Nuclear Factor I (NFI) Isoforms Differentially Activate Simple versus Complex NFI-responsive Promoters*

Promoter-specific differences in the function of transcription factors play a central role in the regulation of gene expression. We have measured the maximal transcriptional activation potentials of nuclear factor I (NFI) proteins encoded by each of the four identified NFI genes (NFI-A, -B, -C, and -X) by transient transfection in JEG-3 cells using two model NFI-dependent promoters: 1) a simple chimeric promoter containing a single NFI-binding site upstream of the adenovirus major late promoter (NFI-Ad), and 2) the more complex mouse mammary tumor virus long terminal repeat promoter. The relative activation potentials for the NFI isoforms differed between the two promoters, with NFI-X being the strongest activator of NFI-Ad and NFI-B being the strongest activator of the MMTV promoter. To determine if these promoter-specific differences in activation potential were due to the presence of glucocorticoid response elements (GREs), we added GREs upstream of the NFI-binding site in NFI-Ad. NFI-X remains the strongest activator of the GRE containing simple promoter, indicating that differences in relative activation potential are not due solely to the presence of GREs. Since NFI proteins bind to DNA as dimers, we assessed the activation potentials of NFI heterodimers. Here, we show that NFI heterodimers have intermediate activation potentials compared with homodimers, demonstrating one potential mechanism by which different NFI proteins can regulate gene expression.

The expression of RNA polymerase II-dependent genes is mediated by a complex set of DNA-protein and protein-protein interactions. General transcription factors directly interact with the RNA polymerase holoenzyme to form a transcriptioncompetent complex which is capable of a basal level of transcription (see Refs. 1 and 2, for reviews). Efficient gene transcription occurs when promoter-specific regulatory proteins (transcriptional activators and repressors) bind to DNA sequences in promoters and enhancers and interact directly, or via adapter proteins, with components of the basal machinery (Refs. 3-6; and see Refs. 7 and 8, for reviews). Transcriptional activators are modular in nature containing separable and functionally distinct DNA binding and activation domains (Refs. 9 and 10; see Ref. 11, for review). Activation domains have been grouped into three classes depending on their amino acid composition: acidic, glutamine-rich, or proline-rich (9,12). We are studying the nuclear factor I (NFI) 1 family of sitespecific DNA-binding proteins which, based on the amino acid composition of NFI-C proteins, is grouped into the proline-rich class of activators (13).
NFI was initially identified as a host-encoded protein required for the efficient initiation of adenovirus (Ad) replication in vitro (14) and was later shown to be required for the correct expression of numerous cellular and viral genes. Cloning of cDNAs encoding NFI proteins from various species (13,(15)(16)(17) has identified a family of four genes (NFI-A, NFI-B, NFI-C, and NFI-X) that are highly conserved from chicken to human. NFI proteins contain a highly conserved NH 2 -terminal 220 amino acid region, which mediates DNA binding, dimerization, and the initiation of Ad replication (18 -20). NFI proteins bind to DNA as both homo-and heterodimers and recognize the consensus binding site, TTGGC(N 5 )GCCAA with the same apparent affinity (21,22). However, considerable variation occurs within the COOH-terminal domains of the NFI proteins which likely encode distinct transcription modulation domains. Additional variation between NFI proteins is generated through differential splicing of transcripts from each of the four genes (23).
The existence of 4 different NFI genes in vertebrates, their differential expression during mouse development (24), and the regulated expression of NFI-dependent genes expressed in multiple organs including brain (25,26), liver (27), muscle (28,29), and other terminally differentiated tissues (30), argues in favor of diverse transcription modulation properties for NFI proteins. Different NFI-C isoforms isolated from HeLa cells and porcine liver have different transcriptional activation potentials (13,15). The precise mechanism of NFI-mediated activation is unknown, however, direct interactions between one NFI-C protein (CTF1) and components of the basal transcriptional machinery have been reported (31)(32)(33). This interaction was shown to be dependent on a sequence motif related to the COOH-terminal heptapeptide repeat (CTD) of RNA polymerase II. However, this is unlikely to be the only mechanism by which NFI proteins activate transcription as some NFI proteins lacking a CTD repeat are potent activators in both yeast (34,35) and mammalian cells (24,36).
Although it is widely accepted that NFI proteins function as * This work was supported in part by the Lerner Research Institute of the Cleveland Clinic Foundation and National Institutes of Health Grant HD34908 (to R. M. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.   1 The abbreviations used are: NFI, nuclear factor I; MMTV-LTR, mouse mammary tumor virus long terminal repeat; GRE, glucocorticoid response element; GR, glucocorticoid receptor; Ad, adenovirus; CTD, carboxyl-terminal domain; ␤-gal, ␤-galactosidase; CMV, cytomegalovirus; HA, hemagglutinin; dexamethasone, 9␣-fluoro-16␣-methyl-11␤,17␣,21-trihydroxy-1,4-pregnadiene-3,20-dione; TBP, TATA-binding protein; TFIIB, transcription factor IIB; TAF, TBP-associated factor; GMS, gel mobility shift; PBS, phosphate-buffered saline. transcription modulators, it is not clear if they exhibit promoter-specific differences in activation potential. To address this question, and to better understand the transcription modulation properties of NFI proteins, we performed transient transfection experiments using two model NFI-dependent reporter constructs. Here, we demonstrate that NFI isoforms representing each of the 4 murine NFI genes exhibit a broad range of transactivation potentials. Our data also show that activation potentials are modulated in a promoter-specific manner, determined by the COOH-terminal regions of the NFI proteins.

MATERIALS AND METHODS
Plasmid Constructs-The reporter plasmids, pNFI␤-gal and p⌬NFI␤-gal, are derivatives of the pB series of vectors which were described previously (37). The bacterial chloramphenicol acetyltransferase gene of pB and pBFB was removed by digestion with BalI and repair of the ends with Klenow polymerase. The bacterial ␤-galactosidase gene (␤-gal) was isolated by digesting pCMV␤-gal (Stratagene) with NotI and repair of the overhanging ends with Klenow polymerase. The blunted NotI ␤-gal fragment was cloned into pB and pBFB creating p⌬NFI␤-gal and pNFI␤-gal, respectively. Orientation was verified by restriction enzyme digestion. pNFI␤-gal differs from p⌬NFI␤-gal by presence of a single NFI-binding site cloned immediately upstream of the Ad major late promoter (Ϫ51 to ϩ33) promoter element. Reporter plasmid pNFIGRE␤-gal was generated by cloning the synthetic oligonucleotide (CTAGTGTACAGGATGTTCTGCGGCCGCTGTACAGGA-TGTTCTCTAG) which consists of 4 GRE half-sites (in bold) into the XbaI site of pNFI␤-gal. pMMTV␤-gal was created by cloning the Hin-dIII/BamHI fragment of pMAMNeo␤-gal (CLONTECH) into HindIII/ BamHI digested pBSIIKSϩ (Stratagene). NFI effector plasmids pCH-NFI-B, pCHNFI-C, and pCHNFI-X have been described previously (24) and express murine proteins homologous to chicken NFI-B2, -C2, and human NFI-X2 (23). pCHNFI-A is identical to the NFI-A effector plasmid described previously (24) except that it has a 92-nucleotide deletion at residue 473 of the coding sequence, which changes the reading frame and generates a murine protein homologous to chicken NFI-A4 (23). Plasmid pCHNFI-B-235, which expresses the DNA-binding domain of the murine NFI-B isoform was made by digesting pCHNFI-B with BstXI/BglII, repair with Klenow and religation. Digestion of pCHNFI-C with BstEII/BglII and pCHNFI-X with AflII/BglII followed by repair with Klenow and religation created pCHNFI-C-240 and pCHNFI-X-210, respectively. pCHNFI-B/X was made by digesting pCHNFI-B with Asp-718, repair with Klenow and digestion with NotI, liberating the fragment containing the murine NFI-B DNA-binding region. This fragment was cloned into pCHNFI-X which had been digested with AflII, blunted, and subsequently digested with NotI removing the NFI-X DNA-binding domain. This chimeric construct contains a deletion of a His residue at position 209, which does not affect DNA binding (data not shown). pCHNFI-X/B was created by digesting pCHNFI-X with AflII, repair with Klenow, and digestion with BglII, removing the COOHterminal region of the murine NFI-X cDNA. The NFI-B COOH-terminal region was isolated as a blunted XhoI/BglII fragment and cloned into the repaired AflII/BglII digested pCHNFI-X generating pCHNFI-X/B. During cloning, a Leu at position 209 is changed to a Ile, which has no effect on the DNA binding or expression of the NFI-X/B chimeric protein (data not shown). pCHNFI-X-VP16 was created by cloning the blunted BglII/BamHI fragment of CRF-1 (generous gift from Steve Triezenberg) containing the acidic activation carboxyl-terminal domain of the HSV-1 transactivator VP16 into the blunted AflII/BglII fragment of pCH-NFI-X. Sequence, orientation, and correct reading frame of all plasmids was verified by automated sequencing (CCF Molecular Biotechnology Core Facility).
Cell Culture Transfection and Assays-JEG-3 choriocarcinoma cells (American Type Culture Collection) were cultured in ␣-minimal essential medium (Mediatech) containing 10% fetal bovine serum. Twentyfour hours prior to transfection, 2 ϫ 10 5 cells were plated onto 60-mm dishes and then transfected using calcium phosphate coprecipitation as described (38). Typically, each coprecipitation consisted of a NFI-dependent ␤-gal reporter construct (5 g), SV-40-luciferase (pGL2-Control, Promega) internal control (2.5 g), and various amounts of the indicated NFI-effector constructs (0.25-5.0 g). Unless otherwise indicated, 1.0 g of a hGR expression vector (gift from Ron Evans) was co-transfected with pMMTV␤-gal and pNFIGRE␤-gal. Carrier DNA (pBSIIKSϩ) was added to a total of 15 g of transfected DNA per plate. Cells were incubated with CaPO 4 /DNA precipitate for 12-16 h, washed with PBS, and incubated for 24 h in culture media with or without 0.1 M dexamethasone. Cells were harvested in 300 l/plate of 1 ϫ Reporter Lysis Buffer (Promega) with luciferase and ␤-gal assays performed as described (24). Transfections were performed in quadruplicate using duplicate precipitates for each point, and all results were confirmed by multiple independent experiments using at least two different CsCl purified preparations of plasmid DNA.
Gel Mobility Shift Assays and Antibody Supershifts-To prepare JEG-3 whole cell extracts, cell monolayers were washed 2 times with ice-cold PBS and scraped into a 1.5-ml Eppendorf tube. Cells were pelleted by spinning at 1000 ϫ g for 5 min at 4°C, the supernatant was removed and the pellet resuspended in lysis buffer (100 mM Tris, pH 7.4, 350 mM NaCl, 10% glycerol, 1% Nonidet P-40, 1 mM EDTA, 1 mM dithiothreitol, 10 g/ml leupeptin, 10 g/ml pepstatin, 10 g/ml aprotinin) and incubated on ice for 15 min with occasional swirling. Cell debris was pelleted by spinning at 16,000 ϫ g for 10 min, with the supernatant removed and stored at Ϫ80°C. Gel mobility shift analysis was performed using the 32 P-labeled FIB 2.6 oligonucleotide or a control mutant oligonucleotide FIB C.2 as described previously (21). The single point mutation present in FIB C.2 abolishes NFI binding in vitro (39). For antibody supershift analysis with the C125A anti-hemagglutinin (HA) antibody (Boehringer Mannheim), binding reactions were carried out for 30 min on ice, followed by the addition of 1 l of the anti-HA antibody (200 g/ml) and further incubation for 20 min at room temperature. As a control, binding reactions not incubated with antibody, were treated with 1 l of PBS and incubated for 20 min at room temperature. The DNA-protein complexes were resolved as described in Ref. 21 and analyzed using a Molecular Dynamics model 400 PhosphorImager.
Immunocytochemistry-To detect the intracellular location of the HA-tagged NFI proteins, HeLa cells were grown on coverslips, transfected with vectors expressing the NFI protein or control vectors, cultured for 48 h, and fixed in chilled methanol at Ϫ20°C for 15 min. Fixed cells were blocked for 1 h at room temperature with 3% bovine serum albumin, 0.1% Tween 20 in PBS (PBST), incubated with 2-3 g/ml anti-HA antibody in PBST for 45 min, washed in PBS, incubated for 45 min with fluorescein isothiocyanate-conjugated secondary antibody, washed in PBS, the coverslips were mounted with Vectashield (Vector Laboratories), the cells were examined using a Nikon fluorescence microscope and the images were captured with an Oncor imaging system (Cleveland Clinic Fluorescent Microscopy Core). For control staining, the specific antibody solution was replaced by PBST alone.

NFI Proteins Exhibit Promoter-specific Differences in Their
Maximal Activation Potentials-We have recently reported the cloning of NFI cDNAs from each of the 4 murine NFI genes, analyzed their embryonic and postnatal expression patterns, and demonstrated their ability to activate the glucocorticoidinducible mouse mammary tumor virus (MMTV) promoter in JEG-3 cells (24). To extend these studies we have used transient transfection assays to determine if the four NFI gene products exhibit promoter-specific differences in activation potential. The NFI effector plasmids and NFI-dependent reporter constructs used in co-transfection experiments are shown in Fig. 1A. NFI effector plasmids contain the human cytomegalovirus (CMV) immediate early promoter/enhancer (40) expressing murine proteins homologous to chicken NFI-A4, NFI-B2, NFI-C2, and human NFI-X2 (23). Each NFI protein contains an NH 2 -terminal human influenza virus hemagglutinin (HA) epitope tag (41) for ease of detection. The NFI-dependent promoters used are a simple chimeric promoter, pNFI␤-gal and the complex viral pMMTV␤-gal reporter construct. The reporter plasmid pNFI␤-gal is derived from the chimeric pB series of vectors which have been described previously (37) and contains a single NFI-binding site upstream of the adenovirus major late promoter (Ϫ51 to ϩ33) driving expression of the bacterial ␤-gal gene. pMMTV␤-gal contains the glucocorticoidinducible MMTV promoter driving expression of ␤-gal. The MMTV promoter has previously been shown to contain an NFI-binding site required for its function (42,43). Transfection experiments were carried out in JEG-3 human choriocarcinoma cells, which contain low levels of endogenous NFI pro-teins and have previously been used to measure activation of the MMTV promoter by porcine NFI-C (43).
To assess the maximal transactivation potential of each of the NFI isoforms, increasing amounts of each effector plasmid were transfected into JEG-3 cells. With both NFI-dependent promoters, increasing amounts of each effector plasmid caused a simple monotonic increase in expression up to apparent saturation, showing no squelching at high levels of effector plasmid (Fig. 1B, i and ii). The maximal level of reporter expression differed for each of the isoforms and was promoter-specific. NFI-X shows the strongest activation of pNFI␤-gal (ϳ17-fold, Fig. 1B, panel i), followed by, in decreasing order of activation, NFI-B (ϳ14-fold), NFI-C (ϳ9-fold), and NFI-A (ϳ3-fold). In contrast, NFI-B is the most potent activator of pMMTV␤-gal (Fig. 1B, panel ii) with maximal activation of ϳ13-fold, followed by NFI-X (ϳ11-fold), NFI-C (ϳ6-fold), and NFI-A (ϳ3-fold). The fold activation is shown only in the presence of dexamethasone as MMTV promoter activity in the absence of hormone induction is undetectable (see Fig. 2B, for example). Differences in activation potential of each NFI isoform are not due to differences in protein level expression as Western blot analysis (data not shown) and gel shift analysis indicate similar NFI protein levels in transfected cells (Fig. 1C, lanes 3-6, arrow B), which are appropriately super-shifted when incubated with the anti-HA antibody (lanes 8 -11, arrow A).
2A, lane 1 versus 6), which is likely due to the low levels of endogenous NFI proteins seen in JEG-3 cells (Fig. 1C, lane 2). As expected, NFI-mediated activation of the MMTV reporter requires co-transfection of a vector expressing the human glucocorticoid receptor (Fig. 2B, lanes 2 20). NFI-mediated activation is dependent on the COOH-terminal domain of the NFI-B, -C, and -X isoforms, as deletion constructs lacking the COOH-terminal domain fail to significantly activate pMMTV␤gal (Fig. 2C, lanes 2 versus 6, 8, and 10) or pNFI␤-gal (lanes 11 versus [13][14][15]. These data indicate that the four NFI gene products exhibit promoter-specific differences in their maximal activation potentials. Nuclear Localization of the HA-tagged NFI Proteins-To assess whether the differences in transactivation potential of the NFI isoforms could be due to differences in cellular localization, we examined the intracellular location of the 4 NFI proteins by immunocytochemistry. When stained with monoclonal antibodies directed against the HA epitope tag, cells transfected with vectors expressing any of the 4 HA-tagged NFI proteins exhibited strong nuclear fluorescence (Fig. 3, panels A-D). Mock transfected cells stained with the anti-HA antibodies, or cells stained with secondary antibody alone showed no nuclear fluorescence (Fig. 3, panel E, and data not shown). While the HA-tagged NFI-A protein appeared to stain with a slightly more punctate distribution than the other isoforms (Fig. 3,  panel A versus B-D), all 4 proteins were strongly localized to the nucleus. These data agree with our observation that all 4 NFI proteins are found predominantly in nuclear extracts rather than cytosolic extracts when cells are lysed under low salt conditions (not shown). Thus the 4 NFI proteins, while they differ in their transactivation potentials and promoter specificity, show a similar pattern of nuclear localization.
Promoter-specific Activation Potential of NFI Proteins Is Determined by Their COOH-terminal Domains-Previous studies have demonstrated that the transcription modulation properties of differentially spliced NFI-C and NFI-X isoforms were regulated by their COOH-terminal domains (13,44). To determine if the promoter-specific differences in the activation potential of the NFI-B and NFI-X isoforms were mediated by their COOH-terminal domains, we swapped these regions, creating the chimeric fusion constructs pCHNFI-B/X and pCH-NFI-X/B. pCHNFI-B/X expresses the NH 2 -terminal 208 amino acid DNA-binding domain of NFI-B fused to amino acids 209 -401 of NFI-X, while pCHNFI-X/B contains the NH 2 -terminal 210 amino acids of NFI-X fused to the COOH-terminal amino acids 209 -420 of NFI-B. Saturating amounts of the NFI effector plasmids were transfected with each NFI-dependent reporter construct (Fig. 4). Maximal activation of the MMTV reporter is observed with the NFI-B or NFI-X/B effector constructs, with NFI-X and NFI-B/X activating less (Fig. 4, lanes 4  and 6 versus 8 and 9). In contrast, the strongest activation of pNFI␤-gal occurs with either NFI-X or NFI-B/X effector plasmids, with reduced levels of activation mediated by NFI-B and 8, 17, and 18), or NFI-X (lanes 9, 10, 19, and 20) into JEG-3 cells which were cultured in the absence or presence of 0.1 M dexamethasone (Ϫ or ϩ Dex) for 24 h. ␤-Galactosidase activity was normalized to luciferase levels and is expressed as fold activation over the CMV control plasmid (lanes 1, 2, 11, and 12). Panel C, reporter plasmid pMMTV␤-gal (5 g) and a hGR expression plasmid (1 g), or pNFI␤-gal (5 g), were transfected with 2.5 g of either CMV control vector (lanes 1, 2, and 11) or vectors expressing full-length NFI-B (lanes 3 and 4); NFI-X (lane 12), or the truncated DNA-binding domains (see "Experimental Procedures" for construction) of NFI-B (lanes 5, 6, and 13), NFI-C (lanes 7, 8, and  14), or NFI-X (lanes 9, 10, and 15). ␤-Gal activity was normalized to luciferase levels and is expressed as fold activation over the CMV control plasmid.  6 -10), and co-transfected with 2.5 g of a CMV control vector (lanes 1 and 6), or vectors expressing NFI-A (lanes 2 and 7), NFI-B (lanes 3 and 8), NFI-C (lanes 4 and 9), NFI-X (lanes 5 and 10), and 2.5 g of the SV-40 luciferase internal control vector. ␤-Gal expression was normalized to luciferase values with the bars representing the mean and range of four measurements from duplicate transfections. Panel B, reporter pMMTV␤-gal (5 g) was transfected alone (lanes 1-10), or with 1.0 g of a hGR expression vector (lanes 11-20), and co-transfected with 2.5 g of: CMV control vector (lanes 1, 2, 11, and 12), CMV vectors expressing NFI- A  (lanes 3, 4, 13, and 14), NFI-B (lanes 5, 6, 15, and 16), NFI-C (lanes 7, NFI-X/B (Fig. 4, lanes 14 and 15 versus 12 and 13). These findings demonstrate that the promoter-specific differences in relative activation potential of NFI-B and NFI-X appear to be mediated solely by their COOH-terminal regions.

Differences in Activation Potential of NFI-B and -X Are Not
Due Solely to the Presence of GREs-While both reporter constructs used in this study are NFI-dependent, the MMTV promoter is considerably more complex, with its activity also dependent on GREs and other sequence motifs (reviewed in Ref. 45). To determine if the promoter-specific differences in relative activation potentials of the NFI-X and NFI-B isoforms were due to the presence of GREs, we compared the activation properties of pNFI␤-gal with those of pNFIGRE␤-gal, which was made by addition of two GREs upstream of the NFIbinding site of pNFI␤-gal (Fig. 5A). In the absence of coexpressed GR, the activity of pNFIGRE␤-gal appears identical to that of pNFI␤-gal with NFI-B activating ϳ10-fold and NFI-X ϳ15-fold (Fig. 5B, lanes 7-12 versus 1-6, see also Fig. 2A). The lack of dexamethasone induction here is due to the absence of either GREs (lanes 2, 4, and 6) or GR (lanes 8, 10, and 12). In the absence of dexamethasone, when GR is coexpressed with pNFI-GRE␤-gal, the reporter expression is identical to that seen with pNFI␤-gal, with NFI-B and NFI-X activating expression ϳ10-fold and ϳ15-fold, respectively, (Fig. 5B, lanes 15 and  17 versus 13). However, a robust dexamethasone-dependent activation occurs when the human glucocorticoid receptor is coexpressed with pNFIGRE␤- gal (ϳ150-fold, lanes 14 versus  13). Dexamethasone-dependent activation is further enhanced when NFI-B is co-transfected (lane 16, ϳ260-fold). However,  1, 2, and 11), or vectors expressing NFI-B (lanes 3, 4, and 12), NFI-X (lanes 7, 8, and 14), NFI-B/X (lanes 9, 10, and 15), or NFI-X/B (lanes 5, 6, and 13). Cells transfected with pMMTV␤-gal were cultured in the absence or presence of 0.1 M dexamethasone for 24 h (Ϫ or ϩ Dex). ␤-Gal activity was normalized to luciferase and is plotted as in Fig. 2. maximum expression of pNFIGRE␤-gal is observed when it is co-transfected with the NFI-X effector construct (lane 18, ϳ350-fold). Therefore, although the addition of GREs to pNFI␤-gal renders it glucocorticoid responsive, the relative order of activation potentials of the NFI isoforms remains unchanged. These findings suggest that promoter-specific differences in the relative activation potential of NFI-X and NFI-B isoforms are not due solely to the presence of GREs.
The VP16 Acidic Domain Is a More Potent Activator Than the NFI Activation Domains-To determine the relative potencies of the NFI activation domains, we compared the activation properties of NFI domains to that of the potent VP16 acidic activation domain (46). For this comparison, the COOH-terminal 78 residues of VP16 were fused to the DNA-binding domain of NFI-X (NH 2 -terminal 210 amino acids) creating effector plasmid pNFI-X-210-VP16. Co-transfecting 2.5 g of pNFI-X-210-VP16 with each of our model NFI-dependent reporter constructs, results in a strong activation of each reporter (Fig. 7, lane 10 versus 2 and lane 15 versus 11). If more than 2.5 g of pNFI-X-210-VP16 plasmid was transfected, a decrease in ␤-gal expression was seen (data not shown). This decrease is probably due to squelching by the strong transactivation domain of VP16 (47), which is not seen with any of the NFI activation domains (Fig. 2B and data not shown). Activation is dependent on the VP16 activation domain as control plasmid pNFI-X-210, which contains only the NFI-X DNA-binding domain fails to significantly activate either NFI-dependent reporter construct (Fig. 7, lane 4 versus 2 and lane 12 versus 11). VP16-mediated activation is significantly stronger than NFI-X or NFI-B-mediated activation of each NFI-dependent reporter construct (lane 10 versus 6 and 8 and lanes 15 versus 14 and 13), showing that the maximum activity of each promoter is greater than that seen with the most potent NFI activation domain used. DISCUSSION In this study, we used two model NFI-dependent promoters to assess the transcription activation properties of murine NFI proteins from each of the four mouse genes (NFI-A, -B, -C, and -X). Transfection experiments in JEG-3 cells demonstrate promoter-specific differences in activation potentials for the NFI isoforms, with NFI-X being the most potent activator of a FIG. 6. Activation potentials of NFI heterodimers. Panel A, reporter pNFI␤-gal (5 g) was transfected into JEG-3 cells with 5 g of a empty CMV vector (lane 1), or with the indicated amounts of expression vectors for NFI-C and NFI-X as indicated below the figure. Panel B, reporter construct pMMTV␤-gal (5 g) and phGR (1 g) were co-transfected into JEG-3 cells with 5 g of a empty CMV vector (lanes 1 and 2) or co-transfected with the indicated amounts of expression vectors for NFI-C and NFI-B as indicated below the figure. Cells were cultured for 24 h with or without dexamethasone (Ϫ or ϩ Dex). Normalized ␤-gal activity in panels A and B is plotted as in Fig. 2, with the bars representing the mean and range of four measurements from duplicate transfections.
simple NFI-Ad promoter and NFI-B showing maximum activation of the complex MMTV promoter (Fig. 1B, i versus ii and  Fig. 2, A versus B). Promoter-specific differences in activation potential are mediated by their COOH-terminal domains, as swapping these regions transpose their relative activation potentials (Fig. 4). NFI-X remains the strongest activator of the simple promoter which has been modified by the addition of GREs, indicating that the promoter-specific differences in relative activation between the NFI-B and NFI-X isoforms, are not due solely to the presence of GREs (Fig. 5). Coexpression of multiple NFI isoforms to assess the transactivation potentials of heterodimers, suggests that activation potentials of heterodimers are the average of the activation seen with homodimers, indicating additional mechanisms by which coexpression of different NFI proteins may regulate gene expression (Fig. 6).
Previous studies indicated that NFI proteins possess a twodomain structure consisting of highly homologous NH 2 -terminal DNA-binding domains and more divergent COOH-terminal domains (16,18,19,48). The different maximal activation potentials of the NFI isoforms used in this study appear to be mediated by the COOH-terminal domains, as deletion constructs containing only the DNA-binding domain fail to significantly activate either promoter (Fig. 2C). These findings support previous studies that suggest the transcription modulation properties of the human NFI-C and NFI-X isoforms reside in their COOH-terminal domains (18,36).
The molecular basis for differential activation by the four NFI proteins studied here is unknown. Previous studies showed that one human NFI-C isoform (CTF1) contains a region homologous to the COOH-terminal domain of RNA polymerase II (CTD), which can physically interact with TBP (33,49), TFIIB (32), and TAFII55 (50) and is critical for CTF1-mediated activation in yeast (35). However, the NFI-B and -X isoforms used here lack a CTD-like repeat, yet they are potent activators. Also, splice variants of NFI-C proteins, CTF5 and CTF7, which have stronger maximal activation potentials in yeast than CTF1, lack the CTD-like repeat (31,34). These data indicate that additional unknown protein domain(s) on the NFI proteins can substitute for the CTD repeat in promoting interactions with components of the transcription apparatus and that NFI gene products may differ in the molecular basis of their transcriptional activation. It has also been proposed, that the activation potential of NFI-C isoforms is determined by the proportion of proline residues in the activation domain (18). Our data indicate that the relative composition of proline residues is not the sole determinant of activation potential, as NFI-X with only 9% (17/192 residues) proline in its COOHterminal domain, is a strong activator of both NFI-dependent promoters. In contrast, the NFI-C isoform has a higher proline content, 13% (30/229 residues), yet has a ϳ2-fold lower maximal activation potential (Fig. 1B, i and ii). Further studies are needed to determine the mechanism(s) by which NFI proteins modulate transcription and to identify potential NFI isoformspecific coactivator proteins.
While previous studies have suggested that NFI proteins regulate the tissue-specific expression of NFI-responsive genes, little is known about the ability of NFI proteins to modulate transcription in a promoter-specific manner. Human NFI-X2 and CTF1 were both shown to be potent activators when bound at either promoter or enhancer positions on the human papilloma type 16 virus (51). In contrast, human NFI-X1 transactivated only at the promoter position (51), suggesting that NFI-binding site location could mediate promoter-specific differences in the activation potential of these isoforms. Also, Krebs et al. (52) have reported that murine NFI-A1.1 is a more potent activator of the neurotropic JC viral promoter than human CTF1, which has been shown to be a strong activator of other NFI-dependent promoters (13). However, the relative activation potentials of NFI-A1.1 and CTF1 have not been directly compared on other NFI-responsive promoters so it is unknown if these two isoforms exhibit promoter-specific differences in activation potential.
Here we report direct evidence of promoter-specific differences in NFI-dependent activation, with NFI-X mediating the strongest activation of pNFI␤-gal, and NFI-B maximally activating pMMTV␤-gal (Fig. 1B, i and ii). The molecular basis for promoter specificity is unknown, but appears to be mediated by the COOH-terminal regions of the proteins, as swapping these regions transposes relative activation potential (Fig. 4). One possible model to explain these differences may be that NFI isoforms differ in their ability to interact with other transcription factors binding to adjacent sites in the promoter. pM-MTV␤-gal differs from pNFI␤-gal by the presence of GREs and Oct-1 recognition sequences flanking the NFI-binding site, which are required for maximal hormone induction (53,54). The addition of GREs to pNFI␤-gal renders it glucocorticoid responsive, but does not alter the relative activation potentials of the NFI-B and NFI-X proteins, with NFI-X being the strongest activator of both pNFIGRE␤-gal and pNFI␤-gal (Fig. 5B). These findings suggest that the promoter-specific differences in relative activation potential are not due solely to the presence of GREs and may be determined by additional sequence element(s). In a previous study, adjacent Oct-1 and NFI-binding sites in the human papilloma virus type 16 enhancer were found to activate cooperatively by the ability of Oct-1 to stabilize NFI binding to the enhancer in epithelial cells (55). However, the composition of NFI isoforms binding to the enhancer in these cells was not determined. It will be of interest to determine if the presence of octamer-binding sites and Oct-1 can influence the relative activation potentials of NFI isoforms.
The NFI-dependent promoters used in this study appear to differ in their activation mechanism, as the MMTV promoter is in a repressed state in the absence of dexamethasone even with FIG. 7. The VP16 acidic activation domain is more potent than the most potent NFI activation domain. JEG-3 cells were transfected with either 5 g of pMMTV␤-gal and 1 g of phGR (lanes 1-10) or 5 g of pNFI␤-gal (lanes 11-15) and with 2.5 g of CMV control vector (lanes 1, 2, and 11), NFI-X-210 (lanes 3, 4, and 12), NFI-B (lanes 5, 6, and 13), NFI-X (lanes 7, 8, and 14), or the chimeric NFI-X-210-VP16 expression vector (lanes 9, 10, and 15). Cells were cultured for 24 h with or without dexamethasone (Ϫ or ϩ Dex). Normalized ␤-gal activity is plotted as in Fig. 2, with the bars representing the mean and range of four measurements from duplicate transfections. co-transfected NFI effector plasmids (Fig. 2B, ϪDex lanes). Activation depends on both NFI and GR (Fig. 2B, lanes 1-10, versus [11][12][13][14][15][16][17][18][19][20]ϩDex), which is consistent with previous reports (43,44). This requirement for both NFI and GR for MMTV promoter activation may result from the proposed nucleosomal structure of the MMTV promoter. The MMTV-LTR is arranged into phased nucleosomes in cells stably transfected with minichromosomes containing MMTV-driven reporter genes (56). The spatial distribution of the phased nucleosomes allows access of GR to the GREs, but preclude access of NFI to its cognate site (57). The proposed mechanism of activation is a two-step model where binding of the GR upon hormone treatment initiates a chromatin remodeling event which then enables NFI to bind the promoter, resulting in transcriptional activation (reviewed in Ref. 45). This model postulates an ordered requirement of GR and NFI for dexamethasone-dependent induction of the MMTV promoter. In contrast, pNFIGRE␤gal expression does not appear to be sequentially dependent on both GR and NFI, as either transcription factor alone activates expression, albeit at different levels (Fig. 5B), suggesting that a GR-dependent chromatin remodeling event may not be required for activation of pNFI-GRE␤-gal. Thus, the potentially different roles for NFI in the activation of pMMTV␤-gal and pNFIGRE␤-gal could account for the promoter-specific differences in relative activation potentials of the NFI-B and NFI-X proteins.
NFI proteins have been shown to form heterodimers both in vitro and in transfected cells (22,58). Our previous observation that NFI genes are expressed in unique but overlapping patterns during both embryonic development and in adult tissues (24) suggest that NFI heterodimers are also likely to be present in vivo. Our data suggests that heterodimers have intermediate activation potentials which are the average of the activation potentials of homodimers. This conclusion is based on the assumption that under conditions where equal amounts of two NFI isoforms are coexpressed, 25% of the molecules are homodimers of one, 50% are heterodimers, and 25% are homodimers of the other. As the NFI-B, -C, and -X proteins are of similar molecular weights, we were unable to distinguish between homo-and heterodimers by gel mobility shift analysis. However, to test the ratio of heterodimers formed between NFI-C and NFI-B, we coexpressed equal amounts of the fulllength NFI-B effector plasmid with the truncated NFI-C-240 effector plasmid, which has DNA binding and dimerization properties similar to the full-length NFI-C protein (data not shown). Under these conditions, the expected distributions of 25% NFI-B homodimers, 50% NFI-B/-C-240 heterodimers, and 25% NFI-C-240 homodimers were observed as determined by gel mobility shift analysis (data not shown).
To calculate the transactivation potentials of NFI heterodimers, we used the following model. Transfecting 1 g of the NFI-C effector plasmid activates pNFI␤-gal ϳ9-fold, while 1 g of the NFI-X plasmid results in a ϳ22-fold activation (Fig.  6A). However, when 1 g each of the NFI-C and NFI-X plasmids are transfected, NFI-C and -X homodimers each comprise 25% of NFI dimers in transfected cells. Their relative contributions to activation can be calculated as: 0.25 (4.5-fold ϩ 4.5-fold) ϭ 2.25-fold for NFI-C and 0.25 (11-fold ϩ 11-fold) ϭ 5.5-fold for NFI-X. NFI-X/-C heterodimers represent ϳ50% of the NFI population, with their relative activation potentials being: 0.50 (4.5-fold ϩ 11-fold) ϭ 7.75-fold. If this model is correct, we would expect maximal activation to be a sum of the relative contributions of homo-and heterodimers: (2.25-fold ϩ 5.5-fold ϩ 7.75-fold) ϭ 15.5-fold, which is nearly identical to the experimentally observed ϳ15-fold activation of pNFI␤-gal when 1 g each of NFI-X and NFI-C effector plasmids are coexpressed in JEG-3 cells (Fig. 6A). Similar intermediate activation levels are observed when 1 g each of NFI-B and NFI-C are coexpressed in transfections with the MMTV reporter construct (Fig. 6B), suggesting that heterodimers behave similarly on both NFI-dependent promoters. Our data suggesting that heterodimers have activation properties which are the average of homodimers is consistent with a model in which the maximal activation seen is the sum of the activation potentials of each activation domain recruited to the promoter. These findings suggest that when dimeric NFI binds to a promoter element, two copies of an activation domain are required for maximal activation and that a single copy of the activation domain leads to ϳ50% of maximal activation. In contrast, if a single copy of an activation domain present at a promoter could maximally activate and if different NFI activation domains function through independent mechanisms to activate transcription, then heterodimers containing one copy of each activation domain would have a greater maximal activation potential than the most potent homodimer. Clearly this is not the case with heterodimers of the NFI isoforms used here. It will be of interest to determine if NFI heterodimers modulate transcription in a similar manner on other NFI-dependent promoters.
It is important to note that the four NFI isoforms tested in this study represent only a small subset of the known differentially spliced isoforms of NFI and that we have only tested two NFI-dependent promoters in this study. It is therefore likely that the differences in maximum activation potential, and promoter specific differences in activation potential reported here, do not reflect the maximal difference between the transcription modulation properties of all NFI isoforms. Also, the stronger activation observed with VP16 (Fig. 7), suggests that the maximal activation values observed with the various NFI isoforms may not necessarily represent the maximal NFImediated activation of all isoforms. As the JEG-3 cells used in this study provide a near null background, it would be of particular interest to test the ability of various NFI isoforms to activate or repress the expanding number of tissue-specific and developmentally-regulated NFI-dependent promoters. The data presented here indicate that both homo-and heterodimers of NFI isoforms would likely exhibit both promoter-specific, and potentially cell-type specific modulation of such developmentally regulated promoters.