Histone H1 Recruitment by CHD8 Is Essential for Suppression of the Wnt–␤-Catenin Signaling Pathway

Masaaki Nishiyama,a,b Arthur I. Skoultchi,c and Keiichi I. Nakayamaa,b Department of Molecular and Cellular Biology, Medical Institute of Bioregulation, Kyushu University, Fukuoka, Japana; CREST, Japan Science and Technology Agency, Saitama, Japanb; and Department of Cell Biology, Albert Einstein College of Medicine, Bronx, NY, USAc

Members of the chromodomain helicase DNA-binding (CHD) family of proteins are thought to regulate gene expression. Among mammalian CHD proteins, CHD8 was originally isolated as a negative regulator of the Wnt–␤-catenin signaling pathway that binds directly to ␤-catenin and suppresses its transactivation activity. The mechanism by which CHD8 inhibits ␤-catenin-dependent transcription has been unclear, however. Here we show that CHD8 promotes the association of ␤-catenin and histone H1, with formation of the trimeric complex on chromatin being required for inhibition of ␤-catenin-dependent transactivation. A CHD8 mutant that lacks the histone H1 binding domain did not show such inhibitory activity, indicating that histone H1 recruitment is essential for the inhibitory effect of CHD8. Furthermore, either depletion of histone H1 or expression of a dominant negative mutant of this protein resulted in enhancement of the response to Wnt signaling. These observations reveal a new mode of regulation of the Wnt signaling pathway by CHD8, which counteracts ␤-catenin function through recruitment of histone H1 to Wnt target genes. Given that CHD8 is expressed predominantly during embryogenesis, it may thus contribute to setting a threshold for responsiveness to Wnt signaling that operates in a development-dependent manner.

he Wnt–␤-catenin signaling pathway plays key roles in devel- contain histone deacetylases and function as transcriptional re- Topment, specification of cell fate, and adult stem cell prolifer- pressors (47). Mutations in CHD7 result in CHARGE syndrome, ation (2, 21, 27, 44). Abnormal activation of this pathway is asso- a multiple-malformation syndrome in humans for which more ciated with various human cancers, including colorectal cancer than 40 alleles have been defined (42). Chd7ϩ/Ϫ mice recapitulate and head and neck squamous cell carcinoma. Many genes associ- several aspects of this human disease, including inner-ear vestib- ated with tumor growth, including those for c-Myc, matrix met- ular dysfunction (13). Molecular studies suggest that CHD7 con- alloproteinases, and cyclin D1, have been identified as Wnt target tributes to transcriptional activation of tissue-specific genes dur- genes, with ␤-catenin functioning as a transcriptional coactivator ing differentiation (38). However, most of the biological functions at such genes (11, 16, 22, 35). In the absence of ␤-catenin binding, mediated by members of the CHD family remain to be elucidated. the promoters of Wnt target genes are occupied by Tcf/Lef family CHD8 exists in two isoforms, short (CHD8S) and long proteins, the corepressor Groucho (also known as TLE1), and (CHD8L), that are generated as a result of alternative mRNA splic- histone deacetylase 1 (3, 5, 9, 34). Wnt signaling results in the ing. CHD8S (also known as duplin) was originally isolated as a accumulation of ␤-catenin in the cytosol and nucleus, and nuclear negative regulator of the Wnt–␤-catenin signaling pathway (36) ␤-catenin displaces Groucho from Tcf by binding Tcf at the pro- and binds directly to the Armadillo repeats of ␤-catenin (36, 41). moters of Wnt target genes. ␤-Catenin then recruits chromatin- The COOH-terminal region of full-length CHD8 (CHD8L) inter- remodeling complexes or other transcriptional coactivators to acts with the insulator-binding protein CTCF, with this interac- stimulate gene transcription (2, 12, 21, 27, 43, 44). Several mem- tion being important for insulator activity (14). CHD8 has also bers of the Snf2 superfamily of ATP-dependent chromatin- been implicated as a positive or negative transcriptional regulator remodeling enzymes, including BRG1 (1), Snf2H, and p400 (39), of various genes (19, 32, 33, 45, 46). We previously showed that were recently found to be recruited by ␤-catenin in the induction Chd8Ϫ/Ϫ mice die early during embryogenesis, manifesting wide- of target gene transcription. Although these findings suggest that spread apoptosis (28), whereas additional deletion of the tumor ATP-dependent chromatin remodeling plays a fundamental role suppressor gene p53 ameliorated this developmental arrest (29). in the regulation of ␤-catenin-dependent transcription, it remains Both isoforms of CHD8 bind to p53 and suppress its transactiva- unclear how such enzymes contribute to the regulation of chro- tion activity by recruiting histone H1, with formation of a p53- matin at Wnt target genes. CHD8-histone H1 trimeric complex on chromatin being required Members of the chromodomain helicase DNA-binding for inhibition of p53-dependent transactivation and apoptosis. (CHD) family of proteins also belong to the Snf2 superfamily of These observations led us to examine whether such histone H1 ATP-dependent chromatin remodelers. Among the nine mam- recruitment by CHD8 might also contribute to the regulation of malian members of this family, CHD1 is thought to play an important role in gene transcription. The tandem chromodomains of human CHD1 thus specifically recognize and bind to the tri- Received 9 October 2011 Accepted 6 November 2011 methylated form of lysine 4 of histone H3 (H3K4me3), a hallmark Published ahead of print 14 November 2011 of actively transcribed chromatin (7), and mediate the recruit- Address correspondence to Keiichi I. Nakayama, [email protected]. ment of transcriptional initiation and pre-mRNA splicing factors Copyright © 2012, American Society for Microbiology. All Rights Reserved. (40). In mammals, CHD3 and CHD4 are subunits of nucleosome- doi:10.1128/MCB.06409-11 remodeling and histone deacetylase (NURD) complexes, which

0270-7306/12/$12.00 Molecular and Cellular Biology p. 501–512 mcb.asm.org 501 Nishiyama et al.

FIG 1 CHD8 interacts with and inhibits transactivation by ␤-catenin. (A) Whole-cell lysates of HEK293T cells were subjected to immunoprecipitation (IP) with mouse antibodies to ␤-catenin or with control mouse IgG. The resulting precipitates, as well as 1% of the original cell lysates (input), were subjected to immunoblot analysis (IB) with antibodies to CHD8 or to ␤-catenin. (B and C) HEK293T cells were incubated in the absence (Ϫ) or presence (ϩ) of LiCl (20 mM) (B) or Wnt3a (20 ng/ml) (C) for 4 h and were then subjected to immunoprecipitation with rabbit antibodies to CHD8 or with control rabbit IgG. The resulting precipitates, as well as 1% of the original cell lysates (input), were subjected to immunoblot analysis with antibodies to ␤-catenin, to CHD8, or to Hsp90 (loading and negative control). (D and E) HeLa cells infected with retroviruses encoding FLAG-tagged forms of CHD8S or CHD8L (or with the empty retrovirus) were incubated in the absence (Ϫ) or presence of LiCl or Wnt3a for 4 h. The cells were then subjected to ChIP with antibodies to FLAG, and the precipitated DNA was quantitated by real-time PCR analysis with primers speciﬁc for Wnt-responsive elements (WRE) or a control region (CR) of the axin2 (D) or c-Myc (E) gene promoter or for the p27 gene promoter. The cells were also subjected to immunoblot analysis with antibodies to CHD8 or to Hsp90 (D). (F and G) HeLa cells infected with a retrovirus for CHD8S or the empty vector were incubated with LiCl or Wnt3a for 4 h and were then subjected to RT and real-time PCR analysis of axin2 (F) and c-Myc (G) mRNAs. All quantitative data are means Ϯ standard deviations (SD) from three independent experiments. other genes such as Wnt target genes. We now show that CHD8 were from Sigma (St. Louis, MO); those to the hemagglutinin (HA) mediates the recruitment of histone H1 to Wnt target genes, re- epitope (HA.11) were from Covance (Princeton, NJ); those to Myc (9E10) sulting in suppression of the expression of these genes induced by were from Roche (Indianapolis, IN); those to glutathione S-transferase activation of Wnt–␤-catenin signaling. Our results thus suggest (GST) were from MBL (Nagoya, Japan); those to histone H1 were from that CHD8 may target multiple transcriptional activators, includ- Abcam (Cambridge, United Kingdom); and those to CHD8 were gener- ing ␤-catenin and p53, for suppression of transcriptional activity ated in rabbits by injection with a recombinant fragment of mouse CHD8 through recruitment of histone H1. (residues 484 to 670). Recombinant human ␤-catenin was obtained from Millipore (Billerica, MA), and recombinant mouse Wnt3a was obtained MATERIALS AND METHODS from R&D Systems (Minneapolis, MN). Alexa Fluor 488-conjugated goat Antibodies and reagents. Antibodies to ␤-catenin and to Hsp90 were antibodies to mouse immunoglobulin G (IgG) were obtained from Mo- obtained from BD Biosciences (San Jose, CA); those to FLAG (M2 or M5) lecular Probes-Invitrogen (Carlsbad, CA).

502 mcb.asm.org Molecular and Cellular Biology Histone H1 Recruitment by CHD8 Inhibits Wnt Signaling

␤ FIG 2 -Catenin mediates the association of CHD8 with chromatin. (A) HeLa cells stably overexpressing CHD8S or infected with the empty vector were incubated in the absence or presence of LiCl (20 mM) or Wnt3a (20 ng/ml) for 4 h. The cells were then subjected to ChIP with antibodies to ␤-catenin, and the precipitated DNA was quantitated by real-time PCR analysis with primers speciﬁc for Wnt-responsive elements or a control region of the axin2 gene promoter or for the p27 gene promoter. (B) HeLa cells infected with retroviruses encoding shRNAs speciﬁc for ␤-catenin mRNA or for EGFP mRNA (control) were incubated in the absence or presence of LiCl or Wnt3a for 4 h, after which the cells were lysed and subjected to immunoblot analysis with antibodies to ␤-catenin, to CHD8, or to Hsp90 (loading control). (C) HeLa cells infected with retroviruses for ␤-catenin or EGFP (control) shRNAs were incubated in the absence or presence of LiCl or Wnt3a for 4 h and were then subjected to ChIP analysis with antibodies to CHD8 and with PCR primers as in panel A. All quantitative data are means Ϯ SD from three independent experiments.

Plasmids. Complementary DNAs encoding wild-type or mutant subcloned into pCGN. His6-tagged proteins were expressed in Escherichia forms of mouse CHD8S tagged with the FLAG or Myc epitope at the NH2 coli strain BL21(DE3)pLys(S) (Merck, Darmstadt, Germany). terminus were subcloned into pcDNA3 (Invitrogen) with the use of the Cell culture, immunoprecipitation, and immunoblot analysis.

Gateway vector conversion system (Invitrogen); those encoding wild-type HEK293T and HeLa cells were cultured under an atmosphere of 5% CO2 or mutant forms of mouse histone H1c with three copies of the FLAG at 37°C in Dulbecco’s modiﬁed Eagle’s medium (DMEM) (Wako, Osaka, epitope at the NH2 terminus were also subcloned into pcDNA3. A cDNA Japan) supplemented with 10% fetal bovine serum (Invitrogen). NIH 3T3 ␤ for mouse -catenin tagged at its NH2 terminus with the HA epitope was cells were cultured under the same conditions in DMEM supplemented

January 2012 Volume 32 Number 2 mcb.asm.org 503 Nishiyama et al.

504 mcb.asm.org Molecular and Cellular Biology Histone H1 Recruitment by CHD8 Inhibits Wnt Signaling

with 10% bovine serum (Invitrogen). Embryonic stem (ES) cells were TGATAA-3= and 5=-CGGCCCCGAAATCCATCGCTCTGA-3= for cultured as described previously (25, 26). In some experiments, HeLa or hAxin-2 (Wnt-responsive element [WRE]), 5=-CTGGAGCCGGCTGCG NIH 3T3 cells were cultured in the presence of LiCl (20 mM) or Wnt3a (20 CTTTGATAA-3= and 5=-TGGCCCCGAAATCCATCGCGAACGG-3= for ng/ml) for 4 h and ES cells were incubated with LiCl (2.5 mM) or Wnt3a mAxin-2 (WRE), 5=-CTGGCTTTGGTGAACTGTTG-3= and 5=-AGTTG (2.5 ng/ml) for 4 h. Cell lysis, immunoprecipitation, and immunoblot CTCACAGCCAAGACA-3= for hAxin-2 (control region [CR]), 5=-ATCT analysis were performed as described previously (15). GTATGTCCTGTCTGCCAGCG-3= and 5=-TGTCCTGATTCCCAAGTC Retrovirus expression system. Complementary DNAs encoding AAGCAC-3= for mAxin-2 (CR), 5=-GCGGGTTACATACAGTGCACTTC wild-type or mutant forms of mouse CHD8S, human CHD8L, or mouse A-3= and 5=-TGGAAATGCGGTCATGCACAAA-3= for hc-Myc (WRE), histone H1c tagged with the FLAG epitope at the NH2 terminus were 5=-CTACCCTCTCAACGACAGCA-3= and 5=-AGAGCAGAGAATCCG subcloned into pMX-puro (kindly provided by T. Kitamura, University of AGGAC-3= for hc-Myc (CR), and 5=-CCGCCGCCGCAACCAATGGA Tokyo, Japan) with the use of the Gateway vector conversion system. The T-3= and 5=-GGAGTCGCAGAGCCGTGAGCA-3= for hp27. resulting vectors were used to transfect Plat E packaging cells and thereby to generate recombinant retroviruses (23). NIH 3T3 cells as well as HeLa RESULTS cells stably expressing the mouse ecotropic retrovirus receptor (mCAT-1) Both forms of CHD8 interact with ␤-catenin on chromatin in a were infected with retroviruses produced by Plat E cells and were then Wnt signaling-dependent manner. Forced expression of CHD8 cultured in the presence of puromycin (Sigma) at 10 ␮g/ml. S or CHD8L in mammalian cells revealed that both isoforms asso- RNA interference. The pMX-puro II vector was constructed by dele- ␤ tion of the U3 portion of the 3= long terminal repeat of pMX-puro. The ciated with -catenin (36, 41). We examined whether endogenous ␤ mouse U6 gene promoter, followed by DNA corresponding to a short CHD8S and CHD8L also interact with endogenous -catenin in hairpin RNA (shRNA) sequence, was subcloned into the NotI and XhoI HEK293T cells. Immunoblot analysis revealed that immunopre- sites of pMX-puro II, yielding pMX-puro II-U6/siRNA. The DNA for the cipitates prepared with antibodies to ␤-catenin contained both shRNA encoded a 21-nucleotide hairpin sequence specific to the mRNA CHD8S and CHD8L (Fig. 1A). In addition, reciprocal analysis also target, with a loop sequence (-TTCAAGAGA-) separating the two com- showed that endogenous ␤-catenin was present in immunopre- plementary domains, and contained a tract of five T nucleotides to termi- cipitates prepared with antibodies to CHD8 (Fig. 1B and C). Both nate transcription. The hairpin sequences specific for human ␤-catenin ␤ ␤ endogenous CHD8S and CHD8L thus specifically interacted with ( -catenin-1 and -catenin-2), human CHD8, and enhanced green fluo- endogenous ␤-catenin in HEK293T cells, and such association rescent protein (EGFP) mRNAs (Clontech, Mountain View, CA) corre- ␤ ␤ was markedly increased in cells that were exposed to LiCl (Fig. 1B) sponded to nucleotides 1774 to 1794 ( -catenin-1), 501 to 521 ( - ␤ catenin-2), 138 to 158 (CHD8), and 126 to 146 (EGFP) of the respective or to Wnt3a (Fig. 1C) as activators of the Wnt– -catenin signaling coding regions. The resulting vectors were used to transfect Plat E cells pathway. Given that CHD8 is localized in the nucleus (29), the and thereby to generate recombinant retroviruses. binding between CHD8 and ␤-catenin may increase with the ac- RT-PCR. Total RNA (1 ␮g) isolated from HeLa, NIH 3T3, or ES cells cumulation and nuclear translocation of ␤-catenin induced by with the use of Isogen (Nippon Gene, Tokyo, Japan) was subjected to activation of Wnt signaling. reverse transcription (RT) with ReverTra Ace ␣ (Toyobo, Osaka, Japan). We next examined whether CHD8 might be recruited to the The resulting cDNA was subjected to PCR with Power SYBR green PCR promoter regions of Wnt target genes with the use of ChIP. ChIP master mix in an ABI-Prism 7000 sequence detection system (Applied with antibodies to the FLAG epitope revealed that FLAG-tagged Biosystems, Foster City, CA). The relative amounts of each target mRNA forms of both CHD8 and CHD8 were specifically associated were calculated as described previously (31). The primer sequences for S L RT-PCR analysis (sense and antisense, respectively) are 5=-CTGGCTTTG with Wnt-responsive elements of the axin2 gene promoter in GTGAACTGTTG-3= and 5=-AGTTGCTCACAGCCAAGACA-3= for HeLa cells only when the cells were exposed to LiCl or Wnt3a (Fig. hAxin-2,5=-ATCTGTATGTCCTGTCTGCCAGCG-3= and 5=-TGTCCT 1D). Similar results were obtained with the promoter of the c-Myc GATTCCCAAGTCAAGCAC-3= for mAxin-2,5=-CTACCCTCTCAACG gene (Fig. 1E). These data thus suggested that both CHD8S and ACAGCA-3= and 5=-AGAGCAGAGAATCCGAGGAC-3= for hc-Myc,5=- CHD8L are recruited to the promoters of Wnt target genes in a TCCTGACGACGAGACCTTCATCAAG-3= and 5=-TGAGAAACCGCT Wnt signaling-dependent manner and that CHD8 and ␤-catenin CCACATACAGTCC-3= for mc-Myc,5=-GCAAATTCCATGGCACCG may interact with each other but are not located at such gene T-3= and 5=-TCGCCCCACTTGATTTTGG-3= for hGAPDH, and 5=-GG promoters in unstimulated cells. ACCCGAGAAGACCTCCTT-3= and 5=-GCACATCACTCAGAATTTCA We next examined whether CHD8 might affect transcriptional = ATGG-3 for mRplp0. Human and mouse genes are indicated by h and m, activation by ␤-catenin in cancer cells. Forced expression of respectively. ChIP. Chromatin immunoprecipitation (ChIP) assays were per- CHD8S in HeLa cells markedly inhibited the LiCl- or Wnt3a- formed with a ChIP assay kit (Millipore) and 1 ϫ 106 cells for each reac- induced upregulation of mRNAs derived from the Wnt target tion. Precipitated DNA was quantitated by real-time PCR analysis as genes for axin2 (Fig. 1F) and c-Myc (Fig. 1G). Collectively, these described previously (31). The primer sequences for ChIP analysis results thus suggested that both CHD8S and CHD8L interact with (sense and antisense, respectively) are 5=-CTGGAGCCGGCTGCGCTT ␤-catenin and inhibit its transactivation activity on chromatin.

␤ FIG 3 CHD8 recruits histone H1 to the promoters of Wnt target genes. (A) Recombinant GST– -catenin, HA-CHD8S, and FLAG-histone H1 (HH1) were mixed as indicated and then subjected to immunoprecipitation with antibodies to FLAG. The resulting precipitates, as well as 10% of the original binding ␤ mixtures (input), were subjected to immunoblot analysis with antibodies to GST, to HA, or to FLAG. (B) HEK293T cells expressing HA– -catenin, Myc-CHD8S, ⌬ ϫ Myc-CHD8S H1, or 3 FLAG-histone H1 as indicated were subjected to immunoprecipitation with antibodies to FLAG. The resulting precipitates, as well as 3% of the original cell lysates (input), were subjected to immunoblot analysis with antibodies to HA, to Myc, or to FLAG. (C and D) HeLa cells stably overexpressing

CHD8S or infected with the empty vector were incubated in the absence or presence of LiCl (20 mM) or Wnt3a (20 ng/ml) for 4 h. The cells were then subjected to ChIP with antibodies to histone H1, and the precipitated DNA was subjected to real-time PCR analysis with primers speciﬁc for Wnt-responsive elements or a control region of the axin2 (C) or c-Myc (D) gene promoter. (E and F) HeLa cells infected with retroviral vectors for CHD8 or EGFP (control) shRNAs were subjected to immunoblot analysis with antibodies to CHD8 or to Hsp90 (E) as well as to ChIP analysis (F) as in panels C and D. All quantitative data are means Ϯ .(P Ͻ 0.01 (unpaired Student’s t test); NS, not signiﬁcant (P Ͼ 0.05 ,ءء .SD from three independent experiments

January 2012 Volume 32 Number 2 mcb.asm.org 505 Nishiyama et al.

We investigated whether the association of CHD8 with chromatin at Wnt target genes is dependent on ␤-catenin. ChIP with ␤ antibodies to -catenin revealed that forced expression of CHD8S did not substantially affect the association of ␤-catenin with chromatin in response to Wnt signaling (Fig. 2A). We also depleted HeLa cells of ␤-catenin with the use of two independent shRNAs targeting nucleotides 1774 to 1794 (shRNA-1) or nucleotides 501 to 521 (shRNA-2) of the coding region (Fig. 2B), with the decrease in the abundance of endogenous ␤-catenin being greater with shRNA-1 than with shRNA-2. ChIP with antibodies to CHD8 revealed that the amount of CHD8 associated with Wnt- responsive elements of the axin2 gene promoter was decreased in ␤-catenin-depleted cells stimulated with LiCl or Wnt3a and that the extent of this decrease was correlated with that in the total abundance of ␤-catenin (Fig. 2C). These results thus indicated that ␤-catenin mediates the association of CHD8 with Wnt- responsive elements of the axin2 gene promoter. Collectively, our observations thus suggested that ␤-catenin binds to the axin2 gene promoter in a CHD8-independent manner, whereas CHD8 association with this promoter is dependent on ␤-catenin. They are also consistent with the results of ChIP analysis showing that CHD8 was not localized to a substantial extent at Wnt target genes in unstimulated cells (Fig. 1D and E). CHD8 recruits histone H1 to Wnt target genes. We previously identified histone H1 as a molecule that associates with CHD8 with the use of a “shotgun” proteomics approach (29). Given that histone H1 was shown to be recruited to p53- responsive elements by CHD8 to suppress the transactivation activity of p53, we examined whether CHD8 also recruits histone H1 to Wnt-responsive elements and forms a ␤-catenin–CHD8–histone H1 trimeric complex. Pulldown assays in vitro with recombinant glutathione S-transferase (GST)-tagged ␤-catenin, HA- tagged CHD8S, and FLAG-tagged histone H1 proteins that had been produced in bacteria revealed that histone H1 directly inter- ␤ acted with CHD8S but not with -catenin; histone H1 thus asso- ␤ ciated with -catenin only in the presence of CHD8S (Fig. 3A). We also examined the association of 3ϫFLAG-tagged histone H1 with ␤ HA-tagged -catenin and with Myc epitope-tagged CHD8S or ⌬ CHD8S H1, a CHD8S mutant that lacks the histone H1 binding ⌬ FIG 4 The CHD8S H1 mutant binds to the promoters of Wnt target genes. domain, in HEK293T cells. Coimmunoprecipitation analysis re- ⌬ (A) HEK293T cells expressing FLAG-tagged CHD8S or CHD8S H1 were sub- vealed that the interaction between histone H1 and ␤-catenin was jected to immunoprecipitation with antibodies to FLAG. The resulting precipitates, as well as 3% of the original cell lysates (input), were subjected to im- markedly increased by forced expression of CHD8S but not by that ⌬ munoblot analysis with antibodies to ␤-catenin or to FLAG. (B and C) HeLa of CHD8S H1 (Fig. 3B). Collectively, these results indicated that cells infected with retroviruses encoding FLAG-tagged forms of CHD8S or ␤-catenin directly interacts with CHD8 and that CHD8 directly ⌬ CHD8S H1 (or with the empty retrovirus) were incubated in the absence or interacts with histone H1, whereas ␤-catenin indirectly interacts presence of Wnt3a (20 ng/ml) for 4 h. The cells were then subjected to ChIP with histone H1 through CHD8. with antibodies to FLAG, and the precipitated DNA was quantitated by real- ChIP with antibodies to histone H1 showed that histone H1 was time PCR analysis with primers specific for Wnt-responsive elements or a control region of the axin2 (B) or c-Myc (C) gene promoter. All quantitative indeed recruited specifically to the axin2 gene promoter in HeLa cells data are means Ϯ SD from three independent experiments. in a manner dependent both on exposure of the cells to LiCl or Wnt3a and on forced expression of CHD8S (Fig. 3C). Similar results were obtained with the Wnt-responsive elements of the c-Myc gene promoter (Fig. 3D). These results thus suggested that upregulation of overexpression, these data indicated that CHD8 recruits histone CHD8 might antagonize transactivation of Wnt target genes by H1 specifically to Wnt-responsive elements. ␤-catenin through recruitment of histone H1. Histone H1 recruitment by CHD8 is necessary for suppres- ␤ We also depleted HeLa cells of both CHD8S and CHD8L with a sion of -catenin function. Our previous study with a series of specific shRNA (Fig. 3E) and then performed ChIP analysis with deletion mutants delineated the region of CHD8S required for antibodies to histone H1 (Fig. 3F). The amount of histone H1 binding to histone H1 as that comprising residues 500 to 600 (29). associated with the Wnt-responsive elements of the c-Myc gene Consistent with the previous finding that ␤-catenin interacts with promoter was specifically and markedly reduced in the cells de- rat CHD8 through residues 667 to 749 of the latter (36), which pleted of CHD8. Together with the results obtained by CHD8 correspond to residues 669 to 751 of mouse CHD8, coimmuno-

506 mcb.asm.org Molecular and Cellular Biology Histone H1 Recruitment by CHD8 Inhibits Wnt Signaling

FIG 5 Interaction of CHD8 with histone H1 is necessary for recruitment of histone H1 to the promoters of Wnt target genes. (A and B) HeLa cells stably ⌬ expressing CHD8S or CHD8S H1 were incubated in the absence or presence of LiCl (20 mM) or Wnt3a (20 ng/ml) for 4 h and then subjected to ChIP assays as ⌬ in Fig. 3C and D, respectively. (C and D) HeLa cells stably expressing CHD8S or CHD8S H1 were incubated in the absence or presence of LiCl or Wnt3a for 4 h, after which the abundance of axin2 (C) and c-Myc (D) mRNAs was determined by quantitative RT-PCR analysis. All quantitative data are means Ϯ SD from three independent experiments. precipitation analysis in HEK293T cells revealed that ␤-catenin tone H1 to the Wnt-responsive elements of the axin2 gene pro- ⌬ ⌬ interacted with the mouse mutant protein CHD8S H1 (which moter, CHD8S H1 failed to do so (Fig. 5A). Similar results were lacks residues 500 to 600) to a similar extent as it did with wild- obtained for the c-Myc gene promoter (Fig. 5B). Furthermore, in ⌬ type CHD8S (Fig. 4A). We also performed ChIP analysis with contrast to wild-type CHD8S, CHD8S H1 did not substantially antibodies to FLAG in HeLa cells expressing FLAG-tagged CHD8S inhibit the upregulation of axin2 (Fig. 5C) or c-Myc (Fig. 5D) ⌬ ⌬ or CHD8S H1. Both CHD8S and CHD8S H1 were recruited to mRNAs induced by exposure of cells to LiCl or Wnt3a. These ⌬ similar extents to the Wnt-responsive elements of the axin2 (Fig. results suggested that the inability of CHD8S H1 to suppress 4B) or c-Myc (Fig. 4C) gene promoters in a Wnt signaling- transcriptional activation by ␤-catenin is indeed attributable to dependent manner. the lack of histone H1 binding, rather than to an altered localiza-

Whereas wild-type CHD8S promoted the recruitment of his- tion of the mutant CHD8 on chromatin. We therefore conclude

January 2012 Volume 32 Number 2 mcb.asm.org 507 Nishiyama et al.

letion of the COOH-terminal domain of histone H1 on the suppression of Wnt-mediated transcriptional activation, we forcibly ex-

pressed wild-type histone H1 or the mutants lacking the NH2- terminal or COOH-terminal domain in NIH 3T3 cells (Fig. 6C). The Wnt3a-induced increase in the abundance of axin2 mRNA was inhibited in cells expressing wild-type histone H1 and the mutant lack-

ing the NH2-terminal domain compared with that in control cells, whereas it was unaffected in cells expressing the mutant lacking the COOH-terminal domain. These results also suggested that the association of histone H1 with CHD8 is essential for the suppression of Wnt target genes. Histone H1 is essential for repression of ␤-catenin- ⌬ dependent transcription. Our observation that the CHD8S H1 mutant is impaired in the ability to repress ␤-catenin function led us to investigate the requirement for histone H1 in such repression. To this end, we adopted two independent approaches. First, we examined the expression of Wnt target genes in HH1cϪ/Ϫ HH1dϪ/Ϫ HH1eϪ/Ϫ triple-knockout (TKO) mouse ES cells. In these mutant ES cells, the abundance of histone H1 is reduced to ϳ50% of that in wild-type ES cells, resulting in marked changes in chromatin structure; microarray analysis revealed that the expression of only a few genes is altered in the mutant cells, however (6). Nevertheless, expression of axin2 (Fig. 7A) and c-Myc (Fig. 7B) genes was markedly increased in the TKO cells by low concentra- tions of LiCl or Wnt3a that do not activate Wnt signaling in wild-type control cells. The poor responsiveness of the wild- type cells to stimulation was likely attributable to the inhibition of Wnt signaling by endogenous CHD8, the expression of which is much greater in ES cells than in any adult tissues examined (29). These results thus suggested that CHD8 regulates cellular sensitivity to Wnt–␤-catenin signaling in early development. Second, we examined the effects both of a dominant negative mutant of histone H1 (EGFP-HH1) that reduces the amount of endogenous histone H1 and of a reciprocal mutant (HH1-EGFP) that does not exhibit dominant negative activity (Fig. 7C and D) (8). ChIP analysis revealed that both FIG 6 The COOH-terminal domain of histone H1 is necessary for repression EGFP-HH1 and HH1-EGFP were equally recruited to the Wnt- of ␤-catenin-dependent transcription. (A) Schematic representation of full- responsive elements of the axin2 gene promoter (Fig. 7E). Ex- length (FL) histone H1 (HH1) and its derivatives. The NH -terminal domain, 2 pression of EGFP-HH1 potentiated the stimulatory effects of central globular domain, and COOH-terminal domain are shown. (B) HEK293T cells expressing 3ϫFLAG-tagged histone H1 derivatives were sub- LiCl and Wnt3a on expression of the axin2 gene, whereas that jected to immunoprecipitation with antibodies to FLAG. The resulting precip- of HH1-EGFP had no such effect (Fig. 7F). Together, these two itates, as well as 3% of the original cell lysates (input), were subjected to im- approaches thus supported our conclusion that CHD8 nega- munoblot analysis with antibodies to CHD8 or to FLAG. (C) NIH 3T3 cells tively regulates the transactivation function of ␤-catenin by stably expressing HH1 derivatives were incubated in the absence or presence of Wnt3a (20 ng/ml) for 4 h, after which the abundance of axin2 mRNA was recruiting histone H1 to the promoters of Wnt target genes determined by quantitative RT-PCR analysis. Data are means Ϯ SD from three (Fig. 8). CHD8 is highly expressed in the early phase of embry- independent experiments. onic development and in cancer cell lines, whereas its expression ceases during late embryogenesis and its abundance is generally low in normal adult tissues. CHD8 may therefore that CHD8 recruits histone H1 to Wnt-responsive elements and serve to set a threshold for responsiveness to the Wnt signaling that such recruitment of histone H1 results in suppression of tran- pathway in a development-dependent manner. scriptional activation by ␤-catenin. The COOH-terminal domain of histone H1 plays a major role in DISCUSSION determination of linker DNA conformation and chromatin conden- With the use of biochemical and genetic approaches, we have sation (10). We generated deletion mutants of histone H1 that lack shown that CHD8 negatively regulates ␤-catenin function by re- the NH2-terminal, central globular, or COOH-terminal domain (Fig. cruiting histone H1 to the promoters of Wnt target genes. Forma- 6A) and tested these mutants for their ability to bind to CHD8 with tion of the ␤-catenin–CHD8–histone H1 complex requires two the use of coimmunoprecipitation analysis (Fig. 6B). We found that conditions: expression of CHD8 and stabilization of ␤-catenin by CHD8 interacted with the COOH-terminal domain of histone H1, activation of Wnt signaling. CHD8 is preferentially expressed in suggesting that the biological relevance of this domain relies, at least embryonic tissues and in cancer cell lines, which are thought to in part, on the interaction with CHD8. To examine the effect of de- reﬂect the embryonic state.

508 mcb.asm.org Molecular and Cellular Biology Histone H1 Recruitment by CHD8 Inhibits Wnt Signaling

FIG 7 Requirement for histone H1 in repression of ␤-catenin-dependent transcription. (A and B) Wild-type (WT) or HH1cϪ/Ϫ HH1dϪ/Ϫ HH1eϪ/Ϫ (HH1 TKO) ES cells were incubated with LiCl (2.5 mM) or Wnt3a (2.5 ng/ml) for 4 h and then subjected to quantitative RT-PCR analysis of axin2 (A) and c-Myc (B) mRNAs. (C) Schematic representation of histone H1 (HH1) tagged at its NH2 terminus (EGFP-HH1) or COOH terminus (HH1-EGFP) with EGFP. Both fusion proteins were also tagged at their NH2 termini with HA. (D) NIH 3T3 cells infected with retroviruses encoding EGFP-HH1 or HH1-EGFP (or with the empty vector) were then subjected to immunoblot analysis with antibodies to HA, to histone H1, or to Hsp90. (E) NIH 3T3 cells infected as in panel D were incubated in the absence or presence of Wnt3a (20 ng/ml) for 4 h. The cells were then subjected to ChIP with antibodies to HA, and the precipitated DNA was quantitated by real-time PCR analysis with primers speciﬁc for Wnt-responsive elements or a control region of the axin2 gene promoter. (F) NIH 3T3 cells infected as in panel D were incubated with LiCl (20 mM) or Wnt3a (20 ng/ml) for 4 h and then subjected to quantitative RT-PCR analysis of axin2 mRNA. All quantitative data are .(P Ͻ 0.01 versus the corresponding value for cells infected with the empty virus (unpaired Student’s t test ,ءء .means Ϯ SD from three independent experiments

CHD8S, also known as duplin, was originally isolated as a neg- ever, our present study suggests that not only CHD8L but also ␤ ␤ ative regulator of the Wnt– -catenin signaling pathway (36). Re- CHD8S is able to inhibit Wnt- and -catenin-dependent transac- cent studies have indicated that CHD8L interacts directly with tivation. Given that CHD8S contains the chromodomain but lacks ␤-catenin and negatively regulates ␤-catenin-dependent gene ex- the Snf2 helicase domain and the CTCF binding domain of pression through the ATP-dependent modulation of chromatin CHD8L, neither the ATP-dependent remodeling activity medi- structure (41). On the other hand, the COOH-terminal region of ated by the Snf2 helicase domain nor CTCF binding through the

CHD8L interacts with the insulator-binding protein CTCF, with COOH-terminal region of CHD8L is necessary for this suppres- this interaction being important for insulator activity (14). How- sive effect of CHD8 on the Wnt–␤-catenin signaling pathway. In

January 2012 Volume 32 Number 2 mcb.asm.org 509 Nishiyama et al.

FIG 8 Model for CHD8 function. CHD8 interacts with ␤-catenin and subsequently facilitates recruitment of histone H1, resulting in chromatin compaction and transcriptional silencing. contrast, this suppressive activity requires the region present in structure or of RNA polymerase II. Several transcriptional complexes both CHD8S and CHD8L that mediates binding to histone H1. A or coactivators, including Bcl-9 (also known as Lgs), Pygopus (Pygo), CHD8S mutant that lacks the histone H1 binding domain thus did polymerase-associated factor 1 (Paf1), and SET1 (trithorax), have not exhibit this inhibitory activity, and either depletion of histone been found to be recruited by ␤-catenin for target gene regulation H1 or expression of a dominant negative mutant thereof increased (17, 24, 39). At active target genes, Pygo binds through its PHD finger the cellular sensitivity to Wnt signaling. Our results thus indicate to methylated H3K4 and, by using Bcl-9 as an adaptor, retains that histone H1 recruitment mediated by CHD8 is fundamental to ␤-catenin near Wnt-responsive elements independently of Tcf (37). negative regulation of the Wnt–␤-catenin signaling pathway. Pygo-dependent binding of ␤-catenin to methylated histones frees Histone H1 molecules are highly mobile, and the interaction of a Tcf to recruit its corepressors, such as Groucho, that counteract the specific H1 molecule with a specific nucleosome is transient (18, 20), activating ␤-catenin-induced chromatin-remodeling processes. We suggesting that the molecules are continuously exchanged among have now uncovered another mode of ␤-catenin regulation mediated chromatin binding sites according to a “stop-and-go” process in by the CHD8-dependent recruitment of histone H1 to the promoters which a histone H1 molecule remains at a binding site for a limited of Wnt target genes. time before dissociating and moving rapidly to another such site. We previously showed that apoptosis mediated by p53- Most chromatin fibers thus likely always contain histone H1, but dependent transactivation is also suppressed by histone H1 re- there is a continuous turnover of histone H1 molecules at the level of cruited by CHD8 (29). Loss of CHD8 induced hyperactivation of the individual nucleosome. Given that histone H1 is implicated in p53, resulting in apoptosis, which was prevented by the additional regulation of the expression of specific genes, we postulated the exis- depletion of p53. Mice deficient in CHD8 die during early em- tence of a factor such as CHD8 that extends the residence time of bryogenesis (around embryonic day 7.5) as a result of aberrant histone H1 at the corresponding specific chromatin sites. It is also apoptosis induced by the unscheduled activation of p53 (28, 29). possible that CHD8 alters the interaction of histone H1 with linker It is likely that such apoptosis dominates and masks possible phe- DNA and represses transcription by enhancing the intrinsic stability notypes induced by hyperactivation of Wnt signaling. Additional of nucleosomes. The detailed mechanisms of transcriptional suppres- deletion of p53 in CHD8-deficient mice ameliorated the develop- sion for specific genes remain to be elucidated. mental arrest, resulting in extension of survival until embryonic The identification of various proteins that associate with day 10.5. Death of the Chd8Ϫ/Ϫ p53Ϫ/Ϫ mice is associated with ␤-catenin has provided insight into its function as a transcrip- severe hemorrhage, probably as a result of a defect in cardiovas- tional regulator. Many of these proteins, including components cular development (M. Nishiyama and K. I. Nakayama, unpub- of histone acetylation and methylation complexes as well as lished observations). Given that mice homozygous for a mutant chromatin-binding proteins, contribute to regulation of chromatin Apc allele encoding a product truncated at position 716 (Apc⌬716)

510 mcb.asm.org Molecular and Cellular Biology Histone H1 Recruitment by CHD8 Inhibits Wnt Signaling die in utero as a result of the unscheduled activation of Wnt sig- 15. Kamura T, et al. 2004. VHL-box and SOCS-box domains determine naling (30) and that Wnt signaling plays an essential role in car- binding specificity for Cul2-Rbx1 and Cul5-Rbx2 modules of ubiquitin diovascular development during embryogenesis (4), the embry- ligases. Genes Dev. 18:3055–3065. Ϫ/Ϫ Ϫ/Ϫ 16. Korinek V, et al. 1997. Constitutive transcriptional activation by a onic death of Chd8 p53 mice may be attributable to a defect ␤-catenin-Tcf complex in APCϪ/Ϫ colon carcinoma. Science 275: in the cardiovascular system resulting from unscheduled activa- 1784–1787. tion of Wnt signaling. 17. Kramps T, et al. 2002. Wnt/wingless signaling requires BCL9/legless- The mode of action of CHD8 in antagonism of Wnt–␤-catenin mediated recruitment of pygopus to the nuclear ␤-catenin-TCF complex. signaling appears almost identical to that for p53 inhibition. The Cell 109:47–60. 18. Lever MA, Th’ng JP, Sun X, Hendzel MJ. 2000. Rapid exchange of histone H1 binding domain that is present in both CHD8S and histone H1.1 on chromatin in living human cells. Nature 408:873–876. CHD8L is thus indispensable for CHD8-mediated inhibition of 19. Menon T, Yates JA, Bochar DA. 2010. Regulation of androgen- transactivation by p53 or ␤-catenin. Furthermore, genetic evi- responsive transcription by the chromatin remodeling factor CHD8. Mol. dence from cells depleted of histone H1 also supports this similar- Endocrinol. 24:1165–1174. ity. Given that CHD8 is preferentially expressed in embryonic 20. Misteli T, Gunjan A, Hock R, Bustin M, Brown DT. 2000. Dynamic binding of histone H1 to chromatin in living cells. Nature 408:877–881. tissues and in cancer cell lines, which are thought to reflect the 21. Moon RT, Kohn AD, De Ferrari GV, Kaykas A. 2004. WNT and embryonic state, CHD8 may counteract p53 and Wnt–␤-catenin ␤-catenin signalling: diseases and therapies. Nat. Rev. Genet. 5:691–701. function through recruitment of histone H1 during early embryo- 22. Morin PJ, et al. 1997. Activation of ␤-catenin-Tcf signaling in colon genesis to set a threshold for induction of apoptosis, cell prolifer- cancer by mutations in ␤-catenin or APC. Science 275:1787–1790. ation, and axis formation. It is also possible that CHD8 may bind 23. Morita S, Kojima T, Kitamura T. 2000. Plat-E: an efficient and stable ␤ system for transient packaging of retroviruses. Gene Ther. 7:1063–1066. to transcriptional factors other than p53 and -catenin and act as 24. Mosimann C, Hausmann G, Basler K. 2006. Parafibromin/Hyrax acti- a more general inhibitor of transactivation activity through re- vates Wnt/Wg target gene transcription by direct association with cruitment of histone H1. ␤-catenin/Armadillo. Cell 125:327–341. 25. Nakayama K, et al. 1996. Mice lacking p27Kip1 display increased body size, ACKNOWLEDGMENTS multiple organ hyperplasia, retinal dysplasia, and pituitary tumors. Cell 85:707–720. We thank T. Kitamura for pMX-puro; J. M. Cunningham and K. Hanada 26. Nakayama K, et al. 2000. Targeted disruption of Skp2 results in accumu- for the mCAT-1 plasmid; F. Ishikawa and R. Funayama for plasmids en- lation of cyclin E and p27Kip1, polyploidy and centrosome overduplica- coding the EGFP-HH1 and HH1-EGFP mutants of histone H1; A. Kiku- tion. EMBO J. 19:2069–2081. chi for Wnt3a; M. Sato, H. Takeda, C. Mitai, N. Nishimura, and K. Oy- 27. Nelson WJ, Nusse R. 2004. Convergence of Wnt, ␤-catenin, and cadherin amada for technical assistance; A. Ohta for help in preparation of the pathways. Science 303:1483–1487. manuscript; and members of our laboratories for discussion. 28. Nishiyama M, et al. 2004. Early embryonic death in mice lacking the A.I.S. was supported by NIH grant CA79057. ␤-catenin-binding protein Duplin. Mol. Cell. Biol. 24:8386–8394. 29. Nishiyama M, et al. 2009. CHD8 suppresses p53-mediated apoptosis REFERENCES through histone H1 recruitment during early embryogenesis. Nat. Cell Biol. 11:172–182. 1. Barker N, et al. 2001. The chromatin remodelling factor Brg-1 interacts 30. Oshima M, et al. 1995. Loss of Apc heterozygosity and abnormal tissue with beta-catenin to promote target gene activation. EMBO J. 20: building in nascent intestinal polyps in mice carrying a truncated Apc 4935–4943. gene. Proc. Natl. Acad. Sci. U. S. A. 92:4482–4486. 2. Bienz M, Clevers H. 2000. Linking colorectal cancer to Wnt signaling. 31. Pfaffl MW. 2001. A new mathematical model for relative quantification in Cell 103:311–320. real-time RT-PCR. Nucleic Acids Res. 29:e45. 3. Billin AN, Thirlwell H, Ayer DE. 2000. ␤-Catenin-histone deacetylase 32. Rodenberg JM, et al. 2010. Regulation of serum response factor activity interactions regulate the transition of LEF1 from a transcriptional repres- sor to an activator. Mol. Cell. Biol. 20:6882–6890. and smooth muscle cell apoptosis by chromodomain helicase DNA- 4. Cohen ED, Tian Y, Morrisey EE. 2008. Wnt signaling: an essential binding protein 8. Am. J. Physiol. Cell Physiol. 299:C1058–1067. regulator of cardiovascular differentiation, morphogenesis and progeni- 33. Rodriguez-Paredes M, Ceballos-Chavez M, Esteller M, Garcia-Dominguez tor self-renewal. Development 135:789–798. M, Reyes JC. 2009. The chromatin remodeling factor CHD8 interacts with 5. Daniels DL, Weis WI. 2005. Beta-catenin directly displaces Groucho/TLE elongating RNA polymerase II and controls expression of the cyclin E2 gene. repressors from Tcf/Lef in Wnt-mediated transcription activation. Nat. Nucleic Acids Res. 37:2449–2460. Struct. Mol. Biol. 12:364–371. 34. Roose J, et al. 1998. The Xenopus Wnt effector XTcf-3 interacts with 6. Fan Y, et al. 2005. Histone H1 depletion in mammals alters global chro- Groucho-related transcriptional repressors. Nature 395:608–612. ␤ matin structure but causes specific changes in gene regulation. Cell 123: 35. Rubinfeld B, et al. 1997. Stabilization of -catenin by genetic defects in 1199–1212. melanoma cell lines. Science 275:1790–1792. ␤ 7. Flanagan JF, et al. 2005. Double chromodomains cooperate to recognize 36. Sakamoto I, et al. 2000. A novel -catenin-binding protein inhibits beta- the methylated histone H3 tail. Nature 438:1181–1185. catenin-dependent Tcf activation and axis formation. J. Biol. Chem. 275: 8. Funayama R, Saito M, Tanobe H, Ishikawa F. 2006. Loss of linker 32871–32878. histone H1 in cellular senescence. J. Cell Biol. 175:869–880. 37. Sampietro J, et al. 2006. Crystal structure of a ␤-catenin/BCL9/Tcf4 9. Graham TA, Weaver C, Mao F, Kimelman D, Xu W. 2000. Crystal complex. Mol. Cell 24:293–300. structure of a ␤-catenin/Tcf complex. Cell 103:885–896. 38. Schnetz MP, et al. 2009. Genomic distribution of CHD7 on chromatin 10. Happel N, Doenecke D. 2009. Histone H1 and its isoforms: contribution tracks H3K4 methylation patterns. Genome Res. 19:590–601. to chromatin structure and function. Gene 431:1–12. 39. Sierra J, Yoshida T, Joazeiro CA, Jones KA. 2006. The APC tumor 11. He TC, et al. 1998. Identification of c-MYC as a target of the APC path- suppressor counteracts ␤-catenin activation and H3K4 methylation at way. Science 281:1509–1512. Wnt target genes. Genes Dev. 20:586–600. 12. He X, Semenov M, Tamai K, Zeng X. 2004. LDL receptor-related 40. Sims RJ, 3rd, et al. 2007. Recognition of trimethylated histone H3 lysine proteins 5 and 6 in Wnt/␤-catenin signaling: arrows point the way. Devel- 4 facilitates the recruitment of transcription postinitiation factors and opment 131:1663–1677. pre-mRNA splicing. Mol. Cell 28:665–676. 13. Hurd EA, et al. 2007. Loss of Chd7 function in gene-trapped reporter 41. Thompson BA, Tremblay V, Lin G, Bochar DA. 2008. CHD8 is an mice is embryonic lethal and associated with severe defects in multiple ATP-dependent chromatin remodeling factor that regulates ␤-catenin developing tissues. Mamm. Genome 18:94–104. target genes. Mol. Cell. Biol. 28:3894–3904. 14. Ishihara K, Oshimura M, Nakao M. 2006. CTCF-dependent chromatin 42. Vissers LE, et al. 2004. Mutations in a new member of the chromodomain insulator is linked to epigenetic remodeling. Mol. Cell 23:733–742. gene family cause CHARGE syndrome. Nat. Genet. 36:955–957.

January 2012 Volume 32 Number 2 mcb.asm.org 511 Nishiyama et al.

43. Willert K, et al. 2003. Wnt proteins are lipid-modified and can act as stem 46. Yuan CC, et al. 2007. CHD8 associates with human Staf and contributes cell growth factors. Nature 423:448–452. to efficient U6 RNA polymerase III transcription. Mol. Cell. Biol. 27: 44. Willert K, Jones KA. 2006. Wnt signaling: is the party in the nucleus? 8729–8738. Genes Dev. 20:1394–1404. 47. Zhang Y, LeRoy G, Seelig HP, Lane WS, Reinberg D. 1998. The 45. Yates JA, Menon T, Thompson BA, Bochar DA. 2010. Regulation of dermatomyositis-specific autoantigen Mi2 is a component of a complex HOXA2 gene expression by the ATP-dependent chromatin remodeling containing histone deacetylase and nucleosome remodeling activities. Cell enzyme CHD8. FEBS Lett. 584:689–693. 95:279–289.

512 mcb.asm.org Molecular and Cellular Biology