-buffered saline (PBS), suspended in 500 mM KCl-1% NP-40 in buffer IP (20 mM Tris-HCl [pH 7.9], 0.2 mM EDTA, 10% glycerol, 0.2 mM PMSF) with 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin, and lysed with a Dounce homogenizer. The cell lysates were cleared by centrifugation (15,806 × g for 15 min at 4°C). Samples were dialyzed against 50 mM KCl in buffer IP and were inverted overnight at 4°C with 500 μl of agarose beads conjugated with anti-Flag antibody M2 (Sigma). Samples were washed sequentially with 10 column volumes (each) of 150 mM KCl in buffer IP, 350 mM KCl in buffer IP, and 150 mM KCl in buffer IP. Samples were eluted with 400 μg/ml Flag peptide (Sigma) and 150 mM KCl in buffer A with 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin.
For in vivo studies of interaction between β-catenin and CHD8, nuclear extracts were prepared from ~1 × 107 HeLa cells stably expressing Flag epitope-tagged β-catenin or from the parental cell line. Nuclear extracts were prepared using the method of Dignam et al. (16). Samples were dialyzed against 50 mM KCl in buffer IP and were incubated overnight at 4°C with 20 μl of anti-Flag antibody M2-conjugated agarose beads (Sigma). Samples were washed sequentially with 1 ml of 150 mM KCl in buffer IP, 350 mM KCl in buffer IP, and 150 mM KCl in buffer IP. Samples were eluted with SDS-loading buffer and were then subjected to SDS-PAGE and Western blot analysis.
For in vivo studies of interaction between WDR5 and CHD8, HEK293 cells were transfected with a construct encoding Flag epitope-tagged WDR5 or the parental vector by using Lipofectamine 2000 as described by the manufacturer (Invitrogen). Cells were washed twice with PBS, and extracts were prepared by lysis with buffer R (150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl [pH 7.4], 1 mM EDTA, 0.2 mM PMSF). Samples were incubated overnight at 4°C with 20 μl of anti-Flag antibody M2-conjugated agarose beads (Sigma). Samples were washed with 1 ml of buffer R, eluted with SDS-loading buffer, and subjected to SDS-PAGE and Western blot analysis.
For studies of interaction between recombinant CHD8 and β-catenin or WDR5, cell lysates containing 10 μg of the indicated GST fusion protein were incubated with 20 μl of glutathione-Sepharose for at least 3 h at 4°C. Samples were washed for 10 min twice with 1 ml of 150 mM KCl and 0.2% NP-40 in buffer A. Samples were suspended in 500 μl of 150 mM KCl and 0.2% NP-40 in buffer A. After the addition of 1 μg of recombinant CHD8, samples were incubated for at least 3 h at 4°C. Samples were washed three times, for 10 min each time, with 1 ml of 0.2% NP-40 and 350 mM KCl (β-catenin) or 150 mM KCl (WDR5) in buffer A. Samples were eluted with SDS-loading buffer and subjected to SDS-PAGE and Western blot analysis.
ChIP experiments were performed essentially as described in the ChIP assay kit (Upstate). Briefly, ~1 × 106 cells per immunoprecipitation were fixed with 2.5% formaldehyde for 10 min at 37°C. Cells were washed with PBS and lysed in ChIP lysis buffer. The lysate was cleared by centrifugation, and the chromatin was sheared by sonication (~200- to 1,000-bp fragments). Samples were precleared with protein A agarose (Repligen) blocked with salmon sperm DNA. The indicated antibodies or protein A-purified normal rabbit serum was added and incubated overnight at 4°C. Antibody-chromatin complexes were precipitated by incubation with protein A agarose blocked with salmon sperm DNA. Samples were washed and eluted from the resin according to the instructions except that wash incubations were 30 min each.
Reactions for the determination of ATPase activity contained 10 nM enzyme, 1 mM ATP, 7.5 μCi of [γ-32P]ATP in 50 mM NaCl, 0.5 mM dithiothreitol, 5 mM MgCl2, and 25 mM Tris (pH 7.9). When indicated, plasmid DNA or nucleosomes purified from HeLa cells were added to a final concentration of 5 ng/μl. Reaction mixtures were incubated 1 h at 30°C. Reaction products (1 μl) were spotted onto polyethyleneimine-cellulose thin-layer chromatography plates (Sigma) and resolved with 0.5 M LiCl in 1 M formic acid. Dried plates were imaged and quantified on a Typhoon Trio+ imager using ImageQuant TL software (GE Healthcare).
Accessibility assays were based on the method of Smith and Peterson (68) with the following changes. Fluorescently labeled 277-bp DNA templates were generated using standard PCR techniques with pGEM3z-601 (45) as a template and primers 601 forward and 601 reverse. A 0.1/0.9 ratio of the fluorescent primer (5′-Alexa Fluor 488-N-hydroxysuccinimide ester) to the nonfluorescent primer was used. Mononucleosomes were reconstituted by salt dialysis with 277-bp end-labeled DNA and HeLa core histones as described elsewhere (46) except that RB high buffer contained 2 M NaCl-1 mM EDTA-0.2 mM PMSF in 10 mM Tris (pH 8.0) and RB low buffer contained 1 mM EDTA-0.2 mM PMSF in 10 mM Tris (pH 8.0). Remodeling buffer contained 3 mM MgCl2-50 mM NaCl-2 mM dithiothreitol-0.1 mg/ml bovine serum albumin in 20 mM HEPES (pH 8.0). Reaction mixtures contained 1 mM ATP or adenylyl imidodiphosphate (AMPPNP); 1× enzyme is equal to 9 nM. Deproteinated samples were resolved on a 3% agarose-0.5× Tris-borate-EDTA gel. Wet gels were imaged and quantified on a Typhoon Trio+ imager using ImageQuant TL software (GE Healthcare).
Nucleosome mobilization assays employed fluorescently labeled 277-bp DNA templates generated by using standard PCR techniques and pGEM3z-601 as described above. Experimental templates without a nucleosome positioning sequence were generated by PCR utilizing primers 601 downstream forward and 601 downstream reverse. Two control templates were also produced to serve as marker mononucleosomes with a central or end position. These were generated by PCR utilizing primers 601 forward and 601 reverse or primers 601 slid forward and 601 slid reverse, respectively. Mononucleosomes were reconstituted by salt dialysis with 277-bp end-labeled DNA and HeLa core histones as described above. Accessibility assays were conducted as described above without the addition of the restriction enzyme. Reactions were terminated by addition to 1× of a 6× stop solution containing 10 mM Tris (pH 7.8), 1 mM EDTA, 334 μg/ml salmon sperm DNA, and 334 μg/ml HeLa nucleosomes in 30% glycerol. Native samples were resolved on a 5% 37.5:1 acrylamide-0.2× Tris-borate-EDTA gel. Wet gels were imaged and quantified on a Typhoon Trio+ imager using ImageQuant TL software (GE Healthcare).
HeLa nuclear extracts were prepared using the method of Dignam et al. (16). Fractionations were performed in buffer A with the indicated KCl concentration. Size exclusion chromatography was performed in buffer A with 350 mM KCl. Resins and columns were obtained from the following sources: P11 phosphocellulose from Whatman, DEAE-FF from Sigma, and Superose 6 HR 10/30 from GE Healthcare. CHD8 was affinity purified using ~10 mg of material obtained from partial fractionation of HeLa nuclear extracts by the P11 and DEAE chromatographic steps. Input material was precleared with 250 μl protein A agarose (Repligen). Anti-CHD8 antibodies or normal rabbit immunoglobulin G (IgG) (~660 μg) was cross-linked to 500 μl protein A agarose (Repligen) by standard techniques (30). Precleared inputs were incubated with the antibody-protein A beads at 4°C overnight in buffer IP containing 150 mM KCl, 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin. Samples were washed extensively with 10 column volumes of buffer IP containing 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin under the following conditions: two washes with 150 mM KCl, two washes with 150 mM KCl plus 1% NP-40, two washes with 150 mM KCl, four washes with 1 M KCl, two washes with 150 mM KCl plus 200 mM guanidine hydrochloride, and two washes with 150 mM KCl. Samples were eluted with 100 mM glycine (pH 3.0) and neutralized with a 1/10 volume of 1 M Tris (pH 7.9). Samples were then subjected to SDS-PAGE. The bands identified were then subjected to in-gel trypsin digestion and tandem mass spectrometry (MS-MS) analyses at the Michigan Proteome Consortium at the University of Michigan.
For preparation of cDNA, total RNA was isolated from the indicated cells using the RNeasy kit (Qiagen) according to the manufacturer's instructions. Reverse transcriptase (RT) reactions employed random decamers (Ambion) and Superscript II (Invitrogen) according to the manufacturers' protocols. Real-time quantitative PCR analysis employed the indicated primers, the iQ Sybr green Supermix (Bio-Rad), and a MyiQ single-color real-time PCR detection system (Bio-Rad). All real-time PCRs were performed in triplicate. For RNA analysis, threshold cycle values were normalized to the levels of polymerase III (Pol III)-transcribed H1 RNA (human) or Pol II-transcribed α-tubulin (Drosophila). For ChIP experiments, DNA levels were expressed relative to input levels.
RNAi experiments with HCT116 cells utilized the UI2-puro SIBR shRNA vectors (12). CHD8 knockdown experiments employed two shRNA cassettes, 493 and 6410. UI2-GFP-SIBR Luc-1601 was used as a control (12). The indicated constructs (10 μg) were transfected into HCT116 cells in 10-cm-diameter dishes using Lipofectamine 2000 (Invitrogen) as described by the manufacturer. Twenty-four hours posttransfection, cells were selected with 5 μg/ml puromycin and grown for an additional 36 h before being harvested. Drosophila RNAi experiments were performed essentially as described elsewhere (79) with a total of 12 μg of the indicated double-stranded RNA. Cells were harvested 4 days after the addition of the double-stranded RNA. Control templates were prepared by using primers to β-lactamase with pBSIISK as a template.
The ATP-dependent chromatin remodeling activity of the CHD1 and CHD3 families of proteins has long been established (29, 49). However, it is not known whether members of the CHD6-9 family of proteins possess chromatin remodeling activity. Based on their placement by phylogenetic analysis, the CHD6-9 family of proteins are most likely ATP-dependent chromatin remodeling factors (data not shown). Indeed, DNA-stimulated ATPase activity for CHD6 and for CHD9/CReMM has recently been demonstrated (48, 63). To gain insight into the function of CHD8, and by inference to the rest of the CHD6-9 family of proteins, assays to address the activity of CHD8 were performed.
For enzymatic assays, a recombinant baculovirus that encodes an N-terminal Flag epitope-tagged version of CHD8 was generated (Fig. (Fig.1A).1A). After purification, recombinant CHD8 was tested for ATPase activity and stimulation of ATPase activity by free DNA or DNA incorporated into nucleosomes. As with other ATP-dependent remodelers, CHD8 possesses an ATPase activity that is stimulated by nucleosomal DNA (Fig. (Fig.1B).1B). This activity is comparable to that observed with the ATP-dependent chromatin remodeling enzyme Snf2H (Fig. (Fig.1B),1B), a known nucleosome-stimulated ATPase (10). Located in the Snf2 helicase domain is an ATP binding site with the conserved sequence GXGKT (77). Mutation of the lysine in this sequence in other ATP-dependent remodeling enzymes results in complete abrogation of ATP hydrolysis and, therefore, of chromatin-remodeling activity (13, 20, 36, 74). The ATPase activity of CHD8 requires the Snf2 helicase domain, as evidenced by the fact that mutation of the conserved GXGKT lysine to arginine (Fig. (Fig.1A,1A, K842R) severely impairs the ATPase activity of CHD8 utilizing nucleosomal templates (Fig. (Fig.1C)1C) and naked DNA templates (data not shown).
To directly test for ATP-dependent remodeling, recombinant CHD8 was assayed using “restriction enzyme accessibility” (48, 63). This assay is based on the fact that reconstitution of DNA into a nucleosomal array impedes its digestion by restriction enzymes, and ATP-dependent chromatin remodeling factors can increase the access of restriction enzymes to the nucleosomal DNA. Mononucleosomes were reconstituted by salt dialysis with a 277-bp end-labeled DNA containing the 601 nucleosome positioning sequence with a unique HhaI site located near the dyad axis (45). As shown in Fig. Fig.2A,2A, lanes 1 to 3, addition of recombinant CHD8 greatly enhances the ability of the restriction enzyme to digest the nucleosomal template. To confirm the requirement for ATP hydrolysis in this reaction, two experiments were performed. First, the ATPase-deficient form of CHD8, CHD8(K842R), was tested for chromatin remodeling activity. As predicted, this mutation severely impairs the access of HhaI to the nucleosomal DNA (Fig. (Fig.2A,2A, lane 6). Second, this reaction also requires the hydrolysis of ATP; the nonhydrolyzable ATP analogue AMPPNP is unable to substitute for ATP (Fig. (Fig.2A,2A, lanes 4 and 5). A time course for access to the nucleosomal DNA was also performed. As shown in Fig. Fig.2B,2B, addition of recombinant CHD8 results in an initial rapid increase in HhaI access to the nucleosomal DNA. After this initial phase, the increase in access appeared linear over time. This activity is similar to the reported increased HhaI access to mononucleosomes catalyzed by yeast Swi/Snf (68).
To gain further insight into mechanisms of nucleosomal mobilization by CHD8, nucleosome sliding assays were performed (22). This type of assay is based on the fact that the position of a mononucleosome on DNA affects its electrophoretic mobility on native polyacrylamide gels. A nucleosome with a central position will migrate more slowly than a nucleosome positioned toward either end. To perform these experiments, three nucleosomal templates were generated and reconstituted into mononucleosomes by salt dialysis. The first two serve as marker mononucleosomes with a central or end position (Fig. (Fig.2C,2C, lanes 1 and 2). These were constructed by placing the 601 nucleosome positioning sequence at a central or an end position. The experimental template lacks the 601 nucleosome positioning sequence and therefore results in random positioning of mononucleosomes on the template (Fig. (Fig.2C,2C, lane 3). As shown in Fig. Fig.2C,2C, lanes 4 and 5, addition of recombinant CHD8 with ATP results in a clear mobilization of the nucleosome on the template. This reaction requires the hydrolysis of ATP, as evidenced by the fact that the nonhydrolyzable ATP analogue AMPPNP is unable to substitute for ATP (Fig. (Fig.2C,2C, lane 6).
After mobilization of the nucleosomes, two predominant species are seen. The minor species has a mobility similar to that of an end-positioned nucleosome (Fig. (Fig.2C;2C; compare lane 5 to lane 2). The major species has a mobility consistent with a more centrally positioned nucleosome (Fig. (Fig.2C;2C; compare lane 5 to lane 1). However, its migration is slightly accelerated, suggesting that this species may be intermediate between an end and a central nucleosome. These results are consistent with those of studies using both human CHD1 and CHD3, which tend to position nucleosomes toward the center of the template (60). Indeed, these results are remarkably comparable to those for CHD3 in that the predominant slid species was a centrally positioned nucleosome and a minor species was an end-located nucleosome. These results suggest similar modes of remodeling for members of the CHD3-5 and CHD6-9 families of proteins. Taken together, the data in Fig. Fig.11 and and22 clearly demonstrate that CHD8 is a bona fide remodeling enzyme capable of altering the nucleosomal structure in an ATP-dependent manner.
Most ATP-dependent chromatin remodeling factors isolated to date exist in high-molecular-weight complexes composed of multiple subunits. It is the association with other polypeptides that may serve to target, regulate, or modify the specificity of the complex. To investigate whether CHD8 functions as part of a multisubunit complex, partial purification of CHD8 from HeLa cells was performed. The nuclear extract was fractionated using phosphocellulose chromatography (P11), followed by Western blot detection with anti-CHD8 antibodies. The majority of CHD8 was found to fractionate into the 0.5 M KCl eluate (Fig. (Fig.3A).3A). Further separation of this fraction was accomplished using DEAE-Sephacel and Superose 6 chromatography (Fig. (Fig.3A).3A). The elution profile from the size exclusion column (Superose 6) is consistent with CHD8 being a component of a complex of ~900 kDa. Because this molecular mass is much greater than that predicted for a globular CHD8 monomer (~290 kDa), it suggests that CHD8 is contained in a complex with other subunits that may regulate the activity of the complex.
To explore the possibility that CHD8 is in a multisubunit complex and to identify any associated polypeptides, an affinity purification strategy was employed. HeLa nuclear extract was chromatographed over several columns, as in Fig. Fig.3A,3A, to enrich for CHD8 complexes. This material was then subjected to affinity purification using affinity-purified anti-CHD8 antibodies coupled to protein A agarose. As a control, this material was also subjected to affinity purification using normal rabbit IgG coupled in an identical manner. After glycine elution and SDS-PAGE, numerous polypeptides were visualized (Fig. (Fig.3B).3B). The predominant polypeptides were then subjected to tryptic digestion and MS-MS identification. The ~290-kDa polypeptide was confirmed to be CHD8 (Fig. (Fig.3B),3B), demonstrating the specificity of the purification. Numerous other polypeptides were also identified (data not shown). Among the polypeptides identified were components of the MLL and CoREST chromatin-modifying complexes and components of the SWI/SNF and NuRD ATP-dependent remodeling complexes, as well as components of the U2 and U5 snRNPs. Since Fig. Fig.3A3A demonstrates a predicted molecular mass for the predominant complex of ~900 kDa, the presence of this numerous list of polypeptides would suggest that the affinity purification has resulted in the isolation of several distinct complexes, all of which contain CHD8. Further experiments, including conventional separation of these potential complexes, need to be performed to address this possibility.
To further characterize the CHD8 complex(es), several polypeptides identified as described above were chosen for further characterization. Initially, we focused on WDR5, because a previous purification of WDR5 identified CHD8 as a component of an MLL histone methyltransferase complex (18), possibly linking chromatin remodeling and chromatin modification in a single complex. However, no peptide sequences for MLL were detected in the affinity-purified CHD8 complex, and extensive Western blotting of various purified fractions failed to detect MLL (data not shown). This suggests that WDR5 may be present outside of the MLL complex. Indeed, recent reports indicate that RbBP5, WDR5, and ASH2L form a subcomplex of the MLL methyltransferase complex and that these components can coexist in the absence of MLL (17, 70).
To verify the association of CHD8 and WDR5, coimmunoprecipitation experiments were performed. These experiments utilized an epitope-tagged version of WDR5, Flag-WDR5, transfected into HEK293 cells. Extracts were prepared from these transfected cells or from cells transfected with the control vector, and Flag-WDR5 was affinity purified with anti-Flag M2 antibodies. After extensive washing, bound proteins were detected by Western blot analysis using affinity-purified anti-CHD8 antibodies. Figure Figure3C3C shows that anti-Flag antibodies immunoprecipitate CHD8 along with Flag-WDR5 from extracts of Flag-WDR5 cells but not from extracts of control cells. This experiment confirms the association of WDR5 in vivo with endogenous CHD8.
A direct association of CHD8 and WDR5 was then tested by using recombinant proteins in vitro. Purified GST-WDR5 fusion proteins bound to glutathione-Sepharose were incubated with purified recombinant CHD8. After extensive washing, bound proteins were detected by Western blot analysis using the affinity-purified anti-CHD8 antibodies. Figure Figure3D3D further validates the in vivo association of WDR5 with CHD8 and also demonstrates that CHD8 interacts directly with WDR5. Since WDR5 is thought to bridge MLL to the WDR5-RbBP5-ASH2L core complex, these results also suggest that this interaction of WDR5 with CHD8 or MLL may be mutually exclusive.
The experiments described above demonstrate that CHD8 is indeed an ATP-dependent remodeling factor and also that CHD8 may function in a complex to remodel chromatin. However, information regarding the in vivo function of CHD8 is lacking. Previous studies have demonstrated that Duplin, an N-terminal fragment of rat CHD8, binds β-catenin in vivo and in vitro (61). However, it is not known whether full-length CHD8 also interacts with β-catenin. To test the association of CHD8 and β-catenin, experiments were performed using recombinant proteins in vitro. These experiments utilized purified recombinant CHD8 and purified GST fused to full-length β-catenin or various fragments thereof (Fig. (Fig.4B).4B). Purified GST-β-catenin fusion proteins bound to glutathione-Sepharose were incubated with purified recombinant CHD8. After washing, bound proteins were detected by Western blot analysis using affinity-purified anti-CHD8 antibodies (Fig. (Fig.4A).4A). While CHD8 was unable to interact with GST-fused N-terminal or C-terminal fragments of β-catenin, CHD8 was able to interact directly with full-length β-catenin, an N-terminal deletion of β-catenin, and β-catenin lacking both the N-terminal and C-terminal domains. These three constructs have in common the presence of the central 12 armadillo repeats of β-catenin. Armadillo repeats are composed of ~42 amino acids, and tandem armadillo repeats form a superhelix of helices proposed to mediate the interaction of β-catenin with its binding partners (32). These data demonstrate that CHD8 interacts directly with β-catenin in vitro, that the interaction is mediated through the central armadillo repeats of β-catenin, and that these central repeats are necessary and sufficient for this interaction.
To determine whether CHD8 and β-catenin associate in cells, a stable HeLa cell line that expresses an epitope-tagged version of activated β-catenin, Flag-β-cateninactive, was created. This activated β-catenin harbors four alanine substitutions in the glycogen synthase kinase 3β recognition site that result in the stabilization and nuclear localization of β-catenin (1, 58, 81). Extracts were prepared from this cell line, and Flag-β-cateninactive was immunoprecipitated with anti-Flag M2 antibodies. Figure Figure4C4C shows that anti-Flag antibodies immunoprecipitate CHD8 along with Flag-β-cateninactive from extracts of the Flag-β-cateninactive cell line, but not from extracts of the parental cell line. These results demonstrate that CHD8 interacts with β-catenin in cells and that the interaction is a direct association of the two proteins.
The experiments described above suggest that full-length CHD8, like the Duplin fragment, may also function in the regulation of β-catenin-mediated transcription, since the interaction of CHD8 with β-catenin could serve to target the chromatin remodeling activity of CHD8 to β-catenin-responsive promoters. To test for a direct association of CHD8 with endogenous β-catenin-responsive promoters, ChIP experiments were performed. Three reported β-catenin-responsive genes, Axin2, Dkk1, and Nkd2, were chosen for analysis (http://www.stanford.edu/~rnusse/pathways/targetcomp.html). These experiments were performed with the colorectal carcinoma cell line HCT116, because this cell line has an activated Wnt signaling pathway (40, 54). Primer pairs were designed to both the 5′ ends (proximal promoter regions) and the encoded 3′ untranslated regions (UTR) of the indicated target genes. Figure Figure5A5A demonstrates that CHD8 is bound to the 5′ ends of the Axin2, Dkk1, and Nkd2 genes and not to the 3′ ends of these genes. In addition, CHD8 was not seen bound to either the 5′ or the 3′ end of the control estrogen-responsive gene PS2. ChIPs were also performed with antibodies to acetyl H4 and trimethyllysine 4 of histone H3, two general marks of active chromatin (8). No correlation between the degree of CHD8 binding and the extent of acetylation of histone H4 was seen at any of the analyzed regions. However, the binding of CHD8 seemed to correlate with the extent of trimethylation of histone H3 lysine 4. Since this methylation mark is enriched at promoter regions surrounding the transcriptional start site (4, 9, 37, 62), these data suggest that CHD8 may bind exclusively to the proximal promoter regions of active genes.
To further investigate the binding of CHD8 to the 5′ ends of genes, additional ChIP experiments were performed on the Axin2 gene. Primer pairs were designed along the Axin2 gene from the proximal promoter region (−1050) to the encoded 3′ UTR (+29519). The experimental results shown in Fig. Fig.5B5B demonstrate the specificity of CHD8 for the promoter, and possibly the 5′ coding region, but not for the remaining coding sequence or the 3′ UTR sequence. This result demonstrates that CHD8 is specifically recruited to the proximal promoter regions of these β-catenin-responsive genes. Taken together with the demonstration of chromatin remodeling activity for CHD8 and the fact that CHD8 interacts directly with β-catenin, these results strongly suggest a role for CHD8 chromatin remodeling in the regulation of β-catenin-mediated transcription.
As stated above, previous work has demonstrated that an N-terminal fragment of Rattus norvegicus CHD8 binds to β-catenin and that expression of this fragment can inhibit β-catenin-mediated transcription from a Tcf-responsive luciferase reporter system (61). As demonstrated in Fig. Fig.4,4, full-length CHD8 can also interact with β-catenin in vivo and in vitro. To address whether CHD8 can modulate the transcription of β-catenin target genes, an shRNA strategy was employed to deplete endogenous CHD8, and several endogenous β-catenin target genes were chosen for analysis. Axin2, Dkk1, and Nkd2 were chosen because in HCT116 cells, CHD8 is bound to the 5′ end of each of these target genes (Fig. (Fig.5).5). HCT116 cells were transfected with control shRNA constructs or with shRNA constructs targeting CHD8 that coexpress a puromycin resistance marker. After selection of transfected cells with puromycin, cDNA was isolated for analysis by quantitative PCR. Given the association of CHD8 with β-catenin and the localization of CHD8 to several endogenous β-catenin targets in a colorectal cell line harboring an activated Wnt signaling pathway, it was predicted that depletion of CHD8 would have a negative impact on transcription. Surprisingly, depletion of CHD8 resulted in a modest but reproducible induction of all three target genes (Fig. (Fig.6A).6A). These results demonstrate that CHD8 does indeed participate in the regulation of endogenous β-catenin target genes but that the normal function of CHD8 is to repress or negatively regulate the transcription of these target genes.
The regulation of human target genes may be complicated by the fact that four paralogs of CHD8 exist in mammalian genomes. To address this concern and to further validate the results described above, a Drosophila S2 cell culture system was utilized, because Kismet is the only homolog of CHD8 in Drosophila. To activate the Wnt signaling pathway, RNAi to axin was utilized. Axin functions along with glycogen synthase kinase 3β and APC in the degradation of β-catenin in the unstimulated state. Therefore, loss of Axin mimics Wnt signaling by promoting the stabilization of β-catenin. nkd, like its mammalian counterpart, is induced upon Wnt stimulation and is upregulated in S2 cells following treatment with axin RNAi (55); therefore, it was chosen as the target for analysis.
As shown in Fig. Fig.6C,6C, treatment with axin RNAi results in approximately 10-fold stimulation of nkd transcription. Consistent with the results for HCT116 cells, treatment of S2 cells with axin and kismet RNAi results in a further fivefold stimulation of transcription. It is noteworthy that treatment with kismet RNAi alone, i.e., in the absence of Wnt signaling, also results in fivefold stimulation over transcription with the control RNAi, suggesting that Kismet also functions in the uninduced state. These results further validate the conclusion that CHD8 functions in the negative regulation of β-catenin target genes.
The CHD6-9 family of proteins has only recently been identified, and very little is known about the function of these proteins (29, 49). The first report concerning CHD8, however, was focused not on full-length CHD8 but on an N-terminal fragment of rat CHD8 termed Duplin. This fragment of CHD8, which lacks the catalytic Snf2 helicase domain and part of the second chromodomain, was found to interact directly with β-catenin. Characterization of this fragment led to the conclusion that Duplin is a negative regulator of the Wnt signaling pathway and functions by blocking the binding of β-catenin to TCF (61). However, further studies also indicated that Duplin could attenuate the Wnt signaling pathway downstream of β-catenin target genes (38). Bioinformatic analysis suggests that Duplin arose from an improper splicing event, and no evidence for the existence of a human counterpart can be found (data not shown).
This information on Duplin suggested that full-length CHD8 may regulate β-catenin target genes. To investigate this possibility, we first examined the binding of full-length CHD8 to β-catenin both in vivo and in vitro. We found that, like Duplin, full-length CHD8 can interact directly with β-catenin, further suggesting that CHD8 may also play a role in the regulation of β-catenin target genes. To strengthen this hypothesis, we investigated the ability of CHD8 to bind β-catenin target genes in vivo. Using ChIP experiments, we demonstrate that CHD8 is localized to the 5′ ends and not the 3′ ends of several β-catenin-responsive genes. We also show that this binding is specific for the promoter region and possibly sequences immediately downstream of the transcriptional start site but is not localized to the coding region of the gene.
To directly test our hypothesis that CHD8 regulates β-catenin target genes, we chose to utilize shRNA techniques to deplete CHD8 and then to examine the transcription of endogenous targets identified by ChIP analysis. We found that CHD8 does indeed function in the regulation of endogenous β-catenin target genes, since depletion of CHD8 results in an increase in the expression of the β-catenin target genes tested. To verify these results in an independent system, we chose to investigate the Drosophila CHD8 ortholog Kismet. This system also circumvents the complication of the three other CHD8 paralogs possibly functioning at these loci. As discovered with CHD8 in human cells, depletion of kismet by RNAi results in an increase in the transcription of the Wnt target gene nkd.
Kismet was originally identified as an extragenic suppressor of Polycomb and was therefore suggested to be a member of the trxG of activators (15). Further studies examining the distribution of Kismet on polytene chromosomes suggested a more general role in transcriptional regulation. Kismet staining primarily overlaps that of RNA Pol II. Also, kismet mutant larvae display reduced levels of elongating polymerase without affecting Pol II recruitment, as determined solely by immunostaining for the phosphorylation status of the carboxy-terminal tail of the large subunit of Pol II and by the loss of immunostaining for SPT6 and CHD1, factors associated with elongating polymerase (69). These data would suggest that Kismet assists an early step in transcriptional elongation and would therefore be a positive regulator of transcription. By analogy, CHD8 should also act to facilitate transcription. Our results suggest that while Kismet may affect the overall levels of elongating polymerase, as determined by immunostaining, CHD8 and Kismet may act conversely at specific target genes, such as the Wnt-responsive genes, and obstruct optimal elongation.
Given the predicted function for Kismet described above, it is noteworthy that our purification of human CHD8 identifies spliceosome components that overlap with a set of CHD1-interacting proteins (66). These are components of both the U2 and the U5 snRNPs. Knockdown of CHD1 in HeLa cells reduces the association of spliceosomal components with transcribed regions and also reduces the efficiency of splicing of these genes (66). This may suggest that, like CHD1, CHD8 may also play a role in regulating the recruitment of spliceosomal components to actively transcribed genes. This function would, however, be upstream of CHD1, since kismet mutant larvae display a failure to recruit CHD1 to polytene chromosomes (69). Experiments to address the roles of CHD1 and CHD8, and the interplay between them, are required to advance our understanding of chromatin, transcriptional elongation, and splicing.
A recent study has identified CHD8 as a factor that associates with the zinc finger transcription factor hStaf/ZNF143 (82). hStaf/ZNF143 is one of two human homologues of Xenopus Staf, a transcriptional activator of snRNA genes (56). CHD8 was identified as a factor in HeLa nuclear extracts that can bind to recombinant hStaf and was also shown by ChIP to bind to both Pol II and Pol III snRNA promoters as well as the hStaf-responsive IRF3 promoter (82). Depletion of CHD8 by small interfering RNA resulted in a modest decrease in transcription at these promoters. These data are possibly consistent with the studies on Kismet as a positive regulator of transcriptional elongation. Again, taken with our results, this suggests that CHD8 can act conversely on different sets of genes and that this specificity may be controlled by the interaction with the underlying transcription factor.
An alternate explanation for the discrepancy between the previous studies on CHD8 and Kismet and our data on CHD8 could lie in the identification of CHD8 as a CTCF binding protein and the finding that CHD8 is required for CTCF-dependent insulator activity (34). CTCF has many roles in transcriptional regulation beyond insulator activity, including transcriptional repression (19). While CTCF has not been identified as a regulator of β-catenin-responsive genes, it is possible that CTCF can act in concert with CHD8 as a repressor at β-catenin-responsive genes, or even as an insulator to protect β-catenin-responsive promoters from proximal enhancer activities. Our affinity purification of CHD8 and MS-MS identification of associated proteins, however, did not identify CTCF in association with CHD8, suggesting that the majority of CHD8 may exist outside of a functional CTCF complex. Further studies to address the possible role of CTCF in β-catenin-mediated gene transcription will need to be performed.
As previously mentioned, CHD8 is closely related to CHD6, CHD7, and CHD9. Recent reports on both CHD7 and CHD9 show functional association with various nuclear hormone receptors. CHD9/CReMM/PRIC320 has been shown to interact specifically with peroxisome proliferator-activated receptor α (PPARα), PPARγ, RXR, ERα, CAR, and GR, and CHD9 also functions as a coactivator for PPARα reporter constructs (51, 72). CHD7 has recently been identified as a component of a corepressor complex that inactivates PPARγ-mediated transcription (73). It has yet to be determined whether CHD8 also interacts with various nuclear hormone receptors performing a functional role in nuclear hormone signaling, and if so, to what extent CHD8 functions as a coactivator or a corepressor.
The identification of CHD8 in a high-molecular-mass complex of ~900 kDa suggests that CHD8 may reside in a multisubunit complex, like members of the CHD3 family (29, 49). These additional subunits may prove to be important for the proper function and localization of CHD8. The identification and characterization of these associated factors will play an important role in understanding the physiological role of CHD8. Previously, CHD8 has been identified as a component of a WDR5-containing complex (18). This complex also contains the histone methyltransferase MLL1, suggesting a possible interplay between the activity of CHD8 and covalent histone modifications. Since this initial purification of CHD8 and the MLL-WDR5 complex was achieved via Flag-WDR5 purification, it has yet to be determined whether CHD8 is a component of this larger complex or whether it represents a new, as yet unidentified WDR5-CHD8 complex. Indeed, our purification could not identify MLL as a component of the CHD8 complex. While future experiments will have to be performed to address the molecular composition of this complex, our current functional characterization of CHD8 at the biochemical and cellular levels provides the required foundation for these studies.
An important question regarding the mechanism for regulation by CHD8 is centered on the prediction that CHD8 is indeed a chromatin remodeling enzyme. While previous reports have demonstrated that members of the CHD6-9 family do indeed possess ATPase activity (48, 63), here we report the first demonstration of bona fide chromatin remodeling activity for a member of the CHD6-9 family. This would then suggest that CHD8 indeed regulates transcription through the mobilization of chromatin structure. Given the similarity of the Snf2 helicase domains of members of this family (26), these data suggest that all members of this subfamily should also share this activity. This result thus has important implications for the study of CHD7. Loss of CHD7 function via mutation leads to CHARGE syndrome in humans. CHARGE syndrome is characterized by various developmental defects, including growth retardation, ear malformations, and genital defects, among other symptoms (76). By characterizing the activities and functions of the CHD6-9 family of enzymes, we can begin to develop a mechanistic link between chromatin structure and human disease.
We thank K. A. Jones, R. C. Trievel, and J. Widom for useful reagents and D. L. Turner and A. B. Vojtek for reagents and advice concerning the SIBR expression vectors. We especially thank R. Shiekhattar for many useful discussions and advice during the initial stages of this project. We also thank the K. M. Cadigan and E. R. Fearon labs for helpful discussions, and especially D. S. Parker for assistance with RNAi in S2 cells. We thank R. P. Kwok, R. C. Trievel, M. D. Uhler, and members of the Bochar lab, especially J. Yates, for critical reading of the manuscript.
Proteomics data were provided by the Michigan Proteome Consortium (www.proteomeconsortium.org), which is supported in part by funds from The Michigan Life Sciences Corridor. This work was funded in part by the National Institutes of Health through the University of Michigan's Cancer Center support grant (5 P30 CA46592), a Porter Physiology Fellowship to B.A.T., and an American Cancer Society Research Scholar grant (RSG-07-189-01) to D.A.B.
Published ahead of print on 31 March 2008.