The DNA Binding Domain of Hepatocyte Nuclear Factor 4 Mediates Cooperative, Specific Binding to DNA and Heterodimerization with the Retinoid X Receptor alpha *

(Received for publication, July 2, 1996, and in revised form, September 17, 1996)

Guoqiang Jiang and Frances M. Sladek Dagger

From the Environmental Toxicology Graduate Program, University of California, Riverside, California 92521

ABSTRACT
INTRODUCTION
EXPERIMENTAL PROCEDURES
RESULTS
DISCUSSION
FOOTNOTES
REFERENCES


ABSTRACT

We recently showed that hepatocyte nuclear factor 4 (HNF-4) defines a unique subclass of nuclear receptors that exist in solution and bind DNA elements as homodimers (Jiang, G., Nepomuceno, L., Hopkins, K., and Sladek, F. M. (1995) Mol. Cell. Biol. 15, 5131-5143). In this study, we show that the dimerization domains of HNF-4 map to both the DNA binding and the ligand binding domain. Whereas the latter is critical for dimerization in solution, the DNA binding domain mediates cooperative, specific binding to direct repeats of AGGTCA separated by one or two nucleotides. Whereas amino acid residues 117-125 (the T-box/third helix region) are insufficient for cooperative homodimerization and high affinity DNA binding, residues 126-142 (encompassing the A-box region) are required. Finally, in contrast to the full-length receptor, the DNA binding domain of HNF-4 is capable of heterodimerizing with that of the retinoid X receptor alpha  but not with that of other receptors. These results indicate that the HNF-4 DNA binding domain is distinct from that of other receptors and that the determinants that prevent HNF-4 from heterodimerizing with RXR lie outside the DNA binding domain, presumably in the ligand binding domain.


INTRODUCTION

Dimerization plays an important role in the function of many different types of proteins. From extracellular receptors to nuclear proteins, the ability to dimerize adds diversity as well as specificity to a variety of biological pathways. This is particularly true for the superfamily of ligand-dependent transcription factors. This family of soluble, nuclear receptors contains over 150 members, including the steroid, thyroid hormone, vitamin A, and vitamin D receptors, and an even larger number of orphan receptors for which ligands have not yet been identified (23).

The nuclear receptor superfamily is characterized by two highly conserved domains. The DNA binding domain (DBD),1 in the N-terminal half of the protein, consists of approximately 60-90 amino acids that form two zinc finger modules followed by a C-terminal extension. The so-called ligand binding domain (LBD) in the C-terminal half of the protein consists of approximately 200 amino acids. This region performs a variety of functions including transactivation, ligand binding, and protein dimerization via heptad repeats of hydrophobic residues (7, 8, 23).

With a few exceptions, the nuclear receptors all have the capacity to bind DNA as dimers. Several, particularly those related to the retinoid receptors, also bind DNA preferentially as heterodimers. Retinoid X receptor (RXR) is the most promiscuous of the receptors, dimerizing with at least 10 different receptors on a variety of DNA elements (23). Full-length RXR, in fact, binds DNA only poorly as a homodimer and not at all as a monomer (42).

Dimerization domains in the receptors have been localized to both the DBD and the LBD using mutagenesis studies (5, 7, 8, 15, 40, 41, 43). Dimerization interfaces have also been visualized in the three-dimensional structures of homo- and heterodimeric complexes of DBDs of various receptors bound to DNA: a homodimer of the estrogen receptor (30), a homodimer of the glucocorticoid receptor (21), and a heterodimer of RXRalpha and the thyroid hormone receptor beta  (TRbeta ) (28). The structure of a homodimer of the LBD of RXRalpha , in the absence of ligand, has also been solved and shown to contain a dimerization interface (3). Nevertheless, the precise role of the different dimerization motifs in DNA binding and other receptor functions has not been clearly defined.

Hepatocyte nuclear factor 4 (HNF-4) is a highly conserved receptor essential for development in organisms ranging from insect to man (4, 34, 44). Found predominantly in the liver, kidney, and intestine, HNF-4 transcriptionally activates a wide variety of genes including those involved in fatty acid and cholesterol metabolism, glucose metabolism, urea biosynthesis, blood coagulation, hepatitis B infections, and liver differentiation (reviewed in Refs. 32, 33). A ligand, however, has not yet been identified for HNF-4. (The term ligand binding domain (LBD), however, will be used for the sake of consistency with other receptors.)

We recently showed that HNF-4, while very similar in DNA binding specificity and amino acid sequence to RXRalpha , does not heterodimerize with full-length RXR or any other receptor analyzed. In fact, full-length HNF-4 bound DNA only as a homodimer and sedimented as a homodimer during glycerol gradient centrifugation. The strong homodimerization activity, as well as the exclusively nuclear localization of HNF-4, led us to conclude that HNF-4 defines a new subclass of receptors (12).

In this study, we analyze in greater detail the mechanism of dimerization of HNF-4 with the goal of deciphering the role and the determinants of homo- versus heterodimerization among the nuclear receptors. The dimerization domains of HNF-4 are localized to both the DBD and the LBD, and the DBD is shown to mediate cooperative, high affinity, specific binding to DNA. The LBD, in contrast, appears to be important for dimerization in solution. It is also shown that, in contrast, to the full-length receptor, the HNF-4 DBD is capable of heterodimerizing with the DBD of RXRalpha , although not with the DBDs of the retinoid acid receptor alpha  (RARalpha ) or thyroid hormone receptor beta  (TRbeta ). These results confirm the unique nature of HNF-4 and implicate the LBD as the major determinant of homo- versus heterodimerization of HNF-4.


EXPERIMENTAL PROCEDURES

Plasmid Constructions

Expression vector pMT7 was constructed by inserting double-stranded oligonucleotide MT7 (top strand: 5'-GTAATACGACTCACTATAGGGCCC<UNL>CTCGAG</UNL>GCG-3'; bottom strand: 5'-AATTCGC<UNL>CTCGAG</UNL>GGGCCCTATAGTGAGTCGTATTACTGCA-3') containing the T7 promoter and unique XhoI sites (underlined) into the parental vector pMT2 (13) and verified by dideoxy sequencing. pMT7 was used for in vivo expression of proteins in COS-7 cells and in vitro expression of proteins using reticulocyte lysate.

cDNAs encoding full-length and truncated rat HNF-4 proteins were prepared by the polymerase chain reaction (PCR) using as template the original HNF-4 clone pf7 (34), appropriate kinased oligonucleotides as primers, and Vent DNA Polymerase (New England BioLabs). The PCR products were gel-purified and subcloned into the pMT7 vector cut with EcoRI, filled-in with Klenow, and dephosphorylated. The sequence of the inserts were verified by dideoxy sequencing. The N-terminal primers, N1 and N45, contain start codons, and the C-terminal primers, C455, C374, C142, C125, C125N123E, contain stop codons. The relative position of the primers are diagrammed in Fig. 1 and defined by the N- or C-terminal-most residue of the region to which they anneal. N1 and C455 have been previously defined as Npf7 and Cpf7, respectively (12). The sequence of the other primers are as follows: N45, 5'-GCCGACAT<UNL>GGATCC</UNL>GCTGGGTGTCAGTGCCC-3'; C374, 5'-CGG<UNL>AAGCTT</UNL>CTAGGGCGCGTCACTGGC-3'; C142, 5'-CGG<UNL>AAGCTT</UNL>CTAGGAGGGTAGGCTGCTG-3'; C125, 5'-CGG<UNL>AAGCTT</UNL>CTACCGCTCATTTTGGACGG-3'; C125N123E, 5'-TT<UNL>GAATTC</UNL>TACCGCTCATTTTGGACGG-3'. Restriction sites used for cloning purpose are underlined.


Fig. 1. Schematic representation of the receptors used in this study. Numbers indicate amino acids in wild-type rat HNF-4 (HNF4.wt) (34), truncated HNF-4 constructs, truncated human RXRalpha , RARalpha , and TRbeta (26) as shown. The conserved zinc finger (Zn2+, first to last conserved cysteine) and the region homologous to the ligand binding domain of other receptors (LBD) are shown as are the T-box (hatched box) and A-box regions (stippled box). The DNA binding domain (DBD) is used as a general term to refer to the zinc finger plus the T-box and, in certain constructs, the A-box region. The T- and A-boxes are regions that have been found to be important for DNA binding in various receptors. See the text and Fig. 5 for more details on the A- and T-boxes, the 3rd helix, and the sequence of HNF4.N123E. The position of the oligonucleotides used for PCR amplification of the various truncated HNF-4 cDNAs are shown and are defined under "Experimental Procedures," C125/N123E, primers C125 and C125N123E.
[View Larger Version of this Image (31K GIF file)]


Protein Expression, Purification, and Electrophoretic Mobility Shift Analysis (EMSA)

Wild-type and truncated HNF-4 proteins were overexpressed in COS-7 cells, and nuclear extracts were prepared as described previously (12). RXR.DBD, RAR.DBD, and TR.DBD were overexpressed in Escherichia coli as glutathione S-transferase fusion proteins using vectors generously provided by T. Perlmann (Karolinska Institute, Stockholm, Sweden). The fusion proteins were purified by glutathione-Sepharose columns (Sigma), and the glutathione S-transferase moiety was removed by cleavage with thrombin as described previously (26). Electrophoretic mobility shift analysis (EMSA) was performed as described previously (12, 26) and as indicated in the figure legends. The double-stranded oligonucleotides used for probes in EMSA are shown in Table I.

Table I.

Shift probes used in this study


Probea Sequenceb

APF1 5'-tcgagcgc<UNL>TGGGCA</UNL>a<UNL>AGGTCA</UNL>cctgc-3'
ApoAI site A 5'-tcgaagggc<UNL>AGGGGT</UNL>ca<UNL>AGGGTT</UNL>cagt-3'
Half-site 5'-gccgaa<UNL>AGGTCA</UNL>cct-3'
DR0 5'-gcgc<UNL>AGGTCAAGGTCA</UNL>cctc-3'
DR1 5'-gcgc<UNL>AGGTCA</UNL>a<UNL>AGGTCA</UNL>cctc-3'
DR2 5'-gcgc<UNL>AGGTCA</UNL>aa<UNL>AGGTCA</UNL>cctc-3'
DR3 5'-gcgc<UNL>AGGTCA</UNL>gaa<UNL>AGGTCA</UNL>cctc-3'
DR4 5'-gcgc<UNL>AGGTCA</UNL>cgaa<UNL>AGGTCA</UNL>cctc-3'
DR5 5'-gcgc<UNL>AGGTCA</UNL>ccgaa<UNL>AGGTCA</UNL>cctc-3'

a  APF1 is from the human APOCIII gene, and ApoAI site A is from the human APOAI gene. Position of the sites within the promoters have been previously described (33). Half-site through DR5 are synthetic elements. DR1, direct repeat with one nucleotide spacing, etc.
b  Sequence of the top strand of the double-stranded oligonucleotides. Hexameric half-sites are in capital and underlined. Not shown are 5'-TCGA overhangs on the APF1 and ApoAI site oligonucleotides.


RESULTS

HNF-4 Contains Dimerization Domains in Both the DBD and the LBD

In order to locate the structural domain(s) responsible for the strong dimerization activity of HNF-4, several constructs were made and are diagrammed in Fig. 1. The constructs were overexpressed in COS cells and examined for DNA binding activity by electrophoretic mobility shift analysis (EMSA) using a variety of probes (see Table I). Since dimerization motifs had been previously mapped to both the DNA binding domain (DBD) and the ligand binding domain (LBD) of other receptors, those two domains were compared in HNF-4.

The results, depicted in Fig. 2, indicate that, as seen previously, the full-length HNF-4 (HNF4.wt) yielded only a dimeric complex with the APF1 probe (Fig. 2, lanes 1-3). In contrast, HNF4.142, which contains the DBD but lacks the LBD, yielded both monomeric and dimeric binding species (lanes 4-5) (see Fig. 3 for verification of the dimer). HNF4.374, on the other hand, which contains both the DBD and the LBD but not the very N or C terminus, bound DNA only as a dimer (lanes 6-7). Furthermore, HNF4.374 dimerized with HNF4.wt, as evidenced by a shift band of intermediate mobility (lanes 9-11), whereas HNF4.142 did not dimerize with either HNF4.wt (lane 8) or HNF4.374 (lane 12). These results indicate that HNF-4 contains at least two dimerization domains, one located between amino acids 45 and 142, corresponding to the DBD, and an additional domain located between residues 143 and 374 (i.e. the LBD), which apparently precludes monomeric binding.


Fig. 2. HNF-4 contains dimerization domains in both the DBD and the LBD. EMSA of COS cell nuclear extracts containing overexpressed HNF4.wt, HNF4.374, or HNF4.142 (0.5-1.0 µg of total protein each) and 32P-APF1 probe (1 ng per 15-µl reaction) was done at RT as described previously (12). Shown is a PhosphorImage (Molecular Dynamics) of the shift gel. alpha -445, antisera raised against the very C terminus of HNF-4 (34); Competition, 50 ng of unlabeled APF1; immune complex, supershift containing probe, receptor, and antisera. Dimers and monomers are identified by the proteins that comprised them: wt, HNF4.wt; 374, HNF4.374; 142, HNF4.142; Free probe, 32P-APF1 not bound by protein. For reactions containing two different HNF-4 proteins (lanes 8-12), extracts from single transfections were mixed and incubated at RT for 15 min prior to the addition of shift mix containing the probe. An endogenous complex from COS cells co-migrates with the HNF4.wt dimer in this gel, as seen in lanes 2-7, 10, and 12.
[View Larger Version of this Image (49K GIF file)]



Fig. 3. HNF4.142 binds DNA in a cooperative, two-step fashion while HNF4.wt binds in a noncooperative, one-step fashion. EMSA was performed as described in Fig. 2 using increasing amounts of COS nuclear extract containing overexpressed HNF4.wt (375 ng to 8.6 µg of total protein) or HNF4.142 (250 ng to 15 µg of total protein) and a constant amount of 32P-labeled APF1 oligonucleotide (0.5 ng) as probe. Total protein was kept constant at 15 µg per reaction by the addition of bovine serum albumin. Reactions were incubated at RT for 30 min to reach equilibrium and then subjected to electrophoresis. Shown are both the autoradiograms of the shift gels (A and B) and a graph of the corresponding densitometric readings (C) of the HNF4.wt dimer-DNA complex and HNF4.142 dimer-DNA complex. Each point in the graph represents the average of two separate experiments. The plots were fitted with either a hyperbolic (HNF4.wt) or sigmoidal curve (HNF4.142) using MicroCal OriginTM (MicroCal Software, Inc., MA).
[View Larger Version of this Image (29K GIF file)]


The HNF-4 DBD Is Responsible for DNA Binding Specificity Whereas the LBD Is Responsible for Dimerization in Solution

Our previous results from glycerol gradient sedimentation and order of dilution experiments indicated that the full-length HNF-4 is a homodimer even in the absence of DNA, i.e. in solution (12). In order to compare the dimerization activity of the HNF-4 DBD with that of the full-length protein, the following two experiments were performed. In the first, HNF4.wt and HNF4.142 were compared for cooperative binding to DNA. A saturation curve was calculated by performing EMSA with increasing amounts of COS cell nuclear extract containing either HNF4.wt or HNF4.142 and a constant amount of 32P-APF1 as probe. The results, shown in Fig. 3, demonstrate that the saturation curve of the HNF4.142 dimer is sigmoidal, whereas that of the HNF4.wt dimer is hyperbolic. The sigmoidal shape indicates that the two subunits of the HNF4.142 dimer bind DNA in a cooperative, and thus two-step, fashion. This verifies the shift complex as a protein dimer bound to the probe, as opposed to two monomers that happen to bind to the same DNA molecule. It also indicates that HNF4.142 exists in solution as a monomer. The hyperbolic shape of the HNF4.wt curve, on the other hand, indicates that the two subunits of the dimer bind DNA in a noncooperative, and thus single-step, fashion, verifying that HNF4.wt exists in solution as a homodimer.

The second experiment to test for dimerization activity in solution also examined DNA binding specificity. The binding activity of HNF4.wt and HNF4.142 on DNA containing either a single half-site or direct repeats separated by an increasing number of nucleotides (DR0-5) was determined. The rationale was that a protein that exists as a dimer in solution might yield a DNA complex migrating as a dimer on a single half-site, whereas a protein that exists in solution only as a monomer would not.

The results, shown in Fig. 4, panel A, indicate that HNF4.wt binds DR1 (lanes 5-6), as expected, as well as DR2 (lanes 7-8). There was also a low level of binding activity on the probe containing a single half-site (lanes 1, 2, 1a, and 2a) but none on any of the other probes, DR3, DR4, or DR5 (lanes 9-14). The low binding activity on the half-site is presumably due to reduced protein-DNA contact and suggests that contacts to both half-sites are important for binding the consensus element (DR1). No monomeric binding of HNF4.wt to the half-site was detected even on very long exposures (not shown). In contrast to HNF4.wt, HNF4.142 bound a single half-site only as a monomer (panel B, lanes 1 and 2). These results, along with those of Fig. 3, confirm that HNF4.wt exists as a homodimer and HNF4.142 as a monomer in solution.


Fig. 4. DNA binding specificity of HNF4.wt versus HNF4.142. EMSA was performed as described in Fig. 2 using the probes indicated and as defined under "Experimental Procedures." Shown are the autoradiograms of shift gels (A and B) and the results of gels quantified by phosphorimaging (C). A, HNF4.wt extracts were subjected to EMSA on a 6% polyacrylamide gel in the presence (+) or absence (-) of alpha 445 as described under "Experimental Procedures." Shown are all the 32P-labeled bands, including the endogenous band from the COS extracts, except for the free probe. Lanes 1a and 2a are enhanced images of lanes 1 and 2. B, HNF4.142 extracts were subjected to EMSA in the presence (+) or absence (-) of 40-fold molar excess self-competition as indicated. Shift complexes due to factors endogenous to COS cells are indicated. No HNF4.142 dimer-half-site complex was detected even upon longer exposure.
[View Larger Version of this Image (47K GIF file)]


The results of Fig. 4 also show that HNF4.142 binds DNA with a specificity similar to that of HNF4.wt. Both HNF4.wt and the HNF4.142 dimer bind significantly only to direct repeats separated by one or two nucleotides (DR1 and DR2, respectively) with an apparent preference for DR1 (lanes 5-8 in Panels A and B). Similar results were obtained from a competition experiment in which HNF4.142 and HNF4.wt were subjected to EMSA using as probe 32P-APF1 and 40-fold molar excess of several different unlabeled oligonucleotides containing HNF-4 binding sites. The results indicated that the oligonucleotides competed the HNF4-APF1 complex in a similar fashion for both the HNF4.DBD dimer and HNF4.wt dimer. The only difference is that the extent of competition appeared to be somewhat less for the HNF4.142 dimer than for the HNF4.wt dimer (data not shown). Since the construct containing just the DBD and the LBD, HNF4.374, showed an identical binding specificity to that of HNF4.wt, these results, together with those of Figs. 2, 3, 4, indicate that the DBD of HNF-4 is primarily responsible for the DNA binding specificity and partially responsible for dimerization activity on DNA. The LBD, on the other hand, is apparently responsible for the strong homodimerization activity seen in the absence of DNA.

Amino Acids 126 to 142 (encompassing the A-box Region) Is Required for Cooperative Homodimerization and Specific and High Affinity Binding to DNA

In addition to the highly conserved zinc finger motifs, the DBD of many of the receptors also contains an A-box and a T-box (see Fig. 1). The A-box, which contacts the nucleotide bases flanking the core recognition sequence, is important for the high affinity binding of NGFI-B monomers and thyroid hormone receptor (TR) monomers and homodimers but not that of receptors such as RXR and retinoic acid receptor (RAR) (16, 28, 38, 40, 41). For these receptors, the T-box appears to be much more important for dimerization and high affinity binding (18, 28, 40, 41). The N-terminal portion of the T-box in RXRalpha in particular, which forms an alpha -helix (the so-called third helix), is required for homodimerization of the RXRalpha DBD on DR1 (18). We therefore wished to determined which was more important for HNF-4 DBD binding, the A-box or the T-box/third helix.

Amino acid sequence alignment, shown in Fig. 5A, indicates that the A-box of HNF-4, like that of the other receptors, is fairly distinct compared with the A-box of other receptors (e.g. RXR versus HNF-4) but relatively conserved among the different species of HNF-4 (i.e. rat versus Drosophila HNF-4). The T-box of HNF-4, on the other hand, is very similar to that of several other receptors, particularly in the third helix region. In fact, HNF-4 is closest to retinoid X receptor (RXRalpha , -beta , and -gamma ) and differs primarily by a single amino acid in this region: a negatively charged glutamic acid (E) in RXR is replaced by an asparagine (N) at residue 123 in HNF-4. In addition to changing the net charge of the region, this single amino acid difference also appears to change the secondary structure of the region by terminating the putative helix in HNF-4 (Fig. 5B).


Fig. 5. Comparison of the T- and A-box regions of HNF-4 and other members of the nuclear receptor superfamily. A, the T- and A-box regions as previously defined (38) are shown for HNF-4 and a variety of other receptors. The amino acid sequences (given in a single-letter code) start from the last conserved cysteine in the zinc finger region and end one amino acid after the A-box. Boxed residues correspond to the region in RXRalpha that forms an alpha -helical structure in solution (the 3rd helix) (18). rHNF-4 (34); dHNF-4 (44); hRXRalpha (22); mRXRbeta (19); mRXRgamma (19); hRARalpha (27); hRARbeta (6); mRARgamma (9); rTRalpha (36); hTRbeta (37); rNGFIB (24). r, rat; d, Drosophila; h, human, m, mouse. B, schematic representation of the 3rd helix region (boxed) of hRXRalpha and the corresponding region in rHNF-4. Charge and polarity are indicated: +, positively charged; -, negatively charged; p, polar but uncharged; o, nonpolar and hydrophobic. Also indicated is the predicted secondary structure as determined by the Chou-Fasman method using the PEPTIDESTRUCTURE program (Genetic Computing Group, GCG): h, alpha -helical structure; -, no apparent structure.
[View Larger Version of this Image (38K GIF file)]


In order to examine the role of the A-box and the third helix region of HNF-4 in DNA binding, the constructs HNF4.125 and HNF4.N123E were prepared (see Fig. 1). Both end with residue 125 which corresponds to the end of the third helix in RXR. HNF4.N123E also contains a glutamic acid (E) in place of an asparagine (N) at position 123, rendering the region essentially identical to that of RXR. To help distinguish the different shift complexes, the HNF4.125 and HNF4.N123E constructs also both contain an intact N terminus, whereas HNF4.142 does not. (As expected, the N terminus did not affect the DNA binding and dimerization properties of HNF-4, data not shown.)

The results of EMSA using HNF4.125 and HNF4.142 are shown in Fig. 6A. Under the conditions used, HNF4.125 binds APF1 as a monomer but not as a homodimer (lanes 1 and 2). In contrast, when HNF4.125 is mixed with extracts containing HNF4.142, a new shift complex appears which is dependent upon the presence of both HNF4.125 and HNF4.142 (lanes 5-14). This indicates that a HNF4.125-HNF4.142 heterodimer is formed and suggests that residues 126 to 142, encompassing the A-box region, contain a dimerization motif that interacts with residues 1-125. Furthermore, the HNF4.142-HNF4.125 heterodimer appears to bind in a cooperative fashion, just as the HNF4.142 homodimer, since increasing amounts of either HNF4.142 (lanes 5-9) or HNF4.125 extract (lanes 10-14) lead to an increasing amount of heterodimer formation.


Fig. 6. The A-box region is required for the homodimerization and specific binding of HNF4.142 to DNA. A, shown is the autoradiogram of EMSA performed as described in Fig. 2 using 32P-APF1 as probe and nuclear extracts from COS cells singly transfected with either HNF4.142 or HNF4.125 and subsequently mixed as indicated. The amount of total nuclear protein used in each shift reaction is indicated. 125, HNF4.125 extract; 142, HNF4.142 extract. Indicated are the HNF4.125 and HNF4.142 complexes. The open arrowhead indicates where a 125-125 homodimer would migrate if it were present. B, EMSA was performed as described in A with a constant amount of HNF4.142 and HNF4.125 extract (1 and 2 µg of total nuclear protein, respectively) in the absence (lane 1) or presence of increasing amounts (25- to 500-fold molar excess) of unlabeled APF1 oligonucleotide as competitor. The open arrowhead is as in A.
[View Larger Version of this Image (37K GIF file)]


To determine whether residues 126-146 are also important in the specificity of binding of HNF-4 to DNA, a competition experiment was performed. The results, shown in Fig. 6B, indicate that the HNF4.142 homodimer and the HNF4.142-HNF4.125 heterodimer can both be efficiently competed by unlabeled competitor DNA, whereas neither the HNF4.142 nor the HNF4.125 monomer can be. This reinforces the notion that the dimerization mediated by residues 126-142 is important for the specific binding of the HNF-4 DBD to DNA.

To assess the role of the A-box in binding affinity, dissociation constants (Kd) for both HNF4.142 and HNF4.125 were determined on both a DR1 and a DR2 probe and compared with that of the RXRalpha DBD. Since the RXRalpha DBD did not homodimerize on any DR2 examined, RARalpha DBD and a native ApoAI site A probe, which can be considered a DR2 (see Table I), were used. The results, shown in Fig. 7, indicate that both the HNF4.142 and the RXR.DBD homodimer bind DR1 with high affinity (Kd = 4.6 and 2.2 nM in panels A and C, respectively). The EMSA for Fig. 7 was done at 4 °C in order to detect the RXR and RAR homodimers. Under these conditions, a HNF4.125 homodimer shift complex was also detected, whereas it was not detected when the reaction was performed at RT (see Fig. 6). The HNF4.125 homodimer complex, however, had a very low binding affinity as evidenced by a nonsaturating curve (panel B). The binding of all the DBD monomers (HNF4.142, HNF4.125, RXR, and RAR) was also nonsaturating, yielding a straight line in each case (data not shown).


Fig. 7. DNA binding affinity of HNF4.142, HNF4.125, RXR.DBD, and RAR.DBD. Shown are the saturation curves of homodimers of HNF4.142, HNF4.125, RXR.DBD, and RAR.DBD on the DR1 and the ApoAI site A probes (a DR2). EMSA was performed at 4 °C as described previously (26) using a constant amount of crude COS cell nuclear extract containing HNF-4 constructs (A, B, D, and E), and RXR.DBD (C) or RAR.DBD (F) purified from bacteria. The binding buffer contained 10 mM Tris (pH 8.0), 40 mM KCl, 0.05% Nonidet P-40, 6% glycerol, 1 mM dithiothreitol, 0.2 µg of poly[d(I-C)]. Increasing amounts of the indicated 32P-labeled oligonucleotide (DR1 or ApoAI, 0.5-30 ng per 15-µl reaction) were used as probe. The reactions were incubated on ice for 30 min to reach equilibrium and then loaded in duplicate onto mobility shift gels. The gels were dried and quantified by phosphorimaging. Shown are the average of the duplicate loadings from one representative experiment, the best fit curve, and the corresponding dissociation constant (Kd) calculated by the MicroCal OriginTM program. Since HNF4.125 binding to DR1 could not be saturated, a Kd could not be calculated (panel B).
[View Larger Version of this Image (38K GIF file)]


In comparison to the DR1, binding to ApoAI (a DR2) occurred at a lower affinity for the HNF4.142 homodimer but was comparable with that of the RARalpha DBD (Kd = 32.3 and 49.3 nM in panels D and F, respectively). Interestingly, the affinity of the HNF4.125 homodimer for ApoAI site A was higher than that for the DR1 (66.4 nM in panel E) but still significantly lower than that of HNF4.142. These results show that, unlike the full-length HNF-4 which binds DNA much more efficiently than does either the full-length RXR or RAR homodimer (12), the HNF-4 DBD binds DNA as a dimer with a similar affinity as the RXR and the RAR DBD. The results also show that while amino acid residues 117-125 (the third helix region) are insufficient for high affinity binding of the HNF-4 DBD, the A-box region is required.

The DBD of HNF-4 Heterodimerizes with the DBD of RXRalpha but Not of Other Receptors

Since the HNF-4 DBD exists as a monomer in solution, we hypothesized that, unlike the full-length receptor, it might be able to heterodimerize with RXRalpha . EMSA was performed as above using the HNF-4 and RXR DBD constructs (see Fig. 1) and DR1 and DR2 as probes. The results, shown in Fig. 8, indicate that whereas RXR.DBD did not heterodimerize with HNF4.142 on DR1 (panel A, lane 3), it did on DR2 (panel B, lane 3). HNF4.125, in contrast, heterodimerized with RXR.DBD on both DR1 and DR2 (lane 6 in panels A and B, respectively). None of the HNF-4 DBD constructs (HNF4.142, HNF4.125, and HNF4.N123E) heterodimerized with full-length RXRalpha (data not shown).


Fig. 8. HNF-4 DBD constructs heterodimerize with RXR.DBD but not with RAR.DBD. EMSA was performed as in Fig. 7 using COS cell nuclear extracts containing HNF4.142 (0.5 ng of total protein), HNF4.125 (5 ng of total protein) or HNF4.N123E (5 ng of total protein), and RXR.DBD or RAR.DBD purified from bacteria. For reactions containing two different proteins, the proteins were mixed and incubated on ice for 15 min on ice prior to the addition of the shift mix containing the 32P-labeled probe (1-2 ng) DR1 (A), DR2 (B), or DR5 (C). After the addition of probe, the reactions were incubated on ice for 30 min, loaded onto an 8% polyacrylamide gel in 0.25 × TBE, and electrophoresed at 4 °C at constant current (12 mA) for 1.5 h. The gels were dried and imaged using a PhosphorImager. The identities of the protein species binding the probe are indicated. Some dimer species co-migrated as indicated (e.g. 125-125 homodimer and N123E-N123E homodimer). Only the portion of the gel containing relevant shift complexes are shown for C. Open and closed arrowheads, positions where an HNF4.142-RAR.DBD and an HNF4.125-RAR.DBD heterodimer, respectively, would migrate if they were present. HNF4-RXR heterodimers were not seen on DR3 or DR4 (data not shown).
[View Larger Version of this Image (61K GIF file)]


In order to determine whether the third helix/T-box region facilitates HNF-4 dimerization, EMSA was performed with RXR.DBD and extracts containing HNF4.N123E, which theoretically contains a third helix analogous to that of RXRalpha (see Fig. 5). The results show that HNF4.N123E does indeed form homo- and heterodimers somewhat more readily than does HNF4.125 on DR1 (Fig. 8A, compare lanes 10 and 11 to 5 and 6, respectively). Therefore, the third helix may facilitate but is evidently not required for homo- and heterodimerization. Interestingly, in contrast to the APF1 probe (Fig. 6), HNF4.125 and HNF4.N123E do not bind DR1 and DR2 as monomers, only as dimers (Fig. 8A, lanes 5 and 10; B, lanes 4 and data not shown). The significance of this difference is not known but could be due to different reaction conditions and/or nucleotide sequence of the probes.

Heterodimerization of the HNF-4 DBD with other receptors was also examined. However, in contrast to RXR.DBD, neither HNF4.142 nor HNF4.125 nor HNF4.N123E heterodimerized with either RAR.DBD (Fig. 8C) or TR.DBD (data not shown). Of potential interest is the observation that the amount of RAR DBD homodimer complex appears to be reduced by the presence of the HNF4.DBD extracts (lanes 6, 9, 12). Whereas the reason for this reduction is not known, it is doubtful that it is due to dimerization of the HNF4 and the RAR DBDs in solution since NMR studies of several different receptor DBDs, including RARbeta and RXRalpha , show the DBDs to exist only as monomers in solution (11, 14, 18, 31).


DISCUSSION

The results of this study show that the DNA binding domain (DBD) of HNF-4 readily heterodimerizes with the DBD of RXRalpha . The results also show that the A-box region at the C-terminal end of the DBD is critical for the cooperative homodimerization and thus high affinity binding of HNF-4 DBD to DNA. Finally, this is the first report of direct evidence that HNF-4 binds DR2 elements as well as DR1 elements. Although HNF-4 has been previously shown to bind native elements comprised of nonconsensus half-sites separated by two nucleotides (2, 25), this is the first published report of HNF-4 binding a synthetic DR2 element.

Determinants of Exclusive Homodimerization of HNF-4 Reside within the LBD

In contrast to the full-length HNF-4 which exists in solution as a stable homodimer and binds DNA only as a homodimer (12), the HNF-4 DBD exists in solution as a monomer (Figs. 3 and 4B) and efficiently heterodimerizes with the DBD of RXRalpha (Fig. 8). These results demonstrate that the determinants responsible for the exclusive homodimerization of HNF-4 reside outside the DBD, presumably in the LBD.

The conclusion that the LBD is crucial in mediating the exclusive homodimerization activity of HNF-4 both in solution and on DNA is not too surprising since the LBD is known to be important in determining the homo- and heterodimerization properties of other members of the nuclear receptor superfamily (reviewed in Ref. 10). For instance, the DBDs of the steroid hormone receptors (i.e. estrogen and glucocorticoid) exist in solution as monomers even at millimolar concentrations (11, 31), whereas the full-length receptors exist in solution as homodimers (7, 20, 29, 39). Mutations in the LBD, especially the conserved ninth heptad region, have also been shown to change the homo- and/or heterodimerization properties of nuclear receptors such as TR, RAR, and RXR (1, 35, 43). However, this is the first example of a DBD of a receptor forming heterodimers when the corresponding full-length receptor forms only homodimers. These results unequivocally implicate the LBD in determining homo- versus heterodimerization of HNF-4. Indeed, the crystal structure of a dimer of the LBD of RXRalpha shows important contacts between amino acids of opposite charge from each of the two monomers at the dimer interface (3). Finally, since HNF-4 defines a new subclass of nuclear receptors (12), we propose that the determinants of the strong homodimerization activity of other receptors that fall into that group, such as germ cell nuclear factor and TAK1, will also reside within the LBD.

HNF-4 DBD Possesses Unique Structural and Functional Properties

Of all the mammalian nuclear receptors, HNF-4 is most similar to RXR in amino acid sequence, especially in the DBD (17). Mouse RXRalpha is 56% identical to rat HNF-4 in the DBD. HNF-4 and RXRalpha also share response elements from at least six different genes as well as a consensus site of a direct repeat of AGGTCA separated by one nucleotide (DR1) (reviewed in Refs. 32, 33). These similarities suggested that the structural and functional properties of the HNF-4 DBD would be similar to that of the RXRalpha DBD. The results from this study show, however, that this is not the case.

First, the A- and the T-boxes appear to play different roles in the homodimerization of the HNF-4 and the RXRalpha DBDs. For example, the A-box is required for the cooperative homodimerization of the HNF-4 DBD (Figs. 6 and 7), whereas the T-box is necessary and sufficient for the cooperative homodimerization of the RXRalpha DBD (18, 40, 41). Second, the DBD of HNF-4 heterodimerizes with the RXRalpha DBD but not with the RAR or the TR DBD, suggesting that the HNF-4 DBD is more like the RAR or the TR DBD than the RXRalpha DBD (Fig. 8 and data not shown).

Since the HNF-4 DBD construct containing the A-box (HNF4.142) dimerizes cooperatively with the HNF-4 DBD construct lacking the A-box (HNF4.125), which by itself does not readily homodimerize (Fig. 6), it is concluded that the dimerization interface of the HNF-4 DBD homodimer is formed through the interaction between the A-box of one HNF-4 DBD and a more N-terminal region of the other HNF-4 DBD. This asymmetric nature of the HNF-4 DBD dimerization interface is consistent with arrangement of the two half-sites of the HNF-4 binding site as a direct repeat.

The finding that the A-box region is required for the cooperative homodimerization of the HNF-4 DBD on DNA is rather unique among nuclear receptors studied (see Table II). Although the A-box has been shown to be important for the monomeric binding of NGFI-B and TR (16, 38), it has been shown to not be critical for the homo- and heterodimerization of RXR and RAR DBDs, for which the T-box is more important (28, 40, 41). The A-box, however, was shown to be important for the high affinity binding of the TR homodimer (16). It thus appears that HNF-4 DBD homodimer functions in a manner similar to a TR DBD homodimer even though it is much more similar to the RXRalpha DBD in amino acid sequence, particularly in the T-box region. This similarity, however, is unexpected since HNF-4 and TR homodimers bind direct repeats with different spacings. Whereas HNF-4 binds both DR1 and DR2 elements, TR binds DR4 elements as well as inverted repeats. And binding of receptor dimers to direct repeats with different spacings are expected to involve dimerization interfaces formed between different regions of the receptors (28).

Table II.

Requirement for the A-box in DNA binding

Information is from this study and a variety of other sources (10, 16, 18, 26, 40, 41).
Monomer Homodimer Heterodimera

RXR No No Not applicable
RAR No No No
TR Yes Yes Yes
HNF-4 No Yes No

a  Heterodimerization with RXR.

Despite the similar role of the A-box in HNF-4 and TR DBD homodimerization, other evidence suggests that the structure of HNF-4 DBD is different from that of the TR DBD. First, the A-box plays apparently different roles in the monomeric binding of the HNF-4 DBD and the TR DBD. Whereas the A-box is absolutely required for the monomeric binding of TR DBD (41), it appears to facilitate but is not necessary for the monomeric binding of HNF-4 DBD (Fig. 6). Second, the A-box of TR is required for heterodimerization with RXR but it is not required for heterodimerization of the HNF-4 DBD (Fig. 8). Third, structural modeling indicates that the A-box of the TR DBD will interfere with the second zinc module of RXR DBD on DR1 and DR2 elements and therefore prevent the two DBDs from heterodimerizing on those elements (28). In contrast, the HNF-4 DBD construct containing the A-box (HNF4.142) heterodimerizes with the RXRalpha DBD on DR2 (Fig. 8B). Potential steric hindrance could, however, explain why the same HNF-4 DBD construct does not heterodimerize with RXR.DBD on DR1 whereas the HNF-4 DBD construct lacking the A-box (HNF4.125) does (Fig. 8A).

In conclusion, the results of this study show not only that the determinants of homo- versus heterodimerization reside within the LBD of the receptors but also that much remains to be learned about the role of the A-box and the T-box in the nuclear receptors. Furthermore, the results indicate that the dimerization interface of HNF-4 DBD is different from that of other nuclear receptors, which could explain why HNF-4, which binds DR1 and DR2, has a different DNA binding specificity than that of other receptors, including RXR, TR, and RAR, which preferentially bind DR1, DR4, and DR5, respectively. Finally, this study provides yet another example of how broad functional diversity among proteins within a highly conserved superfamily can be achieved with a minimal amount of alteration in primary amino acid sequence.


FOOTNOTES

*   This work was supported in part by a Grant-in-Aid from the American Heart Association, CA Affiliate (to F. M. S.), and by the Sigma Xi Society and a Dean's Dissertation Research Fellowship (to G. J.). The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Dagger    To whom correspondence should be addressed: 5419 Boyce Hall, Environmental Toxicology Program, University of California, Riverside, CA 92521. Tel.: 909-787-2264; Fax: 909-787-3087; E-mail: Sladek{at}mail.ucr.edu.
1    The abbreviations used are: DBD, DNA binding domain; HNF-4, hepatocyte nuclear factor 4; RXR, retinoid X receptor; TR, thyroid hormone receptor; LBD, ligand binding domain; EMSA, electrophoretic mobility shift analysis; wt, wild-type; PCR, polymerase chain reaction; RT, room temperature; RAR, retinoic acid receptor.

REFERENCES

  1. Mangelsdorf, D. J., Thummel, C., Beato, M., Herrlich, P., Schutz, G., Umesono, K., Blumberg, B., Kastner, P., Mark, M., Chambon, P., and Evans, R. M. (1995) Cell 83, 835-839 [Medline] [Order article via Infotrieve]
  2. Fawell, S. E., Lees, J. A., White, R., and Parker, M. G. (1990) Cell 60, 953-962 [Medline] [Order article via Infotrieve]
  3. Forman, B. M., Yan, C.-R., Au, M., Casanova, J., Ghysdael, J., and Samuels, H. H. (1989) Mol. Endocrinol. 3, 1610-1626 [Abstract]
  4. Zhang, X.-K., and Pfahl, M. (1993) Receptor 3, 183-191 [Medline] [Order article via Infotrieve]
  5. Kumar, V., and Chambon, P. (1988) Cell 55, 145-156 [Medline] [Order article via Infotrieve]
  6. Dahlman-Wright, K., Siltala-Roos, H., Carlstedt-Duke, J., and Gustafsson, J.-A. (1990) J. Biol. Chem. 265, 14030-14035 [Abstract/Free Full Text]
  7. Zechel, C., Shen, X., Chen, J., Chen, Z., Chambon, P., and Gronemeyer, H. (1994) EMBO J. 13, 1425-1433 [Abstract]
  8. Zechel, C., Shen, X., Chambon, P., and Gronemeyer, H. (1994) EMBO J. 13, 1414-1424 [Abstract]
  9. Zhang, X.-K., Salbert, G., Lee, M.-O., and Pfahl, M. (1994) Mol. Cell. Biol. 14, 4311-4323 [Abstract]
  10. Schwabe, J., Chapman, L., Finch, J., and Rhodes, D. (1993) Cell 75, 567-578 [Medline] [Order article via Infotrieve]
  11. Luisi, B. F., Xu, W. X., Otwinowski, Z., Freeman, L. P., Yamamoto, K. R., and Sigler, P. B. (1991) Nature 352, 497-505 [CrossRef][Medline] [Order article via Infotrieve]
  12. Rastinejad, F., Perlmann, T., Evans, R. M., and Sigler, P. B. (1995) Nature 375, 203-211 [CrossRef][Medline] [Order article via Infotrieve]
  13. Bourguet, W., Ruff, M., Chambon, P., Gronemeyer, H., and Moras, D. (1995) Nature 375, 377-382 [CrossRef][Medline] [Order article via Infotrieve]
  14. Sladek, F. M., Zhong, W., Lai, E., and Darnell, J. E. (1990) Genes Dev. 4, 2353-2365 [Abstract]
  15. Zhong, W., Sladek, F. M., and Darnell, J. E. (1993) EMBO J. 12, 537-544 [Abstract]
  16. Chen, W., Manova, K., Weinstein, D., Duncan, S., Plump, A., Prezioso, V., Bachvarova, R., and Darnell, J. J. (1994) Genes Dev. 8, 2466-2477 [Abstract]
  17. Sladek, F. M. (1993) Receptor 3, 223-232 [Medline] [Order article via Infotrieve]
  18. Sladek, F. M. (1994) in Liver Gene Expression (Tronche, F., and Yaniv, M., eds), pp. 207-230, R. G. Landes Co., Austin, TX
  19. Jiang, G., Nepomuceno, L., Hopkins, K., and Sladek, F. M. (1995) Mol. Cell. Biol. 15, 5131-5143 [Abstract]
  20. Kaufman, R., Davies, M., Pathak, V., and Hershey, J. (1989) Mol. Cell. Biol. 9, 946-958 [Medline] [Order article via Infotrieve]
  21. Perlmann, T., Rangarajan, P., Umesono, K., and Evans, R. (1993) Genes Dev. 7, 1411-1422 [Abstract]
  22. Wilson, T. E., Paulsen, R. E., Padgett, K. A., and Milbrandt, J. (1992) Science 256, 107-110 [Medline] [Order article via Infotrieve]
  23. Kurokawa, R., Yu, V., Naar, A., Kyakumoto, S., Han, Z., Silverman, S., Rosenfeld, M., and Glass, C. (1993) Genes Dev. 7, 1423-1435 [Abstract]
  24. Lee, M., Kliewer, S., Provencal, J., Wright, P., and Evans, R. (1993) Science 260, 1117-1121 [Medline] [Order article via Infotrieve]
  25. Hard, T., Kellenbach, E., Boelens, R., Kaptein, R., Dahlman, K., Carlstedt-Duke, J., Freedman, L. P., Maler, B. A., Hyde, E. I., Gustafsson, J. A., and Yamamoto, K. R. (1990) Biochemistry 29, 9015-9023 [Medline] [Order article via Infotrieve]
  26. Schwabe, J. W. R., Neuhaus, D., and Rhodes, D. (1990) Nature 348, 458-461 [CrossRef][Medline] [Order article via Infotrieve]
  27. Knegtel, R. M. A., Boelens, R., Ganadu, M. L., George, A. V. E., Katahira, M., Bonvin, A. M. J. J., Eib, D., Saag, P. T. V. D., and Kaptein, R. (1993) Ann. N. Y. Acad. Sci. 684, 49-51 [Medline] [Order article via Infotrieve]
  28. Blanchard, K. L., Acquaviva, A. M., Galson, D. L., and Bunn, H. F. (1992) Mol. Cell. Biol. 12, 5373-5385 [Abstract]
  29. Nakshatri, H., and Chambon, P. (1994) J. Biol. Chem. 269, 890-902 [Abstract/Free Full Text]
  30. Glass, C. (1994) Endocr. Rev. 15, 391-407 [Medline] [Order article via Infotrieve]
  31. Wrange, O., Eriksson, P., and Perlmann, T. (1989) J. Biol. Chem. 264, 5253-5259 [Abstract/Free Full Text]
  32. Linstedt, A., West, N., and Brenner, R. (1986) J. Steroid Biochem. 24, 677-686 [CrossRef][Medline] [Order article via Infotrieve]
  33. Rodriguez, R., Weigel, N. L., O'Malley, B. W., and Schrader, W. T. (1990) Mol. Endocrinol. 4, 1782-1790 [Abstract]
  34. Spanjaard, R. A., Darling, D. S., and Chin, W. W. (1991) Proc. Natl. Acad. Sci. U. S. A. 88, 8587-8591 [Abstract]
  35. Au-Fliegner, M., Helmer, E., Casanova, J., Raaka, B., and Samuels, H. (1993) Mol. Cell. Biol. 13, 5725-5737 [Abstract]
  36. Laudet, V., Hanni, C., Coll, J., Catzeflis, F., and Stehelin, D. (1992) EMBO J. 11, 1003-1013 [Abstract]
  37. Mangelsdorf, D. J., Ong, E. S., Dyck, J. A., and Evans, R. M. (1990) Nature 345, 224-229 [CrossRef][Medline] [Order article via Infotrieve]
  38. Leid, M., Kastner, P., Lyons, R., Nakshatri, H., Saunders, M., Zacharewski, T., Chen, J.-Y., Staub, A., Garnier, J.-M., Mader, S., and Chambon, P. (1992) Cell 68, 377-395 [Medline] [Order article via Infotrieve]
  39. Petkovich, M., Brand, N. J., Krust, A., and Chambon, P. (1987) Nature 330, 444-450 [CrossRef][Medline] [Order article via Infotrieve]
  40. de The, H., Marchio, A., Tiollais, P., and Dejean, A. (1987) Nature 330, 667-670 [CrossRef][Medline] [Order article via Infotrieve]
  41. Giguere, V., Shago, M., Zirngibl, R., Tate, P., Rossant, J., and Varmuza, S. (1990) Mol. Cell. Biol. 10, 2335-2340 [Medline] [Order article via Infotrieve]
  42. Thompson, C. C., Weinberger, C., Lebo, R., and Evans, R. M. (1987) Science 237, 1610-1614 [Medline] [Order article via Infotrieve]
  43. Weinberger, C., Thompson, C. C., Ong, E. S., Lebo, R., Gruol, D. J., and Evans, R. M. (1986) Nature 324, 641-646 [Medline] [Order article via Infotrieve]
  44. Milbrandt, J. (1988) Neuron 1, 183-188 [Medline] [Order article via Infotrieve]

©1997 by The American Society for Biochemistry and Molecular Biology, Inc.