Critical Structural Elements and Multitarget Protein Interactions of the Transcriptional Activator AF-1 of Hepatocyte Nuclear Factor 4*

The nuclear receptor hepatocyte nuclear factor 4 (HNF-4) is an important regulator of several genes involved in diverse metabolic and developmental pathways. Mutations in the HNF-4Agene are responsible for the maturity-onset diabetes of the young type 1. Recently, we showed that the 24 N-terminal residues of HNF-4 function as an acidic transcriptional activator, termed AF-1 (Hadzopoulou-Cladaras, M., Kistanova, E., Evagelopoulou, C., Zeng, S., Cladaras C., and Ladias, J. A. A. (1997) J. Biol. Chem. 272, 539–550). To identify the critical residues for this activator, we performed an extensive genetic analysis using site-directed mutagenesis. We showed that the aromatic and bulky hydrophobic residues Tyr6, Tyr14, Phe19, Lys10, and Lys17 are essential for AF-1 function. To a lesser degree, five acidic residues are also important for optimal activity. Positional changes of Tyr6 and Tyr14 reduced AF-1 activity, underscoring the importance of primary structure for this activator. Our analysis also indicated that AF-1 is bipartite, consisting of two modules that synergize to activate transcription. More important, AF-1 shares common structural motifs and molecular targets with the activators of the tumor suppressor protein p53 and NF-κB-p65, suggesting similar mechanisms of action. Remarkably, AF-1 interacted specifically with multiple transcriptional targets, including the TATA-binding protein; the TATA-binding protein-associated factors TAFII31 and TAFII80; transcription factor IIB; transcription factor IIH-p62; and the coactivators cAMP-responsive element-binding protein-binding protein, ADA2, and PC4. The interaction of AF-1 with proteins that regulate distinct steps of transcription may provide a mechanism for synergistic activation of gene expression by AF-1.

Transcription initiation at RNA polymerase II promoters involves the assembly of a basal transcription complex containing RNA polymerase II and general transcription factors (GTFs), 1 including TFIIA, -B, -D, -E, -F, and -H (1). Stimulation of transcription is directed by sequence-specific DNA-binding proteins, termed activators. Typically, activators are modular, containing a DNA-binding domain (DBD) and one or more activation domains that increase the rate of transcription through interactions with other proteins. It is believed that activators recruit chromatin-remodeling proteins and GTFs to the promoter, relieving the repressive effect of chromatin on transcription and affecting the initiation, promoter clearance, and elongation of transcription (1)(2)(3). Activation domains are classified according to their predominant amino acid composition into proline-rich, glutamine-rich, serine/threonine-rich, and acidic activators. However, the structural requirements and mechanisms of function of these activators remain poorly understood.
An emerging theme from mutagenesis studies of acidic activation domains, including those of VP16 (15,16), p53 (17), NF-B-p65 (18), and the glucocorticoid receptor (GR) (19), is that aromatic and hydrophobic residues are essential for activity. In contrast, negative charge, although important for function, is not sufficient. Furthermore, several activation domains, including those of VP16 and NF-B-p65, consist of short modules that act synergistically to activate transcription (16,18). From a structural perspective, however, the molecular basis of acidic activator action remains obscure (4,6). Several contradictory models proposed that these activators function as amphipathic ␣-helices (20), ␤-sheets (21), or unstructured domains (22). However, recent studies support the ␣-helical model. First, the crystal structure of the p53 activator bound to the MDM2 oncoprotein revealed that this activator forms an amphipathic ␣-helix (23). The prominent feature of the p53 interaction interface is a set of three aromatic/hydrophobic residues (Phe 19 , Trp 23 , and Leu 26 ) that insert into a hydrophobic cleft of MDM2. Since the same amino acids are critical for transactivation (17), it was suggested that this domain of p53 may activate transcription also as an ␣-helix, with its hydrophobic residues contacting GTFs (23). Second, nuclear magnetic resonance studies showed that the VP16 activator undergoes an induced transition from random coil to ␣-helix upon interaction with TAF II 31 (24). Thus, an induced amphipathic ␣-helix may be a common structural feature of acidic activators. Additional structural and functional studies of acidic activators are essential for identification of the specificity determinants in these domains and elucidation of their mechanisms of action.
We recently reported that the 24 N-terminal residues of the nuclear receptor HNF-4 constitute an autonomous acidic activator, termed AF-1 (activation function-1) (25). Deletion of AF-1 results in 40% reduction of the HNF-4-mediated activation (25), indicating that this activation domain plays a central role in the function of this nuclear receptor. HNF-4 is an important regulator of several genes involved in diverse metabolic pathways, including the genes for apolipoproteins A-II, B, and C-III (26) and human immunodeficiency virus-1 (27). Moreover, mutations in the HNF-4A gene are responsible for the maturity-onset diabetes of the young type 1, an autosomal dominant early-onset form of non-insulin-dependent diabetes mellitus (28). Therefore, investigations into the mode of action of the HNF-4 activation domains are important for our understanding of the basic mechanisms of gene expression by this nuclear receptor and the pathogenesis of several metabolic diseases, including diabetes.
In this study, we have performed a systematic structurefunction analysis of HNF-4 AF-1 to identify the critical amino acids and molecular targets that mediate its function. We show that aromatic and bulky hydrophobic residues are essential for AF-1 transactivation, whereas acidic residues are not sufficient for activity. We also demonstrate that AF-1 shares common structural motifs and molecular targets with the activation domains of p53, NF-B-p65, and VP16, implying that these activators may function through common mechanisms. Remarkably, AF-1 interacts with multiple proteins that act at distinct steps during transcription, providing a possible mechanism for the functional synergy exhibited by this activator in vivo.

EXPERIMENTAL PROCEDURES
Plasmids-The reporter 5X-Gal-ML-Luc containing five GAL4 DNAbinding sites in front of the adenovirus major late promoter and the luciferase gene (29) and constructs expressing human TBP and TFIIB as glutathione S-transferase (GST) fusions were provided by D. Reinberg (University of Medicine and Dentistry of New Jersey, Piscataway, NJ). Plasmids expressing GST fusions of human TAF II 31 (also known as TAF II 32) (9) and human TAF II 80 (also known as TAF II 70␣) (10,30) were provided by R. Tjian (University of California, Berkeley, CA). The cDNA encoding the p62 subunit of human TFIIH (6) was provided by J. Greenblatt (University of Toronto) and was subcloned into pGEX-2T (Amersham Pharmacia Biotech) using PCR. Plasmids expressing N-terminal, middle, and C-terminal portions of murine CBP fused to GST (11) were provided by T. Collins (Brigham and Women's Hospital, Boston). Construct pRc/RSV-mCB-P.HA expressing hemagglutinin (HA)-tagged murine CBP (31) was obtained from R. Goodman (Oregon Health Science University, Portland, OR). Plasmid pcDNA3-hADA2.HA expressing HA-tagged human ADA2 and a construct expressing GST-hADA2 (32) were provided by S. Berger (Wistar Institute, Philadelphia). The construct expressing the GST fusion of human PC4 (14) was obtained from R. Roeder (Rockefeller University, New York). To create the reporter apoC-III-870.Luc, the human apoC-III promoter (nucleotides Ϫ870 to ϩ24) was excised as an XbaI/XhoI DNA fragment from construct apoC-III-870.CAT (26) and ligated with a KpnI/XbaI linker into the KpnI/XhoI sites of vector pGL3-Basic (Promega), upstream of the luciferase gene.

S]methionine (NEN Life Science Products).
GST Protein-Protein Interaction Assays-GST fusion proteins were expressed in BL21(DE3) cells (Novagen) and purified by standard procedures onto glutathione-Sepharose 4B beads. GST fusion proteins were diluted to give equivalent amounts of bound protein/ml of beads, as monitored by SDS-PAGE. The GST fusion beads (10 l) were incubated with 5 l of in vitro translated proteins and 385 l of binding buffer (0.5% nonfat dry milk, 0.1% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 150 -300 mM NaCl depending on the fusion protein) for 2 h at 4°C with constant mixing. The beads were washed four times at 4°C with 1 ml of binding buffer without nonfat milk, and proteins were resolved by SDS-PAGE, followed by autoradiography.

RESULTS AND DISCUSSION
The Aromatic Residues Tyr 6 , Tyr 14 , and Phe 19 Are Essential for AF-1 Activity-To identify the critical residues of HNF-4 AF-1, we performed an extensive mutagenesis of this activator in which all amino acids were mutated at least once. The resulting missense mutants were fused to GAL4 DBD-(1-147), and their potential to transactivate the reporter 5X-Gal-ML-Luc was tested in transfection assays.
We initially tested the importance of aromaticity for AF-1 activity. Substitution of aspartates for Tyr 6 , Tyr 14 , and Phe 19 reduced the activities of the resulting mutants, Y6D, Y14D, and F19D, respectively, to ϳ10% of the wild-type AF-1 level (Fig. 1A). Western blot analysis showed that the GAL4 fusion proteins were stable and expressed at comparable levels ( Fig.  1B) (data not shown). Thus, despite the overall increase in negative charge in mutants Y6D, Y14D, and F19D, their activities were severely compromised. Replacement of these aromatic residues with leucines in mutants Y6L, Y14L, and F19L also reduced the activity to 35-50% of the wild-type level (Fig.  1A). The inability of negatively charged or bulky hydrophobic residues to substitute for Tyr 6 , Tyr 14 , and Phe 19 shows that aromaticity at these positions is essential for AF-1 function. More important, the aromatic residues tyrosine and phenylalanine were not interchangeable, as mutants Y6F, Y14F, and F19Y had reduced activity, demonstrating that indiscriminate aromaticity at these positions is not sufficient for full activity of AF-1. In contrast, mutants Y6W, Y14W, and F19W exhibited 1.1-, 2.2-, and 1.7-fold higher activity than the wild type, respectively (Fig. 1A), indicating that tryptophan substitution has a position-dependent beneficial effect on AF-1 activity.
The Primary Structure of AF-1 Is Critical for Function-The significance of primary structure for AF-1 function was tested using mutants Y6S/S7Y and A13Y/Y14A, in which Tyr 6 and Tyr 14 and the adjacent Ser 7 and Ala 13 , respectively, were switched. The activities of Y6S/S7Y and A13Y/Y14A were reduced to 20 -30% of wild-type AF-1 activity (Fig. 1A), which is difficult to account for by a model of an unstructured activator. This analysis proves that a defined primary structure is essential for AF-1 optimal activity and that amino acid composition is not sufficient. The positional effects of particular aromatic side chains on AF-1 activity may reflect variations in the binding affinities between AF-1 and its molecular target(s).
Bulky Hydrophobic Amino Acids Are Crucial for AF-1 Activity-To test the contribution of hydrophobicity to AF-1 activity, we replaced the hydrophobic residues with aspartates. In mutants M1D, A4D, and A9D, an increase in acidity resulted in either no change or a small increase in activity (Fig. 1A). In contrast, increasing the net negative charge in mutants M3D, A8D, L10D, A13D, L17D, V22D, and V24D had detrimental rather than beneficial effects on activity, indicating that hydrophobicity at these positions is more important than acidity. The most severe effects were seen in mutants L17D, L10D, and V22D, whose activities were reduced to 10, 18, and 28% of AF-1 activity, respectively (Fig. 1A). We conclude that acidity is preferred to hydrophobicity at positions 1, 4, and 9, whereas bulky hydrophobic residues are critical for AF-1 function, especially at positions 10 and 17.
The Importance of Acidic and Other Residues-Replacement of Glu 18 and Glu 20 with aspartates in mutant E18D, E20D had no significant effect on activity (Fig. 1A), suggesting that acidity at these positions is sufficient for function. In this respect, AF-1 differs from the VP16 and NF-B-p65 activators, where glutamates are poor substitutes for aspartates (15,18). Removal of a single negative charge in D2N, D5N, E18N/E20D, and E18D/E20N reduced the activity to 40 -50% of AF-1 activity (Fig. 1A). Interestingly, removal of a single negative charge in D11N had a more detrimental effect on activity (20% of wild-type activity), indicating that net negative charge is not the sole determinant of activity, but that the position of the acidic residues is also important. Removal of increasing numbers of negative charges progressively decreased the activity to 10% of the wild-type level (Fig. 1A), demonstrating the importance of acidity for AF-1 function.
The contribution of the remaining residues to AF-1 activity was assessed using alanine or aspartate substitutions. Replacement of Ser 7 with alanine reduced the activity to 80% of the wild-type level, whereas aspartate substitution had a beneficial effect, indicating that acidity is preferred at position 7. Interestingly, proline at position 12 proved important for AF-1 activity, as substitution of either alanine or aspartate for Pro 12 reduced the activity (Fig. 1A). In many proteins, a proline inside an ␣-helix bends the helix (33). Since AF-1 is predicted to adopt an ␣-helical structure (25), Pro 12 may introduce a local distortion in the geometry of the helix that is important for function. Mutations of Thr 15 , Thr 16 , and Asn 21 also resulted in significant reduction in activity, indicating that these side chains are required for optimal AF-1 function. In contrast, substitution of alanine or aspartate for Gln 23 led to increased activity, suggesting that Gln 23 is not critical for function.
The Role of Critical AF-1 Residues in the Activity of Fulllength HNF-4 -To determine whether the critical residues identified in the GAL4 fusion experiments are also important for activity of the full-length HNF-4 protein, we introduced single and multiple amino acid substitutions in the AF-1 region of the HNF-4A protein (25). Representative mutations were selected from the GAL4 fusion experiments based on the sever- ity of their effects on AF-1 activity and the type of mutated residues (aromatic: Y6D, Y14D, Y14W, and F19D; hydrophobic: L10D and L17D; and acidic : D2N,D5N,D11N). The resulting HNF-4A mutants were cloned into the eukaryotic vector pcDNA3 and tested for their ability to activate the apoC-III-870.Luc reporter, which contains the enhancer and promoter of the human apolipoprotein C-III gene driving the expression of the luciferase gene. Overall, the phenotypes obtained with the AF-1 mutations in the context of HNF-4A were very similar to those observed with the GAL4 fusion proteins (Fig. 2A). The activities of HNF4-Y6D, HNF4-L10D, HNF4-Y14D, HNF4-L17D, and HNF-F19D were reduced to ϳ34 -52% of wild-type activity ( Fig. 2A). This reduction is consistent with abrogation of AF-1 function because deletion of AF-1 reduces HNF-4 activity by ϳ40% (25). However, the triple substitution D2N,D5N,D11N decreased HNF-4A activity by only 20% (Fig.  2A), whereas it reduced the activity of the corresponding GAL4 fusion protein by 90% (Fig. 1A), suggesting that these acidic residues play a less important role in the intact HNF-4A protein. More important, the Y14W mutation increased the activity of HNF-4A, indicating that tryptophan at position 14 of HNF-4A generates a more potent activator. Taken together, these data validate the results obtained from the GAL4 fusion experiments and confirm the importance of the aromatic and hydrophobic residues for AF-1 function in the context of the intact HNF-4A protein.
Since HNF-4A binds to its cognate DNA element as a dimer and the AF-1 region is not involved in dimerization (25), we tested whether HNF-4A proteins harboring drastic mutations in AF-1 can function as dominant-negative mutants. Cotransfection of a constant amount of HNF-4A with the reporter apoC-III-870.Luc and increasing amounts of HNF4-F19D into HepG2 cells resulted in a dose-dependent decrease of the transactivation obtained by HNF-4A (Fig. 2B). Similar results were obtained with mutants HNF4-Y6D, HNF4-L10D, and HNF4-Y14D (data not shown). These mutants exhibited a weak dominant phenotype as compared with HNF4- (⌬361-465), a potent dominant-negative regulator of HNF-4A (Fig. 2B) (25). The weak dominant phenotype of these mutants is not unexpected since HNF-4 contains a second activator in its ligand-binding domain, and deletion of AF-1 reduces HNF-4 activity by only 40% (25).
It is remarkable that mutants Y14W, F19W, and Q23D are stronger activators than wild-type AF-1 (Figs. 1A and 2A). The physiological significance of why HNF-4 has not adopted these superactivating sequence variations is currently unknown. An intriguing possibility is that AF-1 is still evolving and may become a stronger activator in the future. Interestingly, the rat and human HNF-4 proteins have identical AF-1 domains (25,34), whereas the Drosophila HNF-4 homolog lacks an AF-1-like region (35), supporting the hypothesis that AF-1 is a recently evolved feature of HNF-4, acquired after the arthropod/vertebrate divergence. Alternatively, a more potent AF-1 might be deleterious for the homeostasis of metabolic pathways involving proteins whose genes are regulated by HNF-4. It is also possible that the primary structure of AF-1 represents an evolutionary compromise between creating a strong transactivator (but not the strongest possible) and defining the optimum repertoire of proteins that interact with AF-1. Notwithstanding the reasons why HNF-4 has not adopted these up-regulatory mutations, the identification of these variants may be exploited for the design of short potent activators with applications in gene regulation and gene therapy strategies.
A Tetrad of Aromatic and Hydrophobic Residues Is a Common Structural Feature of Acidic Activators-Sequence comparison of AF-1 with other acidic activation domains revealed that region 1-12 of AF-1 (module A) exhibited significant similarity to region 529 -540 of NF-B-p65, but not with other activators (Fig. 3). In contrast, comparison of AF-1 region 13-24 (module B) with other activation domains revealed a striking conservation of a tetrad of aromatic and bulky hydrophobic residues. The first, third, and fourth of these residues correspond to Phe 19 , Trp 23 , and Leu 26 of p53, respectively, which form the main interface interacting with MDM2. Remarkably, the second hydrophobic residue of this tetrad corresponds to Leu 22 of p53, which also participates in the p53 interaction with MDM2 (23). The critical role of this hydrophobic tetrad in activator function has been documented for HNF-4 AF-1 (Fig. 1A) as well as the activators of VP16 (15,16), p53 (17), NF-B-p65 (18), GR (19), androgen receptor (36), and retinoic acid receptor ␤2 (37). Hence, this hydrophobic tetrad may be a general recognition motif of this class of acidic acti- FIG. 3. A tetrad of hydrophobic and aromatic residues is conserved in many acidic activators. Shown is the manual sequence alignment of the activators of rat HNF-4 AF-1 (residues 1-24), human NF-B-p65 (residues 529 -551) (18), human p53 (residues 6 -28) (17), rat androgen receptor (AR) (residues 148 -170) (36), VP16 (residues 467-487) (15), human GR (residues 187-199) (19), and murine retinoic acid receptor ␤2 (RAR␤2) (residues 1-12) (37). The superactivating mutants Y14W and F19W of HNF-4 AF-1 and E221F of human GR (residues 217-230) (19) are also aligned. Gaps (dashes) were introduced to maximize the alignment of the hydrophobic and aromatic residues (enclosed in shaded boxes). Asterisks denote the first, third, and fourth residues of the tetrad, corresponding to Phe 19 , Trp 23 , and Leu 26 of p53, respectively. Open boxes indicate similar regions between HNF-4 AF-1 and NF-B-p65. Modules A and B are indicated at the top. vators for basal transcription factors, which might form an interaction interface similar to that of the p53⅐MDM2 complex. This is supported by the finding that the VP16 activator binds to TAF II 31 as an ␣-helix, with Phe 479 and Leu 483 on one face of the helix making nonpolar contacts with TAF II 31 (24). The space variation between the four conserved hydrophobic residues could be accommodated by interconverting between ␣-, 3 10 -and possibly -helical forms, thus adjusting the helix phase and maintaining its amphipathic character. Notably, certain up-regulatory mutations introduce large aromatic side chains at the positions of the hydrophobic tetrad. For example, replacement of the phenolic rings of Tyr 14 and Phe 19 with the indole ring of tryptophan in mutants Y14W and F19W, respectively, resulted in higher activity (Fig. 1A). Likewise, replacement of Glu 221 with phenylalanine in human GR (mutant GR E221F in Fig. 3) led to increased activity (19). The higher activities of these mutants could be explained by the formation of stereochemical interfaces with better complementarities and higher affinities for their transcriptional targets.
The identification of a conserved hydrophobic tetrad in many acidic activators that otherwise show little sequence similarity suggests that this tetrad may be a common structural framework around which activator-specific features have been incorporated. Consequently, different activators can interact both with common and activator-specific molecular targets. Consistent with this hypothesis is our finding that HNF-4 AF-1 interacts with the N-terminal domain of MDM2 with much lower affinity than does the p53 activator (data not shown).
Modules A and B of AF-1 Function Synergistically to Activate Transcription-The sequence comparison in Fig. 3 suggested that AF-1 might consist of two discrete modules A and B. To determine the contribution of each module to AF-1 activity, we reevaluated the mutagenesis data of AF-1 described in Fig. 1A. In this analysis, mutant Y6D is considered to have a destroyed module A and an intact module B. The activity of this mutant is only 10% of AF-1 activity (Fig. 1 A), which is attributed to module B. Similarly, the activity of mutant F19D is 10% of wild-type activity ( Fig. 1 A), which is attributed to module A. Therefore, the activity of a single copy of either module A or B is 10% or less of AF-1 activity. Since the activity of AF-1 is larger than additive of the activities of modules A and B, we conclude that modules A and B synergize to activate transcription.
To determine whether modules A and B could substitute for each other, we fused two copies of modules A or B in tandem to GAL4 DBD-(1-94) and tested their activities in cotransfections. Interestingly, GAL-2A and GAL-2B activated the reporter to 40 and 25% of AF-1, respectively, indicating that both modules A and B are necessary for full activity of AF-1 (Fig. 4). Taken together, these results demonstrate that modules A and B do not have equivalent functions, but act synergistically to activate transcription, possibly through interactions with different transcription factors.
AF-1 Interacts with Multiple Components of the Transcrip- tion Machinery: Implications for Transactivation-The finding that AF-1 shares similar structural features with the activators of p53, NF-B-p65, VP16, and GR (Fig. 3) provided a compelling reason to examine whether AF-1 also shares common transcriptional targets with these proteins. For these experiments, we used GTFs and coactivators produced in bacteria as GST fusions and in vitro translated GAL-2(AF-1) because of its high activity (Fig. 4). Mutant GAL-2(Y6D/F19D) was used as a negative control because its activity is severely compromised (Fig. 4). The GAL-2A and GAL-2B proteins were also tested for their abilities to interact with the AF-1 targets. The amount and integrity of in vitro translated proteins were monitored by SDS-PAGE and autoradiography (Fig. 5A). None of these proteins bound to GST (Fig. 5B). In contrast, GAL-2(AF-1) interacted in vitro with TBP, TFIIB, TAF II 31, TAF II 80, TFIIH-p62, the N-and C-terminal regions of CBP, ADA2, and PC4 fused to GST (Fig. 5, C-K, lanes 1), whereas it failed to interact with the middle region of CBP (residues 1069 -1452) (data not shown). These interactions were specific since GAL-2(Y6D/F19D) did not bind efficiently to these proteins (Fig. 5, C-K, lanes 2). Furthermore, the interactions between AF-1 and CBP or ADA2 in intact cells were investigated by cotransfecting plasmids expressing the GAL4 DBD-(1-94), GAL-2(AF-1), or GAL-2(Y6D/F19D) protein with vectors expressing HAtagged CBP or ADA2 into COS-1 cells. An anti-HA antibody immunoprecipitated the CBP and ADA2 proteins with GAL-2(AF-1), but not with GAL DBD-(1-94) or GAL-2(Y6D/F19D) (Fig. 5L), indicating that AF-1 interacted specifically with CBP and ADA2 in vivo. Taken together, these data indicate that AF-1 interacts specifically with TBP, TFIIB, TAF II 31, TAF II 80, TFIIH-p62, CBP, ADA2, and PC4. Remarkably, Tyr 6 and/or Phe 19 of AF-1 appears to mediate the interactions with all these structurally diverse proteins since mutant GAL-2(Y6D/ F19D) did not bind to any of these factors. Although the functional significance of these interactions remains to be established, these results suggest that these proteins may be transcriptional targets of AF-1.
Strikingly, modules A and B exhibited different properties in the protein interaction assays. GAL-2A did not interact efficiently with any of the proteins tested, whereas GAL-2B bound strongly (Fig. 5, C-K, lanes 3 and 4). These differences provide further evidence that modules A and B represent two discrete modules, possibly with distinct structural features. Unexpectedly, although GAL-2B had lower activity than GAL-2(AF-1) and GAL-2A (Fig. 4), it bound more strongly to the tested proteins. The low activity of GAL-2B could be due to the formation of nonproductive complexes between GAL-2B and the transcriptional targets that prevent efficient signaling to the initiation complex. The lack of strong binding of module A to any of the tested proteins, in combination with its requirement for optimal activity of AF-1, raises the question of the role of this module in AF-1 function. One possibility is that module A is a high affinity binding site for a GTF or coactivator not tested here. Alternatively, binding of a target protein to a high affinity site in module B may lead to cooperative binding of the same or a different protein to a low affinity site in module A. Although we cannot discriminate between these possibilities, cooperative binding of GTFs and coactivators to both modules could lead to synergistic activation of transcription. In this regard, the low affinity of module A for transcriptional targets may be pivotal for the combinatorial use of GTFs and coactivators by AF-1 in different promoters.
The multiplicity of AF-1 targets is indicative of a complex and dynamic exchange of interactions resulting in more efficient transcription by RNA polymerase II. Indeed, the multitarget nature of AF-1 interactions suggests that this domain might activate transcription by at least four distinct mechanisms. First, AF-1 could affect the preinitiation step through interaction with CBP and/or the ADA2⅐GCN5 complex, increasing acetylation of histones and rendering the chromatin more accessible to the transcription machinery (31,38). Second, AF-1 could increase the initiation rate through binding to TAF II 31, TAF II 80, TBP, and/or TFIIB and recruitment of the basal transcription machinery. Third, AF-1 could act at a post-initiation step, promoting the opening of the DNA double helix through its interaction with PC4 (39). Fourth, AF-1 could stimulate promoter clearance and transcriptional elongation through its interaction with TFIIH-p62 (2, 3). Thus, it appears that the protein surface of the 24-residue AF-1 activator has evolved a remarkable flexibility to make distinct interactions with multiple transcriptional targets, primarily through a tetrad of bulky hydrophobic residues. Although it is not known whether these interactions are mutually exclusive, it is conceivable that two copies of AF-1 in an HNF-4 dimer could potentially simultaneously bind different components of the transcription machinery, acting at multiple steps of transcription and resulting in synergistic activation of gene expression.