![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Previous Article | Next Article 
Journal of Virology, August 2005, p. 9635-9650, Vol. 79, No. 15
0022-538X/05/$08.00+0 doi:10.1128/JVI.79.15.9635-9650.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.
Departments of Molecular Biophysics and Biochemistry,1 Pediatrics,3 Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut,4 Department of Tumour Virology, German Cancer Research Center, Im Neuenheimer Feld 242, Heidelberg, Germany2
Received 23 February 2005/ Accepted 11 April 2005
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
ZEBRA is sufficient to activate the entire EBV lytic cycle in all cell backgrounds (13, 57, 64). Rta is regularly competent to activate expression of a limited set of early and late viral genes but drives the lytic cycle to completion in only some cell backgrounds (53, 69). The capacity of Rta to initiate the full cascade of lytic-cycle viral gene expression appears to be contingent on its ability to activate ZEBRA and stimulate its own synthesis (53). By using mutant EBV genomes in which the BZLF1 and BRLF1 genes have been disrupted, it has been shown that both immediate-early products are absolutely required for full viral replication and production of mature virions (17). Homologues of Rta are present among all gammaherpesviruses (15, 26, 40, 61, 66). In all these viruses, the Rta proteins play essential roles in stimulating lytic-cycle viral gene expression and DNA replication (53, 61, 66, 67).
EBV Rta is known to carry out many biologic functions. Rta plays a crucial role in initiation of the lytic cycle by reciprocal activation of ZEBRA and by autostimulation (53). Rta and ZEBRA may be expressed from a bicistronic transcript (6, 42); Rta promotes the translation of ZEBRA from this transcript (6). Rta activates transcription of early- and late-lytic-cycle viral genes. Some late-lytic-cycle genes, including gp350 (BLLF1) and BLRF2, are directly activated by Rta in the absence of DNA replication (17, 55). Rta is essential for viral lytic DNA replication, although, unlike ZEBRA, Rta has not been shown to interact directly with the lytic origin of DNA replication. Rta stimulates cells to enter S phase (63).
EBV lytic-cycle genes can be subgrouped into different classes on the basis of their responses to ZEBRA and Rta. One group of genes is activated by ZEBRA, acting alone, and another by Rta. A third group of viral genes is activated synergistically by Rta and ZEBRA. Synergistic activation requires that both ZEBRA and Rta bind DNA (52). A fourth group of viral genes may be activated by Rta and repressed by ZEBRA at early times in the lytic cycle before lytic viral DNA synthesis (55). The repressive action of ZEBRA has recently been shown to be dependent on phosphorylation of ZEBRA by casein kinase 2 (16).
Viral genes are activated by Rta by at least two distinct general modes of action. One group of genes, including those that are stimulated by Rta acting in synergy with ZEBRA, are activated by the binding of Rta to specific DNA sequences, known as Rta response elements (RREs), found in the promoters. Another group of genes that lack RREs in their promoters are activated by indirect mechanisms. At least three examples of the indirect mode of action are known. Rta activates the promoter of the BZLF1 gene (Zp) by activating mitogen-activated protein kinases and phosphoinositol kinases. These kinases in turn lead to activation of the ATF2 protein that binds to the ZII (cyclic AMP response element) site in Zp (1). Rta autostimulates the promoter of BRLF1 (Rp) via an Sp1/Sp3 element (54). Rta activates the EBV DNA polymerase (BALF5) gene via USF and E2F binding sites (39).
Rta is a 605-amino-acid (aa) protein with no known cellular homologues. The N terminus contains an overlapping DNA binding (aa 1 to 280) and dimerization (aa 1 to 232) domain that does not correspond to any previously described DNA binding motif (43). Rta homodimerizes in the absence of DNA. The transcriptional activation domain is found in the C-terminal region of the protein. An obligatory acidic activation domain (aa 520 to 605) contains positionally conserved hydrophobic residues that are predicted to form alpha helices (43). A weaker accessory activating domain contains two proline-rich subregions (aa 352 to 410 and 450 to 500). The activation domain of Rta contacts the TATA binding protein and TFIID in vitro (41). Rta interacts with CREB binding protein (CBP) at multiple sites on both proteins (62). This interaction, and presumably the histone acetyltransferase activity of CBP, is required for Rta's ability to activate some promoters (e.g., BMLF1) but not others (e.g., BZLF1) (62). Rta also interacts with retinoblastoma protein, causing a release of E2F (63). This interaction is thought to have two functional consequences: activation of certain viral promoters such as BALF5 by E2F and activation of host cells to progress through the S phase of the cell cycle (63). Rta is posttranslationally modified by SUMO-1 at several lysines (5). Modification by SUMO-1 is thought to enhance transcriptional activation of RREs by Rta (5).
The first RREs were identified in a region of EBV DNA that contains a bidirectional promoter/enhancer that controls expression of the BHRF1 gene (bcl2 homologue) and the BHLF1 gene whose function remains unknown (22). This region overlaps the origin of lytic viral DNA replication (orilyt) (25). Two contiguous RREs were identified in a stretch of about 70 nucleotides (nt). Both RREs were contacted in vitro by a fragment of Rta protein containing the DNA binding domain (aa 1 to 355). On the basis of guanine methylation interference studies, it was surmised that an Rta dimer contacts two GC-rich core sequences separated by 6 or 7 nt (22). A second RRE was identified in the promoter of the BMLF1 gene that encodes a protein that regulates mRNA expression at the posttranscriptional level (4, 34). This single RRE, located at –374 to –391 relative to the transcription start site of BMLF1, was found to confer a response of BMLF1 linked to ß globin (21). This RRE contains GC-rich sequences separated by 8 bp. A consensus Rta binding element (CRBE), 5'GNCCN9GGNG3', where N is any nucleotide and N9 represents a sequence of nine nucleotides, was defined by the cyclic amplification and selection of targets technique (23). There are more than 300 such sites in the EBV genome; about 30% are near TATA elements and could function as enhancers. In the cyclic amplification and selection of targets experiment, Rta was found to bind with comparable affinities to 30 oligonucleotides that fit the consensus and all sites equally mediated Rta-induced transcriptional activation (23).
In this report we reexamine the nature of the CRBE, using Rta protein that has been expressed in human cells. Until now it has not been possible to study DNA binding by Rta protein expressed in mammalian cells. Previous studies of the protein's ability to bind DNA used Rta that was expressed in Escherichia coli or translated in vitro. Using extracts of human cells, we found that deletions of the C terminus of Rta markedly enhanced its DNA binding capacity. This finding enabled us to compare the relative affinities of the CRBEs found in the promoters of five Rta-responsive genes, BMLF1, BHLF1, BMRF1, BaRF1, and BLRF2. Use of the C-terminally deleted Rta mutant also facilitated a broad strategy of mutagenesis to delineate components of the CRBE that influence binding and response to Rta. Finally, we compared the responses of these five genes to Rta in Raji cells, in which Rta does not activate ZEBRA, and in BZLF1-knockout (BZKO) cells that harbor an EBV genome in which the ZEBRA gene is disrupted.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Transfection and chemical induction. HKB5/B5 and BZKO cells were transfected using the DMRIE-C reagent (Invitrogen) (8). BJAB and Raji cells were transfected by electroporation (58). Cells were subcultured 2 to 3 days prior to transfection. To induce the lytic cycle, Raji cells were treated with 10 ng/ml of tetradecanoylphorbol-13-acetate (TPA) and 3 mM sodium butyrate.
Plasmid construction. pRTS, pRTS/Rta, pE4CAT, and pCMV/Z(S186A) have been described previously (53, 55, 61). To make vectors that expressed C-terminal deletion mutants of Rta, various regions of the Rta gene were amplified by PCR and cloned into pRTS at the XbaI and BglII sites. In some constructs, a DNA fragment encoding the VP16 activation domain (aa 413 to 490) was ligated downstream of portions of the Rta gene corresponding to C-terminal truncations. To make reporter constructs, double-stranded annealed oligonucleotides encompassing nucleotides –401 to –365 of the BMLF1 promoter, –766 to –730 and –788 to –752 of the BHLF1 promoter, –182 to –146 of the BMRF1 promoter, –131 to –95 and –232 to –196 of the BaRF1 promoter, and –261 to –225 of the BLRF2 promoter were cloned into pE4CAT digested with HindIII and XbaI. Double-stranded annealed oligonucleotides with different mutations encompassing nucleotides –401 to –365 of the BMLF1 promoter (see Fig. 3, 4, 5, and 6) were cloned into pE4CAT digested with HindIII and XbaI.
|
|
|
|
EMSAs. Annealed double-stranded oligonucleotides were end labeled with 32P by using T4 polynucleotide kinase (Boehringer Mannheim). Binding reaction mixtures contained 15 µg of cell protein in a solution containing 10 mM HEPES (pH 7.5), 50 mM NaCl, 2 mM MgCl2, 2.5 µM ZnSO4, 0.5 mM EDTA, 1 mM dithiothreitol, 15% glycerol, and 0.5 µg poly(dI-dC) in a total volume of 20 µl. After incubation for 5 min at room temperature, 30,000 to 50,000 cpm of labeled oligonucleotide was added in each reaction. For supershift reactions, antisera were added 10 min following the addition of the probe, and incubation at room temperature continued for 10 min. Antibodies to VP16 (sc-7545; Santa Cruz) and YY1 (sc-7341; Santa Cruz) were obtained commercially. Anti-Rta is a polyclonal rabbit antiserum raised against an N-terminal fragment (aa 1 to 320) of Rta (53). For EMSA competition reactions, increasing amounts (50x and 100x) of nonradioactive competitor DNA were added into the initial reaction mix. The reaction mixtures were loaded onto a 4% native polyacrylamide gel in 0.5x Tris-borate-EDTA buffer and electrophoresed at 200 V. Gels were dried on Whatman 3MM paper under a vacuum and exposed to autoradiography film.
Western blot analysis. Cell extracts were mixed with sodium dodecyl sulfate (SDS) sample buffer and heated to 100°C for 5 min before separation by electrophoresis in a 10% polyacrylamide-SDS gel. Following electrophoresis, the proteins were transferred onto nitrocellulose membranes by electroblotting and blocked in 5% nonfat dry milk overnight at 4°C. The blots were incubated with antiserum at 25°C for 2 h, washed twice for 20 min in 10 mM Tris-HCl (pH 7.5)-200 mM NaCl-5% Tween 20, incubated with 125I-labeled protein A for 1 h, and washed again. The membranes were exposed overnight with intensifying screens to Kodak XAR-5 film at –70°C.
CAT assays. BJAB cells (1.5 x 107) were transfected with 5 µg of reporter DNA plus 5 µg of expression vector pRTS/Rta or pRTS vector. Chloramphenicol acetyltransferase (CAT) assays were performed 48 h following transfection (58). Activation was calculated as percent acetylation of chloramphenicol in the presence of an activator divided by percent acetylation in the presence of a vector and normalized to the activity of the wild-type BMLF1 RRE cloned into pE4CAT. Results presented are averages of results from at least two separate transfections.
Northern blot analysis. Raji cells (1.2 x 107) were transfected with 10 µg of pRTS vector or pRTS/Rta plasmid DNA by electroporation or were chemically induced with TPA and butyrate. BZKO cells (5 x 106) were transfected with 6 µg of plasmid DNA by using the DMRIE-C reagent. Total cellular RNA, prepared using an RNeasy mini kit (QIAGEN), was electrophoresed through a 1% agarose formaldehyde gel and transferred onto nylon membranes (Hybond-N+; Amersham Pharmacia Biotech). All probes were labeled by the random-primed method (55). A 531-bp excised EagI fragment from EBV BamHI-M was used to detect the BMRF1 and BaRF1 mRNA, a 1.3-kb EcoRI/BamHI fragment from BamHI-M was used to detect BMLF1 mRNA, and a 268-bp fragment (nt 79,537 to 79,805) amplified by PCR was used to detect BaRF1 mRNA. The BHLF1 probe consisted of a 1.941-kb HincII/SacI fragment from EBV BamHI-H. A 670-bp XbaI/EcoRI fragment from pCMV/BLRF2 was used to detect BLRF2 mRNA. A radiolabeled 370-bp NcoI/PstI fragment of the H1 component of RNase P was used as a probe to control for RNA loading (3). Hybridization was carried out in 6x SSC (1x SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-5x Denhardt's solution-0.5% SDS-100 µg of salmon sperm DNA per ml at 60°C overnight. Membranes were washed in 2x SSC-0.5% SDS once for 20 min and in 0.1x SSC-0.5% SDS three times for 20 min at 60°C.
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Heterogeneity of the transactivation function of Rta and DNA binding of R550 on different CRBEs. A proposed Rta binding element, with the consensus sequence GNCCN9GGNG, had been found in the promoters of three EBV lytic-cycle genes that are known targets of Rta: BMLF1, a posttranscriptional activator (4, 34); BMRF1, the DNA polymerase processivity factor (35, 36); and BHLF1, a gene whose function is not known (37). The CRBEs are located at –374 to –390 in the BMLF1 promoter and at –155 to –171 in the BMRF1 promoter. The BHLF1 promoter contains two CRBEs at –740 to –756 (BHLF1-I) and –760 to –776 (BHLF1-II). Two additional EBV lytic-cycle genes that are primarily responsive to Rta have been identified previously: BLRF2, encoding a tegument component (58), and BaRF1, encoding a subunit of ribonucleotide reductase (18, 19). These genes are activated following overexpression of Rta in Raji cells (55). Sequences consistent with the Rta consensus binding site were found in the promoters of these genes as well. However, it had not been determined whether the CRBEs in the BLRF2 and BaRF1 promoters were functional. In the promoter of the BLRF2 gene, the CRBE is located at –234 to –250. Two CRBEs are present in the BaRF1 promoter at –104 to –120 (BaRF1-I) and –205 to –221 (BaRF1-II).
To analyze and compare their levels of responsiveness to Rta, the seven CRBEs were cloned into a CAT reporter plasmid upstream of the adenovirus E4 minimal promoter. Transcriptional response to Rta was assessed by CAT assays with the EBV-negative B-lymphoma cell line BJAB. The CRBEs differed dramatically in their levels of transcriptional activation by Rta (Fig. 2A). Three CRBEs, one from the BMLF1 promoter and two from the BHLF1 promoter, responded strongly. Three CRBEs, from the BMRF1, BLRF2, and BaRF1-I promoters, responded weakly, at about 20% maximal activity. The response to Rta of the BaRF1-II CRBE did not differ from the low-level activation of the E4-CAT plasmid which lacks a CRBE. These data showed that not all DNA sequences that fit the consensus for an Rta binding element were equivalently responsive to Rta.
|
Importance of the N9 internal sequence of the CRBE. The next group of experiments systematically explored different regions of the CRBE in an effort to account for the heterogeneity of responses to Rta. The proposed CRBE, GNCCN9GGNG, can be considered to consist of three essential components: a 5' core sequence, a 3' core sequence, and 9 nt of internal sequence which has previously been thought to function only as a spacer region between the two GC-rich core sequences (Fig. 2A). The BaRF1-I CRBE, a weak responder, and the BMLF1 CRBE, a strong responder, share identical 5' and 3' core sequences, but they differ in N9 internal and flanking sequences (Fig. 2A). This observation suggested that, in addition to the core sequences, either the N9 or flanking sequences might account for the marked variations in response. To analyze the relative contribution of the different components of the CRBE, we constructed chimeric mutants (Fig. 3A). One set of four hybrid CRBEs contained the flanking and core sequences derived from the BMLF1 RRE with N9 sequences from different CRBEs inserted into this background. An additional two chimeras contained the N9 sequence of the BMLF1 RRE substituted for the N9 sequences of the BaRF1-I and BaRF1-II CRBEs.
Substitution of N9 derived from the two highly active RREs from the BHLF1 promoter into the background of core and flanking sequences of the BMLF1 CRBE, as in the chimeric mutants ML/HL-I and ML/HL-II, did not significantly reduce transcriptional response or DNA binding (Fig. 3A). Both mutants functioned as strong competitors for binding of R550 to the wild-type BMLF1 RRE (Fig. 3B, lanes 10 to 13). Chimeras containing the N9 sequence from the BaRF1-II CRBE that was deficient in activation and binding of Rta produced intermediate results. The construct ML/aR-II, with the N9 sequence from the BaRF1-II CRBE, was diminished in binding (Fig. 3B, lanes 8 and 9). The construct aR-II/ML, with the N9 sequence from the BMLF1 RRE inserted into the core and flanking sequences of BaRF1-II, was also impaired (Fig. 3C, lanes 12 and 13). These results suggested that the N9 sequence in the BaRF1-II CRBE, as well as core and flanking sequences, could contribute to a reduced response. Both core sequences of the BaRF1-II CRBE, 5'GGCC and 3'GGGG, differ from all the other core sequences. In the 5' core of BaRF1-II, there is a G at position 2, whereas in the other six RREs, there is a T at this position. In the 3' core there is a G at position 3 which is a C or T in the other RREs.
The most dramatic results that implicated the N9 element in the response to Rta were observed with chimeric mutants that contained components of the CRBE from BaRF1-I. Substitution of the N9 element of the weakly responsive BaRF1-I RRE into the BMLF1 RRE background (as in mutant ML/aR-I) abolished the transcriptional response to Rta and eliminated DNA binding by Rta. ML/aR-I failed to compete for binding by R550 in EMSAs (Fig. 3B, lanes 6 and 7). Conversely, substitution of the N9 sequence from the BMLF1 RRE into the core and flanking sequences of the BaRF-I background, as in mutant aR-I/ML, completely restored the transcriptional response to Rta and the capacity to compete for DNA binding by Rta (Fig. 3A). The aR-I/ML RRE competed for binding by R550 as well as wild-type oligonucleotide competitors (Fig. 3C, lanes 10 and 11). These data revealed the essential importance of the N9 nucleotide sequence in Rta binding and activation.
Sequence specificity of the N9 element. The N9 sequence of the highly active BMLF1 RRE, CTCTATCAT, differs from the N9 sequence of the weakly responsive BaRF1-I RRE, CATGGAACC, at eight of nine positions. Only C at position 1 is conserved. Constructs with point mutations were generated in order to evaluate which of the remaining positions in N9 might account for the poor response and lack of binding of R550 by the chimera ML/aR-I. Two or three nucleotide point mutations were installed into the BMLF1 RRE in order to mimic the N9 sequence of BaRF1-I. Mutation of nucleotides at positions 4 to 6, as in mutant two (mt2), did not affect transcriptional stimulation or DNA binding by Rta (Fig. 4A and B, lanes 8 and 9). mt3, with alterations at positions 7 to 9, showed about 30% reduction in activation and binding. However, mt1, with changes at positions 2 and 3, reduced the response to Rta by more than 90% and eliminated the capacity of the CRBE to compete with the wild-type BMLF1 RRE for binding of Rta (Fig. 4B, lanes 6 and 7). mt1 behaved similarly to mt4, in which both core elements of the BMLF1 RRE were altered while the N9 and flanking sequences were unchanged.
The results of the experiments described above suggested that a specific nucleotide sequence in N9 was essential for optimal binding and response to Rta. However, the experiments could not exclude the possibility that the ratio of purines to pyrimidines affected the structure of the N9 element and consequently its response to Rta. When the N9 sequences of two highly active RREs were installed in an inverted and complementary orientation, as in the constructs BMLF1 (IN) and BHLF1-I (IN), stimulation by Rta in a reporter assay and binding by R550 were considerably reduced, although not eliminated (Fig. 5). Both inverted constructs contained the sequence TG at positions 2 and 3 of N9. T is present at position 2 in the wild-type BMLF1 RRE and G at position 3 in the wild-type BHLF1-I RRE. This may explain the retention of some activity by the inverted N9 sequences.
N9 spacing is optimal. The preceding experiments (Fig. 3 to 5) with chimeras and constructs with CRBE point and inversion mutations strongly suggested that the N9 nucleotide sequence per se played a role in direct recognition by Rta. However, the contribution of the length of the N9 sequence to the Rta response needed to be evaluated. While recognition of the CRBE might depend on sequence, it might also require the proper spacer length between the core elements. Alteration of the CRBE by addition or subtraction of a single nucleotide markedly affected the response to Rta (Fig. 6A). The effect of removal of one nucleotide was more deleterious than the consequences of adding one nucleotide. A mutated BMLF1 RRE with one added nucleotide still retained a weak response and a slight ability to bind R550 (Fig. 6B, lane 7, and C, lane 9). However, removal of a single nucleotide from N9 eliminated both activity and binding. Insertion of 2 or 4 nt into the N9 sequence also abolished the transcriptional response and binding by Rta (Fig. 6A and C, lanes 10 to 13). Of interest, alteration of the N9 spacing, by elimination of position 5 of N9 or by insertion of nucleotides after position 5, maneuvers that drastically affected binding by Rta, did not impinge on binding of the cellular protein(s) which formed the prominent complex c. This result suggested that binding of Rta to the RRE was independent of binding of complex c.
Identification of complex c as YY1. The RRE present in the BMLF1 promoter contains an inverted consensus YY1 binding site, CGCCATNTT, that overlaps the 3' core element and positions 5 through 9 of the N9 element (Fig. 7). To determine whether YY1 could bind to the BMLF1 RRE, EMSA competitions were conducted using duplex oligonucleotides containing canonical or mutated YY1 recognition sites (Fig. 7). Complex c was eliminated by the wild-type YY1 competitor (Fig. 7B, lanes 3, 4, 9, and 10) but not by the mutated YY1 site (Fig. 7B, lanes 5, 6, 11, and 12). Antibody to YY1 also altered the electrophoretic mobility of complex c (Fig. 7B, lanes 7 and 13).
|
Heterogeneity of expression of EBV lytic-cycle genes containing RREs of different affinities. In Raji cells, Rta activates some lytic-cycle genes but does not detectably cross-stimulate the expression of ZEBRA (55). Therefore, this cell background permits analysis of Rta-responsive genes. We examined Rta-dependent expression of the five lytic-cycle genes in the promoters of which RREs of heterogeneous response had been identified (Fig. 2). After Raji cells were transfected with Rta expression plasmid, the abundance of lytic-cycle mRNAs was examined at intervals from 12 to 48 h after transfection. The level of lytic-cycle mRNA observed following transfection with Rta was compared to that in cells that had been transfected with control vector and in cells that had been treated with TPA and sodium butyrate, a stimulus that activates expression of both Rta and ZEBRA (20, 70).
Two remarkable findings, illustrated in Fig. 8A, were evident from this experiment. The kinetics of expression of the Rta-responsive genes varied markedly. Near-maximal levels of the BMLF1 and BLRF2 mRNAs were present 12 h after transfection with Rta; maximal expression of the other three genes was delayed to 24 h or later. The expression levels following TPA/sodium butyrate treatment, relative to expression following transfection with Rta, varied among the five genes. The expression of two genes, BaRF1 and BMLF1, was equally high after Rta transfection and after TPA/sodium butyrate treatment. The expression of two genes, BHLF1 and BMRF1, that are known to be synergistic targets of Rta and ZEBRA (55) was greater following chemical induction than following transfection with Rta. Expression of a third class of Rta-responsive genes, represented by BLRF2, was inhibited by TPA/sodium butyrate treatment. Rta-dependent activation of BLRF2 is known to be repressed by ZEBRA (55). These experiments defined three distinct classes of Rta-responsive genes.
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Since these deletions of the DNA binding inhibitory region eliminated a portion of the transcriptional activation domain of Rta, it was possible that competition between interactions with components of the transcription machinery and interaction of Rta with DNA may account for enhanced binding of C-terminally truncated mutants to DNA. However, C-terminally truncated mutants of Rta to which the powerful activation domain of herpes simplex virus VP16 protein had been appended still bound DNA with higher affinity than wild-type Rta (Fig. 1A, lane 4). Thus, it is unlikely that elimination of the transcriptional activation domain alone accounts for restoration of the DNA binding activity measurable by EMSA.
Several cellular transcription factors, including Ets-1, p53, androgen receptor, and interferon regulatory factor 3, also contain autoinhibitory domains that block the capacity of the proteins to bind DNA (24, 30, 32, 38, 50). Recently, a DNA binding inhibitory region has been identified between aa 520 and 535 of ORF50 protein, the Rta homologue in KSHV (7). Basic amino acids in this inhibitory region of ORF50 that play an essential role in blocking DNA binding also affect ORF50 protein stability (7). There are no obvious amino acid similarities between the autoinhibitory regions of KSHV ORF50 and EBV Rta. Furthermore, elimination of amino acids that inhibit the ability of EBV Rta to bind DNA did not affect protein stability (Fig. 1B and data not shown).
At least two mechanisms can be invoked to account for autoinhibition of DNA binding. First is an intramolecular mechanism: the inhibitory region may interact with the DNA binding domain of the protein. In the folded structure of the protein, the inhibitory region may sterically interfere with the interaction of the DNA binding domain with DNA. Second is an intermolecular mechanism: the inhibitory region may interact with another cellular protein that in turn alters conformation or blocks access to DNA. Studies with purified Rta protein will be required to distinguish between these alternatives.
Since Rta operates by several distinct mechanisms, including direct interaction with viral promoters, synergistic interaction with ZEBRA, and activation of signal transduction cascades, it is likely that the DNA binding activity of Rta protein is tightly regulated. The DNA binding inhibitory region that we have identified likely plays an important role in this regulation.
Heterogeneity of CRBEs. The seven CRBEs could be divided into three groups, based on their affinities for Rta in vitro and on their capacities to be activated by Rta in reporter assays. There was close correlation between binding affinity for Rta, as assessed by EMSA, and the capacity of the CRBE to confer a transcriptional response to Rta. The RREs from BMLF1 and BHLF1 (I and II) were of high affinity; the BMRF1, BLRF2, and BaRF1-I RREs were of low affinity; and the BaRF1-II CRBE did not function as an RRE (Fig. 2).
Extensive mutagenesis experiments showed that a significant component of the variation in responses was attributable to the 9 nt which separated the 5' and 3' core elements of the CRBE. This was most convincingly demonstrated by chimeric constructs. For example, ML/aR-I, a chimera which contained the core and flanking sequences from the high-avidity BMLF1 CRBE and the N9 sequence from the low-affinity BaRF1-I CRBE, lost binding activity and could not be transactivated by Rta. Nonetheless, the N9 sequences were not sufficient to confer a response and needed to be accompanied by appropriate core and flanking sequences. For example, the N9 sequences from the BMLF1 RRE could only partially rescue the nonfunctional BaRF1-II CRBE in the chimera aR-II/ML.
Point mutations within N9 showed that specific nucleotides at positions 2 and 3 and to a lesser extent those at positions 7 to 9 were required for maximal response. Total purine and pyrimidine content within N9 was not sufficient to maintain maximal binding or activation since inversion of the N9 sequences of the highly reactive CRBEs from BMLF1 and BHLF1-I significantly reduced activity (Fig. 5). All of these data indicate that the internal 9 nt of the CRBE serve a sequence-specific function and not only act as a spacer. Nonetheless, the spacer function is also critical, since adding or subtracting a single nucleotide from the N9 sequence of the BMLF1 CRBE reduced its response by more than 80%.
Guanine methylation studies have suggested that a dimer of Rta contacts both GC-rich core sequences separated by 9 nt. This result suggests that each Rta monomer binds to adjacent major grooves of DNA (23). Several other transcription factors appear to have a mode of interaction with DNA in which core-contacting sequences are separated by spacer nucleotides. For example, the Saccharomyces cerevisiae transcriptional activator, GAL4, binds to conserved CCG triplets at each end of a 17-bp site (44). There is an 11-nt spacer. The binding site for human papillomavirus type 16 E2 protein consists of two palindromic triplets, ACC and GGT, separated by 6-nt spacer sequences (28). The binding site for E. coli cyclic AMP receptor protein has two symmetrically related inverted recognition elements, TGTGA and TCACA, separated by a spacer that is either 6 or 8 nt long (2, 31). The response element of the androgen receptor contains inverted repeats of TGTTCT with a 3-nt spacer (11, 59). In view of our results, it will be important to reexamine the binding sites for these transcription factors for a sequence-specific function of the spacer nucleotides.
Our results showing that the N9 component of the CRBE provides not only a crucial spacer function but also a sequence-specific function could have one of two interpretations: the crucial bases within N9 may also be bound by monomers of Rta or they may be bound by cellular proteins. Experiments in which one nucleotide was added or subtracted from the N9 sequence (Fig. 6) showed that cellular complex b is exquisitely sensitive to the 9-nt length. This observation favors the idea that the N9 sequences attract a specific cellular protein. An attractive possibility that would merge the two hypotheses is that the N9 sequences bind the HMGB1 protein which has been found in vitro to bend RRE DNA and to enhance the formation of complexes between Rta and DNA (46).
YY1 is one cellular factor that binds to some CRBEs. Complex c was reproducibly the most abundant of several cellular factors that were found to bind to the BMLF1 CRBE. Several lines of evidence proved that complex c contains YY1 (Fig. 7). A consensus YY1 site is embedded in the CRBE from BMLF1; the YY1 element overlaps nt 8 and 9 of N9 and nt 1 and 2 of the 3' core sequence. Mutations in mt3 and mt4 which destroyed the YY1 consensus sequence abolished formation of complex c. Complex c could be eliminated by competition with a wild-type YY1 site but not by competition with a mutant YY1 oligonucleotide. Complex c could be supershifted by antibody to YY1.
Our experiments suggested that YY1 and EBV Rta bind independently to the CRBE of the BMLF1 promoter. Elimination of YY1 binding by competition with YY1 oligonucleotide did not abolish the binding of Rta (Fig. 7B, lanes 9 and 10). Mutants that altered the length of N9 maintained binding of YY1 but markedly reduced or eliminated binding of Rta (Fig. 6B, lanes 7 and 11). We never observed a ternary complex in which the CRBE was occupied by YY1 and Rta. While such a large complex could fail to enter the gel, a more likely explanation is that YY1 and Rta do not directly interact on DNA since the binding sites for the two proteins overlap.
YY1, a zinc finger protein, can repress or activate transcription (48, 56, 60, 65). YY1 has been reported to act as a repressor in the regulation of the EBV BZLF1 and BRLF1 genes (47, 68). Our data do not yet provide conclusive evidence that YY1 either represses or enhances activation of CRBEs by Rta. Both the BMLF1 CRBE and the chimeric aR-I/ML CRBE strongly bind YY1, yet Rta also powerfully activates these constructs (Fig. 3). Thus, YY1 is not likely to function as a repressor in the transactivation assays but may possibly do so in vivo. The two CRBEs from the BHLF1 promoter lack a YY1 consensus binding site (Fig. 2B); these CRBEs bind Rta avidly, and the CRBEs are stimulated strongly by Rta. Thus, binding of YY1 is not required for activation by Rta. Mutant CRBEs that cannot be activated by Rta by virtue of alterations in the length of the N9 sequence nonetheless bind YY1 (Fig. 6). Thus, YY1 binding alone is insufficient to lead to transcriptional activation of an RRE. Conversely, as shown in Fig. 3B, CRBEs with mutant YY1 sites that do not bind YY1 remain maximally responsive to Rta.
Of the seven CRBEs that we studied, YY1 was found to bind strongly to only one high-affinity RRE, namely, that in BMLF1, and weakly to the low-affinity RRE from BaRF1. Since both high- and low-affinity RREs bound or did not bind YY1, our data do not support an essential universal role for YY1 in regulating Rta-mediated activation of gene expression. Nonetheless, YY1 could function as a repressor of basal activity of the BMLF1 promoter and possibly contribute to temporal regulation of this promoter.
Behavior of the Rta-responsive genes in a biologic context. We attempted to relate the binding affinities of the RREs in vitro and the transcriptional responses of reporter constructs to the biologic outcome of lytic gene expression following transfection with Rta in two cell backgrounds, Raji and BZKO (Fig. 8). In Raji cells, where Rta does not detectably activate ZEBRA or stimulate its own expression, we compared expression levels of the target genes after transfection with Rta with their expression levels after treatment with TPA and sodium butyrate, which activate both ZEBRA and Rta (55). All five genes whose promoters contained identifiable RREs (Fig. 1) responded to transfection with Rta but with striking differences in levels of transcript abundance, kinetics of expression, and the synergistic roles of additional expression of ZEBRA. It was notable that the most abundant transcripts were observed in Raji cells for the BMLF1 and BHLF1 genes, whose promoters contain the highest-affinity RREs. Transcript abundance was considerably less for the genes with low-affinity RREs. More quantitative studies of transcript abundance are needed before this correlation can be established. However, the kinetics of expression of Rta-induced genes did not correlate with RRE affinity. Near-maximal activity of BMLF1, with a high-affinity RRE, and BLRF2, with a low-affinity RRE, was observed at 12 h, while expression of the BHLF1 and BMRF1 genes was delayed.
The identification of RREs of different affinities is only the beginning of an analysis that would attempt to understand such complex phenomena as transcript abundance and temporal regulation of the Rta-responsive genes. Our analysis was limited to a reporter assay that measured the response of the RRE fused to a heterologous minimal promoter. Additional positive or negative regulating elements undoubtedly exist in the full-length natural promoters. These promoters are also likely to be regulated by epigenetic mechanisms, including DNA methylation and chromatin modifications. Moreover, our analysis did not examine a number of other important components of mRNA regulation, such as transcriptional initiation rate, elongation rate, control of RNA processing, RNA transport, and RNA stability.
In related studies, it was demonstrated that all of the five lytic-cycle genes were activated by transfection of BZKO cells with Rta. BZKO cells are 293 cells containing an EBV bacmid in which the BZLF1 gene has been insertionally inactivated (17). In these cells, transcripts of the genes with high-affinity RREs, namely, BHLF1 and BMLF1, were more abundant than mRNAs from genes with low-affinity RREs, BaRF1, BMRF1, and BLRF2, following expression of Rta alone. BMRF1 and BHLF1 could be shown to be synergistic targets on the basis of markedly enhanced expression following coexpression of Rta and Z(S186A) (Fig. 8B). Since the activity of Z(S186A) can be revealed only in the presence of Rta, it is useful in defining synergistic targets. The other three genes did not respond synergistically to Z(S186A). All five genes were activated by expression of ZEBRA in BZKO cells (data not shown) because ZEBRA activates Rta in this background (17). BLRF2 was not repressed by ZEBRA in BZKO cells because in this cell background ZEBRA stimulates cell DNA synthesis and BLRF2 is repressed only at early times (16, 55). Otherwise, studies with the two-cell background produced identical results. BMLF1 and BaRF1 are dominantly controlled by expression of Rta. BHLF1 and BMRF1 are synergistic targets of Rta and ZEBRA. This classification of Rta-responsive genes is similar to those in previous studies using reporter constructs (9, 10, 14, 29, 33, 52) and is supported by experiments in which the effects of transfection of Raji cells with Rta alone and with Rta plus the Z(S186A) mutant were compared (55). In summary, in both cell backgrounds, RRE affinity correlated with transcript abundance, but in neither cell background could RRE affinity predict whether the responding gene was a dominant target of Rta or a synergistic target of Rta and ZEBRA.
In conclusion, this is the first report to study the DNA binding function of Rta expressed in mammalian cells. The system we describe has permitted us to elucidate differences in affinities of the RREs and to further define the nature of the RREs and has enabled us to begin to delineate interactions among RREs, Rta, and cellular proteins. An ultimate goal will be to understand how the differences in the RREs that we have described influence biologic parameters of the EBV lytic cycle, such as transcript kinetics, abundance, and cell type-specific controls on the target genes of Rta.
|
ACKNOWLEDGMENTS |
|---|
We thank Jill Countryman and Ayman El-Guindy for helpful comments on the manuscript.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Rta, addition of Rta antibody.
