NMR Spectroscopic Studies of the DNA-binding Domain of the Monomer-binding Nuclear Orphan Receptor, Human Estrogen Related Receptor-2

Unlike steroid and retinoid receptors, which associate with DNA as dimers, human estrogen related receptor-2 (hERR2) belongs to a growing subclass of nuclear hormone receptors that bind DNA with high affinity as monomers. A carboxyl-terminal extension (CTE) to the zinc-finger domain has been implicated to be responsible for determining the stoichiometry of binding by a nuclear receptor to its response element. To better understand the mechanism by which DNA specificity is achieved, the solution structure of the DNA-binding domain of hERR2 (residues 96–194) consisting of the two putative zinc fingers and the requisite 26-amino acid CTE was analyzed by multidimensional heteronuclear magnetic resonance spectroscopy. The highly conserved zinc-finger region (residues 103–168) has a fold similar to those reported for steroid and retinoid receptors, with two helices that originate from the carboxyl-terminal ends of the two zinc fingers and that pack together orthogonally, forming a hydrophobic core. The CTE element of hERR2 is unstructured and highly flexible, exhibiting nearly random coil chemical shifts, extreme sensitivity of the backbone amide protons to solvent presaturation, and reduced heteronuclear {1H-15N} nuclear Overhauser effect values. This is in contrast to the dimer-binding retinoid X and thyroid hormone receptors, where, in each case, a helix has been observed within the CTE. The implications of this property of the hERR2 CTE are discussed.

hERR1 1 and hERR2 are members of the nuclear hormone receptor superfamily of transcription factors based upon sequence homology (1,2). They are derived from human kidney and cardiac cDNA libraries, respectively (3). Members of this family contain a highly conserved DNA-binding domain (DBD) of 66 amino acids, which consists of two zinc fingers, each of which coordinates a zinc ion tetrahedrally with four cysteine thiolates. Much of what is known mechanistically about this family has been derived from functional (4 -6) and structural (7)(8)(9)(10)(11)(12)(13) studies of the glucocorticoid, estrogen, thyroid hormone, and retinoid receptors, which bind as dimers to direct or inverted repeats of an AGGTCA or AGAACA response sequence (called half-site). The hERR differs from these classical nuclear hormone receptors in that (a) it has no known activating ligand, making it an "orphan receptor"; and (b) the intact protein as well as the isolated DBD bind as a monomer to a 5Ј-extended version of the estrogen/retinoid receptor half-site: TCAAG-GTCA. 2 Similar response elements are recognized by a number of key monomer-binding orphan receptors, including murine steroidogenic factor-1/embryonal long terminal repeat-binding protein (14,15), bovine Ad4-binding protein (16), rat nerve growth factor I-B (17), and Drosophila FTZ-F1 (fushi tarazu factor) (18). In particular, steroidogenic factor-1 recognizes CCAAGGTCA, is a major regulator of the expression of steroidogenic enzymes (19), and has been implicated in sex determination (20). hERR1 has recently been found to activate the expression of human lactoferrin in endometrium and mammary gland cell lines and hence must play a role in modulating immune responses and cell growth (21).
Although there is currently no structural data available on monomer-binding nuclear hormone receptors, several functional/mutagenesis studies have identified a carboxyl-terminal extension (CTE) of the zinc-finger modules as a second structural determinant that contributes to DNA binding. Studies of the hERR homolog, FTZ-F1, have identified a 30-residue CTE termed the FTZ-F1 box (residues after Glu-171 in hERR2), which contributes a factor of 1000 to the binding constant; this region is proposed to be responsible for binding the 5Ј-end base pairs in the FTZ-F1 response element (22). In contrast, a shorter 12-amino acid CTE in the RXR termed the T box is helical and has been implicated in providing protein-protein interactions that stabilize the homodimer (12,23). Furthermore, a chimeric protein, generated between the zinc fingers of the RXR and a carboxyl-terminal element in nerve growth factor I-B called the A box, binds as a monomer to the nerve growth factor I-B response element (23). Thus, while the CTE of the RXR is helical and contributes to the binding affinity via protein-protein interactions, the monomer-binding receptors have an as yet structurally uncharacterized CTE that provides additional binding affinity via protein-DNA interactions. The present NMR studies of the hERR2 DBD, which consists of the zinc-finger modules and the CTE, represent a first step toward defining the structural basis of these interactions.

EXPERIMENTAL PROCEDURES
Expression of the Wild-type hERR2 DBD-The hERR2 DBD (residues 96 -194) was expressed in BL21 cells using the pGEX-2T vector (Pharmacia Biotech Inc.), which produces an amino-terminal fusion protein with glutathione S-transferase. Unlabeled protein was expressed at 37°C in Luria-Bertani broth by induction for 4 h with 0.5 mM isopropyl-␤-D-thiogalactopyranoside once cells had reached A 600 ϳ 0.8. 15 N-Labeled protein was obtained using a minimal medium at pH 7.2 containing 44 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , 9 mM NaCl, 1 mM MgSO 4 , 0.1 M CaCl 2 , 0.02 M FeCl 3 , 0.1 mM ZnCl 2 , 1% glucose, 1 mg/liter thiamine, and 1 g/liter ( 15 NH 4 ) 2 SO 4 (Cambridge Isotope Laboratories). Cells were pelleted and resuspended in lysis buffer containing 50 mM Tris (pH 8.0), 250 mM KCl, 5 M ZnCl 2 , 30 mM DTT, 1 mM phenylmethylsulfonyl fluoride, and 1% Triton X-100. After one freeze-thaw cycle, cells were lysed by sonication. The high DTT concentration was necessary to recover the hERR2 fusion protein from the pellet since there was extensive protein cross-linking via disulfide bonds. The sonicate was cleared by centrifugation for 15 min at 15,000 ϫ g. Glutathione-Sepharose 4B beads (Pharmacia) were added to the crude cell lysate along with 100 mM DTT and 100 M ZnCl 2 . After a 1-h incubation at 5°C, the fusion protein⅐glutathione-Sepharose complex was pelleted and washed three times with a phosphate-buffered saline solution containing 100 mM DTT and 1 mM ZnCl 2 . The first wash was done with 0.8 M NaCl to remove DNA. The pelleted protein/beads were then resuspended in 50 mM Tris (pH 8.0), 150 mM NaCl, 2 mM CaCl 2 , 100 mM DTT, and 100 M ZnCl 2 containing 50 units/ml bovine thrombin (Sigma) and incubated at room temperature until all of the hERR2 DBD had been cleaved from the glutathione S-transferase⅐glutathione-Sepharose complex, at which point the proteolysis reaction was quenched with 1 mM phenylmethylsulfonyl fluoride. The eluted protein was concentrated with an Amicon flow cell (YM-3) and purified by HPLC using a C 18 reverse-phase column (Vydac). After an initial wash with 27% acetonitrile and 0.1% trifluoroacetic acid, protein was eluted with a 0.6%/min acetonitrile gradient. Fractions containing the hERR2 DBD were lyophilized and weighed. The yield of pure hERR2 (Ͼ97%) was 2 mg/l of culture for labeled protein and 10 -15 mg/liter for unlabeled protein. Protein was homogeneous based upon SDS-polyacrylamide gel electrophoresis, HPLC, and electrospray mass spectral analysis (calculated M r ϭ 11,416; observed M r ϭ 11,417). Mass spectral analysis of the 15 Nlabeled hERR2 DBD indicated Ͼ97% incorporation of label. The electrospray mass spectrum of the native hERR2 DBD showed an M r of 11,547 as expected for hERR2 with two bound zinc atoms.
Construction of the C163A Mutant-To improve the level of expression of the hERR2 DBD in Escherichia coli and the stability of the protein in vitro, a C163A mutation that removes a non-coordinating sulfhydryl group and a number of silent mutations designed to favor the biased codon usage in E. coli were introduced into the gene for the hERR2 DBD. 3 NdeI and BamHI sites flanking the DBD gene were also incorporated via polymerase chain reaction. The NdeI/BamHI hERR2 DBD fragment was cloned into pET24a behind a T7/lacO promoter (24), and the expression plasmid was then introduced into a lysogenic E. coli strain (BL21(DE3)). Uniformly 15 N-enriched protein sample was prepared using a minimal medium similar to that described above, except that the concentrations of glucose and ( 15 NH 4 ) 2 SO 4 were 3 and 2 g/liter, respectively. 13 C, 15 N-Labeled protein was obtained with a minimal medium in which [ 13 C 6 ]glucose (ISOTEC) was used. Gene expression was induced by adding isopropyl-␤-D-thiogalactopyranoside at a cell density of A 600 ϳ 1.0 -1.2. 1 h after isopropyl-␤-D-thiogalactopyranoside induction, rifampicin was added to the culture to a final concentration of 100 g/ml; the cells were incubated for another 5 h at 37°C. The protein was isolated and purified to homogeneity (Ͼ95%) following a protocol similar to that used for the wild-type hERR2 DBD, with the exclusion of the glutathione-Sepharose chromatography step. The yield of purified DBD improved to 10 mg/liter of minimal medium.
DNA Binding Assays-Gel-shift assays were carried out using 32 Pend-labeled oligonucleotides that represent the recognition site for hERR2, with two additional GC base pairs added on either end for stability (5Ј-GCTCAAGGTCACG-3Ј and the complementary strand). Labeling and assays were performed using standard procedures (25). The gel was 5% acrylamide (1:50 bisacrylamide) in 45 mM Tris borate. Incubations were carried out for 30 min at 25°C in 6% glycerol, 1 mM DTT, 10 mM Tris (pH 8.0), 50 mM KCl, and 0.5 mM MgCl 2 , with DNA at Ͻ3.7 nM, hERR2 (wild-type and C163A) varied from 23 M to 23 nM, and poly(dI⅐dC) at 0.1 g/l.
NMR Sample Preparation-The lyophilized hERR2 DBD samples were resuspended in 6 M urea containing 0.5 mM ZnCl 2 and 100 mM DTT and immediately passed over a PD-10 gel filtration column (Pharmacia) equilibrated with NMR buffer (20 mM potassium phosphate (pH 6.5-6.7), 100 mM KCl, 0.3 mM ZnCl 2 , and 5% D 2 O). After the addition of 10 mM DTT-d 10 , the sample was concentrated with a Centricon 3 membrane to give a final hERR2 DBD concentration of 0.8 -1.5 mM. All solvents used for sample preparation were purged with argon gas. Since the protein precipitated at pH Ͻ6.5, all NMR experiments were carried out at pH 6.5-6.7.
NMR Spectroscopy-All NMR spectra were recorded on Bruker DMX750, AMX600, and AMX500 spectrometers at 300 K. The 1 H carrier was set on H 2 O/HOD in all experiments and shifted to the center of the amide region during acquisition in the HSQC experiments. Phasesensitive detection was achieved using either time proportional phase incrementation (26) or TPPI-States (27) quadrature detection. A 1 H-1 H NOESY spectrum (28) was recorded with a 100-ms mixing time, while jump and return NOESY spectra (29) were recorded with 200-, 100-, and 50-ms mixing times. A NOESY spectrum with an 80-ms mixing time was also recorded in D 2 O. The NOESY spectra included a Hahnecho period at the end of the pulse sequence to improve the base line (30,31). TOCSY spectra (32) were recorded with DIPSI-2 (decoupling in the presence of scalar interactions) (33) spin-lock periods of 30 and 50 ms. Double quantum filtered COSY (34) and double quantum (35,36) experiments were performed according to standard methods. Two-dimensional homonuclear data were recorded with 128 -320 scans, 512 t 1 increments, and 8192 data points/increment and at 500-MHz proton carrier frequency. A spectral width of 12,500 Hz was used in the F 2 dimension, while spectral widths of 6250 -7000 Hz were used in the F 1 dimension (12,500 Hz in the double quantum experiment). Digital oversampling by a factor of 2 was used to reduce base-line distortion (37). Two-dimensional { 1 H-15 N} HSQC and NOESY-HSQC (100-ms mixing time) spectra were acquired using standard methods (38,39). Threedimensional 15 N NOESY-HSQC (40,41) data were recorded with mixing times of 100 and 150 ms, using 128 t 1 ( 1 H) increments, 32 t 2 ( 15 N) increments, and 16 free induction decays of 1024 data points/increment pair. Three-dimensional TOCSY-HSQC data were recorded in the same manner, with a spin-lock period of 40 ms. All multidimensional data were processed using either a modified version of the FTNMR software provided by Dr. Dennis Hare (Hare Research) or the Triad software package (Tripos Associates). Spectra were Fourier-transformed after applying a lorentzian-to-gaussian weighting function in the directly detected dimension and a shifted sine bell weighting function in the indirectly detected dimension(s). The water signal was removed from all spectra by applying a low pass filter to the time domain data (42).
Three-dimensional constant-time HNCA and HNCOCA spectra were collected on the 13 C, 15 N-labeled C163A hERR2 DBD following established methods (43); pulsed-field gradients were incorporated in these schemes for solvent suppression. These spectra were recorded at 600 MHz with 32 t 1 increments ( 15 N carrier at 117.75 ppm, spectral width of 1429 Hz), 64 t 2 increments ( 13 C-␣ carrier at 54 ppm, spectral width of 4225 Hz), and 16 free induction decays of 1024 data points for the directly detected 1 H resonances (carrier at 8.2 ppm, spectral width of 5434 Hz). Heteronuclear { 1 H-15 N} steady-state NOEs for the C163A hERR2 DBD were also measured at 600 MHz using two-dimensional HSQC experiments (44). To minimize magnetization transfer from water to any rapidly exchanging amide protons, a 6-s relaxation/irradiation delay was employed between passes through the pulse sequence, and water was suppressed with flip-back and gradient pulses.

RESULTS
Functional Characterization-The recombinant hERR2 DBD (residues 96 -194) exhibited a high affinity for the 5Јextended half-site sequence, 5Ј-TCAAGGTCA-3Ј, in gel-shift assays. 2 Analysis of the holoprotein by pneumatically assisted electrospray spectrometry gave a molecular mass 130 Ϯ 2 Da higher than that of the apoprotein, indicative of two tightly bound zinc ions/polypeptide (data not shown). These data are consistent with the prediction that hERR2 is a zinc-finger DNA-binding protein (3).
NMR Assignments of the Wild-type hERR2 DBD-NMR studies were carried out on the recombinant DBD peptide, which contains the putative 66-amino acid (ZnCys 4 ) 2 -type zincfinger domain and the requisite 26-amino acid CTE for high affinity monomer binding to the extended half-site sequence (22,45). 87% of the backbone protons, 53% of the side-chain protons, and 86% of the backbone nitrogens have been assigned for Leu-102-Arg-174. Sequence-specific assignments were made using the NOE-based sequential assignment procedures (46) and are summarized in the two-dimensional { 1 H-15 N} HSQC spectrum (Fig. 1). Sequential d NN connectivities in twodimensional NOESY and NOESY-HSQC spectra were used to connect most of the residues, with the help of two-dimensional TOCSY and COSY spectra to identify spin systems. The double quantum spectrum was especially helpful in resolving ambiguities in the assignment of side-chain protons and in assigning glycine C␣H protons. Assignment ambiguities due to overlap could usually be resolved using the three-dimensional NOESY-HSQC and TOCSY-HSQC spectra in conjunction with the higher resolution two-dimensional { 1 H-15 N} HSQC spectrum. In cases where both 15 N and 1 H shifts were overlapped, a high resolution two-dimensional { 1 H-15 N} HSQC spectrum obtained at lower temperature (292 K) proved useful. The sequential and medium-range NOE connectivities, summarized in Fig. 2, clearly identify two helices, but show no sign of helix formation after the Gly-Met termination site for helix-2. Of the 77 back-bone amide cross-peaks observed, 67 could be assigned to 68 residues, while the remainder exhibited no sequential connectivities. The majority of the unassigned residues are in what is a highly basic loop region in finger-2 (residues 149 -155) of other structurally characterized nuclear hormone receptors and in the carboxyl-terminal region. Resonances for 15 of the 20 residues C-terminal of Arg-174 could not be assigned since sequential and medium-range NOEs are either weak or absent.
Secondary Structural Elements and Global Fold of the Wildtype hERR2 DBD-The medium-range NOEs summarized in Fig. 2 show one 13-residue helix (Glu-121-Gly-133) that starts at the carboxyl-terminal end of finger-1 and a second 12-amino acid helix (Gln-156 -Gly-167) that starts at the carboxyl-terminal end of finger-2. The region after helix-2 shows no sign of secondary structure and may be disordered (see below). There is also a short antiparallel ␤-sheet between Cys-103/Leu-104 and Ser-119/Cys-120 based upon strong d ␣N connectivities and long-range NOEs between Cys-103 C␣H and Cys-120 NH and between Ser-119 C␣H and Leu-104 NH. This same ␤-sheet is also present in homologous sequences of the retinoid X (12), retinoic acid (11), and glucocorticoid (7) receptors.
It is also interesting to note that the Val-105 and Cys-106 amide protons are two of the most persistent (along with three others discussed below) in D 2 O (ϳ30 h), suggesting that they are both protected from solvent and probably involved in strong internal hydrogen bonds. The equivalent residues in other zincfinger proteins have been shown, through structural and/or H 2 O/D 2 O exchange data, to be hydrogen-bonded to S-␥ of the first zinc-finger cysteine ligand (7,12,(47)(48)(49)(50)(51)(52)(53). This amide to S-␥ hydrogen bond may be present in hERR2 based upon homology and the slow hydrogen/deuterium exchange rate.
On the basis of long-range NOEs that had been identified ( Fig. 3A), the 66-amino acid zinc-finger domain (i.e. up to the Gly-Met termination signal for helix-2) of hERR2 consists of two helices that are packed against each other to enclose a hydrophobic core consisting of Phe-126, Phe-127, Thr-130, Ile-131, Phe-160, and Leu-164 (Fig. 3B). Long-range NOEs between residues in or near this hydrophobic core and Val-173 indicate that the region just past helix-2 also folds onto this hydrophobic core. A similar overall pattern of long-range NOEs was observed in the RXR (12), suggesting that this part of the structure has a fold very similar to that of the retinoid X, estrogen (9, 10), glucocorticoid (7,8), and retinoic acid (11) receptors. The slow exchange rates for the amide protons of Cys-123, Lys-124, and Ala-125 suggest that these residues may be buried in the protein interior, which is reasonable since they reside on the amino-terminal end of helix-1, near the hydrophobic core.
Role of Cys-163-Alignment of the sequence with those of structurally characterized receptor molecules suggests that Cys-163 of hERR2 is not involved in coordinating zinc ions. The homologous cysteines in these hormone receptors are found close to the hydrophobic core and are partially solvent-accessible. Assuming that Cys-163 has similar spatial orientation, it could facilitate aggregation via disulfide formation at the high protein concentrations and temperatures employed for the NMR experiments and cause sample degradation over extended periods of time. Hence, to increase the intrinsic stability and solubility of hERR2, Cys-163 was replaced with an alanine.
Alanine was chosen because it has comparable hydrophobic character to cysteine since it is probable that Cys-163 is partly buried.
Gel-shift studies indicate that both the wild-type and C163A hERR2 DBDs have comparable affinities for the extended halfsite sequence (data not shown). With the aid of the existing data for the wild-type hERR2 DBD, the proton and nitrogen chemical shifts were reassigned for the mutant protein from two-and three-dimensional 15 N-edited spectra. Based on perturbations of amide proton and nitrogen chemical shifts (Fig.  4), the mutation has little effect on the structure, other than minor perturbations of groups near the hydrophobic core where Cys-163 is located. The perturbations of 15 N shifts for some residues in finger-1 (Fig. 4A) are consistent with long-range NOE data (Fig. 3A) that indicate that this region of finger-1 participates in the formation of the hydrophobic core.
The CTE Is Unstructured-It should be noted that the threedimensional 15 N-edited spectra for the mutant protein did not enable us to extend the backbone resonance assignments beyond what was defined for the wild-type protein. Because of dramatically improved expression levels with the mutant gene construct, a uniformly 13 C, 15 N-labeled sample of the C163A hERR2 DBD was prepared, and triple-resonance experiments were recorded. The combination of three-dimensional HNCA and HNCOCA spectra allowed us to establish inter-residue connectivities using 13 C-␣ chemical shift values and to complete the backbone assignments for the CTE of hERR2. The 13 C-␣ chemical shift provides a reliable index of the local secondary structure; a downfield or upfield shift in ␦( 13 C-␣) is indicative of a helix or ␤-stranded configuration, respectively (54,55). Comparison of the C-␣ chemical shifts in the mutant protein with random coil values (Fig. 5A) reveals the same two helical elements observed in the wild-type DBD. Furthermore, the analysis establishes that the CTE does not adopt a regular secondary structure, but has chemical shifts characteristic of a random coil. The absence of long-range NOEs and most sequential NOEs for residues C-terminal of Arg-174 also provides strong evidence that the CTE is unstructured.
It has been generally surmised that unstructured regions of a protein are characterized by increased main-chain flexibility. The dynamic properties of the DNA-binding domain of hERR2 in solution were evaluated by measuring heteronuclear { 1 H-15 N} NOEs for the C163A variant. The magnitude of the steady-state { 1 H-15 N} NOE is highly sensitive to the dynamics of the 1 H-15 N vector; negative or small positive NOE values are diagnostic of significantly fast motions relative to the overall tumbling rate of the molecule. Residues in the CTE and the N terminus of the C163A hERR2 DBD exhibit diminished heteronuclear NOEs relative to those in the zinc-finger core (Fig. 5B). The loop (Thr-148 -Ser-155) between the sixth and seventh ligating cysteines represents the most flexible region of the zinc-finger core. In the CTE, the heteronuclear NOE begins to decrease C-terminal of Val-173, with near zero values reported for Arg-179 -Arg-187 and negative values observed for the last six residues. This trend in heteronuclear NOE for the CTE is consistent with an unstructured polypeptide tethered to the folded zinc-finger domain. The flexible nature of the CTE also explains the absence of sequential and long-range interactions involving this region that could be identified from the two-and three-dimensional NOESY data sets. The evidence indeed supports that the functionally critical CTE of hERR2 is dynamically disordered. DISCUSSION The region encompassing residues 103-168 of hERR2 shares a strong sequence homology with the zinc-finger domains of numerous members of the nuclear hormone receptor superfam- FIG. 3. Global fold of the hERR2 DBD. A, sequence (3) of the full-length wild-type hERR2 DBD used in this study, showing the zinc ligands, the location of the two zinc fingers, and residues exhibiting long-range NOEs (7). The additional amino-terminal Ser-Gly sequence is a cloning artifact of the pGEX-2T expression system. Asterisks indicate slowly exchanging amides. Circled and boxed residues represent the P and D boxes, respectively. In steroid receptors, the P box provides base contacts to the half-site major groove, whereas the D box residues form the dimer interface. The T and A boxes that comprise the CTE are described under "Discussion." B, structural model of the hERR2 DBD consistent with the observed long-range NOEs. The model was constructed by incorporating the observed secondary structural elements into the reported structure of the retinoic acid receptor (11) and is illustrated here using MOLSCRIPT (58).
ily. This domain and a region (CTE) immediately downstream of it comprise the minimum motif for site-specific 1:1 binding to an extended version of the canonical half-site sequence (5Ј-TCAAGGTCA-3Ј). We have constructed a recombinant form of the DBD of hERR2 and shown that it binds two zinc(II) ions/ polypeptide. NMR structural analysis of the DBD indicates a polypeptide fold consisting of two orthogonally packed helices within the putative zinc-finger domain; this fold is typical of those observed in steroid and retinoid receptors. These data confirm that hERR2 is a member of the nuclear hormone receptor family of zinc-finger transcription factors.
The zinc-finger domains of nuclear hormone receptors contain nine highly conserved cysteines, yet only the eight Nterminal cysteines are involved in coordinating the zinc ions. We have replaced the ninth cysteine (Cys-163) of hERR2 with an alanine. Comparison of the functional and solution NMR properties of the mutant with those of the wild-type protein suggests that Cys-163 is important neither for the proper folding of the polypeptide around the zinc sites nor for binding of the DBD to the cognate DNA. We cannot, however, exclude a functional role for this invariant cysteine in vivo.
It is the 26-amino acid CTE that is of greatest functional interest in hERR2 since the equivalent regions in nerve growth factor I-B (23), steroidogenic factor-1 (19), and FTZ-F1 (FTZ-F1 box) (22) have been shown to provide key interactions for monomeric instead of dimeric binding to DNA (see above). The region has further been proposed to provide sequence-specific contacts with the 5Ј-base pairs in the response element (TCAAGGTCA in hERR2) (45). Prior to this work, no structural information was available on the CTE elements of monomerbinding receptors. The CTE of the hERR2 DBD (Fig. 3A) contains sequences corresponding to the T and A boxes in a number of structurally characterized dimer-binding receptors. In the RXR, the T box is helical and has been implicated in stabilizing the homodimer via protein-protein interactions (12). In the ternary complex of the thyroid hormone receptor, the RXR, and DNA (13), the CTE of the thyroid hormone receptor does not participate in the dimer interface. Instead, the T box serves as a connecting loop to a long helical A box, which makes extensive nonspecific contacts with the DNA backbone. It is evident that the CTE of the thyroid hormone receptor has a different functional role than those of the monomer binders.
Unlike those of the RXR and the thyroid hormone receptor wherein helices have been observed, the carboxyl-terminal T/A box region of hERR2 appears to be disordered, having no identifiable secondary structure. This is based on (a) C-␣ chemical shifts (Fig. 5A) that approach random coil values; (b) reduced heteronuclear { 1 H-15 N} NOEs (Fig. 5B); (c) the appearance of an additional six HSQC peaks when gradient pulses or spinlock purge pulses are used, rather than presaturation, for water suppression (data not shown); and (d) sequential NOEs that are either weak or absent. Since the CTE contributes to both affinity and specificity, it may adopt a defined secondary structure upon DNA binding as in the trp repressor or glucocorticoid receptor (8,56), or it may adopt an ordered but extended structure as in GATA-1 (57). Preliminary NMR studies of the C163A hERR2-DNA complex suggest that ordering of the CTE does occur upon binding to DNA, 3 and studies of the structure of the complex are currently in progress.