Alternatively spliced isoforms of TFII-I. Complex formation, nuclear translocation, and differential gene regulation.

TFII-I is a multifunctional phosphoprotein with roles in transcription and signal transduction. Here we report characterization of three additional alternatively spliced isoforms of TFII-I. Employing isoform-specific antibodies, we show that the isoforms form a stable complex in vivo preferentially in the nucleus compared with the cytoplasm. We further show that both homomeric and heteromeric interactions are possible and that the heteromeric interactions between a wild type and a nuclear localization-deficient mutant result in nuclear translocation of the complex, leading us to postulate that complex formation might aid in nuclear translocation. In functional assays all four isoforms individually bind to DNA and transactivate reporter genes to a similar extent. However, although co-expression of different TFII-I isoforms leads to enhanced basal activity, it results in attenuated signal responsive activity. Thus, TFII-I might differentially regulate its target genes via complex or subcomplex formation.

TFII-I is a multifunctional phosphoprotein with roles in transcription and signal transduction. Here we report characterization of three additional alternatively spliced isoforms of TFII-I. Employing isoform-specific antibodies, we show that the isoforms form a stable complex in vivo preferentially in the nucleus compared with the cytoplasm. We further show that both homomeric and heteromeric interactions are possible and that the heteromeric interactions between a wild type and a nuclear localization-deficient mutant result in nuclear translocation of the complex, leading us to postulate that complex formation might aid in nuclear translocation. In functional assays all four isoforms individually bind to DNA and transactivate reporter genes to a similar extent. However, although co-expression of different TFII-I isoforms leads to enhanced basal activity, it results in attenuated signal responsive activity. Thus, TFII-I might differentially regulate its target genes via complex or subcomplex formation.
TFII-I is an intriguing transcription factor, with broad biochemical and biological activities and may be involved in several genetic disorders (1)(2)(3)(4)(5)(6)(7)(8). TFII-I functions through the Inr element both in vitro (1, 4, 9 -11) and in vivo (1,5,11,12). It also functions through upstream regulatory elements in the adenovirus major late promoter and in c-fos promoter in vivo (2,3,12). Based on its unique physical and functional interactions at both the Inr element and upstream regulatory sites, TFII-I is postulated to be a novel transcriptional cofactor that integrates signals from the regulatory components to the basal machinery (1,12). We have recently shown that TFII-I is phosphorylated at both serine and tyrosine residues and that tyrosine phosphorylation of TFII-I is critical for its transcriptional properties (5). Equally interesting is the observation that a variety of extracellular signals mediating through cell surface receptors, including growth factor receptors, lead to enhanced tyrosine phosphorylation and increased transcriptional activity of TFII-I (3,5,7). TFII-I is a ubiquitous protein partitioned between the cytoplasm and nucleus (7). In the B cell cytoplasm a significant fraction of TFII-I is associated constitutively with Bruton's tyrosine kinase (7), mutations in which lead to Xlinked immune deficiency in humans and mice (14 -16). TFII-I is tyrosine-phosphorylated by Bruton's tyrosine kinase in vitro (13), and upon immunoglobulin receptor cross-linking in B cells it is released from Bruton's tyrosine kinase to enter the nucleus (7). These observations suggest that TFII-I is downstream of several signal transduction pathways and may be a mediator linking signal-responsive activator complexes to the general transcription machinery, perhaps in a cell type-specific fashion.
Recent genetic data suggest that TFII-I belongs to a family of proteins, each having the I-repeat first identified in the founding member TFII-I (8,12,17,18). Interestingly, both TFII-I and the related protein WBSC11 (18) have been mapped to the breakpoint regions of the 7q11. 23 Williams-Beuren syndrome deletion (8,18). Furthermore, genetic analysis by Francke and co-workers suggest the presence of several alternatively spliced isoforms of TFII-I both in humans and in mice (8). Based on these predicted protein sequences (8), we cloned TFII-I isoforms by using a PCR based strategy (see "Experimental Procedures") and ectopically expressed and biochemically analyzed them. Here we show that in addition to the 957-amino acid form of TFII-I (⌬), three other alternatively spliced isoforms exist in human, henceforth referred to as ␣ (977 amino acids), ␤ (978 amino acids), and ␥ (998 amino acids). Compared with the ⌬-isoform, the ␣-isoform contains an additional 20 amino acids (encoded by exon A), the ␤-isoform contains an additional 21 amino acids (encoded by exon B), and the ␥-isoform, which arises by the presence of both exons A and B, contains 41 additional amino acids (8). All four isoforms, when ectopically and individually expressed in COS cells, exhibit similar subcellular distribution. Use of isoform specific antibodies allowed us to demonstrate stable complex formation between the various isoforms either when co-expressed ectopically or present endogenously in eukaryotic cells. The endogenous complex was preferentially located in the nucleus compared with the cytoplasm, and the co-expression of wild type isoform with a nuclear localization-deficient mutant resulted in the localization of the complex to the nucleus. All four isoforms, when expressed in recombinant forms, exhibit similar DNA binding characteristics and bind to both the Inr element of the V␤ promoter and the upstream regulatory site overlapping the serum response element (SRE) 1 of the c-fos promoter (2,3). However, although the isoforms possess similar homomeric transactivation potentials, heteromeric complex formation leads to differential activation of reporter genes. Taken together, these data suggest that the complex formation is a means by which the nuclear localization and the transcriptional activity of TFII-I are regulated.
cDNA Cloning of TFII-I Isoforms-To construct the isoforms of TFII-I, the sequences of exon A and exon B (8) were inserted into pET11-d-II-I (12) and into pEBGII-I (1) by PCR-based insertional mutagenesis. To construct TFII-I␣, the nucleotide sequence of exon A was inserted into pET11-d-II-I or into pEBGII-I. The nucleotide sequence of exon A was PCR-amplified by using the primers 1 (5Ј-GTGGATCCAC-CATGGGCAG-3Ј) and primer 2 (5Ј-GCTGTTTTTCATCAACATCATCA-GTTTCAGAAGGGCCTGCTTGAATGTTATA-3Ј) (reaction 1). A separate PCR reaction was done by using primer 3 (5Ј-GATGTTGATGAA-AAACAGCCCCTATCGAAGCCTTTGCAAGGAAGCCACCATTCTTCA-3Ј) and primer 4 (5Ј-GTGGATCCACCATGGGCAG-3Ј) (reaction 2). The templates used in the above reactions were either pET11-d-II-I or pEBGII-I. In the second round of PCR, 0.25 g each from reactions 1 and 2 were mixed and used as templates for further amplification using primers 1 and 4. The final product was gel isolated, digested with BamHI and SnaB1, and then ligated with pET11-d-II-I or with pEB-GII-I. By using the same strategy, exon B was also inserted into pET11d-II-I and into pEBGII-I. The nucleotide sequence of exon B were PCR-amplified by using the primer 1 (5Ј-GTGGATCCACCATGGGCA-G-3Ј) and primer 2 (5Ј-TCCTCACTTGTTTCTGAAGGGACATGTTGA-GTAGAATCTTCTGCTGGTGCTT-3Ј) (reaction 1). A separate PCR reaction was done by using primer 3 (5Ј-CAGAAACAAGTGAGGACCCT-GAAGTTGAGGTGACTATTGAAGATGATGATTATTCT-3Ј) and primer 4 (5Ј-GTGGATCCACCATGGGCAG-3Ј) (reaction 2). In a second round of PCR, products from the two reactions were mixed and further amplified by using primers 1 and 4. The final product was gel-isolated, digested with BamHI and SnaB1, and then ligated with pET11-d-II-I or with pEBGII-I. The isoform having both exon A and exon B (TFII-I␥) was constructed by inserting the nucleotide sequence of exon A into TFII-I␤ as described above, except the template used in the PCR reactions were TFII-I␤. The authenticity of the final constructs was confirmed by DNA sequencing.
Construction of Tag-less and Green Fluorescence Protein (GFP)-fused TFII-I Isoforms-To remove the hexahistidine tag from the N-terminal end, the region between the amino acids 1 and 134 was PCR-amplified by using the primers (5Ј-TAAGGATCCATGGCCCAAGTTGCA-3Ј) and (5Ј-TGGTACAGGTACCACTGTGGA-3Ј). The PCR product was gel-purified and digested with BamHI and Nsi1 and then ligated with either pEBGII-I⌬, pEBGII-I␣, pEBGII-I␤, or pEBGII-I␥. The positive clones of the isoforms lacking hexahistidine tag were then digested with BamHI and NotI, and the entire cDNAs were subcloned into the BamHI and NotI sites of pEBB plasmid (a gift from Bruce Mayor) lacking the GST tag. Tag-less TFII-I isoforms having the GFP tag at their C-terminal end (pEBBGFPII-I) was constructed by ligating the GFP cDNA sequence from the plasmid pEGFP-N1 (CLONTECH). The stop codons of TFII-I isoforms were removed, and an AccI site was created by PCRbased mutagenesis. The region between the amino acids 590 and the stop codon was PCR-amplified by using the primers (5Ј-CGTCCGCGG-GAGTAATAAA-3Ј) and (5Ј-GGGCGGGCGTCGACCACGTGGG-3Ј). The amplified product was gel-purified, digested with SacII and NotI, and then ligated to the TFII-I isoforms. These constructs were then digested with AccI and NotI and ligated with the GFP cDNA isolated from pEGP-N1 by digesting with the same enzymes. The authenticity of the constructs was confirmed by DNA sequencing.
Construction of Deletion Mutants of TFII-I-The deletion mutants of TFII-I were generated by the following PCR strategy. In the first PCR reaction, a restriction enzyme-specific sequence was used as a forward primer (primer 1), and a TFII-I-specific sequence, just upstream of the deleted sequence, was used as a reverse primer (primer 2). In the second reaction, a TFII-I-specific sequence, just downstream of the deleted sequence, was used as a forward primer (primer 3), and a restriction enzyme-specific sequence was used as a reverse primer (primer 4). The product (0.25 g) of each reaction was mixed, and the second round of PCR was performed using the primers 1 and 4. The PCR product was isolated and digested with the two restriction enzymes, the recognition sequences of which were used as the primers 1 and 4 and ligated to digested TFII-I-expressing vector, pEBGII-I or pEBBGFPII-I. The mutants were confirmed by DNA sequencing. The specific sequences of these primers and the corresponding restriction enzymes are detailed below.
Isoform-specific Antibodies-Antibodies specific to TFII-I␣ and TFII-I␤ isoforms were raised in rabbits (Research Genetics), employing the synthetic peptide DVDEKQPLSKPLQ (␣) and QHVPSETSEDPE-VEV (␤), respectively. The total IgG fraction from these antisera was purified by using protein A-Sepharose according to the method supplied by the manufacturer (Roche Molecular Biochemicals) and used for immnunostaining.
Purification of Bacterially Expressed Recombinant Isoforms of TFII-I-To express TFII-I isoforms in prokaryotes, bacterial cells (BL21trxB (DE3)pLysS (Novagen)) containing the expression plasmids were grown overnight in 3 ml of LB media containing chloramphenicol and ampicillin antibiotics. One ml of overnight culture was added to TBM 9 media (10 g of Bacto tryptone, 4 g of glucose, 5 g of NaCl, 1 mM MgSO 4 , 1 g of NH 4 Cl, 3 g of KH 2 PO 4 , 6 g of Na 2 HPO 4 ⅐7H 2 O/liter of media), grown to an A 600 of ϳ0.3, and induced with 50 mM isopropyl-1thio-␤-D-galactopyranoside for 4 h at 30°C. After 3 h, the cells were harvested by centrifugation at 5000 rpm for 30 min.
The bacterial pellet was snap-frozen on dry-ice and resuspended in 30 ml of BC100 buffer (20 mM Tris-HCl (pH 7.9), 100 mM KCl, 10% glycerol) containing 0.1% Nonidet P-40 and protease inhibitor mixture Complete TM EDTA-free (Roche Molecular Biochemicals). The lysate was clarified by centrifugation for 10 min at 12,000 rpm at 4°C, and the His 6 -tagged TFII-I isoforms were purified over a 1-ml TALON column (CLONTECH) as described for purification of eukaryotically overexpressed protein.
Electrophoretic Mobility Shift Analysis (EMSA)-The EMSA reactions in Fig. 3, panel A, was performed with an Inr probe (Ϫ28 to ϩ12) derived from the V␤ promoter (10), and panel B was done with a c-fos SRE probe 5Ј-AGCTTAACAGGATGTCCATATTAGGACATCTG-3Ј (7). The EMSA was done as described previously (4).
Immunoprecipitation-For immunoprecipitation, nuclear and cytoplasmic extracts were prepared from Ramos cells according to published procedures (29). Cells were harvested and resuspended in 400 l of hypotonic buffer (20 mM Tris (pH 7.9), 5 mM NaF, 2 mM Na 3 VO 4 , and 1 mM Na 2 P 4 O 7 , and protease inhibitor mixture). The resuspended cells were incubated in ice for 15 min and lysed by 5 passages through a 25-gauge needle. The nuclei were collected by centrifugation at 5,000 rpm for 10 min at 4°C. Nuclear proteins were extracted by resuspending the nuclei in 200 l of the lysis buffer containing 420 mM NaCl and 25% glycerol and incubating in ice for 30 min. For the immunoprecipitation assay, either 37.5 g of nuclear or 75 g of cytoplasmic extract was incubated with anti-TFII-I␤ or with preimmune serum in a final volume of 400 l of BC100 at 4°C for 20 min. 10 l of protein A-Sepharose (50% slurry) was then added and incubated further for 1.5 h at 4°C, after which the immune complexes were washed 4 times with buffer BC100 containing 0.1% Triton X-100, and 4ϫ SDS sample buffer was added. The beads were heated to 100°C for 3 min, analyzed by SDS-PAGE, and immunoblotted successively with anti-TFII-I␣, anti-TFII-I␤, and anti-TFII-I antibody (4).
GST Pull-down Assay-Whole cell extracts (200 g) from COS7 cells cotransfected with TFII-I isoforms were subjected to precipitation with glutathione-Sepharose (35 l, 1:1 slurry; Sigma) for 2 h with rocking at 4°C. At the end of the incubation the beads were washed 3 times with 1 ml of buffer BC100 containing 0.1% Triton X-100. After the final wash, 4ϫ SDS sample buffer was added to the beads and the beads were heated to 100°C for 3 min, analyzed by SDS-PAGE, and immunoblotted with anti-GFP antibody (CLONTECH). The blots were then stripped to remove the anti-GFP immunecomplex (5) and reprobed with anti-GST antibody.
Transient Transfection and Immunofluorescence-COS7 cells were transfected with either the GFP construct alone or cotransfected with GFP-tagged TFII-I isoforms as described (1). 30 h post-transfection, cells were fixed with 4% paraformaldehyde and prepared for immunofluorescence. For detection of the splice variants by immunostaining, the protein A-Sepharose-purified TFII-I␣ or TFII-I␤ antibody were used at a dilution of 1:2500 in conjunction with Alexa TM 594 goat anti rabbit IgG (HϩL) (Molecular Probes) at a dilution of 1:20,000. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI) dye (Sigma). Immunofluorescence was detected using a fluorescence microscope (Nikon, E400).
Reporter Assays-Transient transfection and luciferase reporter assays were done essentially as described (1) with the following modifications. One day before transfection COS7 cells at 80 to 90% confluency were trypsinized, and each well of 6-well plates was seeded with 4 -5 ϫ 10 4 cells. Transfection was done with LipofectAMINE (Life Technologies, Inc.). In an Eppendorf tube, either 600 ng of V␤ (3) or c-fos (3) reporter plasmids with or without different TFII-I isoform expression constructs (400 ng when transfected individually and 200 ng each when co-transfected in combinations) and 35 ng of Renilla luciferase plasmid (pRL-TK; Promega Corp.) were mixed. To each transfection reaction, 400 ng of empty vector plasmid (pEBG) was also added to normalize the total amount of DNA. The final volumes were adjusted to 100 l with DMEM. In a separate tube, 6 l of LipofectAMINE was mixed with 94 l of DMEM. The plasmid-containing medium was mixed with LipofectAMINE-containing medium and incubated at room temperature for 45 min. At the end of the incubation period cells were washed two times with DMEM, and 800 l of additional DMEM was added to the cells. The DNA-lipid complex was then added to the cells and incubated overnight in a CO 2 incubator at 37°C. After 12-14 h of incubation, 1 ml of DMEM containing 20% fetal bovine serum was added to each of the wells, and the incubation was continued for another 8 h. In reporter assays having the V␤ promoter, the media was exchanged at the end of incubation with normal media (DMEM containing 10% fetal bovine serum) and incubated for an additional 12-14 h. In experiments with the c-fos reporter plasmid, the transfection media was exchanged with DMEM, and the cells were then serum-starved for 12-14 h. Finally, the cells were stimulated with human EGF (Sigma) at a final concentration of 25 ng/ml for 4 h. The cells were washed twice in PBS and lysed, and the luciferase activities were determined according to manufacture's protocol (dual luciferase assay; Promega Corp).

Analysis of Alternatively Spliced Isoforms of TFII-I by Isoform-specific Antibodies: Preferential Complex Formation in the
Nucleus-Both by Western blot analysis and by immunoprecipitation assays using an anti-TFII-I antibody, TFII-I protein always appeared as a doublet (120 and 128 kDa) in either human or murine cells (4, 6, 7). Limit digestion and microsequencing of the biochemically isolated 128-kDa form from HeLa cells yielded two peptide sequences that matched com-pletely with the published TFII-I sequences (data not shown), suggesting that the 128-kDa isoform arises either by posttranslational modifications of the 120-kDa form or by alternative splicing. However, because the genomic analysis suggested that there exist three additional isoforms of TFII-I that would arise by alternative splicing (8), we pursued this latter idea. To test the existence of these putative isoforms and to biochemically characterize them further, we employed a PCR-based cloning strategy starting with the published sequence of TFII-I as the template (12). We successfully cloned the three additional isoforms of TFII-I and named them ␣ (977 amino acids), ␤ (978 amino acids), and ␥ (998 amino acids) (also see Ref. 8). The shortest form (957 amino acids) henceforth in this paper will be referred to as the ⌬-isoform. The schematic diagram of these isoforms is shown in Fig. 1A.
Antipeptide polyclonal antibodies were prepared against the peptide ␣ and peptide ␤ ("Experimental Procedures"). Note that because both ␣ and ␤ sequences are present in the ␥-isoform, no unique sequence was present in this isoform that could be used to raise an antibody. Hence all the antibodies recognized this isoform. The alternatively spliced isoforms were expressed in COS cells to permit their biochemical characterization and to test the specificity of the antibodies. All four of the isoforms were subcloned into pEBG vector (1,19) and expressed in COS cells as GST-hexahistidine-tagged fusion proteins. The expression of each protein was tested after purification over the TALON column by Western blot analysis using an anti-TFII-I peptide polyclonal antibody that recognizes all four isoforms (Fig. 1B, left panel). Having ensured that the expression of these proteins was similar and the loading was equivalent, we stripped the blot of immune complexes and reprobed with the anti-␣ peptide antibody (Fig. 1B, middle panel). In parallel, an identical blot was probed with an anti-␤ peptide antibody (Fig. 1B, right panel). As expected, although anti-␣ antibody recognized the ␣and the ␥-isoforms but not the ⌬and ␤-isoforms, the anti-␤ antibody recognized the ␤and the ␥-isoforms but not the ⌬and ␣-isoforms, demonstrating isoform/exon specificity of these antibodies. Subsequently, endogenous existence of the ␣and ␤-isoforms was confirmed by the isoform specific antibodies (data not shown).
Because all the isoforms were preferentially found in the nucleus, we wished to test whether they translocate independently or as heteromeric complexes. We further rationalized that if they interact to form a complex then we should be able to immunoprecipitate the complex with any of the isoform-specific antibodies. We chose Ramos cells since the cytoplasmic TFII-I isoforms were readily detectable in this cell line. Because the level of TFII-I is more in nuclear extract compared with cytoplasmic extract, the extracts were first normalized for the ␤-isoform by Western blot (Fig. 1C, lanes 1 and 2). The anti-␤ antibody was then employed to immunoprecipitate the putative complex from normalized amounts of cytoplasmic and nuclear extracts. The blot was first probed with an anti-␣ antibody (upper panel), followed by anti-␤ (middle panel) and anti-TFII-I (lower panel). Although the preimmune control was incapable of bringing down any of the isoforms (lane 3), the anti-␤ antibody was capable of bringing down not only the ␤but also the ␣and the ⌬-isoforms from both cytoplasmic and nuclear extracts (Fig. 1C, lanes 4 and 5), suggesting that these isoforms exist as a complex. However, although we started out with identical amounts of cytoplasmic and nuclear ␣ and ␤, the amount of precipitated nuclear ␣ and ␤ were 2-fold higher than their cytoplasmic counterparts. Furthermore, the amount of ⌬ co-precipitated with ␣ and ␤ was sub-stoichiometric, suggesting that the vast excess of ⌬-isoform in this cell line is not in a complex with either ␣or ␤-isoforms. Similar results were obtained from different human and murine cell lines, suggesting that the complex formation is a cell type-and speciesindependent phenomenon (data not shown). Taken together these results suggest that either the complex formation leads to preferential nuclear localization, that the complex was more stable in the nucleus, or that the isoforms were more readily available to form a complex in the nucleus compared with the cytoplasm. Furthermore, at present we do not know whether the uncomplexed cytoplasmic isoforms remain as monomers, homo-oligomers, or heteromers with other proteins.

Identification of Nuclear Localization Signal (NLS) in TFII-I (⌬) Isoform: in Vivo Interactions (Heteromerization) of Various
Isoforms-TFII-I has been shown to be distributed between the nucleus and cytoplasm, although the nuclear fraction predominates in most cells (7). To better understand the subcellular localization of TFII-I and its isoforms, we first determined the consensus localization signals in TFII-I (⌬) isoform by computer analysis using the PSORT program (20). Two potential NLSs were identified; the first one, termed NLS1, was present between amino acids 278 and 284, whereas the second one, termed NLS2, was present between amino acids 629 and 632. To test the role of these sequences, we deleted them one at a time and analyzed the expression and subcellular localization of TFII-I in COS cells after tagging with GFP (21). Although GFP alone when co-expressed with the vector control did not exhibit any specific localization (expressed equally in both cytoplasm and nucleus) ( Fig. 2A, extreme left panel), both wild type TFII-I⌬ (middle left panel) and the NLS2 mutant, ⌬NLS2 (extreme right panel), exhibited predominant nuclear localization with very small amounts in the cytoplasm. In contrast the NLS1 mutant, ⌬NLS1 (middle right panel), showed exclusively cytoplasmic fluorescence. Hence we conclude that NLS1 is the predominant and functional NLS under the conditions tested and in this cell type. This is also confirmed by generation of a double mutant of both NLS1 and NLS2 that showed exclusively cytoplasmic localization (data not shown).
Given the fact that the NLS1 is present in all isoforms, as a first approximation it was reasonable to assume that all the isoforms would have similar subcellular localization. However, because the extra exons were present very close to NLS1, it was necessary to determine whether they might affect the subcellular localization of these isoforms. To test this, each of the isoforms were GFP-tagged and expressed in COS cells (Fig. 2B). Moreover, in each case, the cells were also stained with DAPI to visualize the nucleus (purple blue fluorescence, middle panels). Each of the TFII-I isoforms were predominantly found in the nucleus under this condition (GFP, top panels). This conclusion is further strengthened by the superimposition of DAPI and GFP images (bottom panels). However, a small fraction of each of these isoforms was expressed in the cytoplasm as well (not shown). Regardless of the exact amount of cytoplasmic and nuclear TFII-I, which appear to differ depending on the cell type and source of cells (data not shown), similar nucleo-cytoplasmic distribution was observed upon the ectopic expression of each of the isoforms. These results strongly suggest that the exons A and B do not significantly alter the nuclear localization of these proteins.
Next we wanted to determine whether these isoforms interact with themselves (homomerization) and with each other (heteromerization) under these conditions. We rationalized that since the ⌬NLS1 TFII-I remains in the cytoplasm, it might be used as a bait to test heteromeric interactions and alterations in subcellular localization. Thus, if the mutant interacts with any of the wild type isoforms and exhibits green fluorescence in the nucleus, then the nuclear translocation is dominant. Alternatively, if the cytoplasmic retention is dominant, then the mutant will retain the wild type isoforms in the cytoplasm. To visualize and/or co-localize the partner protein, we employed the ␣and ␤-isoform-specific antibodies. As expected, ⌬NLS1, when co-expressed with the vector control, remained in the cytoplasm (Fig. 3, panels A and M), and the anti-␤ antibody did not show significant levels of endogenous ␤ isoform (panel E). However, upon co-expression with either ␣ -, ␤-, or ␥-isoform (visualized in red, panels F-H), the mutant migrated to the nucleus in each case, as clearly evidenced by the nuclear green fluorescence (compare panels A with B-D),  1 and 2), the amount of cytoplasmic (50 g) and nuclear (25 g) extracts were adjusted for the TFII-I␤ isoform (lanes 1 and 2). Cytoplasmic (75 g) or nuclear (37.5 g) extracts were subjected to immunoprecipitation with a pre-immune serum (lane 3) or with an anti-TFII-I␤ antibody (lanes 4 and 5). The immunecomplexes were probed first with an anti-TFII-I␣ antibody (top panel) and reprobed with an anti-TFII-I␤ antibody (middle panel) followed by anti-TFII-I antibody (bottom panel). The positions of the ␣, ␤-(128 kDa), ␣Ј-(120 kDa), and the ⌬-(120 kDa) isoforms are indicated. The ␣Ј band could arise either due to degradation of ␣ or could be a related isoform. demonstrating that nuclear migration of ⌬NLS1 is correlated to its interaction with the wild type isoforms. Superimposition of GFP, Alexa 594, and DAPI (compare panels M with N-P) further confirmed that the isoforms, upon co-expression, interact with the ⌬-isoform, and such interactions lead to nuclear residency.
Homo-and Heteromerization between Various Isoforms-Although the above experiments showed heteromerization between the ⌬-and other isoforms, they did not address whether homomerization is also possible and whether the ␣-, ␤-, and ␥-isoforms would interact with each other. To address this issue, we co-expressed either GFP-tagged version of ␣-, ␤-, ␥-, or ⌬-isoforms together with either a GST-tagged ⌬-isoform or its truncation mutant p70 (1) as baits (Fig. 4). A GST pull-down assay was performed and, after the Western blot was visualized with an anti-GFP antibody, the blot was stripped and reprobed with an anti-GST antibody. As a control, only the GST protein was used as a bait and did not pull down the ␤-isoform (lane 1). Identical results were also obtained with other isoforms (data not shown). The specificity of the anti-GFP antibody was evident, as it failed to recognize GST-tagged ⌬-isoform (lane 2). Interaction of truncated p70 and wild type ⌬-isoform (lane 3) indicated that homomerization does happen.
In accordance with the in vivo data (Fig. 1C), interaction between ⌬and ␣-(lane 4) or ⌬and ␤-(lane 5) or ⌬and ␥isoforms (lane 6) occurred. In addition, interactions between the ␣and the ␤-isoform (lane 7) indicated that heteromerization between isoforms can take place. Densitometric scanning revealed that the amount of the interacting partners (GFPtagged) is proportional to the amount of the pulled down proteins (GST-tagged). Hence when the amount of the various GST-tagged proteins were equalized, the amount of co-precipitating GFP-tagged interacting partners became nearly identical in all cases. Thus, although both homo-and heteromeric interactions within the isoforms occur, the extent of these interactions is similar.
Differential Regulation of Target Genes by Complex Formation-To unambiguously test the DNA binding activities of TFII-I isoforms, we expressed them individually in bacteria. For this purpose each of the cDNAs was subcloned into a hexahistidine-containing pET11-d vector (12). These proteins were expressed in bacteria and affinity-purified over a TALON column (Fig. 5) by virtue of their hexahistidine tag. The purified proteins were visualized by an anti-TFII-I polyclonal antibody that recognizes all the isoforms (Fig. 5C). The DNA binding ability of these expressed proteins was tested by using FIG. 2. Identification of functional NLS in TFII-I. A, NLS1 sequence is indispensable for nuclear translocation of TFII-I. Deletion constructs of two putative nuclear localization sequences of TFII-I were made and expressed in COS7 cells as GFP fusion proteins (⌬NLS1 and ⌬NLS2). COS7 cells were transfected with 600 ng of expression plasmid encoding either E-GFP alone (Vector, far left panels), the C-terminally GFP-tagged versions of TFII-I⌬ wild type (middle left panels), ⌬NLS1 (middle right panels), or ⌬NLS2 (far right panels). 30 h after transfection, cells were fixed with 4% paraformaldehyde, and nuclei were stained with DAPI (middle panels). Superimposition (Merge) of GFP and DAPI staining are shown in the bottom panels. Images were obtained by using a Nikon fluorescent microscope (E400) with a 100ϫ objective. B, subcellular localization of TFII-I isoforms. Subcellular localization of TFII-I isoforms were determined by transfecting COS7 cells with 600 ng of plasmids encoding either GFP-tagged TFII-I␣ (left), TFII-I␤ (middle), or TFII-I␥ (right) isoforms. The transfected cells were fixed, and the nuclei were stained with DAPI (middle panels). Superimposition (Merge) of GFP and DAPI staining are shown in the bottom panels. a radiolabeled probe containing either the Inr sequence derived from the V␤ promoter (Ref. 3 and Fig. 5A) or a TFII-I binding site overlapping the SRE derived from the c-fos promoter (Ref. 3 and Fig. 5B). With each isoform, a roughly similar mobility shifted complex was observed that was ablated by the affinity-purified TFII-I antibody, confirming the authenticity of the protein binding to the probe. When measured by densitometric scanning, the extent of V␤ Inr binding by TFII-I ␣-, ␤-, and ␥-isoforms relative to ⌬were as follows: ␣ (3.65), ␤ (0.64), and ␥ (1.42). Similar results were also obtained with the SRE probe, and the specificity of their binding was also very similar to that of TFII-I⌬ (data not shown).
We then tested the transactivation potentials of these isoforms both individually and in combinations on both V␤ and c-fos promoters (Fig. 6). The V␤ promoter was chosen to study the basal transcriptional activity of the TFII-I isoforms and (3), and the c-fos promoter was chosen to study its signal-responsive activation potentials, since TFII-I hyperactivates the c-fos promoter in response to serum and various growth factor stimulation (2, 3). All the individual isoforms showed comparable transcriptional activities (3-3.5-fold) on the V␤ promoter compared with the vector only control (Fig. 6A, compare lanes 1  with 2-5). However, the basal activity was further enhanced by about 2-fold when the isoforms were co-expressed. The ␣ϩ⌬ combination (lane 6) was the most active, but the combinations of ⌬ϩ␤ (lane 7) and ␣ϩ␤ (lane 8) was also more active than the individual isoforms. The comparable expression of the isoforms, either singly or in combination, was demonstrated by Western blot analysis, although a slightly less expression of ⌬ϩ␤ combination was seen (Fig. 6A, bottom panel). In contrast to the V␤ promoter, both the individual and the combinations of isoforms showed quite different activities on the c-fos promoter in response to EGF stimulation (Fig. 6B). The ⌬-isoform was nearly 2-fold more active compared with the ␥ isoform, whereas ␣ and ␤ isoforms were 65% as active as the ⌬-isoform. More importantly, the combination of isoforms showed 20 -40% less activity compared with the individual isoforms. The basal activity (EGF) did not change significantly in any lane. Once again the Western blot analysis showed that the expression levels of these transfected proteins were comparable either in the presence or in the absence of EGF (Fig. 6B, bottom panel). The functional assays were done such that the ectopic expression of isoforms was maintained at an equivalent level. If the assays were normalized with respect to the relative DNA binding activities of the isoforms, the differences between the promoters might be even greater. Nevertheless, these functional data suggest that homomer versus heteromer formation differentially regulates the transcriptional activity of TFII-I on different promoters.

DISCUSSION
The signals generated outside a cell are transduced to the nucleus through a series of complicated biochemical steps, ultimately resulting in spatial and/or temporal activation of specific sets of genes. Thus, there must exist specific proteins to direct signal transduction pathways to cell type-specific genes and, thereby, provide a molecular link between signal transduction and developmental/differentiation programs in a given cell. Transcription factors play a critical role in development and differentiation in general and often serve as links between signal transduction (origin) and cell type-specific gene activation (end point). Included in this group of transcription factors are the Rel family of factors (22,23), signal transducers and activators of transcription (STAT) family of factors (24), and the nuclear factor of activated T cells (NFAT) family of factors (25). Recent data from our laboratory and others suggest that TFII-I is functionally included in this group of factors that link signal transduction events to transcription (2,3,5,7,13).
Why are there four isoforms of TFII-I apparently with similar DNA binding and transcriptional activities? Do they indeed have redundant functions, or might it be that they have cell type-specific expression and function? Of the four isoforms, the ␥-isoform is perhaps expressed predominantly, if not exclusively, in neuronal cells (8). In our analysis of lymphoid and other non-neuronal cells, we have not observed the expression of the ␥-isoform clearly by Western blot analysis. However, in our preliminary Western blot analysis of PC12 cells, a protein of approximately 150 kDa (corresponding to apparent molecular mass of the ␥-isoform) cross-reacts with the anti-TFII-I antibody (data not shown). It remains to be proven whether this band indeed corresponds to the ␥-isoform. Our preliminary data also indicate the presence of at least two other cell typespecific isoforms. 2 We have consistently seen in Western blot analysis of several cell lines and primary cells a doublet corresponding to 120 and 128 kDa (4,6,7). In retrospect, given the reactivity of the antibody and the nearly identical molecular weights and the migration patterns of the ␣and ␤-isoforms, it was impossible to determine whether the 128-kDa isoform represented only one of the isoforms or a combination of both. Now with the generation of the isoform-specific antibodies, we are in a position to distinguish between the expression patterns of ␣versus the ␤-isoforms in various cell types.
We show that the different isoforms of TFII-I interact with each other when co-expressed ectopically or when present endogenously. Interestingly, interactions between different isoforms are also seen with Ikaros protein (26). However, in our case, we do not yet know the precise stoichiometry of these interactions, i.e. whether they form dimers or higher oligomers. Thus, we have maintained the term homomer for interactions between identical isoforms and heteromer for interactions between different isoforms. Regardless of the exact stoichiometry, heteromeric complexes are preferentially found inside the nucleus relative to the cytoplasm. This may be due to the fact that either complex formation facilitates nuclear entry or the heteromeric complex is more stable in the nucleus. However, we do not yet know whether both homo-and heteromeric complexes translocate to the nucleus to the same rate and extent. Moreover, although the "uncomplexed" forms appear to be more in the cytoplasm, from our current data we cannot conclude that homomerization is facilitated in this compartment since the uncomplexed form may not be in a homomeric state and may be in a heteromeric complex with another cytoplasmic component (Fig. 7). In addition, the stability and/or the extent of homomerization of different isoforms may also be different.
Apart from nuclear translocation, which both homomeric and heteromeric complexes can potentially undergo, we surmised that perhaps heteromeric complexes can mediate differential 2 V. Cheriyath and A. L. Roy, unpublished information. gene regulation. Consistent with our expectations, we observed that different combinations of isoforms activate different reporter genes to varying extents. Thus, although the ⌬ϩ␣ het-eromeric combination activates the V␤ basal promoter 2-fold better than either of the individual isoforms, the same combination was 20% less effective than the ⌬and marginally more  (1). The result shown is an average of three independent experiments done in triplicate. Western blot analysis of lysates from transfection assays show comparable expression of TFII-I isoforms either when they are transfected alone (bottom panel, lanes [3][4][5] or when they are cotransfected (lanes 6 -8) either in the absence (Ϫ) or in the presence (ϩ) of hEGF.
FIG. 7 Model: differential regulation of target genes via complex formation. TFII-I isoforms predominantly exist as homomers or heteromers (in association with other proteins) in the cytoplasm. The interaction of other proteins (X, Y, and Z) with isoforms might prevent them from forming either homomeric or heteromeric complexes among the isoforms. It is possible that a small fraction of isoforms can exist as heteromeric complex (with each other) in the cytoplasm. Compared with the cytoplasm, TFII-I exists predominantly in heteromeric complexes in the nucleus, and such nuclear entry might also be regulated via complex formation. In addition, a subpopulation of homomers, whose level may vary depending on the cell type or species, might also co-exist in the nucleus. This simplified model suggests that the nuclear subcomplexes regulate different genes (depicted as A, B, C, and D) perhaps through differential protein-protein and/or protein-DNA interactions. effective than the ␣-isoform in mediating signal-induced activation of the c-fos promoter. Because TFII-I is downstream of a variety of extracellular signals (7), an obvious possibility raised by our result is that the different complexes or subcomplexes mediate different signals and, consequently, activate different sets of genes (Fig. 6). It is also likely that complex formation may modulate the amplitude or duration of the signals via differential phosphorylation of the isoforms, although we have not yet assessed the phosphorylation status of the isoforms in the complex in response to distinct signals. Such a mechanism to regulate nuclear import and subsequent function has been described for a variety of proteins (reviewed in Ref. 27). Might there be different subcomplexes in different cell types or species? We believe so. The ␣-isoform is not predicted to be expressed in mouse cells (28), and by Western blot analysis we do not see the existence of the ␣-isoform in mouse cell lines (data not shown). Thus, in addition to differential regulation of genes in a given cell type, complex formation by TFII-I isoforms might also mediate cell type or species-specific gene regulation. Further analysis is under way to resolve these issues.