The C-terminal extension (CTE) of the nuclear hormone receptor DNA binding domain determines interactions and functional response to the HMGB-1/-2 co-regulatory proteins.

Previously, we and others reported that the high mobility group proteins, HMGB-1/-2, enhance DNA binding in vitro and transactivation in situ by the steroid hormone subgroup of nuclear receptors but did not influence these functions of class II receptors. We show here that the DNA binding domain (DBD) is sufficient to account for the selective influence of HMGB-1/-2 on the steroid class of receptors. Furthermore, the use of chimeric DBDs reveals that this selectivity is dependent on the C-terminal extension (CTE), amino acid sequences adjacent to the zinc finger core DBD. HMGB-1/-2 interact directly with the DBDs of steroid but not class II receptors, and this interaction requires the CTE. This in vitro interaction correlates with a requirement of the CTE for maximal HMGB-1/-2 enhancement of DNA binding in vitro and transcriptional activation in cells. Finally, class II receptor DBDs have a much higher intrinsic affinity for DNA than steroid receptor DBDs, and this affinity difference is also dependent on the CTE. These results reveal the importance of the steroid receptor CTE for DNA binding affinity and functional response to HMGB-1/-2.

which no endogenous ligand has been identified (1)(2)(3)(4). Each of the receptor subclasses is characterized by a unique mechanism of action with respect to dimerization and DNA sequence recognition. Steroid receptors form homodimers that optimally recognize hexameric DNA elements arranged as inverted repeats separated by three unspecified base pairs. PR, GR, AR, and MR bind to the core hexamer AGAACA, whereas ER recognizes AGGTCA (2). Class II receptors preferentially function as heterodimers with RXR and recognize the AGGTCA hexamer arranged as direct repeats. Variable spacing between the direct repeats determines the RXR heterodimer binding specificity. Class II receptors, particularly TR, can also recognize an inverted repeat as homodimers, or half-sites as monomers. Orphan receptors can bind to the AGGTCA hexamer arranged either as a direct repeat, palindrome, or half-site as heterodimers with RXR, homodimers, or monomers (1,(5)(6)(7).
DNA-bound nuclear receptors activate transcription through assembly of a coactivator protein complex (8 -10). Some of these coactivators possess enzyme activities that are thought to facilitate access of general transcription factors to chromatin templates (11,12). Additionally, the coactivator complex may serve as a protein bridge to facilitate assembly of the basal transcription apparatus (13,14). We and others have identified another group of co-regulatory proteins, the high mobility group proteins 1 and 2 (HMGB-1/-2), that facilitate steroid receptor interaction with specific target DNA sequences and appear to be essential for maximal transcriptional activation by this subgroup of nuclear receptors (15)(16)(17)(18)(19)(20).
HMGB-1/-2 2 proteins are ubiquitous, conserved, non-histone chromatin proteins that bind to the minor groove of DNA in a structure-specific, sequence-independent manner (21,22). A clear physiological role for these proteins has not been defined; however, they have been implicated in processes that require manipulation of DNA structure and assembly of higher order nucleoprotein complexes, such as DNA replication and repair, recombination, and transcription (23,24). In addition, HMGB-1/-2 proteins enhance DNA binding and transcriptional activity of a number of eukaryotic transcriptional activators, including octamer transcription factors (Oct-1, Oct-2, and Oct-6) (25), homeodomain protein HOXD9 (26), p53 (27), viral transcription factors (28,29), Rel family members (30), and the steroid hormone receptors (15)(16)(17)(18)(19)(20). However, not all sequence-specific transcription factors are influenced by HMGB-1/-2. Paull et al. showed that only a subset of transcriptional activators were influenced by loss of the HMGB proteins, NHP6A and 6B, in yeast (31). Additionally, we and others have shown that both HMGB-1/-2 enhance DNA binding and transcriptional activity of the steroid hormone subclass of nuclear receptors while not affecting these functions of class II receptors, including RAR, VDR, or RXR (18,20,26).
The nuclear receptor core DBD consists of two zinc fingers and two ␣-helices. Helix1 makes base-specific contacts in the major groove, while helix2 maintains the overall structural fold of the core DBD (32)(33)(34)(35)(36)(37)(38). Both biochemical and structural analyses of the nuclear receptor DBDs have shown that this domain is highly conserved across all nuclear receptors. In addition to the core DBD, sequences located immediately C-terminal to the second zinc finger, termed the C-terminal extension (CTE), directly participate in DNA binding by the class II and orphan receptors. Unlike the core DBD, CTE sequences are not conserved among nuclear receptors, and the CTE adopts different structural motifs (39). However, these divergent structures share a common function to extend the protein-DNA interface beyond that of base-specific contacts in the major groove thus stabilizing DNA binding. Crystallization of a TR␤-RXR␣ DBD heterodimer bound to a direct repeat element demonstrated that the TR␤ CTE forms an additional ␣-helix (helix3) distinct from the core DBD that projects across the minor groove of the DNA helix making contacts along the phosphate backbone (34). NMR analysis of the RXR DBD in solution also revealed a third ␣-helix in the CTE (33). However, the crystal structure of the RXR DBD-DNA complex had a CTE with an extended structure, suggesting that the RXR CTE undergoes a conformational change upon binding to DNA (37,38). The orphan receptor CTE contains a short, conserved amino acid sequence, termed the "GRIP-box", with the consensus RXGRZP, where "X" is any amino acid and "Z" is a hydrophobic residue. Structural analysis of orphan receptor DBDs bound to DNA as either a homodimer (RevErb, (40)) or monomer (NGFI-B, (36)), revealed the CTE lying along the minor groove of DNA in an extended conformation distinct from that observed in either of the class II receptor DBDs. This GRIP-box creates an additional protein-DNA interface beyond that of the core DBD that interacts with specific base pairs. In the absence of DNA the orphan receptor CTE is unstructured suggesting that it also undergoes a conformational change upon association with DNA (35). Biochemical analysis has also demonstrated the importance of the CTE for high affinity DNA binding because point mutations or truncation of the class II and orphan receptor CTEs reduces or abolishes DNA binding (33,(41)(42)(43)(44)(45)(46). These results taken together suggest that the class II and orphan nuclear receptors have a bipartite DBD consisting of the core DBD that makes base-specific contacts with core hormone response elements (HREs) and the CTE that provides additional, largely nonspecific DNA contacts that stabilize the protein-DNA complex.
There is no structural evidence for the existence of an equivalent CTE in the steroid class of nuclear receptors. Structural studies of the ER␣ and GR DBDs complexed to DNA and in solution either lacked the comparable length CTE sequences in the expressed DBD constructs or the CTE was unstructured (32). Aside from an earlier report that sequences C-terminal to the core ER␣ DBD are important for DNA binding stability (47), no other biochemical data are available to suggest that the CTE of steroid receptors directly participates in DNA binding as it does with other nuclear receptors.
Here, we determine the mechanistic basis for the selective influence of HMGB-1/-2 on the steroid subclass of nuclear receptors. Through biochemical analysis of purified DBDs, we show that HMGB-1/-2 selectively influence DNA binding by the DBDs for steroid but not class II receptors. By use of chimeric DBDs, in which the CTE was swapped between steroid and class II receptors, we demonstrate that the CTE is responsible for the differential influence of HMGB-1/-2 on the two classes of nuclear receptors. Furthermore, class II receptor DBDs exhibited a much higher intrinsic affinity for their target DNAs than the steroid receptor DBDs, a difference also attributed to the CTE. Finally, we show that the CTE of steroid receptors is required for direct interaction with HMGB-1/-2 and for maximal HMGB-1/-2-enhanced DNA binding in vitro and transcriptional activation in cells. These results demonstrate that the CTE of steroid receptors plays a role in DNA binding but acts differently than class II DBDs by a mechanism that involves interaction with HMGB-1/-2.

EXPERIMENTAL PROCEDURES
Bacterial Expression Vectors-Expression vectors for RXR␣ (aa. 130 -223) and TR␤ (aa. 97-207) DBDs have been previously described (6,33) and were obtained from Ron Evans, Salk Institute, San Diego, CA, and Thomas Perlmann, Ludwig Institute for Cancer Research, Stockholm, Sweden, respectively. All other DBD vectors were subcloned by PCR amplification of DNA encoding the appropriate DBD and insertion into the BamHI and EcoRI restriction sites of the GST fusion vector, pGEX2T (Amersham Biosciences). PCR-generated fragments were verified by dideoxy sequencing (Sequenase 2.0, USB, Cleveland, OH) and were determined to be in frame with the GST tag. An EcoRI-SalI fragment containing the HMGB-1 cDNA was excised from the pBlueBacHis2B-HMGB-1 plasmid, previously described (18), and inserted into the EcoRI and SalI sites in pGEX4T1 (Amersham Biosciences). Cloning junctures were sequenced by dideoxy sequencing.
Chimeric DNA binding domains were subcloned using "splicing by overlap extension" (48). In brief, this technique involved two consecutive PCR reactions. The first reaction amplified the core DBDs and CTEs separately using primers containing sequences complementary to the "splice junction". In the second step, the PCR products were mixed with primers to the distal ends. Upon denaturation, reannealing, and amplification, the final PCR product contained the core DBD of one receptor fused at a precise junction to the CTE of another. The chimeric PCR products were inserted into pGEX2T (Amersham Biosciences) using BamHI and EcoRI restriction sites in the polylinker and were sequenced to verify that the DBDs were in frame with the GST tag and that the chimeric junction was correct.
Expression and Purification of GST Fusion Proteins-GST fusion proteins were purified from BL21 cells as previously described (49). Purified proteins were dialyzed against 20 mM Tris, pH 8.0, 50 mM NaCl, 1 mM DTT, and 1 M ZnCl 2 , and small aliquots were stored in siliconized, microcentrifuge tubes at Ϫ80°C. DBD concentrations were determined by a Lowry protein assay per manufacturer's instructions (Pierce) and comparison to known concentrations of ovalbumin on a silver stain SDS-gel. The Lowry and silver results were averaged to give the DBD concentration.
Production and Purification of Baculovirus-expressed Proteins-Recombinant baculovirus vectors that express FLAG-tagged human ER␣ and His-tagged human TR␤ and RXR␣ were obtained from W. Lee Kraus, Cornell University, Ithaca, NY, Nancy Weigel, Baylor College of Medicine, Houston, TX, and Dave Clemm, Ligand Pharmaceuticals, San Diego, CA, respectively. Receptors and HMGB-2 were expressed in Sf9 insect cells in Grace's insect medium supplemented with 10% fetal bovine serum as previously described (18). 200 nM T3 for TR␤ and 200 nM estradiol for ER␣ were added to Sf9 cell cultures for the last 24 h of expression. No ligand was added for HMGB-2 or RXR␣.
Electrophoretic Mobility Shift Assays (EMSA)-EMSAs for fulllength TR␤ and RXR␣ were performed as described previously (18). Briefly, components of DNA-binding reactions were incubated for 30 min on ice in 10 mM HEPES, pH 7.8, 50 mM KCl, 4 mM MgCl 2 , 1 mM DTT, and 12% glycerol in the presence of 0.2 g of poly(dI-dC), and 1 g of ovalbumin as carrier protein (25 l of total volume). Labeled, specific oligonucleotide (0.6 nM) was added, and reactions were incubated for 30 min on ice. DNA binding reactions were electrophoresed on non-denaturing 5% polyacrylamide gels (40:1 acrylamide/bisacrylamide ration) in 0.25ϫ TBE buffer (0.02 M Tris borate, pH 8.0, 0.02 M boric acid, 0.5 M EDTA) at 4°C. Antibody supershifts of TR␤/RXR␣ heterodimers were performed with a TR␤ mouse IgG monoclonal antibody (clone 2386/G10) prepared against a synthetic peptide corresponding to amino acids 31-50 in the N terminus. EMSA for various receptor DBDs used similar methods, except that the binding reactions were carried out in 10 mM Tris, pH 8.0, 50 mM KCl, 6% glycerol, 1 mM DTT, 100 ng of poly(dI-dC), and 0.1% Nonidet P-40 or regulatoryCA630 (Sigma) in 20 l of total volume and were electrophoresed on 6% polyacrylamide gels. After electrophoresis, gels were dried, autoradiographed, and free [ 32 P]DNA and [ 32 P]DNA-protein complexes were quantitated by direct scanning of gels for radioactivity by a series 400 Amersham Biosciences Phosphor-Imager. Data were expressed graphically as the normalized fraction of DNA bound versus DBD concentration. The fraction-bound DNA was calculated as 1-(free DNA/(bound DNA ϩ free DNA)). The data were normalized so that the fraction DNA bound at saturation was set to 1.0. All other data were normalized to 1.0 (52). Saturation typically occurred at fraction-bound values of 0.7 to greater than 0.9.
All DNA binding curves were best fit to the following equation: where y is the normalized fraction DNA bound, x is the total DBD concentration and n is the Hill cooperativity coefficient. The Hill cooperativity coefficient varied from n ϭ 1-2, and the curves shown are the best fit (r ϭ 1; Kaleidagraph). The apparent dissociation constant (K d app ) was determined as the DBD concentration at which y ϭ 0.5 from the average curve of at least three independent experiments.
Mammalian Cell Transfection-The PR-B, DHLBD (DNA binding domain, hinge and ligand binding domain) and HMGB-1 expression plasmids as well as the progestin-responsive PRE 2 tk-LUC reporter gene have been described previously (18,53). The TR␤ expression plasmid and DR4-LUC reporter gene were obtained from Ming-Jer Tsai, Baylor College of Medicine. A PRtrPR chimeric receptor construct was created using splicing by overlap extension as described for the chimeric DBDs and inserted into the same vector as the wild-type PR-B. The PRtrPR chimeric receptor consists of full-length PR-B with the PR CTE (aa 633-670) replaced by the TR␤ CTE (aa 190 -207). The DHtrLBD chimeric construct was subcloned by cutting the PRtrPR plasmid with Bsu36I, filling in with Klenow, and cutting with EcoNI. This fragment was inserted into the BglII (Klenow fill-in)-EcoNI sites of the DHLBD plasmid. The DHtrLBD chimeric receptor consists of PR-B (aa 546 -933) with the PR CTE replaced by the TR␤ CTE. Transient transfections were performed by an adenovirus-mediated method as previously described (18,54). Purified, defective adenovirus particles, covalently coupled to poly-L-lysine were provided by Nancy Weigel, Baylor College of Medicine. The cells were incubated for 24 h posttransfection and then treated with or without hormone for another 24 h at 37°C. At 48 h posttransfection, cells were washed in 40 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 1 mM EDTA and harvested in 200 -300 l of lysis buffer (20 mM potassium phosphate, pH 7.8, 5 mM MgCl 2 , 0.5% Triton-X100) per well. The lysates were then centrifuged at 100,000 ϫ g for 5 min, and supernatants were assayed for luciferase and ␤-galactosidase activities. The total amount of plasmid DNA in each transfection was normalized with empty vector (pBlueScript) such that each well received equimolar amounts of plasmid DNA. A constitutively active pRSVϪ␤-galactosidase plasmid was included in each transfec-tion as an internal control for well-to-well variation in transfection efficiency.
Luciferase and ␤-galactosidase activities were analyzed on a Monolight 2010 luminometer as described previously (18). Luciferase values were corrected for variations in transfection efficiency by calculating luciferase/␤-galactosidase ratios. The data were expressed as relative luciferase activity by setting the hormone-induced luciferase/␤-galactosidase values in wells transfected with receptors in the absence of HMGB-1 to 100%. All other values were calculated relative to 100%. For receptor chimeras, data were expressed as "fold coactivation" by setting the hormone-induced luciferase/␤-galactosidase values in wells transfected with receptors in the absence of HMGB-1 to 1. All other values were calculated relative to 1.
GST Pull-downs-Bacterial lysates expressing GST or GST-HMGB-1 were incubated with 25 l of packed glutathione-Sepharose

HMGB-1/-2 Selectively Enhance DNA Binding and Transactivation of a Steroid Receptor, PR, but Not a Class II Nuclear
Receptor, TR␤-In previous studies we showed that HMGB-1/-2 increased steroid receptor DNA binding affinity in vitro and transactivation in intact cells but had no effect on class II receptors (18). The class II receptors analyzed included RXR, RAR, and VDR but not thyroid hormone receptor (TR␤) (18). Because TR␤ has the best characterized CTE of the class II receptors, it was important for the present study to examine the effects of HMGB-1/-2 on the DNA binding and transactivation properties of TR␤. To complete our comparative analysis we examined the effects of HMGB-1/-2 on TR␤ DNA binding and transactivation using PR as a steroid receptor control. TR␤ binds as a heterodimer with RXR␣ to a direct repeat element separated by four intervening base pairs (DR4). In EMSA, purified TR␤ formed a weak protein-DNA complex with the DR4 oligonucleotide, which was unaffected by the addition of HMGB-2 or an unrelated control protein, ovalbumin (Fig. 1A). Addition of purified RXR␣ induced formation of a TR␤-RXR␣ heterodimer complex, which was also unaffected by either HMGB-2 or ovalbumin. The TR␤-RXR␣ heterodimer complex was specific as shown by a supershift with a TR-specific antibody (Fig. 1A). The influence of HMGB-2 on the affinity of the TR␤-RXR␣ heterodimer for DR4 was determined by varying the concentration of the TR␤-RXR␣ heterodimer in the presence and absence of HMGB-2 (Fig. 1B). As shown in Fig. 1B,   FIG. 2. A, schematic of nuclear receptor core DBDs for human PR and TR␤. B, CTE sequences for steroid and class II receptor DBDs used in this study. Expressed DBDs contain the core DBD (A) consisting of the zinc fingers through the highly conserved GM diamino acid motif and enough C-terminal sequence (B) to encompass a CTE based on either the TR␤ or RXR␣ DBD structure. C, purification of PR DBD 651 . The DBD was expressed in bacteria as a glutathione Stransferase fusion protein and purified in a two-step procedure as described under the saturation DNA binding curves for the TR␤-RXR␣ heterodimer are overlapping in the presence or absence of HMGB-2, indicating that HMGB-2 does not affect the affinity of the TR␤-RXR␣ heterodimer for specific DNA. This contrasts with the effect of HMGB-1/-2 on DNA binding of purified PR; as previously reported, HMGB-1/-2 dramatically increase the apparent DNA binding affinity of PR for a PRE by 20 -50-fold (15,18).
To determine the influence of HMGB-1/-2 on transactivation by TR␤ in cells, Cos-1 cells were transfected with expression plasmids for TR␤, HMGB-1, and a TR␤-dependent reporter gene (DR4-LUC); PR-B and a PRE 2 tk-LUC reporter were included as a steroid receptor control. As expected, HMGB-1 enhanced PR-mediated, hormone-dependent transactivation of the PRE 2 tk-LUC reporter gene by as much as 9-fold at the highest concentration of HMGB-1 plasmid (Fig. 1C). Under the same conditions HMGB-1 had no effect on T3-dependent, TRmediated transactivation of DR4-LUC. These results support our previous observation that HMGB-1/-2 stimulate both DNA binding and transactivation by steroid hormone receptors but does not affect these functions of class II receptors (18).
The Selective Influence of HMGB-1/-2 on Steroid Receptors Is Attributed to the DNA Binding Domain-To determine whether the DNA binding domain is sufficient to account for the differential effect of HMGB-1/-2 on steroid versus class II receptors, we next analyzed the effect of HMGB-1/-2 on the DNA binding properties of expressed DBDs from these two classes of nuclear receptors. The DBDs of PR, GR, and ER␣ were used as representative of the steroid subclass of receptors, while the TR␤ and RXR␣ DBDs were analyzed as representative of class II receptors (Fig. 2). Expressed DBDs contained sequences corresponding to the core zinc binding modules through the conserved glycine-methionine (GM) motif ( Fig. 2A) as well as C-terminal sequence sufficient to contain the CTE of class II receptors and equivalent length sequences from steroid hormone receptors (Fig. 2B). The DBDs were purified to near homogeneity as judged by silver stain SDS gels. By way of example, a purification of PR DBD 651 is shown in Fig. 2C (Ͼ90% purity) (49). The purified DBDs were then analyzed for DNA binding to their respective DNA elements in the presence and absence of recombinant, purified HMGB-2. Fig. 3A shows results from gel mobility shift assays using the PR DBD 651 (aa 552-651) as representative of steroid receptors (Figs. 2B and 3A). Increasing concentrations of PR DBD 651 were incubated with a single concentration of [ 32 P]-labeled, PRE/GRE-containing DNA oligonucleotide. The PR DBD 651 exhibited dose-dependent binding to the PRE/GRE, primarily as a dimer, with some monomer complex also apparent (Fig. 3A). Binding was specific as determined by antibody supershift of the complex and competition with a specific oligonucleotide (20) (data not shown). Addition of HMGB-2 caused the DBD to bind DNA at lower concentrations indicating an enhancement of DNA binding. Estimation of the DBD concentration at which half-maximal binding was achieved indicates that the apparent dissociation constants (K d app ) for the PR DBD 651 -PRE complex increased 9-fold from 140 ϫ 10 Ϫ9 M to 16 ϫ 10 Ϫ9 M in the absence and presence of HMGB-2, respectively (Table I). Because the PR DBD 651 has a shorter CTE than that described for the TR␤ DBD used in crystallization (see Fig. 2, A and B), we also examined DNA binding and HMGB-1/-2 effects on a PR DBD construct containing additional C-terminal sequence, PR DBD 670 (aa 562-670), equal in length to that of the TR␤ DBD (Fig. 2B). As summarized in Table I, the longer PR DBD 670 had a DNA binding affinity similar to PR DBD 651 (K d app ϭ 150 ϫ 10 Ϫ9 M and 17 ϫ 10 Ϫ9 M, respectively) and HMGB-1/-2 stimulated DNA binding by approximately the same extent (8.7-fold).
We next analyzed the effects of HMGB-2 on DNA binding properties of a class II receptor DBD, TR␤. Addition of increasing concentrations of purified TR␤ DBD to a constant amount of the DR4 probe yielded dose-dependent formation of both monomer-and dimer-DNA complexes (Fig. 4A). The monomer complexes constituted a greater proportion of bound DNA than dimer. Unlike the steroid receptor DBDs, HMGB-2 did not influence either TR␤ DBD monomer-or dimer-DNA complexes. We quantitated the influence of HMGB-2 on the affinity of the TR␤ DBD for a DR4 element (Fig. 4B) as well as an inverted repeat element, TREpal (Table I). TR␤ bound both elements with a similar K d app (3.8 ϫ 10 Ϫ9 M on DR4 and 2.4 ϫ 10 Ϫ9 M on TREpal, Table I), which were unaffected by HMGB-2 (K d app ϭ 2.6 ϫ 10 Ϫ9 M on DR4 and 2.2 ϫ 10 Ϫ9 M on TREpal, Table I). Interestingly, TR␤ had a significantly higher intrinsic apparent binding affinity on either DR4 or TREpal (K d app ϭ 3.  Table I). The affinity of PR DBDs for the PRE/GRE only approached that of the TR␤ DBD in the presence of HMGB-2 (K d ϭ 16 -17 ϫ 10 Ϫ9 M; Table I). Thus the TR␤ DBD has a substantially higher intrinsic affinity for its target DNA than the PR DBD.
To extend these analyses to other steroid and class II receptor DBDs, we also tested the influence of HMGB-1/-2 on the binding affinity of purified GR and ER␣ DBDs for their respective GRE and ERE palindromic target elements (Fig. 3, C and  D), RXR␣ DBD for a DR1 element (Fig. 4C), and TR␤-RXR␣ DBD heterodimer for a DR4 element (Fig. 4D). Both class II receptor DBDs had substantially higher intrinsic DNA binding affinities than the GR or PR DBDs and a slightly higher affinity than the ER␣ DBD. Addition of HMGB-2 stimulated DNA binding by the steroid receptor DBDs but had no effect on either class II receptor DBD (Table I). These results suggest that high affinity DNA binding is intrinsic to class II receptor DBDs, while steroid receptor DBDs require the accessory protein, HMGB-1/-2 for a comparable affinity. The Selective Effect of HMGB-1/-2 on the Steroid Receptor DBD Is Dependent on the CTE-To determine whether the CTE is responsible for the higher intrinsic affinity of class II receptor DBDs and the selective effect of HMGB-1/-2 on steroid receptors, we constructed chimeric DBDs that swapped the CTE between PR and TR␤ DBDs. The chimeras were subcloned using splicing by overlap extension to create a precise boundary between the core DBD and CTE without introducing alterations in the peptide sequence (48). As illustrated in Fig. 5A, the PR TRCTE chimera contains the core zinc binding modules of PR through the conserved glycine-methionine motif (aa 562-632) fused to the CTE sequences from TR␤ (aa 170 -207) (Fig.  5A). This chimera is expected to retain the DNA binding specificity of PR for an inverted repeat PRE/GRE. Conversely, the TR PRCTE chimeric DBD contains the TR␤ core DBD (aa 97-169) fused to the CTE region of PR (aa 633-670) and should retain the binding specificity of TR␤ for DR4. DNA binding by these chimeric DBDs was assessed by quantitative gel mobility shift assays as was done with wild-type PR and TR␤ DBD constructs in Figs. 3 and 4. The PR TRCTE DBD bound to the PRE/GRE with a 6-fold higher affinity (K d ϭ 24 ϫ 10 Ϫ9 M; Table  I) than the PR DBD 670 (K d ϭ 150 ϫ 10 Ϫ9 ), while addition of HMGB-1/-2 only increased the affinity of the PR TRCTE for a PRE by 2.4-fold as compared with the 8.8-fold effect observed with the PR DBD constructs (Table I and compare Figs. 3B and  5B). Conversely, the TR PRCTE chimera bound to both DR4 and TREpal elements with 13-30-fold lower affinities than the TR␤ DBD construct (Table I and compare Figs. 4B and 5C). Unlike the TR␤ DBD that was unaffected by HMGB-1/-2, the TR PRCTE chimera exhibited a 5-to 6-fold increase in apparent affinity for the DR4 and TREpal in response to addition of HMGB-2 ( Fig.  5C and Table I). The apparent binding affinities of the TR PRCTE for DR4 and TREpal increased from 51 ϫ 10 Ϫ9 M and 75 ϫ 10 Ϫ9 M to 9.9 ϫ 10 Ϫ9 M and 13 ϫ 10 Ϫ9 M in the absence and presence of HMGB-2, respectively (Table I). Thus swapping the CTE between the PR and TR DBDs resulted in a nearly quantitative reversal of the DNA binding properties of these two classes of nuclear receptor DBDs with respect to binding affinity and response to HMGB-1/-2. These data suggest that the CTE is responsible for both the differences in intrinsic DNA binding affinity of the PR and TR␤ DBDs for their target DNAs and for response to HMGB-1/-2.

The CTE Is Required for Protein Interaction with HMGB-1/-2 and for Full Stimulatory Effects of HMGB-1/-2 on Recep-
tor Activities-We previously demonstrated a physical interaction of HMGB-1/-2 with PR in the absence of DNA that was not observed with a class II receptor, VDR (18). To examine protein interaction with other nuclear receptors, we performed in vitro pull-down assays with GST-HMGB-1. As shown in Fig. 6A, both full-length PR (A-and B-form) and ER␣ interacted specifically with GST-HMGB-1, as compared with the minimal nonspecific interaction with GST alone. In contrast, no specific interaction was observed between HMGB-1 and full-length RXR␣ or TR␤ (Fig. 6A). To narrow the region of steroid receptors required for interaction with HMGB-1/-2, similar pulldown assays were performed with various baculovirus-expressed domains of PR. PR domain constructs containing both the DBD and CTE bound to GST-HMGB-1/-2, but those that lacked the DBD and CTE did not (data not shown) indicating that the PR DBD and CTE were required for interaction with HMGB-1. To determine whether the DBD alone was sufficient to mediate HMGB-1/-2 interaction and to test the involvement of the CTE, we performed pull-downs with three different PR DBD constructs (Fig. 6B): the PR DBD 670 , which contains the full-length CTE equivalent to the TR␤ CTE; PR DBD 651 , which contains an intermediate length CTE; and PR DBD 642 , in which most of the CTE was truncated. PR DBD 670 and PR DBD 651 bound to GST-HMGB-1 with similar efficiency, while PR DBD 642 exhibited little or no specific interaction (Fig. 6B). As anticipated, the RXR␣ DBD did not interact with HMGB-1 (Fig. 6B). These results indicate that HMGB-1 interaction is mediated by the DBD and requires the CTE.
We next sought to determine whether the CTE region was also important for mediating the stimulatory effect of HMGB-1/-2 on steroid receptor DNA binding. Using quantitative EMSA, we analyzed the effect of HMGB-1/-2 on the DNA binding affinity of the three PR DBD constructs used in pull-down assays (Fig. 6B). As summarized in Table I, PR DBD 670 and PR DBD 651 exhibited similar intrinsic DNA binding affinities in the absence of HMGB-1/-2 and responded similarly to HMGB-1/-2 addition with an ϳ9-fold increase in apparent DNA binding affinity. HMGB-1/-2 also stimulated DNA binding by PR DBD 642 ; however, the increase (4-fold) was substantially reduced in comparison to the other DBDs (Table I) edly, PR DBD 642 had a significantly higher intrinsic DNA binding affinity than PR DBD 670 or PR DBD 651 (Table I), suggesting that the CTE represses PR-DNA binding in the absence of HMGB-1/-2.
To examine whether the CTE is also involved in HMGB-1/-2 enhancement of PR-B transactivation in situ, a chimeric receptor, PRtrPR, was created that consisted of full-length PR-B containing the TR␤ CTE (aa 97-207) in place of the PR CTE (aa 633-670) (Fig. 7A). The ability of HMGB-1/-2 to influence transactivation by the PRtrPR chimera was directly compared with effects on PR-B in transient transfection assays. As shown in Fig. 7B, cotransfection of HMGB-1 increased hormone-dependent transactivation mediated by PRtrPR; however, the magnitude of the effect was significantly reduced as compared with wild-type PR-B. This reduction in response to HMGB-1 was not due to differences in protein expression of the two receptors because the levels of PR-B and PRtrPR were comparable and unaffected by HMGB-1 co-expression as determined by immunoblot (data not shown). We also constructed a chimeric receptor that lacked the N terminus, but retained the DBD-CTE and thus is the minimal PR construct capable of mediating hormone-dependent gene transactivation (DHLBD and DHtrLBD, Fig. 7A). As shown in Fig. 7C, the chimeric DHtrLBD also had a similar reduced response to HMGB-1/-2 in comparison to the wild-type DHLBD. These results together with the reduced activity of HMGB-1/-2 on PR DBD 642 -DNA binding suggest that interaction of HMGB-1/-2 with the CTE is important for enhancement of PR-DNA binding and transactivation.

DISCUSSION
The present study reveals that the CTE of steroid receptors plays a role in DNA binding but acts by a distinct mechanism from class II nuclear receptors. By use of chimeric DBDs and truncation mutants, we demonstrate that the CTE is responsible for the differential ability of HMGB-1/-2 to selectively increase the DNA binding affinity of steroid receptor DBDs and is required for a direct protein interaction between steroid receptors and HMGB-1/-2 that does not occur with class II nuclear receptors. We also show that the CTE is responsible for a higher intrinsic DNA binding affinity of class II DBDs. These results taken together demonstrate that the CTE of steroid receptors is a site required for interaction with HMGB-1/-2 and for maximal enhancement of steroid receptor-DNA binding and transactivation by HMGB-1/-2.
No other study to our knowledge has directly compared the apparent binding affinities of different classes of nuclear receptor DBDs under the same conditions. This comparison demon-FIG. 5. The CTE is responsible for differences in intrinsic DNA binding affinity and response to HMGB-2 between PR and TR␤ DBDs. A, chimeric DBDs used in EMSA. The CTE was swapped between PR and TR␤ DBDs to create chimeric DBDs that contain the core zinc finger DBD of PR fused to the CTE of TR␤ (PR TRCTE ) and the core zinc finger DBD of TR␤ fused to the CTE of PR (TR PRCTE ). Chimeric DBDs were expressed and purified from bacteria using methods applied for purification of wild-type DBDs. B, quantitative DNA binding of PR TRCTE chimeric DBD to PRE. Varying concentrations of purified, recombinant PR TRCTE (0 -580 nM) were incubated with PRE oligonucleotide (0.6 nM) in the absence (open circles) or presence (closed circles) of purified, recombinant HMGB-2 (300 ng). EMSA gels were quantitated by phosphorimaging analysis, and data was plotted as fraction of bound DNA versus DBD concentration. Data are averages of three independent assays (n ϭ 3, Ϯ S.E.). Curve fits were performed as described under "Experimental Procedures." C, quantitative DNA binding of TR PRCTE chimeric DBD to DR4 as described in B.
FIG. 6. Direct interaction of HMGB with steroid but not class II receptors is dependent on the steroid receptor CTE. A, pulldown assay with full-length nuclear receptors and GST-HMGB-1. Equal amounts of bacterial lysate expressing GST or GST-HMGB-1 were immobilized to glutathione-Sepharose beads and incubated with full-length PR-A, PR-B, ER␣, RXR␣, or TR␤. Interacting nuclear receptors were detected by immunoblot with antibodies specific for each receptor. B, GST pull-down assay with nuclear receptor DBDs were performed as described for the full-length receptors. The small arrowhead indicates a cross-reacting contaminant from the GST-HMGB-1 preparation.
strates that the class II receptor DBDs have a substantially higher intrinsic DNA binding affinity than the steroid receptor DBDs, particularly the PR and GR DBDs (Table I), suggesting fundamental differences in the way the DBDs of these two nuclear receptor subclasses interact with DNA. Fusing the TR␤ CTE to the PR core DBD increased the affinity of the PR DBD for PREs 6-fold, whereas fusing the PR CTE to the core TR DBD resulted in a 13-30-fold reduction in the affinity of the TR␤ DBD for its target DNA (Fig. 5 and Table I). These CTE domain swapping experiments indicate that the observed affinity differences between the steroid and class II receptor DBDs are largely attributable to the CTE.
The ER␣ DBD appears to be unique among the steroid receptor DBDs, because it bound to DNA with a higher apparent affinity than the other steroid receptor DBDs analyzed (Table  I). ER␣ recognizes the core DNA hexamer AGGTCA, a sequence bound by the majority of the nuclear receptor superfamily with the exception of the GR subgroup of steroid receptors, including PR, GR, MR, and AR, which all bind to AGAACA. Comparison of the GR and ER␣ DBD-DNA crystal structures demonstrates that ER␣ makes more direct and water-mediated DNA contacts in the major groove than does GR, a possible mechanism for the observed affinity differences. Additionally, the ER␣ CTE (aa 261-264) contains a sequence motif similar to the GRIP-box sequence in the orphan receptor SF-1. The GRIP-box forms an extended structure that interacts in the minor groove flanking HREs and is important for stabil-ity of orphan receptor-DNA interaction. Other steroid receptor CTEs, such as those in PR or ER␤, do not contain a similar GRIP-box-like sequence. If the ER␣ CTE contains a bona fide GRIP-box this could account for the affinity differences observed between the ER␣ DBD and other steroid receptor DBDs. Nonetheless, the ER␣ DBD behaved like other steroid receptors in that HMGB-1/-2 increased its DNA binding affinity 7-fold (Table I). It is of interest to note that the DNA binding affinity of ER␣ DBD in the presence of HMGB-1/-2 is higher than the intrinsic affinities of the class II DBDs (Table I) further illustrating that ER␣ has some unique properties that do not fit with the steroid or class II receptor classifications.
Whereas the steroid receptor DBDs have a low relative DNA binding affinity, addition of HMGB-1/-2 enhanced this affinity 7-9-fold, such that the K d app approached that of the class II receptor DBDs. In contrast, HMGB-2 had no effect on DNA binding by RXR␣ or TR␤ DBDs (Figs. 3 and 4, and Table I). This selective interaction with the steroid receptor DBDs is not dependent on the DNA target element because HMGB-1/-2 enhanced PR DBD binding to both inverted repeat and half-site elements (Fig. 3) (20) but did not affect binding by the TR␤ DBD to either direct or inverted repeat elements (Table I). These results suggest that a feature of steroid receptor DBDs and not the target DNA is the important determinant for the selective influence of HMGB-1/-2. Indeed, we show that the DBD of steroid receptors mediates a protein interaction with HMGB-1/-2 that was not detected with class II receptor DBDs  (Fig. 6). That HMGB-1/-2 interaction with the steroid receptors is mediated through the DBD is not entirely surprising since HMGB-1/-2 interaction with the OCT and HOX transcription factors, is also mediated by the DBD (25,26). A correlation was observed between the ability of HMGB-1/-2 to stimulate receptor activity and to make a direct contact with the receptor. HMGB-1/-2 directly interacted with the steroid receptors (PR and ER) and stimulated both DNA binding and transactivation by these receptors. The class II receptors (TR␤ and RXR␣) did not interact with HMGB-1/-2, nor did they respond functionally to HMGB-1/-2 (18) (Fig. 6). This correlation highlights the importance of protein-protein interactions in HMGB-1/-2 stimulation of steroid receptor activity and is consistent with previous observations that HMGB-1/-2 selectively increased DNA binding and transactivation by full-length steroid receptors but not class II receptors (18,20,26).
HMGB-1/-2 interaction with the steroid receptor DBD was dependent on the CTE, since truncation of the CTE resulted in a loss of detectable protein interaction (Fig. 6). Whether the CTE is a binding site for HMGB-1/-2 or confers a conformation on the DBD that is required for interaction remains to be determined. Truncation of the CTE also resulted in a significant reduction in HMGB-1/-2 enhancement of PR DBD-DNA binding (Table I). Likewise, the chimeric PR construct, which contained the TR␤ CTE in place of its own CTE, also exhibited a substantial reduction in HMGB-1/-2 stimulation of PR-mediated transactivation in cells (Fig. 7). Thus, we also observed a correlation between the requirement of the CTE for physical association with HMGB-1/-2 and the ability of HMGB-1/-2 to stimulate PR DNA binding in vitro and PR-mediated transactivation in situ. However, some residual stimulatory activity of HMGB-1/-2 in the absence of the PR CTE was still observed in DNA binding and transfection assays suggesting that physical interaction with the CTE does not account for all the HMGB-1/-2 effects on PR activity. The mechanism for this residual HMGB-1/-2 activity in the absence of the CTE is not known but may be explained by interaction of HMGB-1/-2 with DNA or a weak interaction between HMGB-1/-2 and the core DBD that is not detectable under our assay conditions. The CTEs of class II and orphan receptors have clearly been shown to play an important role in DNA binding through extension and stabilization of protein-DNA contacts through interactions with DNA in the minor groove. Only a few reports exist to support a similar role in DNA binding of the steroid receptor CTE. Truncations of the ER␣ CTE led to an acceleration of the off-rate of the ER DBD from DNA and an increased sensitivity to the ionic strength of EMSA buffers (47) suggesting a role of the CTE in stability of the ER-DNA complex. Additionally, truncation of the GR and AR CTEs to a length shorter than twelve amino acids reduced the DNA binding affinity (59). In the same study, the CTE and the second zinc finger of AR were also required for AR recognition of a direct repeat androgen response element; a function attributed to the different dimerization mode required to bind to the direct versus inverted repeat (59).
The present results support a model whereby HMGB-1/-2 selectively interact with steroid receptors through a direct protein interaction with the DBD that is dependent on the CTE and that this interaction somehow increases the DNA-binding affinity of the receptor. The mechanism by which HMGB-1/-2 interaction enhances steroid receptor-DNA binding remains to be determined. One possibility is that the HMGB-1/-2 recruited by steroid receptors stabilizes the receptor-DNA complex by extending the protein-DNA interface thus substituting for the CTE of other nuclear receptors. However, the existence of a stable ternary DBD/HMGB/DNA complex is questionable since it is very difficult to capture HMGB-1/-2 in the final high affinity PR-DNA complex suggesting that HMGB-1/-2 interaction is transient (18). 3 An alternative mechanism favored by our data is that HMGB-1/-2 interaction either induces or stabilizes a conformation that enables the CTE to directly interact with DNA. This model suggests that the steroid receptor CTE serves a function similar to that of other nuclear receptor CTEs but requires the additional step of HMGB-1/-2 interaction to participate in DNA binding. The latter mechanism is supported by the observation that truncation of nearly the entire CTE of PR (PR DBD 642 ) resulted in an increase in the intrinsic affinity of the DBD for DNA and a reduction of the stimulatory effect of HMGB-1/-2. This suggests that the PR CTE in the absence of HMGB-1/-2 has a repressive effect on DNA binding and that HMGB-1/-2 relieve this repression. We believe that the CTE of steroid receptors is structurally unstable and that the role of HMGB-1/-2 is to stabilize the CTE.
Allosteric modification of steroid receptors is a common theme in gene regulation by steroid hormones. Steroid binding induces a conformational change in the receptor to promote dissociation of the heat shock proteins and enable co-activator recruitment. Specific DNA can also allosterically modify steroid receptors to allow for promoter-specific gene regulation and potential recruitment of specific coactivator proteins. The present results suggest that interaction between HMGB-1/-2 and the DBD-CTE of steroid receptors provides another dimension to allosteric regulation of steroid receptors by inducing a conformation that results in higher affinity DNA binding.