Title: Structure of a dominant-negative helix-loop-helix transcriptional regulator suggests mechanisms of autoinhibition
Abstract: Article27 March 2012free access Structure of a dominant-negative helix-loop-helix transcriptional regulator suggests mechanisms of autoinhibition Ryohei Ishii Ryohei Ishii Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Kazunobu Isogaya Kazunobu Isogaya Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Azusa Seto Azusa Seto Department of Basic Medical Sciences, Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan Search for more papers by this author Daizo Koinuma Daizo Koinuma Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Yuji Watanabe Yuji Watanabe Department of Basic Medical Sciences, Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan Search for more papers by this author Fumio Arisaka Fumio Arisaka Department of Biological Information, Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, Yokohama, Kanagawa, Japan Search for more papers by this author So-ichi Yaguchi So-ichi Yaguchi Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Chuo, Yamanashi, Japan Search for more papers by this author Hiroaki Ikushima Hiroaki Ikushima Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Naoshi Dohmae Naoshi Dohmae Biomolecular Characterization Team, RIKEN, Wako, Saitama, Japan Advanced Science Institute, RIKEN, Wako, Saitama, Japan Search for more papers by this author Kohei Miyazono Kohei Miyazono Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Keiji Miyazawa Corresponding Author Keiji Miyazawa Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Chuo, Yamanashi, Japan Search for more papers by this author Ryuichiro Ishitani Ryuichiro Ishitani Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Advanced Science Institute, RIKEN, Wako, Saitama, Japan Search for more papers by this author Osamu Nureki Corresponding Author Osamu Nureki Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Ryohei Ishii Ryohei Ishii Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Kazunobu Isogaya Kazunobu Isogaya Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Azusa Seto Azusa Seto Department of Basic Medical Sciences, Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan Search for more papers by this author Daizo Koinuma Daizo Koinuma Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Yuji Watanabe Yuji Watanabe Department of Basic Medical Sciences, Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan Search for more papers by this author Fumio Arisaka Fumio Arisaka Department of Biological Information, Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, Yokohama, Kanagawa, Japan Search for more papers by this author So-ichi Yaguchi So-ichi Yaguchi Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Chuo, Yamanashi, Japan Search for more papers by this author Hiroaki Ikushima Hiroaki Ikushima Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Naoshi Dohmae Naoshi Dohmae Biomolecular Characterization Team, RIKEN, Wako, Saitama, Japan Advanced Science Institute, RIKEN, Wako, Saitama, Japan Search for more papers by this author Kohei Miyazono Kohei Miyazono Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Keiji Miyazawa Corresponding Author Keiji Miyazawa Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Chuo, Yamanashi, Japan Search for more papers by this author Ryuichiro Ishitani Ryuichiro Ishitani Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Advanced Science Institute, RIKEN, Wako, Saitama, Japan Search for more papers by this author Osamu Nureki Corresponding Author Osamu Nureki Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan Search for more papers by this author Author Information Ryohei Ishii1,‡, Kazunobu Isogaya2,‡, Azusa Seto3,‡, Daizo Koinuma2,‡, Yuji Watanabe3, Fumio Arisaka4, So-ichi Yaguchi5, Hiroaki Ikushima2, Naoshi Dohmae6,7, Kohei Miyazono2, Keiji Miyazawa 2,5, Ryuichiro Ishitani1,7 and Osamu Nureki 1 1Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan 2Department of Molecular Pathology, Graduate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo, Japan 3Department of Basic Medical Sciences, Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan 4Department of Biological Information, Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, Yokohama, Kanagawa, Japan 5Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Chuo, Yamanashi, Japan 6Biomolecular Characterization Team, RIKEN, Wako, Saitama, Japan 7Advanced Science Institute, RIKEN, Wako, Saitama, Japan ‡These authors contributed equally to this work *Corresponding authors: Department of Biochemistry, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, 1110 Shimokato, Chuo, Yamanashi 409-3898, Japan. Tel.:+81 55 273 6784; Fax:+81 55 273 6784; E-mail: [email protected] of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, 2-11-16 Yayoi, Bunkyo-ku, Tokyo113-0032, Japan. Tel.:+81 3 5841 4392; Fax:+81 3 5841 8057; E-mail: [email protected] The EMBO Journal (2012)31:2541-2552https://doi.org/10.1038/emboj.2012.77 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Helix-loop-helix (HLH) family transcription factors regulate numerous developmental and homeostatic processes. Dominant-negative HLH (dnHLH) proteins lack DNA-binding ability and capture basic HLH (bHLH) transcription factors to inhibit cellular differentiation and enhance cell proliferation and motility, thus participating in patho-physiological processes. We report the first structure of a free-standing human dnHLH protein, HHM (Human homologue of murine maternal Id-like molecule). HHM adopts a V-shaped conformation, with N-terminal and C-terminal five-helix bundles connected by the HLH region. In striking contrast to the common HLH, the HLH region in HHM is extended, with its hydrophobic dimerization interfaces embedded in the N- and C-terminal helix bundles. Biochemical and physicochemical analyses revealed that HHM exists in slow equilibrium between this V-shaped form and the partially unfolded, relaxed form. The latter form is readily available for interactions with its target bHLH transcription factors. Mutations disrupting the interactions in the V-shaped form compromised the target transcription factor specificity and accelerated myogenic cell differentiation. Therefore, the V-shaped form of HHM may represent an autoinhibited state, and the dynamic conformational equilibrium may control the target specificity. Introduction The helix-loop-helix (HLH) proteins are central regulators in a wide variety of developmental and homeostatic processes (Olson and Klein, 1994). Especially, basic helix-loop-helix (bHLH) transcription factors play key roles in regulating gene expression, cell cycle control, and developmental processes, by binding to the 'E box' in the promoters of tissue-specific genes through homo- and heterodimer formation (Murre et al, 1989; Blackwell et al, 1990; Kreider et al, 1992; Zebedee and Hara, 2001). The HLH proteins are divided into seven classes, based on the presence of a DNA-binding region as well as other appended motifs/domains (Massari and Murre, 2000). The class-I bHLH proteins, such as E12, E47, HEB, and E2-2, are ubiquitously expressed in many tissues. In contrast, the class-II bHLH proteins, such as MyoD, NeuroD, and Hes, exhibit tissue-specific expression and form heterodimers with the class-I bHLH proteins to regulate distinct developmental pathways, such as myogenesis, neurogenesis and lymphopoiesis (Weintraub et al, 1990; Lassar et al, 1991; Weintraub et al, 1991; Weintraub, 1993; Parkhurst and Meneely, 1994; Lee et al, 1995; Shen and Kadesch, 1995; Ma et al, 1996; Rawls and Olson, 1997). The class-III HLH proteins include the Myc family of transcription factors, which contain a leucine zipper (LZ) adjacent to the HLH motif. The class-IV HLH proteins include Mad, Max, and Mxi, which are capable of dimerizing with the Myc proteins and with each other (Blackwood and Eisenman, 1991; Ayer et al, 1993; Zervos et al, 1993). The class-V HLH proteins are represented by the Id family proteins (Id1, Id2, Id3, and Id4), which do not contain the basic region prior to the HLH motif and lack DNA-binding ability. The Id family proteins form heterodimers with the bHLH proteins, and inhibit their functions in a dominant-negative manner (Benezra et al, 1990; Sun et al, 1991; Yokota and Mori, 2002; Perk et al, 2005). The class-VI HLH proteins specifically contain a Pro residue in their basic region (Klambt et al, 1989; Rushlow et al, 1989). The class-VII HLH proteins are characterized by the bHLH–PAS domain, and include the aromatic hydrocarbon receptor and its nuclear translocator, as well as HIF1α, SIM, AhR, ARNT, and circadian clock-related factors (Crews, 1998). The Id family proteins of the dominant-negative class-V HLH (dnHLH) have attracted strong medical interest, based on the findings that deregulated Id activity is tumourigenic and contributes to malignancy, such as loss of differentiation, unrestricted proliferation, enhanced cell motility, and neoangiogenesis (Perk et al, 2005). Elevated expression of Id genes has been observed in carcinomas of various origins as well as melanomas and leukaemias (Ishiguro et al, 1995; Fong et al, 2004; Han et al, 2004; Wang et al, 2004). While the Id genes themselves are not canonical oncogenes, their overexpression affects key oncogenic pathways involving Ras, Myc, and ETS (Perk et al, 2005). However, the structures of the Id proteins and the mechanisms controlling their activities have remained unclear. Human homologue of murine maternal Id-like molecule (HHM), also known as cyclin D1-binding protein (DIP1), is a member of the dnHLH family, but it is larger than the Ids and includes a putative LZ motif and an acidic C-terminal region (Terai et al, 2000). By definition, HHM associates with cyclin D1 to regulate the G1/S-phase progression of hepatocytes, probably through the Rb pathway (Xia et al, 2000). Intriguingly, HHM exerts opposite effects on cells, depending on the cellular contexts. HHM was shown to be involved in the growth regulation of hepatocytes and the progression of hepatocellular carcinomas. In HHM-knockout mice, liver regeneration after partial hepatectomy was attenuated, as compared with that in wild-type mice (Ma et al, 2006). In addition, HHM expression is increased in the early phases of hepatocarcinogenesis, and HHM accelerates S-phase entry in HepG2 hepatocellular carcinoma cells (Terai et al, 2000). These observations indicate positive regulatory roles of HHM in proliferation of liver cells. In contrast, HHM-knockout mice often develop liver tumours (Sonnenberg-Riethmacher et al, 2007), and transgenic mice overexpressing HHM in the liver are less susceptible to chemical hepatocarcinogenesis (Ma et al, 2006). These observations indicate that HHM has tumour-suppressor functions in the liver. However, the mechanism underlying these opposite roles of HHM in the regulation of cell proliferation and tumour progression, which are dependent on the cellular context, remains to be elucidated. Recently, HHM was found to disrupt the physical interaction of specific transcription factors with R-Smads, to inhibit TGF-β signalling in a cellular response-specific manner (Ikushima et al, 2008). Oligodendrocyte transcription factor 1 (Olig1), a class-II bHLH protein, was identified as one of the Smad-binding transcription factors inhibited by HHM (Ikushima et al, 2008). We have observed that Olig1 regulates the expression of the TGF-β controlled genes that enhance cell motility and migration (Motizuki M & Miyazawa K, unpublished results). In contrast to the Id proteins, which interact with the ubiquitously expressed class-I bHLH transcription factors, HHM associates with the tissue-specific class-II bHLH transcription factor Olig1 to regulate Smad-dependent transcription, suppressing tumour progression. To understand the regulation mechanism of the dnHLH proteins, we solved the crystal structure of full-length HHM at 2.5 Å resolution. This is the first structure of a dnHLH, as well as the first presentation of the free-form of an HLH transcription regulator, devoid of its dimeric counterpart. The HHM structure adopts a V-shaped conformation composed of N- and C-terminal five-helix bundles. The hydrophobic dimerization interfaces of the HLH region are embedded in these helix bundles, and are thereby protected from nonspecific interactions. Combined with biochemical, physicochemical, and cell biological analyses, we propose that the present crystal structure of HHM represents an autoinhibited state, and the slow equilibrium with the partially unfolded conformation enables the fine-tuning of the specificity for the target transcription factor. Results and Discussion Structure determination The crystal structure of HHM (360 residues, molecular weight 40 kDa) was determined by the multiple anomalous diffraction method, using a selenomethionine-labelled crystal. We first obtained experimental phases to 4.0 Å resolution using a SeMet-labelled crystal, as previously described (Seto et al, 2009), but the quality of the resulting electron density map was quite low and the assignment of the amino-acid residues was virtually impossible. The addition of a sulfhydryl-specific reagent, p-chloromercuribenzoic acid (PCMB), to the protein sample prior to crystallization improved the resolution of the crystal to 3.5 Å, which enabled the preliminary interpretation of the electron density map. The exposed Cys residues may disturb the crystal packing, thus reducing the crystal quality. On the basis of this preliminary structural model, we substituted the Cys residues exposed on the molecular surface (Cys198 and Cys300; Supplementary Figure S1) with Ser, which further improved the resolution to 2.5 Å. The final model was refined against the diffraction data extending to 2.5 Å resolution with crystallographic R/R free=22.2/26.1%, by refinement of the individual B-factor and TLS tensor parameters. The model contains one molecule in the asymmetric unit, in which residues 1–15, 41–44, 139–150, 201–228, and 329–333 are structurally disordered. Overall structure HHM adopts an all α-helical structure consisting of two-helix bundles, which are arranged in a V-shaped conformation (Figure 1). Hereafter, we divide the HHM structure into the following three regions: the N-terminal helix bundle (N-bundle, residues 16–138), the HLH region (residues 151–200), and the C-terminal helix bundle (C-bundle, residues 229–360; Figure 1). The HLH region of HHM shares sequence homology with the other HLH proteins and consists of helices α5 and α6, which correspond to the first and second α-helices, respectively, of the canonical HLH motif and are connected by the short loop L5. The N-bundle consists of helices α1 to α4 and forms extensive hydrophobic interactions with helix α5 of the HLH region (Figure 2A). The C-bundle consists of helices α7 to α10 and forms extensive hydrophobic interactions with helix α6 of the HLH region (Figure 2B). The loop regions connecting the N-bundle (residues 139–150) and the C-bundle (residues 201–228) to the HLH region are structurally disordered. Figure 1.Structure of HHM. (A) Schematic representation of the HHM domain architecture. (B) Overall structure of HHM. (C) Schematic diagram of the secondary structures of HHM. In all of the panels, the same colour code as in panel A is used. Download figure Download PowerPoint Figure 2.Intramolecular interactions in free-standing HHM. (A) Hydrophobic interface between α5 and the N-bundle. (B) Hydrophobic interface between α6 and the C-bundle. (C) Intramolecular interactions between the N terminus of α6 and the N- and C-bundles at the edge of the V-shape. (D) van der Waals interactions involving the conserved NKAAA motif, stabilizing the V-shaped conformation to support the interactions between the HLH region and the N- and C-bundles. In all of the panels, the same colour code as in panel A is used. Download figure Download PowerPoint Previous studies suggested that the acidic domain and the putative LZ motif immediately follow the HLH motif, and these regions of HHM may be involved in intermolecular interactions (Hwang et al, 1997; Terai et al, 2000). This acidic domain is included in loop L6, which is between the HLH region and the C-bundle, and is mostly disordered in the present crystal structure. Moreover, the conserved Leu and Ser/Cys residues (Leu240, Leu247, Cys254, and Leu261) in this putative LZ motif on helix α7 participate in hydrophobic interactions with the core of the C-bundle, and do not form a canonical LZ structure (Supplementary Figure S2). Interactions between the HLH and helix bundles in the free-standing HHM In the present structure of HHM, helix α5 and loop L5 of the HLH region only interact with the N-bundle, while the N terminus of helix α6 forms extensive interactions with both the N- and C-bundles (Figure 2A and C, and Supplementary Figure S3). Especially, the N terminus of helix α6 harbours a sequence conserved in the HHM orthologues from various species, representing the 169NKAAA173 motif, which is likely to be important for the interactions between the HLH region and the N- and C-bundles (Figure 2C and D). This conserved NKAAA motif reinforces the interactions between the N- and C-bundles. The Nδ atom of the Asn169 side chain, which is located at the N terminus of helix α6, hydrogen bonds with the side-chain carboxyl group of Asp275 and the main chain carbonyl oxygen of Val271 in helix α8 of the C-bundle (Figure 2C). On the other hand, the Oδ atom of Asn169 hydrogen bonds with the main chain amide group of Leu114 in helix α4 of the N-bundle (Figure 2C). Furthermore, the side-chain amino group of Lys170 hydrogen bonds with the C-terminal carbonyl oxygen atoms of helix α7 in the C-bundle (Figure 2D). Finally, the side-chain methyl groups of the three consecutive Ala residues (Ala171, Ala172, and Ala173), which form the first turn of helix α6, closely pack against the hydrophobic core of the C-bundle (Figure 2D). Therefore, these interactions seem to be important for HHM to adopt the V-shaped conformation. In addition to the interactions described above, minor interactions that are independent of the HLH region occur between the N- and C-bundles (Figure 2C). The side-chain carboxyl group of Asp275 hydrogen bonds with the Oγ atom of Thr113 and the main chain amide groups of Ile112 and Thr113. The Cγ atom of Val271 makes a van der Waals contact with the Cα atom of Gly111. Gly111, Val271, and Asp275 are also conserved in the HHM orthologues from various species. These interactions anchor the N termini of helices α4 and α8, thereby stabilizing the V-shaped structure. Structural comparison between the HLH motifs of HHM and canonical bHLH transcription factors The canonical structures of the bHLH transcription factors reported to date form homo- or heterodimers through the conserved hydrophobic residues on the amphiphilic α-helices, H1 and H2 (Longo et al, 2008; Figure 3A). The C-terminal halves of helices H1 and the N-terminal halves of helices H2 form a short four-helix bundle, while the N-terminal halves of helices H1 provide a basic DNA-binding interface and the C-terminal halves of helices H2 form a two-helix bundle structure (Figure 3B). These dimeric DNA-bound structures of the bHLH transcription factors are considered as the 'active forms', which can activate the transcription of specific genes. Figure 3.Active dimer formation of HLH proteins. (A) The canonical heterodimer structure of the E47 and NeuroD1 bHLH transcription factors bound to DNA. (B) Typical dimerization interface of the E47 and NeuroD1 bHLH transcription factor complex, divided into five sections based on the hydrophobic interactions. (C) Sequence alignment of HLH proteins. Sections in the dimerization interface are coloured grey. The basic residues that interact with DNA are coloured blue. (D) Schematic diagram of each section in the dimerization interface between E47 and NeuroD1. (E) Putative schematic diagram of each section in the dimerization interface between HHM and Olig1. Download figure Download PowerPoint In contrast to these dimeric HLH proteins, the crystal structure of the free-standing HHM revealed that HHM does not form a dimer, and the conserved hydrophobic residues of helices α5 and α6 (corresponding to H1 and H2 in the canonical bHLH, respectively) separately participate in the hydrophobic core formation with the N- and C-bundles (Figure 2A and B). The arrangement of helices α5 and α6 is quite different from that of H1 and H2 observed in the canonical HLH transcription factors (Figures 1B and 3A). There is no contact between helices α5 and α6, whereas helices H1 and H2 of the dimerized HLH transcription factors form intramolecular interactions. Instead, the HLH region of HHM bridges the N- and C-bundles, stabilizing the V-shaped conformation (Figure 1B). Conservation in the HLH region allows active heterodimer formation in HHM It is unlikely that the HLH region in the present V-shaped HHM can interact with another HLH protein without undergoing a structural change, as its molecular surface is mainly hydrophilic, and there is no hydrophobic cluster on the surface suitable for the interactions (Supplementary Figure S4). On the basis of the sequence similarity between the bHLH domains of transcription factors and the dnHLH domains of the Id family proteins, the dnHLH and bHLH domains are considered to form a heterodimer, similar to the active forms of the bHLH transcription factors (Wibley et al, 1996). By analogy, can the HLH region of HHM also form a similar heterodimer to those of the transcription factors? Chavali et al (2001) proposed some of the important positions for the residues involved in the dimeric stability and the functional specificity of the DNA-bound 'active form' of HLH proteins. Especially, positions 8′ and 11′ in helix H1 and 4″, 5″, 8″, 11″, 12″, 15″, and 19″ in helix H2 are important for the stable packing of the core residues, at the interface of the short four-helix bundle and two-helix bundle structures (Figure 3C). Here, we divided the dimerization interface into five sections, from I to V (Figure 3B and C). These sections are depicted schematically in Figure 3D, using the complex structures of NeuroD1 and E47 (Longo et al, 2008) as an example. The HLH region of HHM shares amino-acid sequence similarity with the bHLH transcription factors, as well as other dnHLH transcriptional regulators (Figure 3C). In HHM, sections I to V are also occupied by similar amino acids to those observed in the canonical HLH proteins (Figure 3E). In sections I and II, positions 4″ and 8″ of HHM are replaced by larger residues (Met and Asn) as compared to the canonical HLH proteins, while the corresponding interaction partners (positions 8′ and 11′) are replaced by smaller residues (Val and Ala; Figure 3D and E). Therefore, upon binding to the target transcription factors, the HLH region of HHM can form the heterodimeric 'active' structure, with the interface stabilized by these complementary residues. Again, it should be noted that these putative interface residues of HHM are involved in the core formation of the N- and C-bundles in the present free-standing structure. In addition, position 1″ in helix H2 is highly conserved as Lys or Arg in the bHLH transcription factors (Figure 3C), which coincides with the observations that the basic residue at this position interacts with the backbone phosphates of DNA in the reported crystal structures. In contrast, this position is replaced by Ala (the last Ala173 in the NKAAA motif) in HHM. This substitution seems to be reasonable, as HHM no longer interacts with DNA. In the case of Id3, another member of the dnHLH family, this position is occupied by a non-basic Gln residue. Furthermore, in many HLH proteins, including Myc, Max, and the Id family proteins, the amino-acid sequences of the loop region between helices H1 and H2 also share significant similarity, although they are not directly involved in dimer formation. The third and penultimate positions of the loop region are conserved as small hydrophobic residues (Figure 3C), forming the core of the loop region in the three-dimensional structure (Figure 3A). Thus, these conserved residues are important for the active dimer formation. In HHM, these positions are also conserved as small hydrophobic residues (Ile165 and Ala171). Moreover, the last four residues of the loop region (169NKAA174, the first four residues of the NKAAA motif), which form the N-terminal turn of helix α6 (H2) in the present structure, share considerable sequence similarity with the loop region of the bHLH transcription factors, including v-Myc (EKAA) and Max (EKAS) (Figure 3C). Although HHM is bound to the class-II bHLH transcription factors and thus may correspond to the class-I HLH proteins, we hypothesize that this loop region of HHM may form a similar structure to those of the class-III and IV transcription factors, such as Myc and Max (Nair and Burley, 2003). Therefore, the present helical structure (α6) of these four N-terminal residues of HHM may be restructured into the loop conformation upon binding with the target transcription factors. On the basis of these observations, we constructed a docking model of Olig1–bHLH and the HLH region of HHM (Figure 4A). In HHM, helix H2 is longer than that of the canonical HLH structure, suggesting that the C terminus of helix H2 of HHM may become disordered upon complex formation. Similarly, the basic region of Olig1 in the complex may be disordered without bound DNA. Previous in vitro experiments demonstrated that the HLH region of HHM alone is sufficient to exclusively interact with the HLH region of the class-II bHLH transcription factor, Olig1 (Ikushima et al, 2008), supporting this docking model. Therefore, although the complex structure of the HLH regions of HHM and Olig1 is still not available, we propose that HHM disrupts the Smads–Olig1 complex by forming an HHM–Olig1 heterodimer via the HLH regions, as in this docking model. Furthermore, this model suggests that the V-shaped form of HHM should undergo a drastic structural change to form this heterodimeric complex. Figure 4.Conformational transition of HHM upon association with the target transcription factors. (A) Docking model of Olig1–bHLH and the HLH region of HHM. The regions that are presumably disordered are semitransparent. (B) Model of the conformational transition of HHM upon association with and dissociation from the target transcription factors. Download figure Download PowerPoint Equilibrium between the V-shaped and relaxed conformations To investigate the stability of the V-shaped form of HHM in solution, we incorporated two TEV protease recognition sites into the disordered loops L4 and L6 of the N-terminally GST-tagged HHM (Figure 5A), to enable the separation of the GST-tagged N-bundle, the HLH region, and the C-bundle. We then performed a GST pull-down assay following proteolysis with TEV protease (Figure 5B). If the tertiary interactions between the HLH region and the N- and C-bundles are tight enough to form the stable V-shaped structure, then the C-bundle (and HLH region) will be pulled down together with the GST-tagged N-bundle. Unexpectedly, the HLH region and the C-bundle were not co-precipitated with the