The human transcription factor IID subunit human TATA-binding protein-associated factor 28 interacts in a ligand-reversible manner with the vitamin D(3) and thyroid hormone receptors.

Using coexpression in COS cells, we have identified novel interactions between the human TATA-binding protein-associated factor 28 (hTAF(II)28) component of transcription factor IID and the ligand binding domains (LBDs) of the nuclear receptors for vitamin D3 (VDR) and thyroid hormone (TRalpha). Interaction between hTAF(II)28 and the VDR and TR LBDs was ligand-reversible, whereas no interactions between hTAF(II)28 and the retinoid X receptors (RXRs) or other receptors were observed. TAF(II)28 interacted with two regions of the VDR, a 40-amino acid region spanning alpha-helices H3-H5 and alpha-helix H8. Interactions were also observed with the H3-H5 region of the TRalpha but not with the equivalent highly related region of the RXRgamma. Fine mapping using RXR derivatives in which single amino acids of the RXRgamma LBD have been replaced with their VDR counterparts shows that the determinants for interaction with hTAF(II)28 are located in alpha-helix H3 and are not identical to those previously identified for interactions with hTAF(II)55. We also describe a mutation in the H3-H5 region of the VDR LBD, which abolishes transactivation, and we show that interaction of hTAF(II)28 with this mutant is no longer ligand-reversible.

of TAF II s are present not only in TFIID but also in the complexes (Refs. 6 -10; for review, see Refs. [11][12][13][14]. TAF II function has been studied genetically in yeast and by transfection experiments in mammalian cells. In yeast, a variable requirement for TAF II s has been found. Temperature sensitive mutations in yeast TAF II 145 (yTAF II 145) result in cell cycle arrest and lethality, but the expression of only a small number of genes is affected (15). In contrast, tight temperature-sensitive mutations in yTAF II 17,yTAF II 25, yTAF II 60, yTAF II 61/68, and the TFIID-specific yTAF II 40 strongly affect the transcription of the majority of yeast genes (16 -21).
An increasing body of results also shows that human TAF II 28 (hTAF II 28), hTAF II 135, and hTAF II 105 can act as specific transcriptional coactivators in mammalian cells. For example, distinct domains of hTAF II 135 interact specifically with Sp1, cAMP response element-binding protein, and E1A, and coexpression of these hTAF II 135 derivatives has a dominant negative effect on the activity of these activators (22)(23)(24)(25), whereas coexpression of hTAF II 135 strongly potentiates transcriptional activation by several nuclear receptors (26). Similarly, hTAF II 105 interacts specifically with the p65 subunit of nuclear factor-B, and TAF II 105 expression potentiates activation by nuclear factor-B in mammalian cells (27).
The viral protein Tax interacts directly with hTAF II 28, and coexpression of hTAF II 28 strongly potentiates activation by Tax (28). Expression of hTAF II 28 also potentiates activation by the ligand-dependent activation function-2 of the nuclear receptors (NRs) for vitamin D 3 (VDR), 9-cis retinoic acid (RXR), and estrogen (29,30). Coexpression of TBP and hTAF II 28 shows that they act synergistically to potentiate activation by the VDR or estrogen receptor. This synergism requires specific amino acids of the hTAF II 28 histone fold domain located in the conserved C-terminal half of the protein and can also be abolished by a mutation in the H1Ј helix of TBP (31,32).
Despite its ability to act as a transcriptional coactivator for several NRs, we previously reported that no ligand-dependent interactions with the RXR could be observed (29). In this study, we show that hTAF II 28 does selectively interact with the VDR and thyroid hormone receptor (TR␣) but does so in a ligandreversible manner. Analysis of hTAF II interaction with VDR deletion mutants shows that hTAF II 28 interacts with a 40amino acid region spanning ␣-helices H3-H5 and containing the NR signature and previously shown also to interact with hTAF II 55 (33). Interaction was also seen with the H3-H5 region of the TR␣ but not with the analogous highly related region of the RXR␥. Substituting single amino acids of the RXR␥ H3-H5 region with their counterparts from the VDR induced interactions with hTAF II 28, but comparison with the results previously obtained with hTAF II 55 indicates that the determinants for the interaction of these TAF II s with this re-gion are not identical. We also show that exchange of amino acids in the H3-H5 region between the VDR and RXR leads to a loss of activity in the case of the VDR and a gain of activity in the case of the RXR.

Construction of Recombinant Plasmids-
The hTAF II expression vectors are as described previously (29,(33)(34)(35). The mouse RXR, mouse retinoic acid receptor, Sp1, and AP-2 expression vectors are also as described (26,29,33,36,37). All the G4-VDR, TR␣, and RXR chimeras were constructed by PCR using the appropriately designed oligonucleotides with restriction sites and cloned into the vector pXJ440 encoding the DNA binding domain of the yeast activator GAL4 (38). All plasmids were verified by automated DNA sequencing. Further details of constructions are available on request.
Transfection of COS Cells and Immunoprecipitations-COS cells were transfected by the calcium phosphate coprecipitation technique, and immunoprecipitations were performed as described previously (33)(34)(35). 48 h after transfection the cells were harvested by three cycles of freeze-thaw in buffer A (50 mM Tris-HCl, pH 7.9, 20% glycerol,. 1 mM dithiothreitol, and 0.1% Nonidet P-40) containing 0,5 M KCl. The expression of the transfected proteins was verified on Western blots. For immunoprecipitations cell extracts were incubated for 1 h at 4°C with 1-2 g of the indicated monoclonal antibodies after which time 50 l of protein G-Sepharose was added, and incubation was continued for another 2 h. After extensive washing the precipitated proteins were detected by immunoblotting with the indicated antibodies using an ECL kit (Amersham Pharmacia Biotech). Where indicated, ligands were added (50 nM all-trans-retinoic acid, 9-cis-retinoic acid, and 3,5,3Јtriiodo-L-thyronine and 100 nM 1,25-dihydroxyvitamin D 3 ) at the same time as the DNA-calcium phosphate coprecipitate. For chloramphenicol acetyltransferase (CAT) assays, 3 g of the 17 m5-TATA-CAT reporter plasmid was cotransfected with 2 g of a ␤-galactosidase reporter as an internal control, along with the indicated concentrations of the G4-RXR␥ expression vectors. After correction for transfection efficiency using ␤-galactosidase assays, CAT assays were performed by standard protocols, and the percentage of acetylated chloramphenicol was determined by quantitative phosporimager analysis on a Fujix BAS 2000 apparatus.

Selective Ligand-reversible Interactions between hTAF II 28 and the Ligand Binding Domains of the VDR and TR-We
have previously reported that although expression of hTAF II 28 strongly potentiates activation by several NR activation function-2s, no ligand-dependent interactions with NRs could be observed that would account for its coactivator activity (29). In the course of these experiments we did, however, observe two hTAF II 28-NR interactions. These interactions were seen when vectors expressing chimeras comprising the VDR or TR␣ ligand binding domains (LBDs) (or full-length VDR) fused to the DNA binding domain of the yeast activator GAL4 (GAL4 (1-147), G4) were cotransfected into COS cells along with vectors expressing native or B10-tagged hTAF II 28 (the expression vectors are schematized in Fig. 1).
A series of deletion mutants (34) was used to determine the region of hTAF II 28 required for interaction with the VDR. G4-VDR coprecipitated with wild-type hTAF II 28 (1-211) and with the C-terminal deletion mutant (1-150) (Fig. 2C, lanes  1-4). In contrast, G4-VDR was not coprecipitated with hTAF II 28 (1-114) (Fig. 2C, lanes 5 and 6). Therefore, a region within or overlapping with amino acids 114 -150, which encode the N-terminal portion of the hTAF II 28 histone fold, is required for interaction with the VDR.
The ability of hTAF II 28 to interact with other NRs was investigated. As observed with the VDR, hTAF II 28 could be coprecipitated by the anti-G4 mAbs in the presence of coexpressed G4-TR␣(DE), and conversely G4-TR␣(DE) was coprecipitated by the anti-hTAF II 28 antibodies (Fig. 3A, lanes 1, 2,  and 5). A similar result was observed with the C-terminal deletion mutant hTAF II 28 (1-179) (Fig. 3A, lanes 7 and 8). Strikingly, as observed with the VDR, this interaction is destabilised in the presence of ligand (Fig. 3A, compare lanes 2, 4, 5,  6, 8, and 10). Thus, the presence of ligand dramatically reduces is masked by the IgG(H), lanes 5 and 6 have been revealed with peroxidase-conjugated secondary antibody directed against the light chain (IgG(L)). TAF II 28 can be seen between the two signals generated in this region of the blot using this conjugated antibody. In subsequent figures either the light chain or the heavy chain of the immunoprecipitating antibody is indicated, depending on which peroxidase-conjugated secondary antibody was used. B and C, no interactions between hTAF II 28 and the retinoic acid receptor RXR Sp1 or AP-2 activation domains. D, no interactions between the VDR LBD and hTAF II 18 and hTAF II 20. The layout is as described for the other figures.  1-6, and B, lanes 1 and 2) are control immunoprecipitations showing that the subsequent coprecipitation is hTAF II 28-dependent. the interactions between hTAF II 28 and the TR␣ and the VDR LBDs in transfected COS cells.
In contrast to the above, no interactions were observed with G4-RXR␣(DE) or G4-retinoic acid receptor-␣(DEF) in either the absence or presence of ligand (see Fig. 3B, lanes 1-6; Ref. 29; data not shown) or with several activators that do not belong to the NR superfamily (see Fig. 3C, G4-Sp1 and G4-AP2).
As we have previously observed interactions of hTAF II 55 with the VDR and TR (33), we also verified the specificity of these interactions with respect to the TAF II s. For this, the VDR was cotransfected with hTAF II 20 or hTAF II 18. In contrast to hTAF II 28 and hTAF II 55, no significant interactions were observed between the VDR and hTAF II 20 or hTAF II 18 (Fig. 3D).
The region of the VDR required for interaction with hTAF II 28 was next determined using the previously described series of G4-VDR LBD deletion mutants (33). G4-VDR DE (90 -195) was not coprecipitated with B10-hTAF II 28 (data not shown). In contrast, both the N-terminal half of the E region present in G4-VDR E (196 -311) (Fig. 4A, lanes 1-3 and 7-9) and the C-terminal half of the E region in G4-VDR E (312-427) (Fig. 4A, lanes 4 -6 and 10 -12) were coprecipitated with hTAF II 28. Therefore, hTAF II 28 interacts with two distinct sites in the VDR E domain but not with the VDR D domain.
Selective Interaction of hTAF II 28 with the H3-H5 NR Signature-containing Regions of the VDR and TR␣-The H3-H5 region of the NRs contains the NR signature, which is a group of well conserved amino acids involved in intramolecular interactions required for stabilization of the LBD fold (41,42). As a consequence the H3-H5 region is one the most highly conserved regions within the NR LBDs. Based on the above results, we next investigated hTAF II 28 interactions with the related H3-H5 regions of the chicken TR␣ and the mouse RXR␥. G4 chimeras containing these H3-H5 regions (G4-TR E (220 -260) and G4-RXR␥ E (273-313)) were coexpressed along with hTAF II 28. G4-TR␣ E (220 -260) was specifically precipitated along with hTAF II 28 (Fig. 5A, lanes 4 -6). In contrast, no coprecipitation of G4-RXR␥ E (273-313) was observed (Fig.  5B). The selective interaction of hTAF II 28 with the H3-H5 region of the VDR and TR␣, but not the RXR␥, therefore mimics the specificity seen using the complete LBDs of these NRs.
Determinants in the ␣3 Helix of the VDR Are Required for Interactions with hTAF II 28 -The above results show that hTAF II 28 interacts with the H3-H5 regions of the VDR and TR␣ but not the RXR␥. We have previously shown that hTAF II 55-RXR␥ LBD interactions can be induced by exchanging solvent-exposed amino acids of the RXR␥ H3-H5 region by their VDR equivalents. We next determined which if any of these mutations (Fig. 7A) could induce hTAF II 28-RXR interactions.
In a converse experiment we simultaneously introduced the above three RXR␥ amino acids into the equivalent positions of the VDR LBD to generate a triply mutated G4-VDR(DE) m1 (Q239F,K240T,S256L; Fig. 7A). This triple mutation did not abolish interaction with hTAF II 28, as expected from the fact that hTAF II 28 interacts also with the VDR H8 region. Interestingly, however, the interaction with the mutated VDR LBD was no longer ligand-reversible (Fig. 7B, lanes 4 and 5 compared with 2 and 3).
We next determined whether the triple mutation had affected the ability of the VDR to activate transcription. To do this, increasing quantities of vectors expressing the wild-type or mutated G4-VDR(DE) were cotransfected along with a G4responsive CAT reporter (see "Materials and Methods" and Refs. 26,29). The wild-type VDR LBD strongly activated transcription from this promoter (Fig. 8A, lanes 2-4), whereas the mutated VDR was unable to activate transcription (Fig. 8A,  lanes 5-7). As described previously (29,33,43), little or no activation of transcription from this promoter is observed with the wild-type G4-RXR␥(DE), whereas the mutant bearing the three VDR amino acids strongly activates transcription (Fig.  8A, lanes 8 -10 and 11-13, respectively). Equivalent expression of each of the chimeras was detected by immunoblotting (Fig.  8B). Consequently, replacing the three amino acids of the VDR with their RXR␥ equivalents leads to a loss of function, whereas in the converse experiment there is a gain of function. These three amino acids therefore play a critical role in transactivation.

DISCUSSION
Ligand-reversible hTAF II 28-NR Interactions-We describe here novel interactions between the LBDs of the VDR and the TR␣ and the hTAF II 28 component of transcription factor TFIID. After coexpression in COS cells, hTAF II 28 could be specifically coprecipitated with the LBDs of the VDR and TR␣ in the absence of the appropriate ligands, whereas coprecipitation is dramatically reduced in the presence of ligand. The selectivity of the interactions is shown by the observation that under the same conditions no interactions between hTAF II 28 and other NRs were observed, nor did hTAF II 20 or hTAF II 18 interact with the VDR or TR LBDs.
Interaction with the VDR requires the evolutionary conserved C-terminal domain of hTAF II 28, which contains the histone fold motif (31). Interaction is seen with a deletion mutant (1-150) in which the ␣2and ␣3-helices of the histone fold are deleted, but not with , in which the ␣Nand ␣1-helices are deleted. Therefore, interaction with the NRs does not require the integrity of the histone fold domain but does require determinants in the ␣N-␣1 region. In support of this, interaction with the VDR is abolished by mutation of exposed residues in the ␣1-helix, further suggesting that this helix plays an important role in the interaction. 2 hTAF II 28 interacts with two independent sites in the VDR LBD. We have characterized in detail the interaction of hTAF II 28 with one of these sites located between amino acids 234 and 274 spanning ␣-helices H3-H5 and containing the NR signature. The equivalent region of the chicken TR␣ also interacts with hTAF II 28, whereas no interaction is seen with the equivalent region of the RXRs. Two amino acids in helix H3 are critically involved in the ability of hTAF II 28 to discriminate between the VDR and RXR␥ LBDs. In the context of the full RXR␥ DE region, replacement of Phe 278 or Thr 279 with Gln or Lys, respectively, of the VDR induces a ligand-reversible interaction with hTAF II 28. No significant interaction, however, was seen with mutant L295S located in helix H4.
We have previously shown that position Leu 295 is an important determinant for interaction with hTAF II 55, which also interacts with the H3-H5 region of the VDR and TR. The differential interaction of hTAF II 55 and hTAF II 28 with the L295S mutation indicates that, although both TAF II s interact with the H3-H5 region, the determinants for their respective interactions are not identical. TAF II 28 may interact only with the H3 helix, whereas hTAF II 55 requires determinants in both the H3 and H4 helices.
As discussed previously (33), the hTAF II 28 and hTAF II 55 interaction sites are close to, but not identical to, with those required for interaction with the LXXLL motif in several transcriptional intermediary factors (44 -46). The close proximity of these two interaction sites may nevertheless explain the ligand reversibility observed with the hTAF II 28 interaction, because there may be steric hindrance between hTAF II 28 and the TIF(s) whose association with the LBD is ligand-induced (see below). It is not clear, however, how the ligand-induced conformational change in the LBD, which brings helix H12 in proximity to H3 and H4 to create the hydrophobic TIF interaction surface, would affect the interaction of hTAF II 28 with its second interaction site in helix H8.
Possible Roles of the VDR/TR-hTAF II 28 Interaction-Our previous results indicated that there is a correlation between transcriptional activation potential and interaction with hTAF II 55. The wild-type RXR LBD, which has negligible transactivation potential on a minimal promoter, does not interact with hTAF II 55. In contrast, the single amino acid changes in the RXR␥ LBD induce weak interactions with hTAF II 55 and an increase in transactivation, whereas the F278Q,L295S double mutation induces a strong interaction with hTAF II 55, and this derivative activates transcription to a level equivalent to that seen with the VDR itself (33).
On the other hand, the NR interaction with hTAF II 28 does not correlate with activation, since it is ligand-reversible. We therefore do not suggest any obvious role for the NR-hTAF II 28 interaction in the transcriptional activation described here. Ligand-reversible interactions with NRs are not observed with transcriptional coactivators but are characteristic of those seen with the corepressors N-Cor and SMRT (47,48). At present we have no direct evidence that NR-hTAF II 28 interactions are involved in the transcriptional repression seen with the unliganded TR or VDR, and the RXR mutants that interact with hTAF II 28 do not gain repressor properties, at least not on the promoters that we have tested.
Transcriptional activation is thought to involve both the recruitment of histone acetylase complexes and interactions with the basal transcription apparatus. It is therefore tempting to speculate that, in addition to recruitment of histone deacteylase complexes, interactions such as those described here between the VDR and TR and the basal transcription apparatus may contribute to transcriptional repression. Indeed, previous in vitro experiments have suggested that the unliganded TR may target TBP itself and block preinitiation complex formation leading to transcriptional repression (49,50). It has also been shown that TBP/TFIID may be a target of other repressors such as the homeodomain protein even-skipped (51,52). Furthermore, ligand-reversible interactions between the TR and TFIIB have also been suggested to contribute to transcriptional silencing (53) Our results identify amino acids in helices H3 and H4 critical for transcriptional activation by the VDR. These amino acids, which all lie on the surface of the receptor and are not involved in ligand binding (54), cannot be replaced by their RXR homologues, whereas their incorporation into the RXR yields a potent transcriptional activator. Because the interaction of hTAF II 28 with the VDR m1 mutant, which does not activate transcription, is ligand-independent rather than ligand-reversible, this would suggest that the introduction of the RXR amino acids has abrogated the ability of the VDR to interact with coactivators (one of which may be TAF II 55), which normally displace hTAF II 28 in the presence of ligand. Indeed, a similar constitutive interaction is seen when the VDR activation function-2-AD core in helix H12 is deleted. 3 In other receptors deletion of the activation function-2-AD core does not impair ligand binding but does abolish interaction with TIFs. TAF II 28 interacts with the mutated RXR LBDs in a ligandreversible manner. Thus, in this context, which is the converse of that observed with VDR m1, it is likely that the introduction of the VDR amino acids allows interactions not only with hTAF II 28 but also with coactivators that displace hTAF II 28 in the presence of ligand. Together, these results show that the H3-H5 region contains receptor-specific determinants for transactivation.