Characterization of Functional Domains within the Multifunctional Transcription Factor, YY1*

YY1 is a multifunctional transcription factor capable of either activation or repression of transcription. Using a series of mutant proteins, we have characterized do- mains responsible for activation or repression. We found that the YY1 transcriptional activation domain lies near the amino terminus and requires amino acids 16–29 and 80–100 for maximal activity. The region be- tween residues 16 and 29 has the potential to form an acidic amphipathic helix, whereas residues between 80 and 100 are rich in proline and glutamine. The YY1 repression domain lies near the carboxyl terminus and is embedded within the YY1 zinc finger region necessary for binding to DNA. Deletion of YY1 amino acids, which include zinc fingers 3 and 4, abolishes repression. How-ever, site-directed mutagenesis, progressive deletion, and internal deletion mutant analyses indicate that the normal structures of zinc fingers 3 and 4 are not re- quired for repression.

YY1 is a multifunctional transcription factor capable of either activation or repression of transcription. Using a series of mutant proteins, we have characterized domains responsible for activation or repression. We found that the YY1 transcriptional activation domain lies near the amino terminus and requires amino acids 16 -29 and 80 -100 for maximal activity. The region between residues 16 and 29 has the potential to form an acidic amphipathic helix, whereas residues between 80 and 100 are rich in proline and glutamine. The YY1 repression domain lies near the carboxyl terminus and is embedded within the YY1 zinc finger region necessary for binding to DNA. Deletion of YY1 amino acids, which include zinc fingers 3 and 4, abolishes repression. However, site-directed mutagenesis, progressive deletion, and internal deletion mutant analyses indicate that the normal structures of zinc fingers 3 and 4 are not required for repression. YY1 (variously called NF-E1, ␦, or UCRBP; Refs. [1][2][3][4], is a multifunctional transcription factor that can either activate or repress transcription. Repression has been observed in the context of the immunoglobulin 3Ј enhancer, the Moloney murine leukemia virus long terminal repeat, the adeno-associated virus P5 promoter, the skeletal ␣-actin promoter, the ␤-casein promoter, ⑀and ␥-globin genes, the serum amyloid A1 promoter, the human immunodeficiency virus promoter, and the human papilloma virus type 18 promoter (1, 4 -13; reviewed in ref. 14). In contrast YY1 can activate the c-Myc promoter, the ribosomal protein L30 and L32 promoters, and the intracisternal A-particle upstream promoter element (15)(16)(17)(18). Interestingly, YY1 can either activate or repress some promoters depending upon either promoter architecture or intracellular milieu. For instance, YY1 typically represses the adeno-associated virus P5 promoter, but can be converted into a potent transcriptional activator in the presence of adenovirus E1A protein (1). YY1 can also either activate or repress the c-Fos promoter based on the orientation or the position of a YY1 binding site within the promoter (19). Finally, YY1 can either repress or activate the human papilloma virus type 18 pro-moter depending upon the presence of an adjacent DNA sequence that binds to a distinct nuclear factor (6).
The mechanism of YY1 function is presently unclear. In some cases, YY1 binding appears to preclude the binding of activator proteins. For instance, YY1 binding competes with the binding of NF-B to an overlapping -sequence in the serum amyloid A1 promoter (13). Similarly, YY1 competes with serum response factor binding in the ␣-actin promoter, GATA-1 binding in the ⑀-globin promoter, and with binding of a lactation-associated factor in the ␤-casein promoter (7, 9 -11). YY1 function may also relate to its ability to bend DNA (19). In other cases, the function of YY1 appears to be controlled by interaction with other proteins such as adenoviral E1A, c-Myc, or switch binding protein (1,6,20). Interestingly, loss of YY1 binding sites in the human papillomavirus type 16 silencer region is implicated in the overexpression of viral oncogenes and in tumor progression (21,22). YY1 may also contribute to transcriptional regulation by serving as an initiator binding protein during the formation of the basal transcription complex. YY1 can activate transcription in vitro in the context of the adeno-associated virus P5 initiator or the cytochrome c oxidase subunit Vb initiator elements (23,24). YY1 can also support basal transcription from supercoiled plasmid templates in vitro in the presence of only TFIIB 1 and RNA polymerase II (25). The multiple functions of YY1 suggest that it can interact with many proteins and that it contains multiple functional domains. Given its diverse functional properties and its implication in oncogenesis, it would be interesting to more precisely characterize the functional domains of the YY1 protein.
We show here that deletion of only 43 amino acids from the carboxyl terminus of YY1 can convert the protein into a strong transcriptional activator. Through progressive deletion and internal deletion analyses, we show that the transcriptional activation domain of YY1 is composed of two regions requiring amino acids 16 -29 and 80 -100 for maximal activity. The transcriptional repression domain lies within the zinc finger region near the carboxyl terminus. We show that YY1 sequences between 333 and 397 are included in the repression domain. These sequences include zinc fingers 3 and 4. However, point mutation and deletion studies show that normal zinc finger structure of fingers 3 and 4 is not necessary for repression.

MATERIALS AND METHODS
Plasmid Constructions-CMVGAL-DBD was created by subcloning a BglII-XbaI fragment of pSG424 (26) containing the GAL4 DNA binding domain (amino acids 1-147) into the corresponding sites of CMV expression plasmid pCB6ϩ. The restriction sites upstream of the GAL4 sequence were destroyed by digesting with BglII and HindIII, followed by blunting with Klenow polymerase and religation. The polylinker for subcloning downstream of the GAL4 sequences is identical to that of the parent plasmid pSG424 except for a distal BamHI site contributed by the pCB6ϩ polylinker. Plasmid CMVGAL-YY1 was created by subcloning a BglII-XbaI DNA fragment from plasmid GAL4-E1 (2) containing the GAL4 DNA binding domain fused to amino acids 2-414 of YY1 into the corresponding sites of pCB6ϩ. This construct contains pCB6ϩ restriction sites upstream of the GAL4 encoding region of the plasmid and a BamHI site downstream of the insert. CMVGAL-1-102, CMVGAL-1-143, CMVGAL-1-188, CMVGAL-1-256, and CMVGAL-1-341 were generated by exonuclease III digestion using the Promega Erase-a-Base system. The parent plasmid, CMVGAL-YY1, was digested at the BamHI site 3Ј of the cDNA sequences, and the recessed 3Ј end was filled in with 40 mM each dNTP and Klenow polymerase (50 units/ml) to protect from exonuclease III digestion. A 5Ј protruding end susceptible to exonuclease III digestion was created by digestion of the XbaI site just 5Ј to the BamHI site. Progressive exonuclease III deletions through the YY1 cDNA sequence were performed using various incubation times at 30°C, followed by treatment with S1 nuclease to remove single-stranded DNA. Blunt-ended molecules were religated and transformed into competent JM109 cells. For preparation of CMVGAL-1-200, CMVGAL-1-370, and CMVGAL-1-397 an EcoRI DNA fragment containing the YY1 cDNA sequences was cut with SmaI, FspI, or HincII, respectively. The appropriate DNA fragments were gel-purified and cloned into the EcoRI-XbaI/blunted sites of plasmid CMVGAL-DBD. To prepare CMVGAL-201-414, a SmaI-EcoRI DNA fragment containing the coding sequence 201-414 was isolated by gel purification. The fragment was blunted and cloned in frame into the EcoRI-blunted site of CMVGAL-DBD. CMVGAL-333-414 was made by digesting CMVGAL-YY1#2 (see below) with HindIII and XbaI. The HindIII-XbaI fragment was gel-purified, blunted, and cloned in frame into the blunted EcoRI site of CMVGAL-DBD. Plasmids CMVGAL-1-15, CMVGAL-1-42, and CMVGAL-1-69 were generated by the polymerase chain reaction (PCR) using the GAL forward sequencing primer and appropriate reverse primers with the CMVGAL-YY1 plasmid as template. The Invitrogen PCR Optimizer kit was used to establish optimum PCR buffer conditions for these sets of primers. Standard assays contained 1x appropriate buffer, template DNA (10 ng), oligonucleotide primers (500 ng each), 200 M dNTPs, and 2.5 units of Taq polymerase in a final reaction volume of 100 l. The PCR cycle was 95°C, 1 min; 50°C, 5 min; and 70°C, 2 min for 30 cycles. Amplified DNAs were digested with EcoRI and BglII and subcloned into the EcoRI and BamHI sites of CMVGAL-DBD. YY1 internal deletions ⌬16 -29, ⌬43-53, ⌬70 -80, and ⌬70 -99 were generated by triple ligation of PCR fragments coding 5Ј and 3Ј YY1 regions with the CMVGAL-DBD vector. The 5Ј fragment was generated as indicated above and digested with EcoRI and BglII. The 3Ј fragment was generated using the CMV reverse primer and the appropriate forward primer. Amplified products were digested with BglII and XbaI and the 5Ј and 3Ј fragments were ligated with EcoRI-and XbaI-digested CMVGAL-DBD. To generate carboxylterminal truncations of the internal deletions, the mutant plasmids were digested with EcoRI and XbaI, the resulting fragment was purified by gel electrophoresis, and subsequently digested with SmaI. The 5Ј fragments containing the YY1 sequence 1-200 with the appropriate deletion mutation were purified and subcloned into the EcoRI and XbaI/blunted sites of CMVGAL-DBD. To prepare CMVGAL-YY1#2, an EcoRI fragment containing the full-length YY1 sequence except the initial starting methionine was subcloned into the EcoRI site of CMV-GAL-DBD producing an in-frame chimeric fusion protein. This plasmid differs from CMVGAL-YY1 in that it lacks the pCB6ϩ polylinker sites upstream of the GAL4 coding sequence. CMVGALYY1-C360S and CMVGALYY1⌬371-380 were generated using the Muta-Gene (Bio-Rad) oligonucleotide-mediated in vitro mutagenesis kit. The HindIII-XbaI fragment of CMVGAL-YY1#2 containing sequences encoding the last 83 amino acids of YY1 was subcloned into the phagemid vector pBluescript II/KSϩ (Stratagene). Preparation of uracil containing single-stranded DNA, annealing of the appropriate mutagenic primer, and subsequent in vitro second strand synthesis in the presence of dNTPs was performed according to the manufacturer's specifications. The resulting double-stranded plasmid DNAs were transformed into Escherichia coli HB101. The HindIII-XbaI mutagenized fragment was subcloned into the corresponding sites of plasmid CMVGAL-YY1#2, replacing the original wild-type sequences. The identities of all constructs were confirmed by dideoxynucleotide sequence analysis.
Cell Culture, Transfections, and CAT Assays-S194 cells were grown and transfected using the DEAE-dextran method as described previously (27). 20 ϫ 10 6 log phase cells were transfected with 2.5 g of reporter plasmid, 2.5 g of effector plasmid, and 1 g of a pCB6ϩ-based ␤-galactosidase expression plasmid. 3T3 cells were grown in Dulbecco's modified Eagle's high glucose medium (Life Technologies, Inc.) supplemented with 10% heat inactivated fetal bovine serum (HyClone), 2 mM L-glutamine (Life Technologies, Inc.), 100 units/ml penicillin, and 0.1 mg/ml streptomycin (Life Technologies, Inc.). Cells were transfected using the calcium phosphate coprecipitation method (28). Unless specified, cells were transfected with 5 g of reporter plasmid, 5 g of effector plasmid, and 1 g of ␤-galactosidase expression plasmid. After 48 h cellular extracts were prepared by three freeze-thaw cycles. CAT assays of these extracts were performed as described by Gorman et al. (29), using ␤-galactosidase activity to standardize for transfection efficiency.
Preparation of Nuclear Extracts-Mini-nuclear extracts were prepared from transfected 3T3 cells by the method of Schreiber et al. (30), with minor modifications. Transfections were done in triplicate using 5-10 g of effector plasmid/plate. Cells from three plates were pooled, and two-thirds of the cells (approximately 5-7 ϫ 10 6 cells) were transferred to a separate tube and used to prepare nuclear extracts. The remaining third was processed for CAT assays. Pooled cells were washed once with PBS, and resuspended in 400 l of cold Buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EGTA, 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin, 1 mM leupeptin). Following incubation on ice for 15 min, 25 l of 10% Nonidet P-40 (Sigma) was added, and each sample was vortexed for 10 s. Each sample was centrifuged for 30 s to 1 min in a microcentrifuge to collect the nuclear pellet, the supernatant was aspirated, and the pellet resuspended in 100 l of cold Buffer C (20 mM HEPES, pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 2 mM phenylmethylsulfonyl fluoride, 2 mM pepstatin, 2 mM leupeptin). Samples were incubated for 15 min on ice with frequent gentle mixing. The nuclear extracts were centrifuged for 5 min at 4°C to remove insoluble material, and the supernatant was stored at Ϫ70°C.

SDS-Polyacrylamide Gel Electrophoresis and Western
Blotting-Nuclear extracts (12 l) from cells transfected with various effector plasmids were electrophoresed on 10% SDS-polyacrylamide gels. Transfer of proteins to nitrocellulose (Scheicher & Schuell) was performed at 4°C for 1 h at 100 mV using a 25 mM Tris, 192 mM glycine, 20% (v/v) methanol, pH 8.3 transfer buffer (Bio-Rad Mini Trans-Blot). Following blocking in 5% milk (1-2 h), blots were incubated with affinity-purified rabbit polyclonal YY1 or GAL4 (Santa Cruz Biotechnology) antibody (1:500) overnight at 4°C. After washing, the blots were incubated 2 h at room temperature with a 1:1500 dilution of secondary antibody, a donkey anti-rabbit IgG (Amersham). Blots were developed using the ECL detection system (Amersham). To prepare affinity-purified rabbit polyclonal YY1 antibody, a Sepharose-YY1 affinity matrix was prepared according to the manufacturer's protocol (Pharmacia Biotech Inc.). CnBr-activated Sepharose (1 ml) was coupled to 70 g of purified bacterially produced histidine-tagged YY1 (Qiagen). Sera (10 ml) from YY1 immune rabbits was precipitated with an equal volume of 4 M (NH 4 ) 2 SO 4 . After 30 min, the precipitate was recovered by centrifugation (2000 ϫ g, 10 min). The pellet was dissolved in 2 ml of PBS and dialyzed against 2 liters of PBS overnight at 4°C. The dialyzed fraction was brought up to 10 ml with TBST (.01 M Tris-HCl, 0.15 M NaCl, 0.05% (v/v) Tween) and incubated with the affinity matrix overnight at 4°C with rocking. The matrix was washed three times for 30 min with 20 ml of TBST. Bound antibody was eluted with 1 ml of 100 mM glycine, pH 2.5. Buffer exchange was done over a NAP-5 column (Pharmacia) equilibrated with 10 volumes of PBS (0.1% bo-vine serum albumin, 0.02% sodium azide). Purified antibody was stored at Ϫ70°C.

YY1 Can Function as Either an Activator or a Repressor in 3T3
Cells-Previously, we showed that YY1 can function as either a repressor or an activator (2, 18) of transcription, depending upon the cell type in which it is expressed (3T3 versus PYS-2 cells, respectively). These results suggested that the function of YY1 may be influenced by protein interactions within differing cellular environments. To determine whether YY1 could function under certain conditions as a transcriptional activator in 3T3 cells, we performed transfection experiments with increasing amounts of YY1 expression plasmid. We reasoned that if YY1 activity is influenced by interaction with cellular proteins, differing quantities of YY1 within the cells might show different transcriptional responses because protein-protein interactions may be influenced by the quantity of YY1 within the cell. The reporter plasmid for these experiments contained four copies of the YY1 binding site upstream of the herpesvirus thymidine kinase promoter driving expression of the bacterial chloramphenicol acetyltransferase gene (YY1TKCAT). The effector plasmid contained the intact YY1 cDNA sequence driven by the CMV promoter in plasmid pCB6ϩ (CMV-YY1). Transfection of various quantities of the YY1 expression plasmid showed that low doses of YY1 indeed caused transcriptional activation in 3T3 cells, while higher doses repressed transcription (Fig. 1A). Maximal activation was 4.4-fold at 0.5 ng of effector plasmid. Transfection of 3 and 6 g of effector plasmid resulted in repression to a level of 14%. These studies show that YY1 is capable of functioning as either a transcriptional activator or a repressor in 3T3 cells.
Previously, we studied the YY1 protein linked to the DNA binding domain of a heterologous protein (GAL4) in 3T3 or PYS-2 cells (2,18). These studies showed that GALYY1 repressed transcription of a GALTKCAT reporter plasmid in 3T3 cells, while it activated transcription in PYS-2 cells. To determine whether, like the wild-type YY1 protein, GALYY1 might activate transcription at certain doses of effector plasmid in 3T3 cells, the CMV-GALYY1 plasmid was transfected at various doses into 3T3 cells. Indeed, low doses of GALYY1 effector plasmid activated GALTKCAT activity (Fig. 1B). Maximal activity (7-fold) occurred at 100 ng of effector plasmid. The higher quantity of GALYY1 effector plasmid necessary for maximal activity as compared to wild-type YY1 may reflect the presence of endogenous YY1 protein within the cells. The fact that the GALYY1 fusion protein was capable of activating transcription in 3T3 cells afforded us an assay system for identifying functional domains of the YY1 protein.
Identification of the YY1 Activation Domain-To determine the region of YY1 responsible for transcriptional activation, 11 effector plasmids were constructed consisting of progressive carboxyl-terminal truncations of YY1 linked to the GAL4 DNA binding domain. These constructs were tested for their ability to activate GALTKCAT in 3T3 cells. Our initial experiments indicated that removal of the carboxyl-terminal 83 amino acids converted GALYY1 into a constitutive activator even at high doses of effector plasmid. 2 Therefore, we tested function of each GALYY1 deletion construct at high doses of effector plasmid (5 g). The CAT activity for each construct relative to the GAL4 basal activity is summarized in Fig. 2 acids from the carboxyl terminus (construct 2; GAL-1-397) did not convert GALYY1 into a constitutive activator. However, removal of an additional 27 amino acids (construct 3; GAL-1-370) showed a 4-fold increase in CAT activity over the GAL4YY1 full-length construct (GAL-1-414) and a 9-fold increase over basal CAT activity. Carboxyl-terminal truncations to amino acids 341, 256, 200, 188, 143, or 102 fused to GAL4 (constructs 4 -9) produced proteins that were potent activators of CAT transcription, ranging from 12-to 33-fold activity above basal CAT activity. CAT activity diminished with progressive deletions to amino acids 69 or 42 and returned to basal level upon deletion to amino acid 15.
Similar results were obtained in S194 plasmacytoma cells, although there were some quantitative differences (Fig. 2). For instance, the full-length GALYY1 protein was a more potent activator in S194 cells as compared to 3T3 cells, and the various deletions did not have as dramatic of an activation over the full-length protein. This can be seen by comparing the -fold induction over wild-type YY1 for each construct in 3T3 versus S194 cells (Fig. 3). Whether this represents differences in the cellular milieu of 3T3 and S194 cells is uncertain. However, others have shown that YY1 can be a potent activator in plasmacytoma cells (31). In any case, similar to the case in 3T3 cells, truncation to amino acids 69 and 42 reduced activity in S194 cells, and all activation was lost upon deletion to amino acid 15 (Fig. 2).
To assure that the protein products of the various deletion constructs were being made within the transfected cells, nuclear extracts were prepared from 3T3 cells transfected with each deletion construct. These nuclear extracts were evaluated by EMSA with a GAL4 DNA binding site probe. Representative EMSAs of the carboxyl-terminal deletion constructs are shown in Fig. 4A. Carboxyl-terminal deletions containing YY1 amino acids 15-256 (with the exception of the GAL-1-42 construct) showed similar amounts of probe shifted and complexes of the appropriate size based on the predicted size of the fusion proteins (Fig. 4A, lanes 1-9). Numerous attempts have failed to demonstrate a complex with nuclear extracts isolated from 3T3 cells transfected with construct GAL-1-42 (lane 2). However, this construct must be expressed within the cells because it consistently yielded modest activation of transcription (Fig. 2). Somewhat unusual EMSA complexes were observed with carboxyl-terminal deletion constructs GAL-1-341, GAL-1-370, and GAL-1-397, as well as the full-length fusion protein GAL-1-414 (lanes 10 and 12-14). To assure that these fusion proteins were synthesized in transfected cells, nuclear extracts were subjected to Western blot analysis with polyclonal YY1 antibody (Fig. 4B). This analysis showed that the GAL-1-341, GAL-1-370, GAL-1-397, and GAL-1-414 (wild-type) proteins are indeed expressed in transfected cells. The endogenous YY1 protein was also detected in these extracts and served as a loading control.
EMSA studies were also performed with nuclear extracts prepared from transfected S194 cells. However, no complexes were observed perhaps due to the lower transfection efficiency of plasmacytoma cells.
The Activation Domain Contains Multiple Functional Segments-Results obtained with our progressive carboxyl-terminal deletion mutants indicated that the transcriptional activation domain of YY1 resides within its amino terminus, approximately between amino acids 16 and 100. Several amino acid sequence motifs reside within this region: a segment with the potential to form an amphipathic acidic helix between amino acids 16 and 29, a stretch of 11 consecutive acidic residues between amino acids 43 and 53, a stretch of 11 consecutive histidines between amino acids 70 and 80, and an area enriched in proline and glutamine between amino acids 80 and 100. Internal deletion mutants were prepared that deleted each of these regions either individually or collectively. These mutants were prepared in the context of the YY1 protein truncated at amino acid 200 and linked to the GAL4 DNA binding domain. This truncation increases YY1 activation potential in 3T3 cells (see Fig. 2) thereby facilitating the identification of sequences important for activation. Results are shown in Fig. 5, and values are represented as the percent of CAT activity observed with GAL-1-200. As expected, deletion of amino acids 16 -99 caused a great decrease in CAT activity (5 and 14% of maximal activity in 3T3 and S194 cells, respectively). Deletion of amino acids 16 -80 decreased activity as well, but to only 40 and 62% of maximal CAT activity in the two cell lines, respectively. This suggested that sequences between amino acids 80 and 99 contribute to activation function. However, deletion of amino acids 70 -99 had minimal effects on activation (97 and 91%, respectively). Interestingly, deletion of the consecutive histidine (⌬70 -80) or acid (⌬43-53) stretches had no deleterious effect on transactivation. However, deletion of the amphipathic helix region between amino acids 16 and 29 reduced activation to 33 and 26% of maximal CAT activity in 3T3 and S194 cells, respectively. The above results suggest that the YY1 activation domain is bipartite. Sequences between amino acids 16 and 29 are clearly important for activity. Sequences between 80 and 100 contribute to maximal activity since the ⌬16 -80 construct shows higher activity than the ⌬16 -100 construct. However, this region is somewhat redundant because deletion of residues 70 -99 alone did not decrease transcriptional activity.
To assure that the internal deletion proteins were synthesized in vivo, nuclear extracts were made from the transfected cells and subjected to EMSA with the GAL4 DNA binding probe (Fig. 6). All constructs produced the expected complexes indicating that a loss in transcriptional activity was not simply due to a lack of efficient production of mutant proteins (lanes 1-8).
Characterization of the Repression Domain-Experiments with our carboxyl-terminal mutants indicated that YY1 is a strong activator of transcription when the zinc finger region is deleted. This suggested that sequences important for transcriptional repression reside within the carboxyl-terminal region of YY1. To determine the effect of the carboxyl-terminal domain alone on transcription, the GAL4 DNA binding domain was fused to the last 214 amino acids of YY1 to produce GAL-201-414. This construct contains all four YY1 zinc fingers. This fusion protein repressed transcription in 3T3 cells to a level of 31% (Fig. 7A). To more narrowly define the YY1 repression domain, we prepared a construct containing amino acids 333-414. This construct repressed expression to 39% (Fig. 7A). Western blots of tranfected cell nuclear extracts with anti-GAL4 antisera (because of the limited YY1 sequences in construct GAL-333-414) indicated that the fusion proteins were expressed in vivo (Fig. 7B). These data indicate that YY1 repressive sequences lie between amino acids 333 and 414.
Our carboxyl-terminal truncation experiments (Fig. 2) indicated that there was a transition from a weak transcriptional activator to a strong transcriptional activator when the third zinc finger was disrupted. Thus, the GAL-1-397 mutant protein, which has a disrupted fourth zinc finger, is relatively inactive, while the GAL-1-370 mutant protein, which has a disrupted third zinc finger, is 4 times more active when compared to the full-length GAL-YY1 fusion. To determine if the third zinc finger is critical in possibly masking strong transcriptional activity mediated by the amino-terminal activation domain, two additional GAL4-YY1 fusion mutants were pre- pared. One mutant has a 10-amino acid deletion removing amino acids 371-380 (GAL-⌬371-380), which encompasses the second half of the third zinc finger including the histidines. The other mutant creates a point mutation at amino acid 360 (GAL-C360S), changing the second cysteine of the third zinc finger to a serine. A stable third zinc finger is not formed by either of these mutations, because these mutant proteins are incapable of binding to DNA (data not shown). Neither internal zinc finger mutation mimicked the strong transcriptional activity of the carboxyl truncated third zinc finger mutant GAL-1-370 (Fig. 7C). The latter is on average 4-fold more active as a transcriptional activator than the full-length GAL-1-414 (Fig.  2). The serine-360 mutant was approximately half as active as full-length YY1, while the 10-amino acid deletion activated transcription 1.5-2-fold.
To assure that the fusion proteins were made in vivo, we performed Western blots on nuclear extracts prepared from the transfected cells. The two third zinc finger mutants (GAL-C360S and GAL-⌬371-380) were readily detectable on Western blots with YY1 antibody and were expressed at an abundance similar to the GAL-1-370 truncation (Fig. 7D).

DISCUSSION
The YY1 Activation Domain-We have used a series of progressive carboxyl-terminal deletions and a series of internal deletions to characterize the YY1 transcriptional activation domain. The progressive deletions showed that the activation domain resides within the amino-terminal 100 amino acids of YY1, consistent with the results of Lee et al. (32,33). This region of YY1 contains sequences with the potential to form an acidic amphipathic helix (residues 16 -29), as well as 11 consecutive acid residues (43-53), 11 consecutive histidines (70 -80), and a region rich in glutamine and proline (81-100). Interestingly, deletion of the consecutive acid stretch or the consecutive histidine stretch had no affect on transcriptional activation by YY1 (Fig. 5), indicating that these sequences are not required for transcription. On the contrary, deletion of amino acids 16 -29 caused significant decreases in transcription and residues 80 -100 contribute to maximal activation potential. Therefore, the YY1 transcriptional activation domain appears to be bipartite and includes sequences with the potential to form an acidic amphipathic helix and sequences rich in proline and glutamine. These features are commonly observed in activation domains of other transcription factors (34). It is interesting that the Xenopus laevis YY1 protein lacks the consecutive histidine stretch but contains both amino acid segments that we determined to be involved in transcriptional activation (35). It will be interesting to determine whether the Xenopus YY1 homologue can function as a transcriptional activator similar to the human form.
The YY1 Repression Domain-Previously, the YY1 repression domain was shown to reside within the carboxyl-terminal region of the protein (1, 32, 33). Our studies in 3T3 fibroblasts are consistent with these results and more precisely identify sequences important for repression. We found that deletion of the carboxyl-terminal 17 amino acids (to residue 397) was not sufficient to convert YY1 into a strong activator. Deletion of an additional 27 amino acids to residue 370 resulted in strong activation in 3T3 cells. These results indicate that sequences within the 27 amino acids between residues 370 and 397 inhibit YY1 activation potential. Consistent with these results, fusion of YY1 sequences 201-414 or 333-414 to the GAL4 DNA binding domain resulted in repression of a GAL4 responsive reporter. These results suggest that the repression domain lies within the 65 amino acids between residues 333 and 397.
Our deletion studies showed that the YY1 repression domain includes sequences in the region of the third and fourth zinc fingers. Carboxyl-terminal deletion to residue 397, which does not disrupt the repression domain, interrupts the fourth zinc finger indicating that function of an intact fourth zinc finger is not necessary for repression. To determine whether repressor function requires the intact structure of the third zinc finger, we prepared mutations C360S and ⌬371-380, which disrupt the structure of the third zinc finger. Neither the C360S nor the ⌬371-380 mutants mimicked the strong transcriptional activity of the 370 deletion mutant. These results indicate that repression does not absolutely require the normal structure of either the third or the fourth zinc fingers.
While zinc fingers are typically considered DNA binding motifs, it is interesting to note that some zinc finger regions have been implicated in protein-protein interactions. For instance, the zinc finger of the adenovirus E1A protein has been implicated in interaction with transcription factors (36) and is necessary for physical interaction with TATA-binding protein (37). Similarly, the zinc finger region of ATF2 can interact with the leucine zipper region of cAMP response element-binding protein, whereas the Sp1 zinc fingers can interact with RelA (38,39). A short region containing a zinc finger in TFIIB is necessary for efficient interaction with TFIIF (40), and the zinc finger-containing LIM domain has been implicated in proteinprotein interactions (41). In addition, the GATA-1 zinc finger region can self-associate (42) and can interact with other proteins (43).
Protein Interactions with YY1-Our effector plasmid titration experiments (Fig. 1) suggest that YY1 function may be influenced by protein interactions. Similar concentration-dependent function of transcription factors has been observed (58,59) and is believed to result from protein-protein interactions (reviewed in Ref. 60). YY1 has been observed to physically interact with Sp1, and this interaction apparently is important for stimulation of transcription (61,62). In addition, a twohybrid screen in yeast showed that c-Myc can interact with YY1 and that repressive affects of YY1 can be relieved by c-Myc expression. Interestingly, a similar two-hybrid screen identified the nucleolar phosphoprotein B23, which can also relieve YY1-induced transcriptional repression (63). The physical interaction between YY1 and Sp1 requires YY1 amino acid residues 260 -331, and YY1 residues 201-343 are required for interaction with c-Myc (20). Neither of these protein interaction sequences maps to the YY1 repression domain that we have characterized here. However, it is interesting that the adenovirus E1A protein, which can relieve repression mediated by YY1 (1) can physically interact with YY1 residues 332-414 (64). Recently, it was also determined that E1A-mediated relief of YY1 repression requires the cAMP response element-binding protein-related factor, p300 (65). This protein is also capable of physically interacting with YY1 (65). Further studies will be necessary to identify additional cellular proteins that interact with YY1 and influence the mechanisms of YY1 repression and activation.