Antagonism between members of the CNC-bZIP family and the immediate-early protein IE2 of human cytomegalovirus.

The HCMV IE2 protein negatively autoregulates its own expression as well as represses the transactivation activity of p53. Using the repression domain of IE2 as bait in the yeast two-hybrid system, Nrf1 and Nrf2, members of the CNC-bZIP family, were found to be IE2-interacting proteins. Residues 331-448 encompassing the DNA-binding and the dimerization domains of Nrf1 are sufficient for the interaction. The interaction was further confirmed in vitro by a glutathione S-transferase pull-down assay and in vivo by co-immunoprecipitation. In transient transfection studies, transcription driven by six copies of an NF-E2 site or by chimeric proteins between the DNA-binding domain of LexA and members of the CNC-bZIP family is repressed by IE2. Importantly, the DNA binding activity of the Nrf1/MafK heterodimer is not impeded by IE2. In a parallel study, CNC-bZIP factors attenuate the negative autoregulation of IE2. The attenuation could be explained by the finding that Nrf1 functions alone and synergistically with its heterodimerization partner, MafK, in inhibiting the DNA binding activity of IE2. Taken together, these results demonstrate the existence of antagonism between members of the CNC-bZIP family and IE2.

Human cytomegalovirus (HCMV), 1 a member of the ␤ subgroup of herpesviruses, has a double-stranded DNA genome of 229,354 base pairs with a potential to encode more than 200 proteins (1). A number of immediate-early (IE) proteins of HCMV are made following entry of the virus into cells (2). Among them the IE2 86-kDa protein (referred hereafter as IE2) is most studied. IE2 appears to be a promiscuous transactivator of viral and cellular gene expression (3)(4)(5). IE2 stimulates transcription by interacting with general transcription factors and/or gene-specific factors (6 -9). In addition, IE2 negatively autoregulates its own expression by binding to a short nucleotide sequence, termed the cis repression signal (CRS), located immediately downstream of the TATA box of the HCMV major immediate early promoter (MIEP) (10,11). Binding of IE2 to CRS prevents the recruitment of RNA polymerase II to the promoter via steric hindrance (12). Thus, IE2-mediated, CRS-dependent transcriptional repression is both passive and position-dependent. The efficiency of IE2-mediated repression of MIEP seems to be cell-dependent (4,10,13), strongly suggesting that the IE2 activity may be modulated by post-translational modification and/or by interaction with cellular proteins. Indeed, phosphorylation has been shown to be one of these factors (14).
Furthermore, recent studies demonstrate that IE2 can be converted into a trans-repressor following binding to p53 (15,16). A transcriptional repression domain has been mapped to the C terminus of IE2 (16), which is also required for its transcriptional activation, DNA binding, and interaction with CREB, c-Jun, RB, TBP, and TFIIB (6 -9). The DNA-binding and the repression domains of IE2 seem to be overlapping, because an IE2 mutant devoid of DNA binding activity also loses its repression activity (16).
A subset of bZIP proteins sharing a conserved structural domain, called the CNC domain, were recently cloned (17)(18)(19)(20). These CNC-bZIP proteins are involved in the developmental processes as well as participate in the regulation of gene expression in the adults (21)(22)(23)(24)(25)(26). Members of the CNC-bZIP family bind specifically to a DNA sequence, the NF-E2 binding site, through heterodimer formation with small Maf proteins (MafK, MafF, and MafG) (21,(27)(28)(29). The C terminus of CNC-bZIP proteins contains a region with heptad repeats of leucines proceeded by a basic region that classifies the proteins as members of the bZIP family of transcription factors (30). A high degree of similarity is found in the bZIP and in the region, so-called CNC homology domain, immediate N-terminal to the basic domain among Nrf1, Nrf2, Nrf3, NF-E2, and CNC, the Drosophila segmentation protein cap and collar (17,20). As is usually the case among related transcription factors, CNC-bZIP factors are completely different in the activation domain.
To further our knowledge about how host cells modulate the autoregulation and the transcriptional repression activities of IE2, we isolated several cellular factors, including a member of the CNC-bZIP family. In this report, we characterized the functional interaction between members of the CNC-bZIP family and IE2.

MATERIALS AND METHODS
Plasmid Constructions-Plasmids pSG424, pET11d, pIE2, pSV2CAT pBXL1, pLexA-Vp16, and pL 6 E1bCAT have been described (16,(31)(32)(33). pGAL4DB-IE2C was constructed by inserting a DNA fragment encoding residues 290 -579 of IE2 between the EcoRI and the BamHI sites of pAS2-1 (CLONTECH). pGAL4DB-IE2C(H446L) was constructed similarly by inserting a corresponding DNA fragment from an IE2 mutant with histidine replaced by leucine at residue 446. pGAL4AD-NF-E2 was cloned by inserting a DNA fragment encoding residues 1-373 of the NF-E2 p45 between the NcoI and BamHI sites of pACT2 (CLONTECH). pGAL4AD-Nrf1 was cloned by inserting a DNA fragment encoding residues 1-447 of Nrf1 between the NcoI and the SacI sites of pACT2. pGAL4AD-Nrf2 was constructed by inserting a DNA fragment encoding residues 1-589 of Nrf2 between the BamHI and the SacI sites of pACT2. pGAL4AD-Nrf1(1-330), pGAL4AD-Nrf1(331-447), pGAL4AD-Nrf1(331-394), pGAL4AD-Nrf1(331-356), and pGAL4AD-Nrf1(395-447) were cloned by inserting DNA fragments encoding the corresponding peptide of Nrf1 residues 1-330, 331-447, 331-394, 331-356, and 395-447, respectively, between the BamHI and the EcoRI sites of pACT2, respectively. pGEM4-NF-E2 and pGEM4-Nrf1 were cloned by inserting, respectively, the NF-E2 p45 and Nrf1 cDNAs between the EcoRI and the XbaI sites of pGEM4 (Promega). pGEM3-Nrf2 was constructed by inserting the Nrf2 cDNA between the BamHI and the SacI sites of pGEM3 (Promega). Plasmids for the expression of histidine-tagged IE2 and MafK were constructed by inserting the corresponding cDNA fragment between the NdeI and the BamHI sites of pET11d. The plasmid for the expression of histidine-tagged Nrf1 was constructed by inserting the Nrf1 cDNA between the NdeI and the XhoI sites of pET11d. pNrf1 was constructed by inserting the Nrf1 cDNA between the HindIII and the XbaI sites of pSG424. pNrf2 was constructed by inserting the Nrf2 cDNA between the HindIII and the SacI sites of pSG424. pSVTATACAT was constructed by replacing the BamHI/XbaI fragment of pE1bCAT with an oligonucleotide consisting of the TATA box (5Ј-ACTAATTTTTTTTATTTATGCAG-3Ј) of SV40 promoter. pN6CAT was constructed by inserting six copies of an NF-E2 binding site (5Ј-GCACAGCAATGCTGAGTCATGATGAGTCATGCTG-3Ј) into the BamHI site of pSVTATACAT. pRK5F-LexA-NF-E2, pRK5F-LexA-Nrf1, and pRK5F-LexA-Nrf2 were cloned by inserting a corresponding DNA fragment encoding fusion protein LexA-NF-E2, LexA-Nrf1, and LexA-Nrf2, respectively, between the SmaI and the XbaI FIG. 2. The C terminus of Nrf1 is required and sufficient for the interaction with IE2. Left column, the GAL4 DNA-binding domain hybrid, GAL4DB-IE2C. Middle column, the GAL4 activation domain hybrid, GAL4AD-Nrf1, and its derivatives. Right column, yeast colony color after transformation; the relative color is in parentheses. The Nrf1 fragment fused to GAL4 activation domain is indicated on the right of the diagrams. The relative positions of CNC (stripes), basic (vertical dashes), and ZIP (diamonds) domains of Nrf1 are also depicted.
sites of pRK5F (34), which adds a FLAG epitope to the C terminus of the fusion proteins.
Yeast Two-hybrid Screen-Because the expression of the full-length IE2 fusion protein, GAL4DB-IE2, was detrimental to yeast, fusion proteins of C-terminal IE2, GAL4DB-IE2C and GAL4DB-IE2C(H446L), respectively, were used to screen a human lymphocyte activation domain cDNA library (MATCHMAKER two-hybrid system; CLON-TECH). Approximately 9 ϫ 10 5 transformants were screened according to the manufacturer's protocol. cDNAs from the activation domain library encoding proteins that interacted with wild-type but not with mutant IE2 were isolated and sequenced.
In Vitro Transcription and Translation-In vitro transcription and translation were performed with the TNT SP6 Quick kit (Promega) according to the manufacturer's protocol. [ 35 S]Methionine (Amersham Pharmacia Biotech) was included in the reaction so that the synthesized proteins were labeled. The templates were pGEM4-NF-E2, pGEM4-Nrf1, and pGEM3-Nrf2.
GST Fusion Proteins and Pull-down Assay-The expression and purification of GST and GST-IE2 were previously described (16). The ligand concentrations, using bovine serum albumin as a standard, were 1.97 and 0.41 mg/ml of resin for GST and GST-IE2, respectively. Aliquots (25 l) of the GST and GST-IE2 were incubated for 4 h at 4°C with in vitro translated, [ 35 S]methionine-labeled NF-E2, Nrf1, and Nrf2. After being washed with buffer D (35), bound proteins were eluted from beads with buffer D containing 0.1 M glutathione and analyzed by electrophoresis on a 10% SDS-polyacrylamide gel.
Antibody Preparation and Immunoprecipitation-Polyclonal antibodies against Nrf1, MafK, and IE2 were produced in rabbits according to standard protocols, using purified histidine-tagged Nrf1, MafK and IE2 proteins as antigens, respectively. Protein A-Sepharose beads (CL-4B, Sigma) were pre-equilibrated overnight in IP buffer (10% glycerol, 50 mM HEPES-KOH, pH 7.3, 100 mM potassium glutamate, 0.5 mM DTT, 6 mM magnesium acetate, 1 mM EGTA, 0.1% Nonidet P-40, and 0.5 mg/ml bovine serum albumin). Bead slurry (40 l) was spun in a microcentrifuge, and the beads were resuspended in 40 l of fresh IP buffer. Cell lysate (700 l) and 2 l of anti-Nrf1 sera were added to the beads, mixed and rotated for 4 h at 4°C. The reactions were then spun in a microcentrifuge, and the supernatant was removed. Beads were washed with 1 ml of IP buffer three times. The beads were then resuspended in 40 l of loading dye. The presence of IE2 in the immunoprecipitates was detected by Western blotting using antibodies against IE2 after being resolved on a 10% SDS-polyacrylamide gel.
Western Immunoblotting-Proteins from the immunoprecipitates or from crude extracts (containing approximately 50 g of proteins) of transfected H1299 cells were boiled in a sample buffer (125 mM Tris-HCl, pH 6.8, 100 mM dithiothreitol, 2% SDS, 20% glycerol, 0.005% bromphenol blue) for 5 min and then loaded onto a 10% SDS-polyacrylamide gel. Following electrophoresis, proteins were transferred to an Immobilon membrane (Millipore). IE2 and LexA derivatives were detected by antibodies against IE2 and the FLAG epitope (Kodak), respectively, with the ECL system (Amersham Pharmacia Biotech) according to the manufacturer's instructions.
Cell Culture, Transfection, and CAT Assay-H1299 and WI38 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. Approximately 1 ϫ 10 6 cells were seeded 12 h before transfection. Calcium phosphate-mediated DNA transfections (for H1299 cells) were performed as described (16), and LipofectAMINEmediated transfections (for WI38 cells) were performed according to the manufacturer's suggestion (Life Technologies, Inc.). CAT activity was measured and quantified according to the method described by Carey et al. (36).
Purification of Histidine-tagged Proteins and Electrophoretic Mobility Shift Assay-The His-tagged IE2, Nrf1, and MafK were expressed in and purified from Escherichia coli JM109(DE3) strain according to the manufacturer's suggestions (Qiagen). Electrophoretic mobility shift assay was performed as described previously (33,37), except that a 32 P end-labeled double-stranded oligonucleotide (5Ј-GCACAGCAATGCT-GAGTCATGATGAGTCATGCTG-3Ј) containing an NF-E2 binding site (19) or a 32 P end-labeled double-stranded CRS oligonucleotide (5Ј-AGCTCGATACAATAAACGCCGGCC-3Ј) (33) was used as the probe. When indicated, 0.1 g of individual recombinant histidine-tagged proteins and approximately 10 g of serum proteins were used.

Isolation and Identification of Members of the CNC-bZIP
Family as IE2-interacting Proteins-In a search for proteins that specifically interact with the autoregulation and repression domain of IE2, we screened a human lymphocyte MATCH-MAKER activation domain library (CLONTECH) using a chimeric protein composed of the C terminus (residues 290 -579) of IE2 fused to the GAL4 DNA binding as bait (GAL4DB-IE2C). Fusion protein GAL4DB-IE2C(H446L) containing a mutant IE2 domain lacking both the autoregulation and the repression activities (16) was also included to eliminate false positive clones. Approximately 9 ϫ 10 5 colonies were screened with the two-hybrid system. Among the several clones identified and sequenced, three of them contained cDNAs that encode different versions of the bZIP region of Nrf1 (17).
To examine whether full-length Nrf1 interacts with IE2 and to further determine whether members of the CNC-bZIP family all behave similarly with regard to IE2, the interaction between several CNC-bZIP proteins and IE2 was measured. As summarized in Fig. 1, Nrf1, consistent with data obtained with the truncated versions of Nrf1 during the screening, interacted quite well with IE2 (row 2). Nrf2 also interacted with IE2, although its affinity toward IE2 was lower than that of Nrf1 (compare row 3 with row 2). Interestingly, NF-E2, another member of the CNC-bZIP family, appeared not to interact with IE2 (row 1). Control experiments confirmed the specificity of the Nrf1-IE2 and Nrf2-IE2 interactions: the introduction of plasmid pGAL4DB-IE2C alone into yeast produced no colony (row 6). Similarly, no colony was observed upon co-transformation of pGAL4DB with either pGAL4AD-Nrf1 (row 7) or pGAL4AD-Nrf2 (row 8). Although co-transformation of pGAL4DB-IE2C(H446L) with pGAL4AD-Nrf1 (row 4) or pGAL4AD-Nrf2 (row 5) did generate some yeast colonies, no LacZ expression was detected. Thus, two conclusions could be drawn from the above experiments. First, not all the members of the CNC-bZIP family interacted with IE2. Second, the bZIP domain of Nrf1 was involved in the binding to IE2. Further supports for the conclusions were provided below (Figs. 2, 3A,  and 4B).
Mapping the IE2 Interaction Domain of Nrf1-Because Nrf1 exhibited relatively high affinity toward IE2, we set out to map FIG. 4. The CNC-bZIP factor-mediated transcriptional activation is repressed by IE2. A, transcription driven by NF-E2 sites is repressed by IE2. Transient transfections were performed in WI38 cells (lanes 1-6) or H1299 cells (lanes 7-12), using pSVTATACAT ( lanes  1, 2, 7, and 8), pN6CAT (lanes 3, 4, 9, and  10), or pSV2CAT (lanes 5, 6, 11, and 12) as the reporter. pSVTATACAT is driven by the SV40 TATA box; pSV2CAT is driven by the SV40 promoter/enhancer. pN6CAT was constructed by inserting six copies of an NF-E2 site before the SV40 TATA box of pSVTATACAT. The absence (Ϫ) or presence (ϩ) of IE2 is indicated above each track of the autoradiogram. The CAT activity of the reporter in the absence of IE2 was set at 100. The relative CAT activity (Ϯ standard deviation) was 100 (Ϯ 0), 91.0 (Ϯ 11.7), 100 (Ϯ 0), 31.5 (Ϯ 3. the IE2-binding domain of Nrf1 with a yeast two-hybrid interaction assay using GAL4AD-Nrf1 and a panel of its deletion constructs as bait. As summarized in Fig. 2, the N-terminal region spanning residues 1-330 alone failed to interact with IE2 (row 2). In contrast, residues 331-447 of Nrf1, including the basic region, the zipper region and the most C-terminal region proceeded by the zipper, constituted the minimal IE2binding domain (compare row 3 with rows 4 -6). Obviously, the N terminus of Nrf1, including the CNC region, did not interact directly with IE2 (row 2). However, it might contribute indirectly to the interaction, because full-length Nrf1 interacted with IE2 stronger than the C terminus of Nrf1 (compare row 1 with row 3). Interestingly, residues 395-447 encompassing the region C-terminal to the zipper of Nrf1 were absolutely required for the interaction with IE2. Because this region is completely divergent among CNC-bZIP factors, it was postulated that the requirement of the most C-terminal region of Nrf1 for binding to IE2 might account for the observed differences in the affinity toward IE2 among Nrf1, Nrf2, and NF-E2.
Interaction of CNC-bZIP Factors with IE2 in Vitro and in Vivo-To test whether CNC-bZIP transcription factors interact with IE2 in vitro, an affinity matrix consisting of the GST-IE2 fusion protein as a ligand was prepared. Full-length, radiola-beled NF-E2, Nrf1, and Nrf2 were incubated with the matrix, and bound CNC-bZIP proteins were analyzed by SDS-PAGE. Fig. 3A shows that, consistent with data obtained with the yeast two-hybrid system (Fig. 1), both Nrf1 (lane 6) and Nrf2 (lane 9) bound to GST-IE2, whereas NF-E2 (lane 3) failed to be retained under the same experimental conditions. Moreover, Nrf1 had higher affinity toward IE2 than Nrf2 did (compare lane 6 with lane 9), also in agreement with the results obtained from the yeast two-hybrid experiments (Fig. 1).
To determine whether CNC-bZIP factors associate with IE2 in human cells, a plasmid expressing IE2 was transfected into H1299 cells that had been selected as a system to study the repression activity of IE2 (15,16,33) for co-immunoprecipitation. Further, the transfection efficiency was about 30 -40% for H1299 cells at our hands (data not shown). Thus, H1299 cells seemed to provide a desirable system for the experiments. Cell extracts were prepared from the transfected cells and then immunoprecipitated with an antibody against Nrf1. IE2 in immunoprecipitates was detected by Western blot analysis using an antibody against IE2. Although expressed to similar levels (Fig. 3B, panel II, lanes 2 and 4), IE2 was only coprecipitated by an anti-Nrf1 antibody but not by a preimmune serum (panel I, compare lane 4 with lane 2), indicating that Nrf1 and IE2 appeared to co-exist in a complex in vivo.
Repression of the Transactivation Activity of CNC-bZIP Factors by IE2-Because CNC-bZIP factors interacted with IE2 in vivo (Fig. 3), it was important to ask whether the transcriptional activity of CNC-bZIP factors is modulated by IE2. To address this question, two complementary experiments were performed. First, the effect of IE2 on transcription driven by six copies of an NF-E2 site was measured. As shown in Fig. 4A 2 and 8 with lanes 1 and 7) or a substitution of the NF-E2 sites with the SV40 promoter/enhancer (compare lanes 6 and 12 with lanes 5 and 11) rendered the resultant reporters unresponsive to IE2. In other words, the IE2-mediated repression was dependent on the presence of the NF-E2 sites. Second, because multiple CNC-bZIP factors can interact with an NF-E2 site, the identity of CNC-bZIP factors involved in the repression was determined by measuring the effect of IE2 on transcription activated by LexA-CNC-bZIP fusion proteins. As shown in Fig. 4B, the transactivation activity of LexA-Nrf1 and LexA-Nrf2 was largely reduced in the presence of IE2 in both WI38 and H1299 cells (compare lanes 6, 8, 16, and 18 with lanes 5, 7, 15, and 17). The reduction was specific, because IE2 had little effect toward LexA-VP16 (compare lanes 10 and 20 with lanes 9 and 19). Moreover, although LexA-NF-E2 exhibited a relatively low transactivation activity (lanes 3, 4, 13, and 14), an autoradiogram of longer exposure revealed that IE2 had little effect on LexA-NF-E2 (data not shown). A control experiment demonstrated that IE2 had little effect on the expression of LexA-CNC-bZIP fusion proteins (Fig. 4C, compare lanes 2, 4, and 6 with lanes 3, 5, and  7), eliminating the possibility that a difference in the concentration of fusion proteins might account for the observed IE2mediated repression (Fig. 4B). As expected, IE2(H446L) did not inhibit transcription driven either by the NF-E2 sites or by the LexA-CNC-bZIP fusion proteins (data not shown). Thus, on the basis of data obtained from the aforementioned experiments, it was concluded that a direct interaction between members of the CNC-bZIP family and IE2 was required for the observed repression.
Retention of the DNA Binding Activity of Nrf1/MafK in the Presence of IE2-It is known that small Maf proteins lack a FIG. 5. The DNA binding activity of Nrf1/MafK is not impeded by IE2. A 32 P end-labeled double-stranded oligonucleotide containing an NF-E2 site was used as a probe in the bandshift assay. The presence (ϩ) or absence (Ϫ) of individual protein is indicated above each track of the autoradiogram. Ab, ␣Nrf1, ␣MafK, and ␣IE2 are abbreviations for antibody, anti-Nrf1 antibody, anti-MafK antibody, and anti-IE2 antibody, respectively. DNA-protein complexes were separated from the probe by eletrophoresis on a 5% nondenaturing polyacrylamide gel. Positions of the DNA-protein complexes are indicated as following: asterisk for the Nrf1/MafK-DNA complex; arrowhead for the Nrf1/ MafK-IE2-DNA complex; dot for the ␣MafK-Nrf1/MafK-IE2-DNA complex; and square for the ␣IE2-Nrf1/MafK-IE2-DNA complex. In consistence with previous reports (27,45), the MafK homodimer failed to bind this particular NF-E2 site (lane 3). transactivation domain, and homodimers of small Maf proteins have been shown to act as transcriptional repressors by binding to certain NF-E2 sites (27). Conceivably, IE2 inhibits the transactivation activity of CNC-bZIP factors by blocking the DNA binding activity of CNC-bZIP/Maf heterodimers, therefore favoring the occupation of the NF-E2 sites by homodimers of small Maf proteins. This hypothesis gains support from the observation that the DNA-binding domain of Nrf1 was involved in the interaction with IE2 (Fig. 2). Alternatively, IE2 does not interfere with the DNA binding activity of CNC-bZIP/Maf heterodimers. Rather, it inhibits CNC-bZIP factors, as in the case of p53 (16), by tethering a repression domain to them. The observations that transcription driven by LexA-CNC-bZIP proteins was nonetheless repressed by IE2 (Fig. 4B) is consistent with this idea.
To distinguish between these possibilities, an electrophoretic gel mobility shift assay was performed to investigate whether the DNA binding activity of CNC-bZIP factors is blocked in the presence of IE2. Because Nrf1 (17) and MafK (28,29) were ubiquitously expressed, these two factors are likely to be partners in vivo. We therefore examined whether binding of the Nrf1/MafK heterodimer to the NF-E2 site was affected by IE2. When Nrf1, MafK, or IE2 alone was incubated with the NF-E2 probe, no specific DNA-protein complex was observed (Fig. 5,  lanes 2-4). A complex was formed by co-incubation of the probe with Nrf1 and MafK (lane 5), which was efficiently competed out with the addition of the unlabeled double-stranded NF-E2 DNA oligonucleotide but not with a nonspecific one (data not shown). Importantly, inclusion of IE2 in the reaction did not block the formation of DNA-protein complex. Instead, the mobility of the complex was further retarded (compare lanes 5 and 6), suggesting that IE2 participated in the formation of the new complex. The identity of individual components in the new complex was examined by the addition of specific antibodies to the incubation: an antibody against Nrf1 eliminated the complex (lane 7), whereas those against MafK and IE2 further reduced the mobility (lanes 8 and 9). In contrast, a preimmune serum had little effect on the mobility of the IE2-Nrf1/MafK ternary complex (lane 10). Taken together, these data demonstrate that IE2 did not interfere with the DNA binding activity of the Nrf1/MafK heterodimer. Instead, the resultant ternary complex consisting of Nrf1, MafK, and IE2 was still able to bind the NF-E2 site, favoring the model that IE2 represses the transactivation activity of CNC-bZIP factors by tethering a repression domain to them.
CNC-bZIP Factors Attenuate the Autoregulation Activity of IE2 by Preventing It from Binding to CRS-Reciprocally, because CNC-bZIP factors interacted with the autoregulation domain of IE2, their influence on the IE2 activity was investigated. As shown in Fig. 6, in the absence of IE2, both Nrf1 and Nrf2 had little effect toward the expression of a reporter driven by the HCMV MIEP (compare lanes 2 and 3 with lane 1) or by a control promoter (compare lanes 8 and 9 with lane 7). IE2 negatively autoregulated MIEP (compare lane 4 with lane 1). However, the IE2-mediated repression of MIEP was largely relieved by co-transfection of a plasmid expressing either Nrf1 or Nrf2 (compare lanes 5 and 6 with lane 4). The antagonistic effect of CNC-bZIP factors was specific for the IE2-mediated autoregulation, because it had little effect toward an IE2-unresponsive reporter driven by the SV40 promoter/enhancer (compare lanes 11 and 12 with lane 10).
It is well known that IE2-mediated autoregulation is caused by binding of IE2 to a specific DNA sequence, called CRS, located immediately downstream of the MIEP TATA box. Thus, the observation that CNC-bZIP factors reduced the autoregulation activity of IE2 strongly suggests that association of IE2 with a CNC-bZIP factor may result in the disruption of its CRS binding activity. To test this idea, an electrophoretic gel mobility shift assay was performed. As shown in Fig. 7

DISCUSSION
Based on the results presented here and previous studies on the repression of p53 by IE2 (15,16), we propose a model for the antagonism between Nrf1 and IE2. IE2 loses its DNA binding activity when complexed with Nrf1/MafK (Fig. 7), which probably underlies the mechanism of how members of the CNC-bZIP family attenuate the autoregulation activity of IE2 (Fig.  6). In contrast, binding of IE2 to Nrf1/MafK has little effect on the DNA binding activity of the heterodimer (Fig. 5). Rather, the resultant IE2-Nrf1/MafK ternary complex can still bind to the NF-E2 site and inhibits the transactivation activity of Nrf1/MafK (Fig. 4) possibly by, as in the case of IE2-mediated inhibition of p53 activity (16), tethering a repression domain to Nrf1/MafK. Moreover, the IE2-mediated inhibition may not be restricted to the Nrf1/MafK heterodimer on the basis of the following evidence. First, IE2 also interacts with Nrf2 (Fig. 1), and LexA-Nrf2 is repressed by IE2 (Fig. 4B). Second, CNC-bZIP proteins appear to form obligatory heterodimers with one or another of small Maf proteins (21,(27)(28)(29).
IE2 is an important regulator of HCMV, and thus its activities need to be strictly controlled. Negative autoregulation is one of the most studied functions of IE2. Nonetheless, little is known about how this IE2 activity is regulated. Recent work implicates that phosphorylation plays an important role in controlling the DNA binding activity of IE2 (14). The current studies demonstrate the antagonism between CNC-bZIP factors and IE2 and thus uncover another potential mechanism for host cells to modulate the autoregulation activity of IE2. Nonetheless, it is interesting to note that the efficiency of autoregulation by IE2 seems to be cell-dependent (4,10,13). In this regard, IE2-mediated autoregulation was prominent in H1299 cells, and the expression of exogenous CNC-bZIP factors was required to attenuate the repression (Fig. 6). Perhaps, the relative expression levels of IE2 and CNC-bZIP factors and/or a difference in the phosphorylation status of IE2 may account for the cell-dependent autoregulation. Because of the resolution limit of the SDS-PAGE gel, it was difficult to tell which IE2 species interacted with CNC-bZIP factors (Fig. 3B). Future work is required to solve this issue. Furthermore, several cellular genes have a CRS-like element strategically positioned around the TATA box. 2 It would be important to determine whether the expression of these genes is subjected to the modulation by the IE2-CNC-bZIP antagonistic interaction.
Previous studies demonstrate that IE2 negatively autoregulates its own expression by a passive mechanism (12). The current work, in conjunction with previous studies (15,16), provides strong evidence to support the notion that IE2 can actively repress gene transcription as well by tethering a repression domain to DNA-binding proteins, for example, p53, Nrf1, and Nrf2. This type of IE2 repression activity is observed with many different cells, including primary human coronary smooth muscle cells (15), the osteosarcoma cell line Saos-2 (16), as well as the lung cancer cell line H1299 and an HCMV permissive cell line WI38 (this work). Because IE2 contains a trans-repression domain (16), it is presumed that IE2 represses transcription by interacting with a general transcription factor(s) in a nonproductive manner (38) and/or recruiting a putative co-repressor(s) (39,40) to the promoter.
Many cellular genes, including proinflammatory cytokines tumor necrosis factor and interleukin 4, contain NF-E2-like sites in their promoters (41)(42)(43). By interacting with CNC-bZIP factors, IE2 may shift the balance between activation and repression of these cellular genes, resulting in the modulation of immune responses to HCMV infection. However, it must be pointed out that IE2 is both a positive (3-5) and a negative (10,11,15,16,33) regulator of gene expression. The presence of an NF-E2 site(s) in the promoter of a gene does not necessarily imply that the gene is under negative control by IE2. In fact, the promoter of NQO2 (quinone oxidoreductase 2) (44) is activated by IE2; however, elimination of the NF-E2 site from its promoter renders the mutant promoter more responsive to the IE2-mediated activation. 2 In light of the dichotomy of IE2, the SV40 TATA box, which is unresponsive to IE2 (Fig. 4), was chosen in purpose to investigate the relatively weak repression of CNC-bZIP factors by IE2.
As an additional level of complexity, CNC-bZIP factors were reportedly able to interact with members of Jun/Fos and ATF families (42) as well as with nuclear hormone receptors (21). Besides, IE2 has activities other than autoregulation and trans-repression, such as transcriptional activation and interaction with CREB, c-Jun, RB, TBP, and TFIIB (6 -9). However, the influence of the IE2-CNC-bZIP interaction on the aforementioned activities and the relative IE2-binding affinity among those factors are currently unknown. Further investigation into the antagonism between IE2 and CNC-bZIP factors should shed light on the role of IE2 in modulating viral and cellular gene expression and thus contribute to our knowledge of the biology and pathology of HCMV infection.