Transcriptional activation by hepatocyte nuclear factor-1 requires synergism between multiple coactivator proteins.

Hepatocyte nuclear factor-1 (HNF-1) plays an important role in the regulation of a large number of genes expressed in the liver, kidney, and pancreatic beta-cells. In exploring the molecular mechanism involved in HNF-1-dependent gene activation in the in vivo chromatin context, we found that HNF-1 can physically interact with the histone acetyltransferases (HATs) CREB-binding protein (CBP), p300/CBP-associated factor (P/CAF), Src-1, and RAC3. The transcriptional activation potential of HNF-1 on a genome integrated promoter was strictly dependent on the synergistic action of CBP and P/CAF, which can independently interact with the N-terminal and C-terminal domain of HNF-1, respectively. Moreover, the HAT activity of both coactivators was important, as opposed to the selective requirement for the HAT activity of P/CAF in activation from a transiently transfected reporter. Interaction of CBP with the N-terminal domain of HNF-1 greatly increased the binding affinity for P/CAF with the C-terminal activation domain, which may represent the molecular basis for the observed functional synergism. The results support a model that involves the combined action of multiple coactivators recruited by HNF-1, which activate transcription by coupling nucleosome modification and recruitment of the general transcription machinery.

Tissue-specific expression of hepatic genes is accomplished by the concerted action of a small number of liver-enriched regulatory proteins, including the HNF-1 1 (1)(2)(3)(4), the HNF3 (5), the HNF-4 (6), and the C/EBP (7) families of transcription factors. Although the expression of these factors is not restricted to hepatocytes, only these cells express them simultaneously at high levels (8). This pattern is achieved by a complex cross-regulatory network that has been postulated to determine the hepatic phenotype (9 -11). HNF-1 plays a central role in the coordination of this network, via positive regulation of a large number of downstream target genes (12) and via negative regulation of genes activated by HNF-4, including the HNF-1 gene itself (13,14). The key role of HNF-1 in nonhepatic tissues, such as in pancreatic ␤-cells and the kidney, has been recently highlighted by the complex phenotype of HNF-1 Ϫ/Ϫ mice (15)(16)(17) and the association of various HNF-1 mutations with an early onset form of noninsulin-dependent diabetes in human subjects (18).
HNF-1 contacts DNA via an extra large atypical homeodomain exhibiting distant homology to other known homeoproteins (2,3). Two features distinguish HNF-1 from other homeodomain transcription factors: it contains an extra 21-amino acid loop within the DNA-binding domain and dimerizes via the N-terminal dimerization domain (2,3). The dimerization domain can associate with DcoH (19), an 11-kDa protein that has been suggested to be involved in dimer stabilization (20). The C-terminal part of HNF-1 contains at least three regions, ADI, ADII, and ADIII, that have been shown to be indispensable for transcription activation (3,12).
Recent studies on the regulatory region of phenylalanine hydroxylase (21) and the ␣ 1 -antitrypsin-CBG gene cluster (22) using an HNF-1 Ϫ/Ϫ mouse model and hepatoma cell variants, respectively, raised the possibility of the potential role of HNF-1 in controlling chromatin remodeling. One route to the transcription factor-dependent chromatin modification is the recruitment of coactivators that possess intrinsic histone acetyltransferase activities. Several such intermediary proteins, including GCN5 (23), CBP (24,25), P/CAF (26), and the p160 family of proteins (Src-1 (27) and ACTR/RAC3 (28,29)) have been shown to interact with a wide variety of transcription factors and to stimulate transcription, acting as bridging proteins between activators and the general transcription factors and via acetylation of the surrounding nucleosomes to allow greater accessibility of other factors to the promoter (23, 26, 28 -34).
In the present study we show that HNF-1 can physically interact with CBP, P/CAF, Src-1, and RAC3. In functional assays, these coactivators increase HNF-1-dependent transcription in a synergistic manner. Substantial differences in coactivator usage or requirements for their HAT functions were observed between transiently transfected and chromosomally integrated reporters. We present evidence for a potential mechanism involved in CBP and P/CAF-mediated synergistic coactivation of HNF-1, which provides new insights into the mode of HNF-1-dependent target gene activation in the context of chromatin.
Cell Culture and Transfections-NIH3T3 and Cos-1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Transfections and chloramphenicol acetyltransferase assays were performed as described previously (13).
NIH3T3 cell lines containing integrated copies of the 3xAlbPE-CAT reporter were generated by cotransfection of linearized 3xAlbPE-AdML-CAT and pCB6 (36,37) containing the neomycin resistance gene. After selection with 0.6 mg/ml G418 (Life Technologies, Inc.), individual colonies were expanded and analyzed.
Co-immunoprecipitation Analysis-Nuclear extracts from transfected Cos-1 cells were prepared as described (35), and the lysates were adjusted to 25 mM Hepes, pH 7.9, 150 mM KCl, 10% glycerol, 0.1% Nonidet P-40, 0.2 mM EDTA, 1 mM dithiothreitol, 10 mM NaF, 0.5 mM phenylmethylsulfonyl fluoride, and 10 g/ml approtinin. After a preclearing step with protein A-Sepharose (Amersham Pharmacia Biotech), the extracts were incubated with 4 g/ml rabbit polyclonal CBP (Santa-Cruz Biotechnology) or mouse monoclonal FLAG antibody (Sigma) at 4°C, followed by adsorption to protein A-Sepharose. After extensive washing with the above buffer, the complexes were resuspended in SDS sample buffer, separated by 10% SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes, and immunostained with goat polyclonal HNF-1 antiserum (Santa Cruz Biotechnology).
In Vitro Protein-Protein Interaction Assays- 35 S-Labeled, recombinant proteins were synthesized in vitro using the TNT-coupled reticulocyte lysate system (Promega) in accordance with the manufacturer's instructions. 2 g of GST fusion proteins coupled to glutathione Sepharose column (Amersham Pharmacia Biotech) were incubated with the in vitro translated proteins in a buffer containing 20 mM Hepes, pH 7.9, 200 mM NaCl, 5 mM MgCl 2 , 0.1% Nonidet P-40, 0.2% bovine serum albumin, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml approtinin, at 4°C with constant agitation. After excessive washing with the same buffer lacking bovine serum albumin, the beads were resuspended in SDS sample buffer, and the proteins were separated by electrophoresis in 6% SDS-polyacrylamide gels.

RESULTS
In vivo protein-protein interactions between HNF-1 and the coactivator proteins, CBP, P/CAF, Src-1, and RAC3, were investigated by co-immunoprecipitation assays. Cos-1 cells were cotransfected with HNF-1 and CBP or FLAG epitope-tagged P/CAF, Src-1 and RAC3 expression vectors. Lysates from the transfected cells were immunoprecipitated with either ␣CBP or ␣FLAG antibodies, and the presence of HNF-1 in the precipitates was detected in Western blots using ␣HNF-1 antibody. As shown in Fig. 1, substantial amounts of HNF-1 were coprecipitated with either the ␣CBP or ␣FLAG antibody from the different extracts, indicating that HNF-1 can interact with CBP, P/CAF, Src-1, or RAC3 in vivo To determine whether these interactions are direct and to map the relevant domains of HNF-1 involved in HNF-1-coactivator interactions, we performed in vitro GST pull-down assays. As shown in Fig. 2A, CBP specifically interacts with the N-terminal region of HNF-1. The minimal region required for this interaction was mapped between amino acids 95-295, which includes the POU-homeo-related DNA-binding domain of HNF-1 (Fig. 2B). On the other hand, P/CAF, Src-1, and RAC3 interacted with the C-terminal activation domain (ADI), situated between amino acids 493 and 628. The reverse experiment using GST-CBP and GST-P/CAF fusion proteins and 35 S-HNF-1 revealed that HNF-1 can interact with two distinct domains of CBP: the N-terminal aa 1-706 region and the aa 1620 -1877 region (Fig. 2C), both overlapping with the previously characterized CBP domains interacting with P/CAF (Fig.  2D). The minimal region of P/CAF required for interaction with HNF-1 was found to overlap with the RAC3 interaction domain (653-736 aa) but to be distinct from the CBP and Src-1-binding region ( Fig. 2, C and D).
To assess the functional significance of these interactions, we coexpressed HNF-1 together with CBP, P/CAF, Src-1 or RAC3 in NIH3T3 cells, and their effects on a synthetic HNF-1-dependent reporter (3xAlbPE-CAT) were measured. NIH3T3 cells were chosen for these assays because they lack endogenous HNF-1 and contain the lowest levels of endogenous CBP and P/CAF compared with other commonly used nonhepatic cell lines (data not shown). Under conditions where HNF-1 activated transcription 5-fold above basal levels, cotransfection of CBP, PCAF, Src-1, and RAC3 expression vectors resulted in 20-fold, 44-fold, 26-fold, and 19-fold further activation of HNF-1-dependent transcription, respectively (Fig. 3A). Coexpression of CBP with P/CAF, Src-1, and RAC3 resulted in 135-fold, 121-fold, and 89-fold induction, respectively, of the activity obtained with HNF-1 alone (Fig. 3B). Similarly, coexpression of P/CAF with Src-1 or RAC3 increased the HNF-1 induced activity 132-fold and 152-fold, respectively (Fig. 3B). These values are at least 2-fold higher than the sums of the superactivations observed by either coactivator alone and thus can be considered as synergistic effects. This synergism can be explained by a mechanism that involves simultaneous interactions of two coactivators with HNF-1 directly or indirectly through interactions of coactivators with each other when one is bound to HNF-1.
To investigate these possibilities we first tested two C-terminally truncated constructs containing the 1-280 and the 1-440 amino acid regions of HNF-1. Proteins expressed from these vectors have previously been shown to retain the ability to enter the nucleus and bind DNA (14), whereas the data presented in Fig. 2 indicate that they can interact in vitro with CBP but not P/CAF, Src-1, or RAC3. Our analysis focused on the synergism between CBP and P/CAF. In transient transfection assays, as expected, both HNF-1 (1-280) and HNF-1 (1-440) were inactive and were not activated by P/CAF, because they lack the essential ADI region (Fig. 3C). Although they do contain the CBP interacting region, neither of them was activated by CBP alone, pointing to the importance of the ADI domain. When CBP and P/CAF were coexpressed along with HNF-1, substantial transactivation was detected which corresponded to 16 -18-fold activation above basal levels (Fig. 3C). This activation, however, has in neither case exceeded more than 2-fold the induction observed with wild type HNF-1. These results suggest that although the CBP-mediated recruit- The immunoprecipitates were resolved in 10% SDS-polyacrylamide gels, transferred to nitrocellulose membranes, and probed with goat polyclonal HNF-1 antibody. The first lanes of each panel (Ϫ) represent nonimmunoprecipitated lysate corresponding to about 2% of the input. ment of P/CAF onto HNF-1 may take place independently of the ADI region, the presence of ADI is necessary for the high level of synergism observed in Fig. 3B. The ability of HNF-1bound CBP to recruit P/CAF has also been observed in vitro.
GST-HNF-1 (1-295), containing the CBP interaction domain of HNF-1, retained some 35 S-PCAF only when excess baculovirus expressed CBP was included in the binding reaction (Fig. 4A,  left panel). To test whether CBP bound to the N-terminal region of HNF-1 may increase the binding of P/CAF to the C-terminal activation domain of HNF-1, we performed pulldown experiments using GST-HNF-1 (1-628), which contains the full-length protein. P/CAF efficiently interacted with fulllength HNF-1 both in the presence and absence of excess cold CBP. However, when the stringency of the washing conditions was increased, the HNF-1-P/CAF complex was detectable only in the presence of CBP (Fig. 4B). This effect was observed only in the context of the full-length HNF-1 protein and not with the fusion protein HNF-1 (493-628) containing the isolated P/CAF interaction domain. In fact, in this latter case CBP slightly reduced the binding of P/CAF to the HNF-1 ADI domain, which may be due to competition between HNF-1 ADI and CBP for interaction with P/CAF (Fig. 4A, right panel). The above data strongly suggest that binding of CBP to the N-terminal region of HNF-1 increases the affinity of P/CAF binding to the Cterminal ADI domain, and the increased stability of this trimeric complex may be responsible for the synergistic activation observed in the transfection assays.
Next, we investigated whether synergism between CBP and P/CAF is important for HNF-1-dependent activation in hepatic cells. Overexpression of CBP or P/CAF in HepG2 cells resulted in a 2.1-and a 1.8-fold increase of HNF-1-mediated transactivation, respectively (Fig. 5A). Coexpression of both CBP and P/CAF induced HNF-1 activity about 9-fold (Fig. 5A), suggesting that CBP and P/CAF may act synergistically with HNF-1 in hepatic cells. The above values of induction are at least 1 order of magnitude smaller compared with those obtained with NIH3T3 cells. The difference can be explained by the much higher endogenous levels of CBP and P/CAF in HepG2 cells. To prove unequivocally that both CBP and P/CAF are important for HNF-1-mediated transcription in hepatic cells, we performed squelching experiments with E1A and its mutant derivatives defective in either interaction with CBP (E1A CBP-Mut) or with P/CAF (E1A P/CAFMut) or both (E1A ⌬N) (35). Overexpression of wild type E1A efficiently inhibited HNF-1dependent activation (Fig. 5B). A similar degree of inhibition was observed with either E1A CBPMut or E1A P/CAFMut, but not with E1A ⌬N (Fig. 5B), suggesting that the simultaneous interaction of HNF-1 with both CBP and P/CAF is necessary for high level activation in hepatic cells.
To investigate the contribution of the HAT activities of CBP and P/CAF on their coactivation function on HNF-1, we compared the activities of wild type coactivator constructs with the activities of their mutant forms, CBP(L1690K/C1691L) and P/CAF(Y616A/F617A), which lack HAT activity (38,39). The HAT-deficient (HAT Ϫ ) CBP mutant coactivated HNF-1dependent transcription and synergized with P/CAF to a similar extent as its wild type counterpart (Fig. 6, A and C). On the other hand, diminished coactivation and no synergism were observed with the HAT Ϫ P/CAF mutant (Fig. 6, B and C). Overexpression of E1A, which can both inhibit the HAT activity of CBP and P/CAF and sequester them from transcription factors (35,40,41), inhibited CBP-and P/CAF-mediated coactivation on HNF-1 to different degrees, with 25 and 6% of CBP-and P/CAFmediated activity remaining, respectively (Fig. 6, A and B). Treatment of cells with TSA did not further induce coactivation by CBP, P/CAF, and the CBP HAT Ϫ mutant but substantially affected the diminished coactivation by P/CAF HAT Ϫ (Fig. 6, A  and B). These results suggested that the HAT function of P/CAF but not that of CBP is required for maximal transcription.
The importance of nucleosome modifying activities for HNF- 1-dependent transcriptional activation prompted us to study the function of coactivators using a genome-integrated promoter, because the chromatin structure adopted by the transiently transfected template could differ from the typical chromatin organization of the eukaryotic genome (42). To this end, NIH3T3 cells were transfected with 3xAlbPE-CAT, and cell lines containing stably integrated copies of the transgene were obtained by G418 selection. Individual cell lines were analyzed for HNF-1-and coactivator-induced transactivation. Overexpression of CBP, P/CAF, Src-1, and RAC3 only marginally induced HNF-1-dependent transcription from the integrated template by factors of 2.5-, 3.0-, 1.5-, and 1.7-fold, respectively (Fig. 7). Interestingly, the HAT deficient mutants of either CBP or P/CAF did not coactivate HNF-1 in this system. High level activation (23-fold above that of HNF-1 alone) was observed only by the simultaneous overexpression of wild type CBP and P/CAF. Although cotransfection of Src-1 or RAC3 together with either CBP or P/CAF resulted in increased activation, the values obtained (5.4-fold for Src-1/CBP, 5.1-fold for RAC3/CBP, 5.0-fold for Src-1/P/CAF, and 5.2-fold for RAC3/P/CAF) correspond to an additive rather than synergistic effect. Similar additive effects were observed when wild type CBP and the HAT Ϫ derivative of P/CAF, or wild type p/CAF and the HAT Ϫ mutant of CBP were coexpressed (Fig. 7). These results suggest that in the integrated promoter context the HAT activity of both CBP and P/CAF are equally required. Moreover, the actual rate of HNF-1-dependent transactivation highly depends on the composition of the coactivator complex. In this reporter system, the combined action of CBP and P/CAF was necessary to achieve the highest transcription rates. DISCUSSION Previous studies employing in vitro transcription and transfection assays identified a large number of liver-specific genes activated by HNF-1 (12). These studies provided important information with respect to the crucial roles of HNF-1 on specific promoters and its cooperation with other transcription factors but did not reveal how HNF-1 could be involved in the activation of these genes in the in vivo chromatin context. Addressing this question, two recent reports suggested that the mechanism of HNF-1-mediated target gene expression in vivo might involve HNF-1-dependent remodeling of chromatin. First, DNase I-hypersensitive site mapping of the phenylala-

FIG. 5. The simultaneous interaction of CBP and P/CAF with HNF-1 is required for HNF-1-dependent transactivation in hepatic cells.
HepG2 cells were transfected with 2 g of 3xAlbPE TK-CAT, 0.2 g of HNF-1 and 2 g of CBP or P/CAF (A) or 0.2 g of HNF-1 together with 0.2 g of E1A, or E1A CBPMut, or E1A P/CAFMut or E1A ⌬N (B) expression vectors as indicated. The bars represent normalized CAT activities and standard errors from two independent experiments and expressed as fold induction above the levels obtained with HNF-1 alone (A) or as a percentage of the CAT activity obtained with HNF-1 (B). nine hydroxylase promoter revealed significant differences between wild type and HNF-1 Ϫ/Ϫ mice (21). Second, studies on the ␣ 1 -antitrypsin-CBG gene cluster have shown altered patterns of DNase I-hypersensitive sites in hepatoma variants depending on the expression of HNF-4 and HNF-1 (22). In both cases, a positive correlation between chromatin structure and target gene expression was observed. The above observations raised the question of whether the changes in chromatin structure are an indirect consequence of HNF-1 expression, e.g. via the activation of other genes, or whether HNF-1 could be directly involved in chromatin remodeling.
The results presented in this paper demonstrating physical interactions of HNF-1 with proteins possessing intrinsic nucleosome modifying activities could provide the molecular basis for the second scenario. We found that CBP, P/CAF, Src-1, and RAC3 can independently interact with distinct functional domains of HNF-1 and dramatically increase HNF-1-mediated transactivation in a synergistic manner. The molecular basis of this synergism could lie in the ability of the coactivators to interact with each other. The importance of such interactions for the recruitment of multiple coactivators by various transcription factors has been demonstrated before (38,39,43,44). For example, in the case of nuclear hormone receptors, P/CAF and Src-1 seem to provide a molecular platform for the recruitment of CBP and thus to create a highly active coactivator complex (38,39). Our results suggest that, in the case of HNF-1, a novel type of mechanism may apply that involves the interaction of CBP with the N-terminal domain of HNF-1, leading to increased affinity of P/CAF for the C-terminal domain of HNF-1. This could be achieved either by a CBP-induced conformational change in HNF-1 enhancing the affinity of P/CAF for HNF-1 or by the additional protein-protein interaction interface provided by CBP to P/CAF independently recruited by HNF-1. These two possibilities are probably not mutually exclusive, and the trimeric complex formed via multiple interactions (HNF-1 (N terminus)-CBP; CBP-P/CAF and P/CAF-HNF-1 (C terminus)) should be highly stable and represent the most active configuration of the HNF-1-coactivator complex. Although we have focused our investigation on CBPand P/CAF-mediated activation, we speculate that the synergism observed with CBP and Src-1 or RAC3 may involve a similar mechanism. It is more difficult to explain the synergism observed with P/CAF and Src-1 or RAC3, because they contact the same domain of HNF-1. Although it is not known whether these interactions are mutually exclusive, it is possible that the two ADI domains in an HNF-1 dimer could provide interfaces for simultaneous binding of two different coactivator molecules. Another possibility is that HNF-1-bound P/CAF may recruit Src-1 or RAC3 via direct interactions with them.
The fact that HNF-1 is able to recruit four different HAT proteins indicates that histone acetylation may play a role in HNF-1-dependent gene activation. Indeed, the histone deacetylase inhibitor TSA (45) strongly potentiated HNF-1 transcriptional activity, although it did not further enhance CBP-or P/CAF-mediated superactivation in transient transfection experiments. This indicates that histone deacetylation, which is inhibited by TSA, confers a repressed state to the promoter, which can be overcome by HNF-1-mediated recruitment of endogenous or transfected CBP and P/CAF. TSA-induced activation, however, was always smaller than that observed by overexpression of CBP or P/CAF, suggesting that their "adaptor" or "bridging" activities between HNF-1 and the general transcription machinery are also important in coactivator function. In this way, the processes of histone acetylation and FIG. 6. The histone acetylase activity of P/CAF but not that of CBP is required for coactivation of HNF-1-dependent transcription from a transiently transfected reporter. NIH3T3 cells were co-transfected with 2 g of 3xAlbPE TK-CAT, 0.2 g of HNF-1 together with combinations of 0.2 g pRSV E1A or 2 g of CBP, P/CAF expression vectors, or their HAT Ϫ mutant forms (CBPL1690K;C1691L and P/CAF Y616A;F617A). Where indicated, transfected cells were treated with 1 M TSA 12 h before harvest. The data represent normalized CAT activities and standard errors from at least four independent experiments and expressed as fold increase above the activity obtained with HNF-1 alone.
FIG. 7. Activation of a chromosomally integrated HNF-1-dependent promoter requires both the CBP and P/CAF histone acetyltransferase functions. NIH3T3 cells with stably integrated 3xAlbPE TK-CAT reporter were transfected with 1 g of cytomegalovirus HNF-1 and 2 g of the indicated expression vectors. The data represent normalized CAT activities and standard errors from at least four independent experiments performed in the same cell line. Essentially, the same effects with small variations in the absolute activities were observed in two other individual stable cell lines. The activity obtained by HNF-1 alone was set to 1. All other data are expressed as fold increase above the activity obtained with HNF-1 alone. transcriptional activation are closely linked.
Different classes of transcription factors may not only recruit complexes with different coactivator compositions but may also induce different configurations of the specific coactivator components leading to both differential requirements for or modulation of their HAT activities. For example, the HAT activity of P/CAF but not that of CBP appears to be important for nuclear hormone receptor, MyoD or NFB-dependent activation (38,39,46,47), whereas the HAT activity of CBP is required for CREB or STAT-1 function (37,38). In addition, it has been shown that the Src-1 homologue pCIP can regulate the function of CBP by either stimulating or inhibiting its HAT activity in a substrate-specific manner (48). In the case of HNF-1 we found that the HAT activity of both CBP and P/CAF were equally important for high levels of activation when examined on genome integrated promoter as opposed to selective requirement for the HAT activity of P/CAF in the transiently transfected reporter. Moreover, in the integrated promoter context, only the combined action of CBP and P/CAF on HNF-1 was sufficient to induce high level transcription, whereas either CBP or P/CAF alone produced only marginal activation. These observations indicate that in the context of genomic DNA, the promoter is organized into a chromatin structure that imposes a stronger barrier for transcription, compared with that imposed by the structure formed on transiently transfected plasmid DNA, which may not be arranged in a typical array of nucleosomes (42). The simultaneous requirement for multiple HAT activities with different substrate specificities (CBP and P/CAF) (26,29,32) could thus be directly related to the increased repressive effect of the chromatin structure adopted by the integrated template, whereas participation of proteins in the coactivator complex that possess weaker HAT activities (Src-1 and RAC3) (28, 33) may not be sufficient to create the most active configuration.
In this work we used a synthetic HNF-1-dependent promoter construct to avoid complications arising from the action of other DNA-binding proteins. This has allowed us to identify and analyze the functions of multiple coactivator proteins, which may provide a mechanistic ground to decipher the complex mechanism involved in HNF-1-dependent chromatin remodeling and gene activation in vivo. In the context of natural regulatory regions, however, the HNF-1-coactivator complexmediated disruption of neighboring nucleosomes may also increase the access of other transcription factors, which may in turn alter the composition and configuration of the coactivator complex in a promoter-specific manner. In addition, the relative expression levels of the different coactivators in a given cell type may also play a determinative role in the actual composition of the coactivator complex recruited by HNF-1. In accordance with the differential effects of different combinations of coactivators observed in this study, we propose that the actual composition of the coactivator complex formed on HNF-1 may have a direct consequence on promoter strength and the contribution of HNF-1 in the transcriptional activity of a given gene in a given cell type.