Decapentaplegic-responsive Silencers Contain Overlapping Mad-binding Sites
2006; Elsevier BV; Volume: 281; Issue: 35 Linguagem: Inglês
10.1074/jbc.m603371200
ISSN1083-351X
Autores Tópico(s)Developmental Biology and Gene Regulation
ResumoSmad proteins regulate transcription in response to transforming growth factor-β signaling pathways by binding to two distinct types of DNA sites. The sequence GTCT is recognized by all receptor-activated Smads and by Smad4. The subset of Smads that responds to bone morphogenetic protein signaling recognizes a distinct class of GC-rich sites in addition to GTCT. Recent work has shown that Drosophila Mad protein, the homologue of bone morphogenetic protein rSmads, binds to GRCGNC sites through the same MH1 domain β-hairpin interface used to contact GTCT sites. However, binding to GRCGNC requires base-specific contact by two Mad proteins, and here we provide evidence that this is achieved by contact of the two Mad subunits that overlap across the two central base pairs of the site. This topology is supported by results indicating that His-93, which is located at the tip of the Mad β-hairpin, is in close proximity to base pairs 2 and 5. Also consistent with the model is disruption of binding by mutation of Glu-39 and Glu-40, which are predicted to lie at the interface of the two overlapping Mad MH1 domains. As predicted from the overlapping model, binding is disrupted by insertion of 1 bp in the middle of the site, whereas insertion of 2 bp creates abutting sites that can be bound by the Mad-Medea heterotrimer without requiring Glu-39 and Glu-40. Overlapping Mad sites predominate in decapentaplegic response elements, consistent with a high degree of specificity in response to signaling. Smad proteins regulate transcription in response to transforming growth factor-β signaling pathways by binding to two distinct types of DNA sites. The sequence GTCT is recognized by all receptor-activated Smads and by Smad4. The subset of Smads that responds to bone morphogenetic protein signaling recognizes a distinct class of GC-rich sites in addition to GTCT. Recent work has shown that Drosophila Mad protein, the homologue of bone morphogenetic protein rSmads, binds to GRCGNC sites through the same MH1 domain β-hairpin interface used to contact GTCT sites. However, binding to GRCGNC requires base-specific contact by two Mad proteins, and here we provide evidence that this is achieved by contact of the two Mad subunits that overlap across the two central base pairs of the site. This topology is supported by results indicating that His-93, which is located at the tip of the Mad β-hairpin, is in close proximity to base pairs 2 and 5. Also consistent with the model is disruption of binding by mutation of Glu-39 and Glu-40, which are predicted to lie at the interface of the two overlapping Mad MH1 domains. As predicted from the overlapping model, binding is disrupted by insertion of 1 bp in the middle of the site, whereas insertion of 2 bp creates abutting sites that can be bound by the Mad-Medea heterotrimer without requiring Glu-39 and Glu-40. Overlapping Mad sites predominate in decapentaplegic response elements, consistent with a high degree of specificity in response to signaling. TGF 2The abbreviations used are: TGF, transforming growth factor; BMP, bone morphogenetic protein; MH, Mad homology; GFP, green fluorescent protein; oligos, oligonucleotides; Dpp, decapentaplegic; rSmad, receptor-activated Smad; SBE, Smad-binding element.2The abbreviations used are: TGF, transforming growth factor; BMP, bone morphogenetic protein; MH, Mad homology; GFP, green fluorescent protein; oligos, oligonucleotides; Dpp, decapentaplegic; rSmad, receptor-activated Smad; SBE, Smad-binding element.-β signaling contributes to numerous developmental processes, including axis formation, organogenesis, and limb development (1De Robertis E.M. Kuroda H. Annu. Rev. Cell Dev. Biol. 2004; 20: 285-308Crossref PubMed Scopus (542) Google Scholar, 2Raftery L.A. Sutherland D.J. Trends Genet. 2003; 19: 701-708Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar, 3Rusten T.E. Cantera R. Kafatos F.C. Barrio R. Development (Camb.). 2002; 129: 3575-3584Crossref PubMed Google Scholar, 4Niswander L. Nat. Rev. Genet. 2003; 4: 133-143Crossref PubMed Scopus (212) Google Scholar, 5Hogan B.L. Kolodziej P.A. Nat. Rev. Genet. 2002; 3: 513-523Crossref PubMed Scopus (259) Google Scholar), and is a factor in cancer (6Kang Y. He W. Tulley S. Gupta G.P. Serganova I. Chen C.R. Manova-Todorova K. Blasberg R. Gerald W.L. Massague J. Proc. Natl. Acad. Sci. U. S. A. 2005; 102: 13909-13914Crossref PubMed Scopus (429) Google Scholar, 7Mishra L. Derynck R. Mishra B. Science. 2005; 310: 68-71Crossref PubMed Scopus (169) Google Scholar). A TGF-β ligand initiates signaling by binding type I and type II receptor serine/threonine kinases at the cell surface (8Hart P.J. Deep S. Taylor A.B. Shu Z. Hinck C.S. Hinck A.P. Nat. Struct. Biol. 2002; 9: 203-208PubMed Google Scholar). Consequently, activated type I receptors phosphorylate Smad proteins at C-terminal serines, and these receptor-activated Smads (rSmads) then form a complex with Smad4 (9Massague J. Seoane J. Wotton D. Genes Dev. 2005; 19: 2783-2810Crossref PubMed Scopus (1880) Google Scholar, 10Feng X.H. Derynck R. Annu. Rev. Cell Biol. 2005; 21: 659-693Crossref Scopus (1507) Google Scholar). The activated Smad complexes translocate into the nucleus where they regulate transcription of target genes, through physical interaction and functional cooperation with co-activators and co-repressors (9Massague J. Seoane J. Wotton D. Genes Dev. 2005; 19: 2783-2810Crossref PubMed Scopus (1880) Google Scholar, 10Feng X.H. Derynck R. Annu. Rev. Cell Biol. 2005; 21: 659-693Crossref Scopus (1507) Google Scholar). Decapentaplegic (Dpp) is a well characterized TGF-β ligand in Drosophila that regulates transcription through the Smad proteins, Mothers Against Dpp (Mad), and Medea (2Raftery L.A. Sutherland D.J. Trends Genet. 2003; 19: 701-708Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar, 11Affolter M. Marty T. Vigano M.A. Jazwinska A. EMBO J. 2001; 20: 3298-3305Crossref PubMed Scopus (92) Google Scholar). Mad is homologous to the BMP-specific rSmads (Smad1, Smad5, and Smad8), and Medea is the homologue of Smad4. Discrete patterns of Dpp expression during many phases of development (12Posakony L. Raftery L. Gelbart W. Mech. Dev. 1991; 33: 69-82Crossref Scopus (141) Google Scholar, 13Masucci J.D. Miltenberger R.J. Hoffmann F.M. Genes Dev. 1990; 4: 2011-2023Crossref PubMed Scopus (208) Google Scholar) result in graded patterns of phosphorylated Mad (pMad) across developing tissues (14Dorfman R. Shilo B.Z. Development (Camb.). 2001; 128: 965-972Crossref PubMed Google Scholar, 15Tanimoto H. Itoh S. Ten D.P. Tabata T. Mol. Cell. 2000; 5: 59-71Abstract Full Text Full Text PDF PubMed Scopus (316) Google Scholar), which are interpreted by individual cells as positional information that instructs cell-type specification (2Raftery L.A. Sutherland D.J. Trends Genet. 2003; 19: 701-708Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar, 11Affolter M. Marty T. Vigano M.A. Jazwinska A. EMBO J. 2001; 20: 3298-3305Crossref PubMed Scopus (92) Google Scholar). pMad levels are read out differentially by target genes with various thresholds for activation or repression by pMad, in combination with Medea and tissue-specific cofactors (16Xu X. Yin Z. Hudson J.B. Ferguson E.L. Frasch M. Genes Dev. 1998; 12: 2354-2370Crossref PubMed Scopus (219) Google Scholar, 17Knirr S. Frasch M. Dev. Biol. 2001; 238: 13-26Crossref PubMed Scopus (87) Google Scholar, 18Halfon M.S. Carmena A. Gisselbrecht S. Sackerson C.M. Jimenez F. Baylies M.K. Michelson A.M. Cell. 2000; 103: 63-74Abstract Full Text Full Text PDF PubMed Scopus (268) Google Scholar, 19Lee H.H. Frasch M. Development (Camb.). 2005; 132: 1429-1442Crossref PubMed Scopus (44) Google Scholar, 20Xu M. Kirov N. Rushlow C. Development (Camb.). 2005; 132: 1637-1647Crossref PubMed Scopus (35) Google Scholar). Considerable progress has been made in understanding how Smad proteins interact with cofactors to activate or repress transcription, and it is expected that response thresholds will be determined by their combined affinities for particular cis elements. In Drosophila this is complicated by a double-negative mechanism, in which Dpp represses the brinker (brk) gene, and the Brk protein then represses other Dpp targets, in some instances by competing for Mad-binding sites (21Marty T. Muller B. Basler K. Affolter M. Nat. Cell Biol. 2000; 2: 745-749Crossref PubMed Scopus (98) Google Scholar, 22Sivasankaran R. Vigano M.A. Muller B. Affolter M. Basler K. EMBO J. 2000; 19: 6162-6172Crossref PubMed Scopus (72) Google Scholar, 23Jazwinska A. Kirov N. Wieschaus E. Roth S. Rushlow C. Cell. 1999; 96: 563-573Abstract Full Text Full Text PDF PubMed Scopus (224) Google Scholar, 24Campbell G. Tomlinson A. Cell. 1999; 96: 553-562Abstract Full Text Full Text PDF PubMed Scopus (232) Google Scholar, 25Minami M. Kinoshita N. Kamoshida Y. Tanimoto H. Tabata T. Nature. 1999; 398: 242-246Crossref PubMed Scopus (190) Google Scholar, 26Rushlow C. Colosimo P.F. Lin M.C. Xu M. Kirov N. Genes Dev. 2001; 15: 340-351Crossref PubMed Scopus (98) Google Scholar, 27Saller E. Bienz M. EMBO Rep. 2001; 2: 292-297Crossref PubMed Scopus (36) Google Scholar, 28Kirkpatrick H. Johnson K. Laughon A. J. Biol. Chem. 2001; 276: 18216-18222Abstract Full Text Full Text PDF PubMed Scopus (64) Google Scholar, 29Zhang H.L. Levine M. Ashe H.L. Genes Dev. 2001; 15: 261-266Crossref PubMed Scopus (72) Google Scholar). Repression of brk by Dpp is through a small silencer element (brkS) that contains a high affinity Mad-Medea-binding site (30Muller B. Hartmann B. Pyrowolakis G. Affolter M. Basler K. Cell. 2003; 113: 221-233Abstract Full Text Full Text PDF PubMed Scopus (141) Google Scholar, 31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar). A nearly identical silencer controls repression of bag of marbles (bam) in germ line stem cells in response to Dpp secreted from an adjacent somatic layer of the ovary or testis (32Chen D.H. McKearin D. Curr. Biol. 2003; 13: 1786-1791Abstract Full Text Full Text PDF PubMed Scopus (293) Google Scholar, 33Chen D.H. McKearin D.M. Development (Camb.). 2003; 130: 1159-1170Crossref PubMed Scopus (289) Google Scholar). Alignment of these sites and mutational analysis revealed that these silencers are bipartite, with a GRCGNC Mad-binding site spaced 5 bp from a GTCT Medea-binding site (30Muller B. Hartmann B. Pyrowolakis G. Affolter M. Basler K. Cell. 2003; 113: 221-233Abstract Full Text Full Text PDF PubMed Scopus (141) Google Scholar, 31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar, 32Chen D.H. McKearin D. Curr. Biol. 2003; 13: 1786-1791Abstract Full Text Full Text PDF PubMed Scopus (293) Google Scholar, 33Chen D.H. McKearin D.M. Development (Camb.). 2003; 130: 1159-1170Crossref PubMed Scopus (289) Google Scholar). Once bound, Mad and Med together recruit Schnurri, a large zinc finger transcription factor with a repression domain responsible for silencing of brk and bam (31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar). Smad proteins contact DNA with a β-hairpin contained in the MH1 domain. β-Hairpin side chains make specific major groove contact with bases at the 1st, 3rd, and 4th positions of GTCT (34Shi Y. Wang Y. Jayaraman L. Haijuan Y. Massague J. Pavletich N. Cell. 1998; 94: 585-594Abstract Full Text Full Text PDF PubMed Scopus (602) Google Scholar), and each of these amino acids is required in both Mad and Medea for silencer binding (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). However, an analysis of stoichiometry revealed that the GRCGNC site is contacted by two Mad MH1 domains and that all three base-contact residues are required in both subunits (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). This then leads to the question of how two Mad MH1 domains can squeeze onto a 6-bp site using the same contacts involved in binding of a single MH1 to a 4-bp site. The silencer GRCGNC Mad-binding site resembles the previously identified GCCGNCG Mad consensus deduced from alignment of Dpp-activated enhancers (16Xu X. Yin Z. Hudson J.B. Ferguson E.L. Frasch M. Genes Dev. 1998; 12: 2354-2370Crossref PubMed Scopus (219) Google Scholar, 36Kim J. Johnson K. Chen H.J. Carroll S. Laughon A. Nature. 1997; 388: 304-308Crossref PubMed Scopus (446) Google Scholar), and which also occurs in vertebrate BMP-response elements (37Benchabane H. Wrana J. Mol. Cell. Biol. 2003; 23: 6646-6661Crossref PubMed Scopus (76) Google Scholar, 38Ishida W. Hamamoto T. Kusanagi K. Yagi K. Kawabata M. Takehara K. Sampath T.K. Kato M. Miyazono K. J. Biol. Chem. 2000; 275: 6075-6079Abstract Full Text Full Text PDF PubMed Scopus (230) Google Scholar, 39Kang Y. Chen C.R. Massague J. Mol. Cell. 2003; 11: 915-926Abstract Full Text Full Text PDF PubMed Scopus (427) Google Scholar, 40Karaulanov E. Knochel W. Niehrs C. EMBO J. 2004; 23: 844-856Crossref PubMed Scopus (114) Google Scholar, 41Korchynskyi O. ten Dijke P. J. Biol. Chem. 2002; 277: 4883-4891Abstract Full Text Full Text PDF PubMed Scopus (696) Google Scholar, 42Kusanagi K. Inoue H. Ishidou Y. Mishima H.K. Kawabata M. Miyazono K. Mol. Biol. Cell. 2000; 11: 555-565Crossref PubMed Scopus (154) Google Scholar, 43Lopez-Rovira T. Chalaux E. Massague J. Rosa J.L. Ventura F. J. Biol. Chem. 2002; 277: 3176-3185Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar, 44Seki K. Hata A. J. Biol. Chem. 2004; 279: 18544-18549Abstract Full Text Full Text PDF PubMed Scopus (53) Google Scholar). Here we provide evidence that two Mad MH1 domains bind to GRCGNC by overlapping across the two central base pairs of the site. Plasmids—Mad, Med, TkvQD, ShnCT, and 2XGFP-Mad/-Med effector plasmids for expression in human 293T cells have been described previously. Site-directed mutagenesis of Mad and Med effector plasmids was carried out using a method described previously (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar, 45Sawano A. Miyawaki A. Nucleic Acids Res. 2000; 28: 78Crossref PubMed Scopus (298) Google Scholar). Plasmid constructs and products of site-directed mutagenesis were verified by DNA sequencing. For reporter assays, the brkS sequence or derivatives were inserted between the EcoRI and NotI sites of 3xSu(H)-HSC (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). pPacTkvQD, pPacSu(H), and pPacNAct have been described previously (28Kirkpatrick H. Johnson K. Laughon A. J. Biol. Chem. 2001; 276: 18216-18222Abstract Full Text Full Text PDF PubMed Scopus (64) Google Scholar). Gel Shift and Reporter Assays—Gel shift assays and preparation of DNA probes and lysates from transfected human 293T cells were performed as described previously (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). Gel images were collected with a Storm 870 PhosphorImager (Amersham Biosciences), and the amount of probe present in bands was determined using ImageQuant software. Reporter assays were performed by transfection of Drosophila S2 cells as described previously (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar) with pPac-luc as an internal control. S2 cells were harvested 2 days after transfection, and chemiluminescent β-galactosidase and luciferase assays were performed on cell extracts using the GalactoStar assay system (Tropix, Inc.) according to the supplier's instructions. Reporter assays were performed in triplicate with results normalized to the internal control. Western blotting was performed as described previously (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). Molecular Modeling—The Swiss-Model feature of Deep View (46Guex N. Peitsch M.C. Electrophoresis. 1997; 18: 2714-2723Crossref PubMed Scopus (9396) Google Scholar) was used to thread amino acids 26-146 of the Mad MH1 domain sequence onto the Smad3 MH1-DNA co-crystal structure of Chai et al. (47Chai J. Wu J.W. Yan N. Massague J. Pavletich N.P. Shi Y. J. Biol. Chem. 2003; 278: 20327-20331Abstract Full Text Full Text PDF PubMed Scopus (62) Google Scholar). Compared with Smad3, Mad has a 3-residue gap in loop L1 between helices H1 and H2, as do the homologous BMP rSmads. By using the alignment feature, this gap was manually positioned between Lys-35 and Gln-36, allowing the Mad sequence to thread along the complete lengths of helices H2 and H3. Manual alignment was also used to position a single residue gap between Smad3 Lys-44 and Thr-45 in the turn between helices H2 and H3. The threaded alignment was submitted as a modeling request to the Swiss Model server. The resulting improved Mad MH1 domain model was manually docked to a B-DNA model of the brkS element generated using the model.it program of the DNAtools website (48Vlahovicek V. Kajan L. Pongor S. Nucleic Acids Res. 2003; 31: 3686-3687Crossref PubMed Scopus (111) Google Scholar). The Mutate feature of DeepView/Swiss-Pdb-Viewer (46Guex N. Peitsch M.C. Electrophoresis. 1997; 18: 2714-2723Crossref PubMed Scopus (9396) Google Scholar) was used to explore possible hydrogen bonding between side chains at the interface between the two Mad MH1 domains. The same homology modeling and docking procedures were used to generate the Medea MH1 domain model shown in Fig. 5A. The images shown in Fig. 5 were generated by POV-Ray version 3.5 from screens created using DeepView. Binding of Two Mad MH1 Domains to a 6-bp Site Suggests an Overlapping Model for DNA Contact—Alignment and mutational analysis of the bam and brk silencer elements identified a Mad-binding site spanning 6 bp but in which only 4 positions are rigidly specified (31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar, 35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). These Mad sites are contacted by two Mad MH1 domains, each involving all three canonical β-hairpin contact residues. In Mad these base-contact amino acids are Arg-88, Gln-90, and Lys-95 (Fig. 1A). Fig. 1B shows three models for how two Mad MH1 domains might contact GRCGNC. In each model, Arg-88 and Lys-95 contact guanines spaced 2 bases apart on opposite strands, as they do in the Smad3MH1-DNA structure. In the first model, the two MH1 domains abut facing outward so that 1 additional base on each end of the site (positions -1 and 7) would be contacted by Gln-90, a side chain that has been shown to be critical for binding of Mad to brkS (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). Pyrowolakis et al. (31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar) showed that substitution of G at -1 or C at +7 had little effect on binding of the Mad-Medea complex, inconsistent with critical base-specific contact at either position. However, contact by Gln-90 might be compatible with more than 1 base, which appears to be the case for the corresponding Gln-111 of Medea, because either T or G is tolerated at the 4th position of its binding site in brkS (i.e. GTCT or GTCG) (31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar, 35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). To explore this further, we measured the effects of all possible single base substitutions of the flanking base pairs using a competitive gel shift assay (Fig. 1C). In this experiment excess unlabeled wild-type and mutant brkS DNAs were added to gel shift binding reactions containing labeled wild-type brkS probe. The observed effects on binding to brkS were minor, as judged by comparison of 3-fold steps in the concentration of cold competitor DNAs. Furthermore, according to model 1, -1A and 7T would be expected to improve binding because they match the Smad box consensus, but instead diminished binding affinity. Thus, these results do not lend support for model 1. For the same reasons, the results also weigh against a second model in which the MH1 domains bind pointed in the same direction with 1 bp of overlap, such that the predictions would apply to one flanking side or the other. The results do not rule out a third model in which the two Mad MH1 domains point toward each other with their β-hairpins overlapping across the two central base pairs in the major groove (Fig. 1B, model 3). According to this model, all base contact by β-hairpin side chains occurs within the 6-bp site, and the observed minor effects of substitutions at flanking positions would be indirect (e.g. perhaps reflecting constraints on DNA conformation). Consistent with this model, we note that the symmetry of -1T and 7A matches the conserved T adjacent to the Medea-binding site of brkS or bamS (TGTCT). Silencer Binding Is Sensitive to Mutations Affecting the Tip of the Mad β-Hairpin—To further distinguish between the Mad-binding site models, we sought a way of determining how the β-hairpins of the two Mad subunits are positioned with relation to specific base pairs within the brkS-binding site. At the tip of the β-hairpin, where it makes the turn between β-sheets B2 and B3, rSmads contain a histidine residue (His-93 in Mad) whose imidazole side chain is pointed toward the neighboring MH1 domain in the Smad3-DNA structure (34Shi Y. Wang Y. Jayaraman L. Haijuan Y. Massague J. Pavletich N. Cell. 1998; 94: 585-594Abstract Full Text Full Text PDF PubMed Scopus (602) Google Scholar, 47Chai J. Wu J.W. Yan N. Massague J. Pavletich N.P. Shi Y. J. Biol. Chem. 2003; 278: 20327-20331Abstract Full Text Full Text PDF PubMed Scopus (62) Google Scholar). The overlapping model for binding of Mad MH1 domains to brkS positions His-93 near the DNA-contacting Arg-88 side chain of the opposite Mad subunit (Fig. 2A). The predicted close proximity of these residues suggests that binding might be affected by substitution of a bulky residue in place of His-93. To test this a H93W mutation (MadH93W) was made and found to cause substantial reduction in binding affinity for brkS (Fig. 2B). By co-expressing MadH93W together with 2XGFP-Mad, it was found that complexes in which just one Mad subunit contains the H93W mutation were relatively unaffected compared with wild type (Fig. 2C). The corresponding position in Medea and Smad4 is a glycine instead of histidine (Gly-114 in Medea), and it was found that a MedeaG114W mutation had a weaker effect on silencer binding (Fig. 2B), consistent with the binding of Medea to the unpaired GTCT and indicating that a Trp substitution at this position in the β-hairpin does not drastically interfere with DNA contact if the site is nonoverlapping. In separate experiments it was noticed that substitutions at position 2 or position 5 of the overlapping site caused reductions in binding affinity (Fig. 2D). Position 2 shows a bias for G and position 5 shows the same bias on the opposite strand, consistent with a completely symmetrical Mad-binding site of optimal sequence GGCGCC. In the overlapping model, bases at positions 2 and 5 are not predicted to make direct contact with the MH1 domains but are near the His-93 side chain. To test whether an unfavorable interaction with His-93 might be responsible for the reduced binding affinity, MadH93A was tested for binding to brkS, whose Mad site sequence is GGCGAC, and to the position 2 and 5 mutations, GCCGAC and GGCGGC (Fig. 2E, numbering as in Fig. 1B). The MadH93A mutation had little or no effect on binding to the GGCGAC site or to the optimal GGCGCC but enhanced binding to the 2C and 5G probes, consistent with proximity of these 2 bases to the corresponding His-93 side chains. Western blot analysis showed that neither H93A nor H93W detectably affected the level of Mad protein in the cell lysates used for gel shift experiments (Fig. 2, B and E, lower panels). These results support an overlapping model in which the tip of the β-hairpin of one Mad subunit is close to position 2 and on the other subunit is close to position 5. The brkS Mad Site Is Disrupted by Interrupting Symmetry in the Proposed Overlap—The overlapping model predicts that a 1-bp insertion at the center of the GGCGAC site will disrupt overlap of the two sites, creating an asymmetric 1-bp overlap or two tandem sites (Fig. 3A). A varying degree of disruption by such insertions was observed (Fig. 3B), the least damaging of which (the 3(A)4 construct) correlated with the creation of a GTCT site. In contrast, a series of 2-bp insertions left binding largely intact (3(GC)4, 3(TA)4, and 2SB_RL probes in Fig. 3C), whereas further separation of the sites reduced binding affinity (Fig. 3C, right 4 lanes) suggesting that proximity of the Mad MH1 domains is required for stable binding. These effects were confirmed in a competitive gel shift assay (Fig. 3D). Binding was also disrupted when the GGCGAC Mad-binding site was replaced with other arrangements of two SBEs (2SB_LR, 2SB_LL, and 2SB_RR probes in Fig. 3E), indicating that stability of the bound Mad-Medea complex depends upon a specific topology of the Mad MH1 domains. The failure to disrupt binding by 2-bp insertions might be due to the creation of two abutting sites in the same configuration as the Smad3/Smad4 consensus binding site (49Zawel L. Dai J.L. Buckhaults P. Zhou S.B. Kinzler K.W. Vogelstein B. Kern S.E. Mol. Cell. 1998; 1: 611-617Abstract Full Text Full Text PDF PubMed Scopus (884) Google Scholar), (indicated by the arrows marking each sequence in Fig. 3A). In support of this idea, binding was not disrupted by substitution of GTCTAGAC (the Smad3/4 consensus) in place of GGCGAC; rather, slight enhancement of binding was observed (Fig. 3, C and D, 2SB_RL). The resulting gel shift complex had a slightly slower mobility than the brkS complex and was still dependent on the Medea site for high affinity binding (Fig. 3F, compare brkS, 3(TA)4, and 2SB_RL to corresponding GTCTm probes). Stoichiometry of the complex was assessed using 2xGFP fusions to Mad or Medea (Fig. 3G). The substitution of 2xGFP-Mad for Mad caused a supershift consistent with the presence of two Mad subunits (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar), and an intermediate band was visible when Mad and 2xGFP-Mad were both present in the complex. In contrast, the supershifted 2xGFP-Medea aligns with the immediate Mad+2xGFPMad band and caused no obvious intermediate band in combination with Medea, consistent with the presence of just a single Medea subunit. Together, these results indicate a 2:1 Mad:Medea heterotrimer complex, as was found for the brkS Mad-Medea complex (35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). Reporter assays were carried out in Drosophila S2 cells to determine whether an abutting site still allows silencer function. In this assay the brkS and bamSE silencers can over-ride reporter activation by Notch and Suppressor Hairless (30Muller B. Hartmann B. Pyrowolakis G. Affolter M. Basler K. Cell. 2003; 113: 221-233Abstract Full Text Full Text PDF PubMed Scopus (141) Google Scholar, 31Pyrowolakis G. Hartmann B. Muller B. Basler K. Affolter M. Dev. Cell. 2004; 7: 229-240Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar, 35Gao S. Steffen J. Laughon A. J. Biol. Chem. 2005; 280: 36158-36164Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). By using this assay, we found that when abutting Mad-binding sites matched the GC-rich Mad consensus, i.e. GGCGCGCC, then ShnCT still bound (2GC_RL probe in Fig. 3H) and the element repressed reporter expression in response to Dpp signaling (2GC_RL reporter in Fig. 3I). If the abutting site was changed to GTCTAGAC, then ShnCT bound was reduced (2SB_RL probe in Fig. 3H), and repression was lost (Fig. 3I), perhaps reflecting a situation where, in vivo, the binding of co-activators wins out when Shn binding affinity drops below a critical threshold, despite its ability to bind weakly in vitro. Repression was also disrupted when the overlapping Mad sites were replaced with other arrangements of two SBEs (Fig. 3I). The results described above show that Mad binds to GTCT or GGCG with similar affinity despite a difference in the base apparently contacted by Mad Gln-90. In the overlapping site the sequence is constrained to GNCGNC, but in abutting sites it seemed possible that GTCT and GGCG might be interchangeable. However, although the 2SB_RL and 2GC_RL probes were bound with similar affinity, a composite of the two, GTCTCGCC, was bound only weakly (GC_SB probe in Fig. 3H). In addition, we tested the alternative Smad1-binding site, GCAT (50Henningfeld K.A. Rastegar S. Adler G. Knochel W. J. Biol. Chem. 2000; 275: 21827-21835Abstract Full Text Full Text PDF PubMed Scopus (69) Google Scholar), and we observed
Referência(s)