Article - Cell Press

Structure

Article DNA Recognition Mechanism of the ONECUT Homeodomain of Transcription Factor HNF-6 Daisuke Iyaguchi,1,3 Min Yao,1 Nobuhisa Watanabe,1 Jun Nishihira,2 and Isao Tanaka1,* 1

Faculty of Advanced Life Sciences, Hokkaido University, Kita-10, Nishi-8, Sapporo, 060-0810, Japan Genetic Lab Co., Ltd., Kita-9, Nishi-15, Sapporo, 060-0809, Japan 3 Present address: Faculty of Pharmaceutical Sciences, Health Sciences University of Hokkaido, 1757 Kanazawa, Ishikari-Tobetsu, 061-0293, Japan. *Correspondence: [email protected] DOI 10.1016/j.str.2006.11.004 2

SUMMARY

Hepatocyte nuclear factor-6 (HNF-6), a liverenriched transcription factor, controls the development of various tissues, such as the pancreas and liver, and regulates the expression of several hepatic genes. This protein belongs to the ONECUT class of homeodomain proteins and contains a bipartite DNA-binding domain composed of a single cut domain and a characteristic homeodomain. This transcription factor has two distinct modes of DNA binding and transcriptional activation that use different coactivators depending on the target gene. The crystal structure of the bipartite DNAbinding domain of HNF-6a complexed with the HNF-6-binding site of the TTR promoter revealed the DNA recognition mechanism of this protein. Comparing our structure with the DNAfree structure of HNF-6 or the structure of Oct-1, we discuss characteristic features associated with DNA binding and the structural basis for the dual mode of action of this protein, and we suggest a strategy for variability of transcriptional activation of the target gene. INTRODUCTION Hepatocyte nuclear factor-6 (HNF-6) is expressed in tissues, such as the liver, pancreas, brain, and spinal cord (Landry et al., 1997), and controls differentiation and morphogenesis (Bouzin et al., 2003; Clotman et al., 2002; Odom et al., 2004; Samadani and Costa, 1996; Servitja and Ferrer, 2004). This protein also regulates the transcription of genes that are required for glucose metabolism (Lannoy et al., 2002; Lemaigre et al., 1993; Streeper et al., 2001; Tan et al., 2002) and genes that encode plasma transport proteins (Samadani and Costa, 1996), and it mediates the effects of growth hormones (Lahuna et al., 1997; Rastegar et al., 2000; Sasaki et al., 1999). HNF-6 belongs to the ONECUT class of homeodomain proteins, a family of proteins with a bipartite DNA-binding

domain composed of a single cut domain and a homeodomain (Lannoy et al., 1998; Lemaigre et al., 1996). Homeodomain proteins constitute a large family, which can be divided into several classes based on the proteins’ amino acid sequences as well as their association with other DNA-binding domains. The ONECUT class is characterized by the presence of a cut domain and by a homeodomain that contains an F48M50 dyad (Lannoy et al., 1998), consisting of phenylalanine and methionine residues at positions 48 and 50, respectively. In other homeodomains, position 48 forms part of the hydrophobic core, and position 50 is essential for sequence-specific DNA binding (Ades and Sauer, 1994; Laughon, 1991). The homeodomain of HNF-6 has a dual role depending on the target gene (Lannoy et al., 1998, 2000). It is involved in binding to the transthyretin (TTR) promoter, whereas it is not required for binding to the HNF-3b promoter; it is also involved in transcriptional activation. The cut domain is involved in both DNA binding and transcriptional activation at both promoters. The cut domain of the ONECUT proteins contains the LXXLL motif, which is involved in interactions with coactivators (Gu et al., 1997; Heery et al., 1997). In the HNF-3b promoter, transcriptional activation involves both the LXXLL motif of the cut domain and the F48M50 dyad of the homeodomain, and HNF-6 interacts with CREB-binding protein (CBP) as a coactivator via the LXXLL motif and the F48M50 dyad. In contrast, in the TTR promoter, HNF-6 interacts with p300/CBP-associated factor (p/CAF) as a coactivator, and neither the LXXLL motif of the cut domain nor the F48M50 dyad of the homeodomain is involved in the interaction with p/ CAF (Lannoy et al., 2000). Thus, the HNF-6 protein has a dual mode of DNA binding and transcriptional activation depending on the target gene. Recently, the DNA-free structure of HNF-6a was determined by NMR (Sheng et al., 2004). It was shown that HNF-6 has two distinct domains, the cut domain and the homeodomain, which are connected by a long, flexible linker. The cut domain folds into a topology homologous to that of the POU-specific domain of Oct-1, another bipartite DNA-binding protein (Assa-Munt et al., 1993; Klemm et al., 1994). However, little information is available about the interaction with DNA and the structural basis for the dual mode of action of HNF-6 depending on the target gene.

Structure 15, 75–83, January 2007 ª2007 Elsevier Ltd All rights reserved 75

Structure Crystal Structure of HNF-6 Complexed with DNA

Figure 1. Structural Overview of the DNA-Binding Domain Complexed with DNA (A) An overview of the HNF-6-DNA complex. The cut domain and the homeodomain of HNF-6 are shown in light blue and yellow, respectively. Helices 1–9 are labeled. (B) Two overviews of the Oct-1-DNA complex related by 70 rotation about the axis parallel to the helical axis of the DNA showing that the POU domain (left figure) and the homeodomain (right figure) can be superimposed with the cut domain and the homeodomain of (A). The POU-specific domain and the homeodomain of Oct-1 are shown in green and orange, respectively. (C) HNF-6 and Oct-1 are shown in the same colors as in (A) and (B). HNF-6 (left) and Oct-1 (right) are related by 90 rotation about the horizontal axis to (A) and the right view of (B), respectively.

The crystal structure of the HNF-6a DNA-binding domain bound to its cognate DNA reveals how HNF-6 discriminates between two types of target genes. Furthermore, our findings suggest the structural basis for the strategy by which the transcription factor with a bipartite DNA-binding domain generates variability of transcriptional activation depending on the target gene. RESULTS Overall Structure of HNF-6 Complexed with DNA HNF-6 consists of two distinct domains, the cut domain and the homeodomain, which are connected by a long linker. In the complex structure, these two domains, together with the linker region, wrap around DNA and make contact with each other (Figure 1). The linker region, which tethers the two DNA-binding domains, could not be observed in the electron density map, indicating that the linker has no interaction with the two domains or with DNA. The direct interactions between the two domains are limited (several van der Waals contacts). However, the two domains bind side by side to the same major groove of the DNA, and it seems that both domains affect

DNA binding by modifying the DNA structure (as described later). We believe that this is the origin of the cooperative binding observed in the previous experiments (Lannoy et al., 1998). Oct-1 also has a bipartite DNA-binding domain consisting of a POU-specific domain and a homeodomain (Assa-Munt et al., 1993). Figure 1 shows a structural comparison of the DNA-bound structures of HNF-6 and Oct-1 (Klemm et al., 1994). Both the POU-specific domain and the homeodomain of Oct-1 bind directly to independent sites on DNA in a way similar to that of HNF-6. Indeed, the binding of each domain to each binding site is identical in a general sense. Thus, each domain can be superimposed, as shown in Figure 1. However, the relative positions of the two domains on DNA (the cut domain and the homeodomain in HNF-6, and the POU-specific domain and the homeodomain in Oct-1) are different by about 70 (Figure 1B), and the two domains in HNF-6 come into closer apposition than in Oct-1 (Figure 1C). From the rotation angle of 70 around B-DNA (ten bases per turn), it is estimated that the recognition sites of the two domains in HNF-6 are closer by two base pairs than in Oct-1. In the HNF-6-DNA complex, adenine-4 and adenine-8 are

76 Structure 15, 75–83, January 2007 ª2007 Elsevier Ltd All rights reserved


Figure 2. Sequence of the DNA Duplex Used for Cocrystallization and of the DNA-Binding Domain of HNF-6 (A) DNA duplex used for cocrystallization. Base pairs 1–13 are from the rat TTR promoter. The 50 -half subsite and the 30 -half subsite are boxed. Asterisks indicate the positions replaced by 5-bromouridines for phasing. (B) DNA duplex used for cocrystallization of Oct-1. The ATGC subsite (50 half) and the AAAT subsite (30 half) of an octamer site are boxed (Klemm et al., 1994). (C) The amino acid sequence of the DNA-binding domain of HNF-6 is shown; the single-letter amino acid code is used. The regions corresponding to the cut domain and the homeodomain are boxed, and the linker segment is labeled. Helices seen in the crystal structure are indicated below the sequence.

recognized by Gln-36 and Asn-147, the most highly conserved residues located within the recognition helices of the cut domain and the homeodomain, respectively (Figure 2A, also see Figure 5). Correspondingly, adenine1 and adenine-7 of the Oct-1-DNA complex are recognized by the conserved residues Gln-44 and Asn-150 of the POU-specific domain and the homeodomain of Oct-1 (Klemm et al., 1994) (Figure 2B). These residues correspond to Gln-36 and Asn-147 of HNF-6, respectively. The distances between the two adenine bases recognized by the two domains are thus different by two bases between HNF-6 and Oct-1 (four bases in HNF-6 and six bases in Oct-1). This finding agrees with the estimate from the rotation value mentioned above. Structure and DNA Recognition of the Cut Domain The cut domain consists of five well-folded a helices (helices 1–3 and helices 5–6) and a small 310 helix (helix 4) (Figures 1A and 2C). Helix 3 fits into the major groove of the DNA; this is the ‘‘recognition helix’’ of the cut domain. The hydrophobic central core is formed mainly by residues Val-9, Ala-10, Ile-13, and Leu-17 from helix 1 and residues Phe-59, Met-62, Trp-65, and Leu-66 from helix 5. Helix 2 is mounted to the central core with residue Phe-27 from helix 2 and residues Ile-22, Val-31, and Leu-32 from the loops. The DNA-free NMR structure reported previously has almost the same features in this hydrophobic core (Sheng et al., 2004). By comparing our DNA-bound structure of the cut domain with the DNA-free NMR structure, we find that extensive rearrangement of the main chain was observed upon DNA binding at the regions of helices 1, 2, and 3

(Figure 3A). Compared to the DNA-free structure of the cut domain, helix 2 is shifted by about 5 A˚ toward helix 3, and the C terminus of helix 1 is straighter and slightly longer in our DNA-bound structure. Helix 3, the recognition helix, is slightly longer in the DNA-bound form, and its direction is changed by about 70 . It is noteworthy that the POU-specific domain of Oct-1 does not undergo such a large structural change upon DNA binding (AssaMunt et al., 1993; Klemm et al., 1994) (Figure 3B). The structure of the cut domain after the extensive conformational changes upon DNA binding can be superimposed on the structure of the POU-specific domain (Figure 3C). However, helix 4 and helix 6 were observed only in our DNA-bound structure (Figure 3C). These two helices were not observed in either the DNA-bound structure of the POU-specific domain of Oct-1 or the DNA-free structure (Figure 3B). Thus, the structural changes induced by DNA binding, such as the positional change of helix 2 and the directional change of helix 3 as well as the generation of helices 4 and 6, as mentioned above, are specific to the cut domain. There is almost no amino acid sequence identity between the cut domain and the POU-specific domain. However, careful comparison of the two structures and alignment of their amino acid sequences revealed some identical residues in these domains (Figure 4). Most of these identical residues are involved in the formation of the hydrophobic core or in DNA recognition. Ala-10, Trp65, and Leu-66 (Ala-13, Trp-70, and Leu-71 in the POUspecific domain) form the hydrophobic core in both the cut domain and the POU-specific domain. These residues are conserved in both HNF-6 and members of the POU



Figure 3. Structural Comparison of the Cut Domain and the POU-Specific Domain (A) Structural comparison of the cut domains of the DNA-bound and DNA-free structures. The cut domain of the DNA-bound crystal structure is shown in light blue, and that of the DNA-free NMR structure is shown in pink. Helices 1–6 of the cut domain are labeled. (B) Structural comparison of the POU-specific domains of the DNAbound and DNA-free structures. The POU-specific domain of the DNA-bound crystal structure of Oct-1 is shown in green, and that of the DNA-free NMR structure is shown in red. (C) Structural comparison of the cut domain and the POU-specific domain. The cut domain of the DNA-bound crystal structure of HNF-6 is shown in light blue. The POU-specific domain of the DNA-bound crystal structure of Oct-1 is shown in green. Helices 1–6 of the cut domain are labeled.

family; Ala-10 is invariant in the POU family, and Trp-65 is completely invariant in both families (Assa-Munt et al., 1993; Sheng et al., 2004). The residues of the cut domain involved in the interaction with DNA are shown in Figure 5. The cut domain binds to DNA by many hydrogen bonds, directly or indirectly via water molecules, and by van der Waals interactions with the bases and sugar phosphate backbone. Of note, the DNA sequence involved in the contact with the cut domain

is wide and covers the entire 10 bp recognition site by itself (Figure 5). The 10 bp recognition site can be divided into the 50 -half subsite containing TATTG and the 30 -half subsite containing ACTTA (Figure 2A). The sequence-specific base contacts occur through two regions of the cut domain—helix 3 and the loop after helix 4 (Figures 4 and 5). For the 50 -half subsite, helix 3 is inserted into the major groove of the DNA, allowing 3 residues (Ser-35, Gln-36, and Gly-37) located on the N-terminal half of this helix to make hydrogen bonds with three bases (adenine-60 , adenine-4, and adenine-50 ); the interactions between Ser-35 and Gln-36 with the bases are especially direct and are not mediated by water. These residues play a major role in recognizing the 50 -half subsite. Three residues (Ser40, Asp-41, and Arg-44) from the C-terminal half of helix 3, 2 residues (Gln-24 and Ala-25) from the N terminus of helix 2, and Lys-18 from the C terminus of helix 1 make contact with the backbone in this 50 -half subsite. The extended recognition of DNA by the cut domain can be understood by investigating the recognition of the 30 half subsite. In the POU-specific domain complexed with DNA, the N-terminal portion of the C-terminal helix recognizes the 30 terminus of the 50 -half subsite (Klemm et al., 1994). In the case of the cut domain of helix 4 (Lys-51), the loops flanking this helix (Lys-47 and Lys-53) and the C-terminal helix 5 (Gly-55 and Thr-58) make contact with the 30 -half subsite of DNA (Figure 5). It is clear that the generation of helix 4 makes it possible for contacts to be made with the distant region from the site recognized by helix 3. This DNA recognition mechanism is specific to the cut domain of HNF-6. It is notable that only Lys-53 has water-mediated base contacts with adenine-100 and thymine-11, and that all other interactions in the 30 -half subsite by the cut domain are phosphate and sugar backbone contacts, indicating that the recognition of this 30 -half subsite by the cut domain is accomplished mainly by backbone contact. A comparison of the residues involved in DNA recognition between the cut domain and the POU-specific domain is shown in Figure 4. The identical residues, which make base and/or backbone contacts in both domains, are Gln-36, Ser-40, Gln-24, and Ser-35. These 4 residues have the same roles in DNA recognition as those of the POU-specific domain. In the interactions of the cut domain with DNA, the side chain of Gln-36 forms a hydrogen bond with adenine-4. The side chain of Ser-40 makes hydrogen bonds with the phosphodiester oxygen of adenine-4. Gln24 makes contact with the phosphates of adenine-4 and thymine-3 located on the 50 side (Figure 5). In the case of the POU-specific domain, residues Gln-44, Ser-48, and Gln-27, which are counterparts of Gln-36, Ser-40, and Gln-24, respectively, in the cut domain, recognize the adenine located at the 50 terminus of the octamer site and the thymine located on the 50 side of this adenine (Klemm et al., 1994) in a manner similar to that of the cut domain. Ser-35 of the cut domain and the corresponding Ser-43 of the POU-specific domain also recognize the same phosphate, which is located at the 30 terminus of the 50 -half subsite. It is notable that these 4 residues are highly conserved in both



Figure 4. Sequence Comparison of the Cut Domain of HNF-6 and the POU-Specific Domain of Oct-1 Residue numbers and secondary structures in the DNA-binding structure of the cut domain and the POU-specific domain are shown above and below each sequence, respectively. Residues involved in the formation of the hydrophobic core are shown in red. Residues hydrogen bonding to backbone atoms and bases of the DNA are shown in light blue and green, respectively. Residues hydrogen bonding to both backbone atoms and bases are shown in purple. Residues identical between the two domains are boxed.

the cut and POU-specific domains; Gln-24 and Gln-36, especially, are invariant residues in CUT superclass proteins, including the CUX, SATB, and ONECUT class proteins (Assa-Munt et al., 1993; Jacquemin et al., 2003; Lannoy et al., 1998; Sheng et al., 2004). Structure and DNA Recognition of the Homeodomain The homeodomain of HNF-6 consists of three a helices (helices 7–9) and docks with DNA by inserting helix 9, the ‘‘recognition helix,’’ into the major groove of the DNA (Figures 1A and 2C). The N-terminal arm also makes contact in the minor groove. The overall folding arrangement and DNA docking of HNF-6 are similar to what is seen in other homeodomains, such as those of antennapedia (Otting et al., 1990), engrailed (Kissinger et al., 1990), MATa2 (Wolberger et al., 1991), Oct-1 (Klemm et al., 1994), and paired (Xu et al., 1995). However, DNA base contacts are mediated by the C-terminal portion of helix 9, in contrast to what occurs in other homeodomains, in which the N-terminal portion of the recognition helix makes base contacts. The interactions between the homeodomain and DNA are shown in Figure 5. The base contacts by helix 9 are accomplished by Arg-151, Asn-147, Asn-143, and Arg-150, which hydrogen bond to guanine-7, adenene-8, cytosine-9, and guanine-90 , respectively. The phosphate contacts by Asn-147, Thr-140, and Asn-143 with guanine-7, adenine-8, and cytosine-9, respectively, stabilize the interactions with DNA in this region. The N-terminal arm of the homeodomain fits into the minor groove of the DNA, as observed in other homeodomain-DNA complexes, and forms water-mediated base contacts with thymine-5 and thymine-6 and backbone contacts with thymine-6 and thymine-40 . All of these contacts are made only by Arg-101. In the homeodomains, base contacts are typically concentrated in the N-terminal arm and the recognition helix. In many cases, residue 5 from the N-terminal arm and residues 47, 50, 51, and 54 from the recognition helix (numbered according to the conventional homeodomain scheme [e.g., Billeter, 1996]) are involved in the base contacts. All of these residues are also involved in the recognition of the bases in the homeodomain of HNF-6 (corresponding to Arg-101, Asn-143, Met-146, Asn-147, and Arg-150, respectively) (Figure 6). The interaction of Asx51 (Asn-147 in HNF-6) with an adenine located at the third

position in the TAAT consensus is conserved in homeodomains (Billeter, 1996; Kissinger et al., 1990; Klemm et al., 1994; Wolberger et al., 1991). Although the DNA sequence recognized by HNF-6 contains the canonical TAAT consensus sequence at the 50 -terminal end, Lannoy et al. (1998) mentioned that the sequence is not crucial for HNF-6 binding. Indeed, in our complex structure, the binding site of the homeodomain is not located on the TAAT sequence. Asn-147 makes contact with adenine-8 (Figures 2 and 5). Met-146 corresponding to position 50 makes contact with N6 of adenine-100 (Figure 6). To our knowledge, this recognition of an adenine base by a methionine residue has never been found elsewhere and is specific to HNF-6. In HNF-6, the interactions with DNA are concentrated on one chain of the DNA, and contacts with the other chain are made only by residue Arg-101 from the N-terminal arm and residue Arg-150, located at the C-terminal end of helix 9. This is due to structural features, in that the recognition helix of the homeodomain is docked in a position shifted slightly to one side of the major groove. In other homeodomains, the recognition helix is docked at the center of the major groove, and the N-terminal portion is buried in the groove. Such characteristic docking influences the relative position of helix 8, which interacts with DNA at the N-terminal portion in other homeodomains, and prevents it from making contact with DNA. In other homeodomains, several regions, including the N terminus, the loop connecting the first and second helices, the start of the second helix, and residues from all parts of the recognition helix, are involved in backbone contacts and increase the stability of DNA binding (Wilson et al., 1996). However, in the homeodomain of HNF-6, backbone contacts are formed only by Arg-101 from the N-terminal arm and by some residues from the recognition helix, indicating that the DNA binding of the homeodomain in HNF-6 is less stable compared with other typical homeodomains. It was reported that the homeodomain of HNF-6 is unable to bind to DNA on its own. This may be the result of the loss of sufficient backbone contacts, which play an important role in stabilization of the complex with DNA in other homeodomains. It is notable that the DNA site recognized by the homeodomain of HNF-6 is located at the center of the cut domain recognition site, and that interactions by both domains clearly overlap (Figure 5). In our structure, DNA



Figure 6. Detailed View of the Homeodomain-DNA Interactions Interactions of Asn-143, Met-146, Asn-147, and Arg-150 with DNA are shown. Dashed red lines indicate hydrogen bonds. van der Waals interactions are indicated as dashed blue lines. Water molecules are shown as green spheres. Residue numbers and base numbers are labeled.

homeodomain and helix 3 of the cut domain. This indicates that two domains bind DNA cooperatively via the conformational changes of DNA. Lannoy et al. (1998) reported that neither the cut domain nor the homeodomain can bind to TTR promoter DNA on its own. In their experiments, it was shown that HNF-6 deletion mutants of either of the two domains abolished binding to DNA by EMSA (Lannoy et al., 1998). Our conclusion based on the structure was supported by the results of their experiments. DISCUSSION

Figure 5. Protein-DNA Contacts Schematic diagram of protein-DNA contacts. Residues from the cut domain are shown in light blue, and the homeodomain is shown in yellow. The bases, phosphates, and sugars contacted by each domain are indicated in the corresponding colors. The bases contacted by both domains are shown in red. Solid black lines indicate hydrogen bonds. van der Waals interactions are indicated as dashed blue lines. Water molecules are shown as green spheres.

backbones around base pairs adenine-4 and thymine-40 or adenine-8 and thymine-80 are distorted, and the minor groove, to which the N-terminal arm of the homeodomain is inserted, is expanded by about 1.5-fold (Figure 7). The DNA region around adenine-4 and thymine-40 is recognized by helix 3 of the cut domain and the N-terminal arm of the homeodomain, and the region around adenine-8 and thymine-80 is recognized by helix 9 of the

HNF-6 protein is expressed mainly in the liver and regulates the transcription of a number of genes. The TTR and HNF-3b genes are representative genes whose transcription is regulated by HNF-6. HNF-6 was reported to bind to the promoters of these genes in different ways even though the recognition sequences are very similar; for the TTR promoter, both the cut domain and the homeodomain bind to DNA, whereas only the cut domain binds to DNA for the HNF-3b promoter. The different binding of HNF-6 to the two promoters was analyzed by EMSA experiments. A deletion mutant of the cut domain abolished binding to both promoters. However, deletion of the homeodomain did not prevent binding to the HNF-3b promoter, in contrast to what occurs with the TTR promoter, in which the homeodomain is required for DNA binding (Lannoy et al., 1998). It was also reported that in transcriptional activation HNF-6 interacts with different coactivators in the two promoters; HNF-6 interacts with the coactivators p/CAF and CBP for the TTR and HNF-3b promoters, respectively. Lannoy et al. transfected rat hepatoma FTO2B cells with the HNF-6 expression vector in the presence of the p/CAF or CBP expression vectors and with a reporter construct that contains the HNF-6 site of either the TTR or HNF-3b promoters. In these experiments,



The results of the present study provide a structural basis for activation of various promoters by transcription factors with bipartite DNA-binding domains. In the case of HNF-6, the cooperative DNA binding by two domains, the conformational change of the cut domain, and the variable interactions with different coactivators make it possible for the protein to generate variability of transcriptional activation depending on the target promoter sequence. EXPERIMENTAL PROCEDURES

Figure 7. Conformational Change in DNA Backbone structures of DNA complexed with HNF-6 and Oct-1 are shown. HNF-6 (left) and Oct-1 (right) are shown in the same colors as in Figure 1.

p/CAF activated transcription of the TTR-type reporter, whereas CBP activated that of the HNF-3b-type reporter (Lannoy et al., 2000). In the present study, we determined the crystal structure of HNF-6 complexed with the TTR promoter, and we showed that the cut domain and the homeodomain bind to DNA cooperatively via a conformational change of the DNA. By comparison with the previously reported DNA-free structure, we also found that the cut domain undergoes extensive structural change upon DNA binding. In our complex structure, DNA recognition of HNF-6 covers from position 3 to position 11. There is only a single base pair difference between this recognition sequence and the HNF-3b promoter sequence (cytosine-9 and guanine-90 in the TTR promoter are replaced by thymine and adenine, respectively, in the HNF-3b promoter). This base pair in the TTR promoter is recognized by the homeodomain. HNF-6 protein was shown to recognize this single base pair difference and changes the mode of DNA binding with a different coactivator. For the HNF-3b promoter, the homeodomain of HNF-6 does not bind to DNA, and the cut domain binds on its own. The LSDLL sequence of the cut domain and the F48M50 dyad (Phe-144 and Met-146 in this paper) of the homeodomain interact with CBP as a coactivator for the HNF-3b promoter. This was verified by GST pulldown experiments of mutants (Lannoy et al., 2000). The LSDLL sequence and the F48M50 dyad are located in helix 3 of the cut domain and helix 9 of the homeodomain, respectively. In the complex structure with the TTR promoter, both helices are inserted into the major groove of the DNA and are not available for interaction with any coactivator. However, on the HNF-3b promoter, the F48M50 dyad of the homeodomain can access the coactivator because only the cut domain binds to DNA for the HNF-3b promoter. Furthermore, the cut domain binds to the HNF-3b promoter without any cooperative effect of the homeodomain, suggesting a different manner of DNA binding from that of the TTR promoter, and that the LSDLL sequence may become available for interaction with a coactivator. This is supported by the large conformational change of the cut domain observed in the present DNA-complex structure.

Preparation of Proteins and Formation of the Complex with DNA The DNA-binding domain of rat HNF-6a (residues 289–444) was expressed in Escherichia coli B834 (DE3) cells by using the plasmid pET22b (Novagen). The histidine-tagged proteins were purified by three steps of column chromatography by using HiTrap Chelating HP charged with nickel ions, HiTrap SP HP, and HiLoad Superdex 75 pg columns (all from Amersham Pharmacia Biosciences). The fractions containing the DNA-binding domain of HNF-6a were dialyzed against buffer containing 10 mM HEPES (pH 7.5) and 100 mM KCl and were concentrated to 10 mg ml1. Synthesized oligodeoxyribonucleotides used for complex formation were obtained commercially (Hokkaido System Science). Equimolar amounts of the complementary strands were mixed at a final concentration of 4 mM in annealing buffer containing 10 mM Tris-HCl (pH 7.5), 100 mM KCl. The mixture was heated at 90 C for 5 min and then cooled slowly to 4 C overnight for annealing. The double-stranded DNA solution was then added to the HNF-6a solution, yielding a final DNA:protein ratio of 1.4:1.0. Crystallization and X-Ray Data Collection Crystal Screen Kits I and II (Hampton Research) were used to determine the crystallization conditions. Cocrystals were screened at 20 C by using the sitting-drop vapor-diffusion method, in which 1 ml complex solution was mixed with 1 ml reservoir solution and allowed to equilibrate against 0.1 ml reservoir solution. A derivative for solving the structure by using the MAD/SAD method was prepared simply by substituting two thymine positions (thymine-2 and thymine-11) with 5bromouridine (Figure 2A). After optimization, both native and derivative crystals suitable for X-ray diffraction measurement grew to dimensions of 0.05 3 0.1 3 0.5 mm within 1 week from 0.1 M sodium acetate (pH 4.6), 0.4–0.7 M 1,6-hexanediol, and 0.01 M CoCl2 by the hanging-drop vapor-diffusion method at 20 C. X-ray diffraction data of native crystals were collected to a resolution of 2.0 A˚ on beamline BL6A at the Photon Factory (Tsukuba, Japan). The diffraction data from the bromo derivative were collected on beamline BL44B2 at SPring-8 (Harima, Japan). These diffraction experiments were performed under cryocooling conditions (100 K) after the crystals were soaked in cryoprotectant consisting of 20% glycerol in the reservoir solution. The diffraction images were analyzed and processed by using the program HKL2000 (Otwinowski and Minor, 1997). The native crystals belonged to the space group P2 with unit cell parameters of a = 73.0 A˚, b = 39.0 A˚, c = 106.5 A˚, and b = 107.6 , indicating two protein-DNA complexes per asymmetric unit with a VM value of 2.6 (A˚3/Da) based on an estimated molecular mass of 28,000 Da (Matthews, 1968). Data collection statistics are summarized in Table 1. Structure Determination and Refinement The structure of the DNA-binding domain of HNF-6a complexed with DNA was determined by using the Br-SAD technique. All four bromine sites were identified, and the initial phase was calculated by using the program SOLVE (Terwilliger, 1994). The initial model containing 70% of the residues was traced automatically with the program RESOLVE (Terwilliger, 2000). After manual fitting of the DNA structure, the complex structure was further refined automatically by using the program LAFIRE (Yao et al., 2006) with CNS (Brunger et al., 1998). After a final



Table 1. Statistics of the Structure Analysis Native

Br Peak

Wavelength (A˚)

0.97799

0.91850

Unique reflections

39,224

31,717

Resolution (A˚)

50.0–2.0 (2.07–2.00)

50.0–2.15 (2.23–2.15)

Completeness (%)

99.7 (99.8)

98.7 (96.6)

Redundancy

5.6 (5.3)

7.1 (6.6)

Rmergeb

5.7 (34.8)

5.7 (30.9)

Data Collection Statisticsa

(%)

Refinement Statistics Resolution (A˚)

10.0–2.0

R factorc/Rfreed (%)

24.5/26.9

Rmsd bond lengths (A˚)

0.0057

Rmsd bond angles ( )

1.2

Average B factor (A˚2)

36.7

Rmsd, root-mean-square deviation. a Values in parentheses are for the highest-resolution shell. b Rmerge = (ShSij I(h)ij)/ShSi I(h)i, where I(h)i is the ith observation of reflection h, and is the mean intensity of all observations of reflection h. c R factor = (SkFoj jFck)/SjFoj, where jFoj and jFcj are observed and calculated structure factor amplitudes, respectively. d Rfree was calculated for 10% of the randomly selected reflections of data sets that were not used in the refinement. check and manual interventions in model modification with the program O (Jones et al., 1991), the structure was refined to 2.0 A˚ resolution with an R factor of 24.5% and an Rfree of 26.9% by using CNS. The refined model of the DNA-binding domain of HNF-6a complexed with DNA contained 270 amino acid residues, 195 water molecules, 2 acetate ions, and 2 glycerols. The residues in the loop region and the C terminus in the DNA-binding domain of HNF-6a (80–99, 155, and fused histidine tags in both molecules) could not be built because the electron densities were poor. Adenine-14 of one DNA molecule and guanine-13 and adenine-14 of another were also disordered. Refinement statistics are summarized in Table 1.

ACKNOWLEDGMENTS We thank Dr. Y. Maegawa for fruitful discussions, her critical comments, and her assistance in the preparation of proteins, and we thank the staff of SPring-8 BL44B2 and the Photon Factory BL6A beamlines for help during the X-ray diffraction experiments. This work was supported by a research grant from the National Project on Protein Structural and Functional Analyses from the Ministry of Education, Culture, Sports, Science and Technology of Japan. Received: September 14, 2006 Revised: November 22, 2006 Accepted: November 25, 2006 Published: January 16, 2007

Assa-Munt, N., Mortishire-Smith, R.J., Aurora, R., Herr, W., and Wright, P.E. (1993). The solution structure of the Oct-1 POU-specific domain reveals a striking similarity to the bacteriophage lambda repressor DNA-binding domain. Cell 73, 193–205. Billeter, M. (1996). Homeodomain-type DNA recognition. Prog. Biophys. Mol. Biol. 66, 211–225. Bouzin, C., Clotman, F., Renauld, J.C., Lemaigre, F.P., and Rousseau, G.G. (2003). The onecut transcription factor hepatocyte nuclear factor6 controls B lymphopoiesis in fetal liver. J. Immunol. 171, 1297–1303. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. Clotman, F., Lannoy, V.J., Reber, M., Cereghini, S., Cassiman, D., Jacquemin, P., Roskams, T., Rousseau, G.G., and Lemaigre, F.P. (2002). The onecut transcription factor HNF6 is required for normal development of the biliary tract. Development 129, 1819–1828. Gu, W., Shi, X.L., and Roeder, R.G. (1997). Synergistic activation of transcription by CBP and p53. Nature 387, 819–823. Heery, D.M., Kalkhoven, E., Hoare, S., and Parker, M.G. (1997). A signature motif in transcriptional co-activators mediates binding to nuclear receptors. Nature 387, 733–736. Jacquemin, P., Pierreux, C.E., Fierens, S., van Eyll, J.M., Lemaigre, F.P., and Rousseau, G.G. (2003). Cloning and embryonic expression pattern of the mouse Onecut transcription factor OC-2. Gene Expr. Patterns 3, 639–644. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119. Kissinger, C.R., Liu, B.S., Martin-Blanco, E., Kornberg, T.B., and Pabo, C.O. (1990). Crystal structure of an engrailed homeodomain-DNA complex at 2.8 A˚ resolution: a framework for understanding homeodomain-DNA interactions. Cell 63, 579–590. Klemm, J.D., Rould, M.A., Aurora, R., Herr, W., and Pabo, C.O. (1994). Crystal structure of the Oct-1 POU domain bound to an octamer site: DNA recognition with tethered DNA-binding modules. Cell 77, 21–32. Lahuna, O., Fernandez, L., Karlsson, H., Maiter, D., Lemaigre, F.P., Rousseau, G.G., Gustafsson, J., and Mode, A. (1997). Expression of hepatocyte nuclear factor 6 in rat liver is sex-dependent and regulated by growth hormone. Proc. Natl. Acad. Sci. USA 94, 12309–12313. Landry, C., Clotman, F., Hioki, T., Oda, H., Picard, J.J., Lemaigre, F.P., and Rousseau, G.G. (1997). HNF-6 is expressed in endoderm derivatives and nervous system of the mouse embryo and participates to the cross-regulatory network of liver-enriched transcription factors. Dev. Biol. 192, 247–257. Lannoy, V.J., Burglin, T.R., Rousseau, G.G., and Lemaigre, F.P. (1998). Isoforms of hepatocyte nuclear factor-6 differ in DNA-binding properties, contain a bifunctional homeodomain, and define the new ONECUT class of homeodomain proteins. J. Biol. Chem. 273, 13552–13562. Lannoy, V.J., Rodolosse, A., Pierreux, C.E., Rousseau, G.G., and Lemaigre, F.P. (2000). Transcriptional stimulation by hepatocyte nuclear factor-6. Target-specific recruitment of either CREB-binding protein (CBP) or p300/CBP-associated factor (p/CAF). J. Biol. Chem. 275, 22098–22103. Lannoy, V.J., Decaux, J.F., Pierreux, C.E., Lemaigre, F.P., and Rousseau, G.G. (2002). Liver glucokinase gene expression is controlled by the onecut transcription factor hepatocyte nuclear factor-6. Diabetologia 45, 1136–1141.

REFERENCES

Laughon, A. (1991). DNA binding specificity of homeodomains. Biochemistry 30, 11357–11367.

Ades, S.E., and Sauer, R.T. (1994). Differential DNA-binding specificity of the engrailed homeodomain: the role of residue 50. Biochemistry 33, 9187–9194.

Lemaigre, F.P., Durviaux, S.M., and Rousseau, G.G. (1993). Liver-specific factor binding to the liver promoter of a 6-phosphofructo-2-kinase/ fructose-2,6-bisphosphatase gene. J. Biol. Chem. 268, 19896–19905.



Lemaigre, F.P., Durviaux, S.M., Truong, O., Lannoy, V.J., Hsuan, J.J., and Rousseau, G.G. (1996). Hepatocyte nuclear factor 6, a transcription factor that contains a novel type of homeodomain and a single cut domain. Proc. Natl. Acad. Sci. USA 93, 9460–9464.

Streeper, R.S., Hornbuckle, L.A., Svitek, C.A., Goldman, J.K., Oeser, J.K., and O’Brien, R.M. (2001). Protein kinase A phosphorylates hepatocyte nuclear factor-6 and stimulates glucose-6-phosphatase catalytic subunit gene transcription. J. Biol. Chem. 276, 19111–19118.

Matthews, B.W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491–497.

Tan, Y., Adami, G., and Costa, R.H. (2002). Maintaining HNF6 expression prevents AdHNF3b-mediated decrease in hepatic levels of Glut-2 and glycogen. Hepatology 35, 790–798.

Odom, D.T., Zizlsperger, N., Gordon, D.B., Bell, G.W., Rinaldi, N.J., Murray, H.L., Volkert, T.L., Schreiber, J., Rolfe, P.A., Gifford, D.K., et al. (2004). Control of pancreas and liver gene expression by HNF transcription factors. Science 303, 1378–1381. Otting, G., Qian, Y.Q., Billeter, M., Muller, M., Affolter, M., Gehring, W.J., and Wuthrich, K. (1990). Protein–DNA contacts in the structure of a homeodomain–DNA complex determined by nuclear magnetic resonance spectroscopy in solution. EMBO J. 9, 3085–3092. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. Rastegar, M., Lemaigre, F.P., and Rousseau, G.G. (2000). Control of gene expression by growth hormone in liver: key role of a network of transcription factors. Mol. Cell. Endocrinol. 164, 1–4. Samadani, U., and Costa, R.H. (1996). The transcriptional activator hepatocyte nuclear factor 6 regulates liver gene expression. Mol. Cell. Biol. 16, 6273–6284. Sasaki, Y., Takahashi, Y., Nakayama, K., and Kamataki, T. (1999). Cooperative regulation of CYP2C12 gene expression by STAT5 and liver-specific factors in female rats. J. Biol. Chem. 274, 37117–37124. Servitja, J.M., and Ferrer, J. (2004). Transcriptional networks controlling pancreatic development and b cell function. Diabetologia 47, 597–613. Sheng, W., Yan, H., Rausa, F.M., 3rd, Costa, R.H., and Liao, X. (2004). Structure of the hepatocyte nuclear factor 6a and its interaction with DNA. J. Biol. Chem. 279, 33928–33936.

Terwilliger, T.C. (1994). MAD phasing: treatment of dispersive differences as isomorphous replacement information. Acta Crystallogr. D Biol. Crystallogr. 50, 17–23. Terwilliger, T.C. (2000). Maximum-likelihood density modification. Acta Crystallogr. D Biol. Crystallogr. 56, 965–972. Wilson, D.S., Sheng, G., Jun, S., and Desplan, C. (1996). Conservation and diversification in homeodomain-DNA interactions: a comparative genetic analysis. Proc. Natl. Acad. Sci. USA 93, 6886–6891. Wolberger, C., Vershon, A.K., Liu, B., Johnson, A.D., and Pabo, C.O. (1991). Crystal structure of a MAT a 2 homeodomain-operator complex suggests a general model for homeodomain-DNA interactions. Cell 67, 517–528. Xu, W., Rould, M.A., Jun, S., Desplan, C., and Pabo, C.O. (1995). Crystal structure of a paired domain-DNA complex at 2.5 A˚ resolution reveals structural basis for Pax developmental mutations. Cell 80, 639–650. Yao, M., Zhou, Y., and Tanaka, I. (2006). LAFIRE: software for automating the refinement process of protein-structure analysis. Acta Crystallogr. D Biol. Crystallogr. 62, 189–196.

Accession Numbers The coordinates have been deposited in the Protein Data Bank under accession number 2D5V.


Article - Cell Press

Article - Cell Press

Suggest Documents