INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Epstein-Barr virus (EBV) is the best known of the 1 herpesviruses, a group of closely related B-lymphotropic agents of primates which have evolved a unique strategy through which to access, disseminate, and persist within the B-lymphoid system. The essential features of that strategy are threefold. First, EBV can efficiently access target B lymphocytes through an interaction between the major viral envelope glycoprotein gp340 and the complement receptor CR2/CD21, a cell surface molecule expressed preferentially though not exclusively on B-lymphoid cells (16, 38, 57). Second, the virus carries a set of latent cycle genes, encoding six nuclear antigens (EBNA1, EBNA2, EBNA3A, EBNA3B, EBNA3C, and EBNA-LP) and three latent membrane proteins (LMP1, LMP2A, and LMP2B), whose coordinate expression in newly infected B cells activates cell growth (26); this growth transformation, first observed in vitro, where experimental infection of resting B cells leads to the outgrowth of permanent lymphoblastoid cell lines, is also seen during primary EBV infection in vivo as a virus-driven expansion of the latently infected B-cell pool (58). Third, following this expansion, the transforming program of latent gene expression can be down-regulated in vivo as the infected cells move out of cycle and enter the resting memory B-cell pool (3, 35, 43); the latter is the reservoir upon which successful virus persistence in the immune host appears to depend.

This report concerns the mechanisms whereby EBV specifically activates its growth-transforming program of gene expression. In experimentally infected B cells, the first viral transcripts are driven from a viral promoter, Wp, localized within the BamHI W repeat region of the viral genome (2, 64). These transcripts lead to the expression of two nuclear antigens, EBNA2 and EBNA-LP, both of which are critical for efficient transformation. EBNA2, acting alone or in cooperation with EBNA-LP, serves to activate a number of cellular promoters for growth response genes (29, 52, 61) as well as the LMP promoters (15, 62, 66) and Cp, an alternative EBNA promoter 3 kbp upstream of Wp, which can drive expression of all six EBNA transcripts (24, 56). It is clear that activation of this growth-transforming program of viral gene expression is in some way dependent on the B-cell environment because experimental infection of other cell types, including epithelium (39, 53), endothelium (25), T cells (63), and monocytes (49), does not lead to full latent protein expression or autonomous cell growth. In the best studied of these examples, CR2 gene-transfected epithelial cells exposed to the virus in vitro showed transient low-level transcription from Wp and Cp but no detectable EBNA2 or EBNA-LP expression; many cells became actively infected, however, and selectively expressed the virus genome maintenance protein EBNA1 from an alternative EBNA1-specific promoter, Qp, 17 kbp downstream of Wp (28, 30).

In B-cell infection, much work has focused on the mechanisms whereby EBNA2, through interaction with cellular factors such as recombination signal binding protein J (19, 21, 67), can activate the Cp and LMP promoters and thereby lead to expression of the full range of virus latent proteins. What is less clear, however, is the mechanism whereby Wp is first activated in B cells to initiate the transformation process. There have been a number of studies with Wp reporter constructs identifying potential regulatory elements in the promoter (23, 47), but apart from a long-range role for EBNA1 through its binding to the oriP region of the viral genome 5 kbp upstream (42), the factors governing Wp activity have only recently begun to be explored. Notably, we showed that the low basal activity of Wp that is seen in a variety of cell types was dependent on sequences more than 250 bp upstream of the transcription start site, in particular on a region called upstream activation sequence 2 (UAS2; 264 to 352), the activity of which was primarily dependent on the ubiquitously expressed transcription factor YY1. By contrast, the much higher activity shown by Wp in B-cell lines mapped to promoter-proximal region UAS1, which contained at least three regulatory sites identified by mutational analysis (6). One of these sites bound ubiquitously expressed proteins of the CREB/ATF family (27), but the identity of factors binding elsewhere in the region and the basis of the promoter's preferential activation in B-cell lines remained to be determined. Here we show that this activation requires members of the RFX family of transcription factors and the B-cell-specific activator protein BSAP/Pax-5.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell lines. A number of human B-cell and non-B-cell lines were used for the preparation of nuclear extracts and for transient transfection assays. Established B-cell lines were the EBV-negative Burkitt's lymphoma-derived cell lines DG75 and Ramos, the EBV-positive Burkitt's lymphoma-derived cell line Akata, and the EBV-transformed lymphoblastoid cell line IB4. The non-B cells included the T-cell leukemic lines CEM and Jurkat, the proerythroleukemic line K562, and the simian virus 40-transformed epidermal keratinocyte line Rhek. All cells with the exception of Rhek were maintained as suspension cultures in RPMI 1640 supplemented with 10% (vol/vol) selected fetal calf serum, 2 mM glutamine, and 100 mg of gentamicin per liter. Rhek cells were grown in Joklik's medium supplemented with 8% (vol/vol) fetal calf serum, 2 mM glutamine, and 0.4 µg of hydrocortisone per ml.

Plasmid constructs. The Wp440/GL2 reporter plasmid, in which the luciferase reporter gene is under the control of Wp sequence positions 440 to +175 (relative to the Wp RNA start site), has been described previously (6). The site B, site C, and site D mutant reporter constructs were made using appropriate oligonucleotides and the Sculptor mutagenesis system (Amersham Pharmacia). The BSAP expression vector containing the human BSAP cDNA cloned into pSG5 (44) was kindly provided by Andreas Reimold (Harvard Medical School, Boston, Mass.).

Bandshift assays. The preparation of nuclear extracts and the in vitro binding assays have been described previously (6); binding reactions were carried out with either poly(dI-dC) (Amersham Pharmacia) or sheared herring sperm DNA (Sigma) to reduce complex formation due to nonspecific DNA binding proteins. In the initial bandshift assays, long probes carrying Wp sequence positions 352 to 264, 264 to 135, and 135 to 70 were generated by digestion of Wp440/GL2 Basic with SacI/AvrII, AvrII/ApaI, and ApaI/MunI, respectively. Two additional overlapping probes, 316/135 and 170/70, were generated by SacI/ApaI digestion of a truncated Wp reporter plasmid Wp316/GL2 Basic produced as one of a series of promoter truncations in earlier work (6) and by NcoI/MunI digestion of a Wp440 derivative in which an NcoI site had been introduced at 173 to 168 by site-directed mutagenesis, respectively. The appropriate fragments were then dephosphorylated using alkaline phosphatase and gel purified before labeling with [-³²P]ATP and polynucleotide kinase. The short double-stranded oligonucleotide competitors (Alta Biosciences, University of Birmingham) used for Fig. 2 were MIF-1 (GATCTAGAGTAGTTATGGTAACTGGG), MIF-1 m (GATCTAGAGTAGTTATGATTACTGGG), HBV enh-I (GATCCGTTGCTCGGCAACGGCCTA), HBV enh-I m (GATCCCAACCTCGGCAACGGCCTA), and HLA DRA X (GATCCCCTTCCCCTAGCAACAGATGA) (mutations are underlined); designations are explained in Results).

Transient transfection and reporter assays. Cell cultures were transiently transfected with 8 µg of luciferase reporter containing the relevant Wp sequences and 2 µg of a constitutively expressed -galactosidase reporter (CMV-gal) as previously described (6). Wp reporter activities were measured by quantifying luciferase expression in whole-cell extracts at 16 to 24 h posttransfection; the luciferase activity in each sample was then normalized for variations in transfection efficiency by measuring the level of -galactosidase expression from the cotransfected CMV-gal plasmid. In the case of the BSAP transactivation experiments, 2 µg of BSAP/pSG5 or 2 µg of empty pSG5 vector was included with the appropriate luciferase reporter constructs and CMV-gal as above.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Binding of lineage-independent and B-cell-specific factors to Wp sequences. Figure 1a shows a schematic representation of Wp in its genomic context, downstream of oriP and Cp, and identifies the main lineage-independent regulatory region UAS2 with its constituent YY1 site and the B-cell-specific region UAS1 with its constituent CREB/ATF site. The additional sites (B, C, and D) within UAS1 were either known from earlier work (6) or identified in the present study. Since our previous analysis of factor binding to Wp had used short oligonucleotide probes representing only a few selected sequences within the promoter, we first carried out a more systematic screening using a series of longer radiolabeled probes designed to cover the entire 352 to 70 region.

View larger version (43K): [in this window] [in a new window]   FIG. 1.   Binding of lineage-dependent and lineage-independent factors to Wp. (a) Schematic illustration of the region of the EBV genome containing the origin of plasmid replication oriP and the viral latent cycle promoters Cp and Wp. Below this is a detailed map of Wp showing its main lineage-independent (UAS2) and B-cell-specific (UAS1) regulatory elements and their constituent binding sites for either known (YY1 and CREB) or unknown (sites B, C, and D) cellular factors. Also shown are horizontal bars representing the three principal Wp restriction fragments used as bandshift probes to screen for factor binding. (b to e) Patterns of protein-DNA complexes obtained by incubating these Wp probes with nuclear extracts prepared from the indicated B- and non-B-cell lines. (b) Complexes obtained using the Wp 352/264 probe in the presence of poly(dI-dC). Lanes 1 to 8, probe plus the indicated nuclear extract; lanes 9 and 10, probe plus Akata nuclear extract in the presence of an oligonucleotide competitor (comp) containing either the minimal YY1 binding sequence within UAS2 (lane 9) or the variant sequence mut YY1 (lane 10). (c) Complexes obtained using the Wp 264/135 probe in the presence of poly(dI-dC) plus the indicated nuclear extract. Note that the unusually strong intensity of the site D complex in lane 1 is due to differences in protein loading rather than a difference in expression of the site D binding factor. (d) Complexes obtained using the Wp 135/70 probe in the presence of poly(dI-dC). Lanes 1 to 7, probe plus the indicated nuclear extract; lanes 8 to 10, probe plus Akata nuclear extract in the presence of minimal CREB binding sequence as competitor, a site B competitor, and a site C competitor, respectively. (e) Complexes obtained using the same Wp 135/70 probe but in the presence of herring sperm DNA. Lanes are as in panel d.

Identification of RFX proteins binding to site C. Focusing first on site C (140 to 99), we searched the TRANSFAC database to identify potential cellular factors that might bind to this sequence. As shown in Fig. 2A, this revealed a significant homology between nucleotides 122 to 110 and the consensus binding motif for the RFX family of transcriptional regulators (14). Of these, RFX1, RFX2, and RFX3 have been implicated in the regulation of viral and cellular genes and shown to bind target sites as both homodimers and heterodimers (22, 31, 45, 46, 48, 50, 51), while RFX5 specifically associates with the promoters of HLA class II genes as a complex with two other proteins, RFX-AP and RFX-ANK/RFX-B (12, 34, 37, 55). In addition, another cellular factor, MIBP1 (myc intron binding factor 1), first identified as a regulator of c-myc gene expression (65), has been implicated in the regulation of RFX-dependent promoters (45) and can form heterodimers with at least one member of the RFX family, RFX1 (7).

View larger version (48K): [in this window] [in a new window]   FIG. 2.   EBV Wp site C binds factors belonging to the RFX/MIBP1 family. (A) Sequence comparison of Wp nucleotides 122 to 110 within site C and known RFX/MIBP1 binding sites. The top line shows a published consensus RFX1 binding site (14) as an imperfect inverted repeat comprising two half-sites (arrowed) separated by a variable spacer region. Shown below are the relevant sequences from the wild-type Wp site C and from a mutant derivative (site C m1) previously shown to have lost factor binding (6). Also shown are the sequences of several published RFX/MIBP1 binding sites (HBV enh I, MIF-1, and HLA DRA X [45]) which were used as competitors in bandshift assays. HBV enhI m and MIF-1 m, mutant sequences previously shown to no longer bind RFX/MIBP1 (45), were included as controls; nucleotide substitutions are shaded. (B) Protein-DNA complexes obtained by incubating a site C probe with DG75 or Jurkat nuclear extracts. Lane 1, probe plus nuclear extract alone; lanes 2 to 8, probe plus nuclear extract in the presence of the indicated competitor oligonucleotide. (C) Characterization of site C complexes produced with DG75 or Jurkat nuclear extracts using an antibody (Ab) directed against specific RFX or MIBP1 polypeptides. Lane 1, probe plus nuclear extract alone showing three major complexes, c', c", and c; lanes 2 to 6, probe plus nuclear extract in the presence of an antibody against the indicated proteins.

Effect of mutations in site C on RFX binding and Wp activity. Although the RFX consensus binding motif contains a palindromic sequence, it has been reported that in certain cases RFX proteins can bind as monomers rather than dimers and that one half of the binding motif may be sufficient for the interaction (8, 10). We therefore investigated the sequence requirements for RFX complex formation at Wp by introducing a series of base substitutions throughout site C and then using each of these mutant sequences (m1 to m4 [Fig. 3A]) as competitor in a bandshift assay. The results in Fig. 3B show that m1 and m2 were unable to compete for factor binding, m3 led to a partial reduction in binding, while m4 competed as effectively as the wild-type sequence. This suggested that Wp sequences in the half-site between 125 and 116 were the more important for RFX binding. To further investigate if binding to site C requires one or both half-sites, we synthesized two additional competitors carrying truncated site C sequences: LH, which contains the 5' half-site sequence (125 to 113); and RH, which contains the 3' sequence (119 to 107). The results in Fig. 3C show that the LH, but not the RH, sequence was sufficient to compete for RFX binding to site C. This result is consistent with the bandshift data in Fig. 3B and also accounts for our earlier observation that the minimum sequences for site C binding are located between 125 and 111 (6).

View larger version (33K): [in this window] [in a new window]   FIG. 3.   Mutational analysis of site C. (A) Nucleotide sequences of wild-type and mutated (m1 to m4) site C competitors used in bandshift experiments; nucleotide substitutions are shaded. Also shown are two truncated site C fragments which carry either the 5' (LH) or 3' (RH) half-site sequence. (B) Effect of mutations m1 to m4 on RFX/MIBP1 binding to site C. Lanes 1 to 5, protein-DNA complexes obtained by incubating a site C probe with DG75 nuclear extract in the presence of the indicated mutant (lanes 1 to 4) or wild-type (lane 5) competitor sequence; lane 6, probe plus nuclear extract alone. (C) Experiment similar to that shown in panel B but including the truncated site C fragments LH and RH as competitors. (D) Effect of site C mutations on Wp activity. Promoter activity was assayed by quantifying luciferase expression (as relative light units [RLU]) in DG75 and in K562 cells transiently transfected with a wild-type Wp reporter (Wp440) or with mutant derivatives carrying m1 to m4 shown above. The reported luciferase values are the means ± standard deviations of three independent experiments.

Site B and site D are binding sites for the same B-cell-specific cellular protein. We then analyzed the two sites within UAS1 which bound B-cell-specific factors in the initial bandshift studies: site B (already identified as 115 to 86) and the newly identified site D, which lay within the Wp 264/135 probe used in the first experiments (Fig. 1c). Further bandshift studies were carried out using this latter probe and a series of shorter oligonucleotide competitors. As shown in Fig. 4A, this localized site D to the 242 to 215 region, that being the shortest sequence capable of fully competing out binding of the B-cell-specific factor.

View larger version (40K): [in this window] [in a new window]   FIG. 4.   Binding of cellular factors to sites D and B. (A) Determination of the minimal sequence required for factor binding to site D. Lanes represent protein-DNA complexes obtained by incubating DG75 nuclear extract with the Wp 264/135 probe in the presence of competitor oligonucleotides carrying the Wp sequences indicated. (B and C) Sites D and B interact with the same B-cell-specific factor. (B) Minimal nucleotide sequence required for factor binding to sites D and B, along with sequences of mutant derivatives of site D (Dm1 and Dm2) and site B (Bm) known to have lost factor binding. Nucleotide substitutions are shaded. (C) Protein-DNA complexes obtained by incubating the minimal site D probe with the indicated nuclear extracts either alone or in the presence of wild-type or mutant competitor oligonucleotide added at 1,000-, 100- or 10-fold, molar excess, as shown (top), and a parallel experiment using a site B probe (bottom).

Identification of sites B and D as binding sites for B-cell-specific activator protein BSAP/Pax-5. Searching databases of known transcription factor binding motifs revealed a limited homology between the site D sequence and a consensus motif (9) for the B-cell-specific transcription factor BSAP/Pax5 (Fig. 5A). There was also a lesser degree of sequence homology between this BSAP consensus and site B, although this would not have been considered significant had we not already known that sites B and D bound the same protein. BSAP, a mammalian homologue of the sea urchin tissue-specific activator protein (4), was originally identified by its ability to interact with conserved regulatory sequences upstream of late histone genes (5). We therefore carried out bandshift assays to determine if these site B and site D sequences could compete for BSAP binding to the sea urchin histone promoter-derived oligonucleotide probe H2B 2.1 (4). The results in Fig. 5B (top) show that the H2B 2.1 probe formed a specific complex in the presence of nuclear extract prepared from the B-cell lines Ramos and DG75, but not in extracts from the T-cell lines CEM and Jurkat, consistent with the B-cell-restricted expression of BSAP (5). Moreover, formation of this complex was abrogated by either site B or site D competitor, implying that the same B-cell-restricted factor interacted with all three sites. This was confirmed in parallel experiments, in which incubation of B-cell nuclear extract with a site D (Fig. 5B, middle) or site B (Fig. 5B, bottom) probe gave rise to a complex with electrophoretic mobility similar to that of the complex formed in the presence of the H2B 2.1 probe. Supershift assays further confirmed that the complexes formed by all three probes contained BSAP; thus, addition of a BSAP-specific polyclonal antibody led to the inhibition of complex formation in all three cases, whereas a control YY1-specific antibody had no effect. Finally, we demonstrated that IVT BSAP formed a single complex with the H2B 2.1, site D, and site B probes, which in each case comigrated with that formed in the presence of B-cell nuclear extract. As expected, these complexes with IVT protein were abrogated by the addition of either site B or site D competitor or an anti-BSAP antibody, confirming that the B-cell-specific factor which binds to these Wp sites is BSAP.

View larger version (58K): [in this window] [in a new window]   FIG. 5.   Interaction of BSAP with sites D and B. (A) Sequence comparison of a published consensus BSAP binding site (9), the known BSAP site present at the sea urchin late histone promoter (H2B 2.1), and EBV sites D and B. Shaded nucleotides indicate positions matching the consensus sequence. (B) Protein-DNA complexes obtained by incubating the H2B 2.1 (top), site D (middle), and site B (bottom) probes with either nuclear extracts (lanes 1 to 8) or IVT BSAP protein (lanes 9 to 13). Lanes 1 to 4, probe plus the indicated nuclear extract; lanes 5 and 6, probe and DG75 nuclear extract in the presence of the indicated competitor oligonucleotide; lanes 7 and 8 probe plus DG75 nuclear extract in the presence of BSAP-specific or YY1-specific antiserum; lane 9, probe and IVT BSAP; lanes 10 to 13, same as lanes 5 to 8 but with IVT BSAP replacing DG75 nuclear extract.

Effect of BSAP binding on Wp activity. Having demonstrated that BSAP binds to two sites in Wp, we next determined the contribution of these sites to Wp activity in different B and non-B-cell lines. The Bm (site B) and Dm1 and Dm2 (site D) mutations, known to abrogate BSAP binding in vitro (Fig. 5), were introduced either singly or in combination into the Wp440 luciferase reporter. The activities of the wild-type and mutant promoter constructs were then compared in transient transfection assays (Fig. 6A). In the three B-cell lines tested (Akata, Ramos, and DG75), mutation of either site B or site D led to a three- to fourfold drop in luciferase expression compared to cells transfected with the wild-type Wp 440 reporter. However, there was an even greater reduction of 5- to 10-fold when both BSAP binding sites were mutated. By contrast, when the experiments were repeated in the non-B-cell lines K562 (Fig. 6A) and CEM (data not shown), mutation of either one or both BSAP binding sites had no significant effect on the low basal levels of promoter activity.

View larger version (23K): [in this window] [in a new window]   FIG. 6.   Contribution of BSAP to Wp activation. (A) Effects of site B and site D mutations on Wp activity in B and non-B cells. Promoter activity was assayed by quantifying luciferase expression in cells transiently transfected with the wild-type Wp reporter (Wp440), a truncated reporter lacking both UAS1 and UAS2 sequences (Wp87), or mutant derivatives of Wp440 carrying Bm, Dm1, and Dm2, alone or in combination (Bm+Dm1 and Bm+Dm2). The reported luciferase values (relative light units [RLU]) are means ± standard deviations of three independent experiments. (B) Effect of ectopic BSAP expression on Wp activity in B and non-B cells. Promoter activity was assayed by quantifying luciferase expression in cells transiently transfected with the BSAP expression vector BSAP/pSG5 and wild-type or mutant Wp reporters as in panel A. In each case, the results (means ± standard deviations of three independent experiments) are represented as the fold activation of luciferase activity seen in cells cotransfected with BSAP/pSG5 and the relevant Wp reporter over that seen in cells cotransfected with pSG5 and the relevant reporter.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

EBV is not exclusively B lymphotropic but can access a range of other cell types, albeit inefficiently, by CR2-dependent or -independent routes (25, 39, 49, 53, 63). However, the growth-transforming program of viral gene expression does not appear to be activated in these other environments. Therefore, what determines the B-cell specificity of the transformation process at a postreceptor/postviral entry stage? This report addresses what may be one of the central aspects of this question, namely, the cell lineage-specific controls governing the activation of Wp, the viral promoter that initiates the growth-transforming program in B cells but which either is silent or shows only transient low-level activity following experimental infection of other cell types. The first set of experiments used large Wp sequence probes in bandshift assays to provide a comprehensive view of the factors binding to the 350 bp of Wp encompassing the main lineage-independent region (UAS2) and B-cell-specific region (UAS1) of the promoter. The only binding detectable within UAS2 mapped to the known YY1 site and involved two complexes, one formed by YY1 alone (6) and a larger complex of YY1 and a second, as yet unidentified protein which is presumably recruited to the site through an interaction with YY1 itself. Both factors are found in a variety of cell types, consistent with the lineage-independent nature of UAS2. Moving to the B-cell-specific region UAS1, the assays identified a new binding site, site D, upstream of the previously described interactions at the CREB site (27) and at the adjacent sites B and C. Most importantly, while site C bound ubiquitously expressed proteins, both site B and the new site D bound B-cell-specific factors (Fig. 1). The main thrust of the work was to identify these UAS1 binding proteins and assess their role in Wp activation.

Focusing first on site C, the evidence from bandshift assays (Fig. 2B), from supershift assays with specific antibodies (Fig. 2C), and from mutational analysis (Fig. 3) strongly suggests that members of the RFX family of proteins bind at this site and that this interaction is an important determinant of Wp activity in B cells. The RFX proteins are a novel family of transcriptional regulators with a conserved 76-amino-acid DNA binding domain (13). Of the five family members (RFX1 to RFX5) known in mice and humans, RFX1, -2, and -3 can bind as either homo- or heterodimers to similar inverted repeat sequences in DNA, in some cases along with another cellular factor, MIBP1. The present antibody shift assays indeed suggest that the complexes found at site C in vitro predominantly contain homo- and heterodimers of RFX1, MIBP1, and RFX3 (Fig. 2C). Interestingly, similar complexes have been observed in in vitro binding assays with several other viral enhancer sequences, notably from cytomegalovirus, polyomavirus, and HBV, and in the latter two cases there is direct functional evidence that such binding has a regulatory role (7, 10, 45, 46, 51). However, in contrast to the polyomavirus and HBV enhancers, where activation requires RFX binding to the full inverted repeat sequence (10), it appears that only the 5' half of Wp site C needs to be conserved both for RFX complexes to form and for optimal Wp activity to be maintained (Fig. 3). In this context, we noted that IVT RFX1 binds to Wp site C both as a monomer and as a dimer (H. Kirby, unpublished observations) and that the recently published structure of the RFX1-DNA complex (17) suggests that the two RFX1 monomers bind independently to their cognate half sites. In the in vivo situation, it is possible that binding of an RFX monomer to the 5' half of site C either is itself sufficient for Wp activity or facilitates the binding of a second monomer to the adjacent half site.

The realization that site C binds RFX proteins also highlighted a potential parallel between Wp and the intensively studied major histocompatibility complex class II promoter (MHCIIp) (32). As illustrated diagrammatically in Fig. 7, in addition to an NF-Y site, MHCIIp possesses functionally important RFX and CREB sites (usually termed the X and X2 boxes, respectively) in similar orientation and position upstream of the transcription start as seen in Wp (32, 36). In the case of MHCIIp, however, the RFX site binds a different family member, RFX5, in vivo; the consensus motif of RFX5 is related to that of RFX1 to -3 but differs in that it forms a heterotrimeric complex with two other ubiquitously expressed proteins, RFX-AP and RFX-B/RFX-ANK (12, 34, 37, 55). The cell lineage restriction over MHC class II gene expression is determined by the presence of the class II transactivator (CIITA), which binds to one or more of the above components, leading to promoter activity (54, 60). Genetic loss of any one of these four components abrogates MHC class II expression and is manifest clinically as an immunodeficiency disease, bare lymphocyte syndrome (32). Evidence to date suggests that these factors which are important at the MHC class II locus are not involved in Wp regulation. Thus, we found no evidence in antibody shift assays for RFX-5 binding to the site C probe (Fig. 2), and transient transfection assays in CIITA-deficient B-cell lines showed Wp to be fully active in such cells (Kirby, unpublished observations).

View larger version (23K): [in this window] [in a new window]   FIG. 7.   Schematic representation of the principal cis-acting functional domains and their cognate binding factors at the EBV Wp, the MHCIIp, and HBV enh-I. (A) Factors binding the B-cell-specific element UAS1 within Wp include the ubiquitously expressed RFX and CREB/ATF proteins and the B-lineage-restricted protein BSAP; sequence numbers shown are relative to the Wp RNA start site. (B) Factors binding at MHCIIp include ubiquitously expressed components of the RFX complex (RFX5 and its associated proteins RFX-AP and RFX-B/ANK) plus ubiquitously expressed CREB and NF-Y proteins, binding to the X, X2BP, and Y box sequences, respectively (32, 36); sequence numbers shown are relative to the MHCIIp RNA start site. Lineage-restricted activity of MHCIIps is determined by expression of CIITA, which binds to one or more above factors, leading to promoter activation. (C) Factors binding to HBV enh-I include ubiquitously expressed RFX, CREB, and NF1 proteins, plus hepatocyte-specific HNF3, HNF4, RXR, and C/EBP proteins; sequence numbers refer to coordinates on the HBV genome. In each case the arrow denotes the relevant transcription start site; note that Xp is the nearest of several HBV promoters activated by HBV enh-I.

A number of other RFX-dependent genes have been reported to be regulated in a cell-type-specific manner (22). Most interesting in this regard is HBV enh-I, which is selectively active in liver cells. Like Wp, the activity of HBV enh-I is dependent on the binding of RFX and CREB proteins at similar locations upstream of the transcription start site (Fig. 7C). However, liver-specific activity requires the binding of additional proteins, including hepatocyte-specific factors HNF3, HNF4, RXR, and C/EBP, to nearby sites in the HBV enh-I sequence (18, 20, 51). The present work reveals a similar situation with respect to Wp, where the promoter's preferential activity in B cells requires two additional sites, B and D, both of which interact with the B-cell-specific protein BSAP/Pax-5. BSAP, a member of the Pax family of transcription factors, is highly lineage restricted in its expression, being found only in B lymphocytes, in the developing brain, and in adult testis (1). Expression in the B lineage begins in the earliest B-cell precursor and continues through all stages of differentiation except the end-stage plasma cell (5). In fact, targeted gene disruption in mice has shown that BSAP is required for B-cell differentiation to proceed beyond the pro-B-cell stage (40, 59). This reflects a key role for BSAP as a determinant of B-cell commitment through its ability both to activate B-cell-specific genes such as CD19 and to suppress the transcription of genes specific for other hemopoietic lineages (41). This versatility as a transcriptional regulator appears to reflect the existence of both activating and inhibitory domains in the protein's C-terminal regulatory region (11). Like all members of the Pax family, BSAP binds DNA through a paired domain, itself composed of two subdomains that each interact with one half of a DNA recognition motif. The ability of one strong subdomain-half-site interaction to compensate for weaker binding at the other half-site (9) helps to explain the degenerate nature of the 18-bp consensus sequence for BSAP binding. We found that Wp site D and particularly site B show only limited homology to this consensus, yet clearly both bind BSAP in vitro, with the interaction at site D being the stronger (Fig. 4 and 5). Most importantly, mutations in site B and/or D which abrogated BSAP binding substantially reduced Wp activity in B-cell lines but hardly affected the promoter's low baseline activity in non-B cells. Conversely, ectopic expression of BSAP in a non-B-cell environment significantly enhanced Wp activity (Fig. 6), implying that BSAP is the main if not the only B-cell-specific cellular protein involved in Wp activation.

It has to be stressed that our experiments identifying roles for CREB, RFX, and BSAP in Wp activation rely on transient transfection assays of Wp reporter constructs in established B-cell lines, and further work will be required to determine how many of these regulatory controls are important in the natural cellular environment in which Wp is activated, i.e., in mature resting B cells. Importantly, however, we have demonstrated that all of these factors are present in nuclear extracts of resting B cells and are therefore presumably available for binding and activation of Wp during virus infection of such cells in vivo. Note also that these factors were present in all EBV-negative and EBV-positive B-cell lines tested, including EBV-positive BL lines such as Akata in which both Wp and Cp on the resident viral genome are transcriptionally silent; this finding is consistent with previous reports (6, 23, 33) that all B-cell lines support the expression of transiently introduced Wp reporter constructs, irrespective of EBV status. Indeed, in cell lines where the endogenous viral Wp is silent, it is clear that the promoter is maintained in an inactive state by CpG methylation rather than by the absence of any essential DNA binding factors (23, 33).

In summary, this work provides the first evidence for the coinvolvement of CREB and RFX family members with BSAP in promoter regulation. Since the loss of a binding site for any one of these factors severely impairs the B-cell-specific activity of Wp, it would appear that transcriptional activation requires the formation of a specific multiprotein complex. In this regard, we note that in in vitro bandshift assays, CREB, RFX, and BSAP were each capable of binding independently to the relevant short oligonucleotide containing their cognate sequence as well as to longer probes containing all three sites. Furthermore, using the longer probes, we saw no clear evidence of cooperative binding as reflected by higher complex formation. However, there may well be cooperative interactions in vivo that stabilize the multiprotein complex at UAS1 and/or enhance its engagement with the transcriptional machinery. The important point is that even though CREB and RFX factors are ubiquitously expressed, the critical requirement for BSAP means that Wp will be efficiently activated only in B lymphocytes. In the infected host, virus dissemination within the B-cell system appears to require a period of virus-driven growth transformation followed by a down-regulation of the transforming genes, thereby allowing latently infected cells to persist in the resting state. This tight physiological control might well be lost, however, if the growth-transforming program were to be activated indiscriminately in other cell types. Arguably, EBV has evolved to exploit the B-cell-restricted nature of BSAP, possibly as one of several strategies, to ensure that viral transformation is initiated only in the appropriate cell environment.