Plasma Cell-Specific Transcription Factor XBP-1s Binds to and Transactivates the Epstein-Barr Virus BZLF1 Promoter

Abstract

Epstein-Barr virus (EBV) in vivo is known to establish persistent infection in resting, circulating memory B cells and to productively replicate in plasma cells. Until now, the molecular mechanism of how EBV switches from latency to lytic replication in vivo was not known. Here, we report that the plasma cell differentiation factor, XBP-1s, activates the expression of the master regulator of EBV lytic activation, BZLF1. Using reporter assays, we observed that XBP-1s was able to transactivate the BZLF1 promoter, Zp, in a plasma cell line and other lymphoid cell lines but, interestingly, not in epithelial cell lines. We have identified an XBP-1s binding site on the ZID/ZII region of Zp, which when abolished by site-directed mutagenesis led to abrogation of XBP-1s binding and promoter activation. Using the chromatin immunoprecipitation assay, we observed direct binding of XBP-1s to endogenous Zp in an EBV-infected plasma cell line. Finally, in the same cell line, we observed that overexpression of XBP-1s resulted in increased expression of BZLF1, while knockdown of XBP-1s with short hairpin RNA drastically reduces BZLF1 expression. We suggest that EBV harnesses the B-cell terminal differentiation pathway via XBP-1s as a physiological signal to reactivate and begin viral replication. We are currently investigating other signals, such as the endoplasmic reticulum stress response proteins, which act upstream of XBP-1s, to identify other interacting factors that initiate and/or amplify the lytic switch.

Epstein-Barr virus (EBV) is a human gammaherpesvirus that establishes persistent latent infection in long-lived memory B cells (48). Recent studies have led to a model of EBV persistence whereby the virus uses different transcription programs within the context of the normal biology of B lymphocytes in order to carry out its life cycle (48). Paralleling the effects of antigen on B-cell activation and differentiation, EBV expresses the growth, default, and latency transcription programs at the B-cell blast, germinal-center B cell, and memory B-cell stages, respectively, in order to gain access to and persist within memory B cells. It is here that the virus shuts down all viral protein encoding genes to evade cytotoxic-T-cell immunosurveillance and achieve long-term persistence (18). Once in the memory B cell, the virus may replicate its latent genome as the cell undergoes cellular division (19, 34), or it can reactivate from latency by replicating and producing infectious progeny for transmission.

Immunohistochemical staining of infectious mononucleosis and healthy tonsils suggested that the virus replicates in cells that have a plasma cell morphology (1, 11, 35, 36). We have previously shown, by cell surface phenotyping of tonsil B-cell subsets and sensitive PCR-based detection of viral lytic gene expression that the EBV lytic program does indeed occur in plasma cells in vivo (29). Furthermore, we observed that neither signals involved in memory B-cell activation nor signals from the virus were sufficient to induce reactivation. Rather, plasma cell differentiation signals are required to initiate EBV reactivation. Precisely what these signals are remains to be resolved.

EBV reactivation is characterized by the expression of the immediate-early gene, BZLF1. It encodes the viral lytic regulator, Zta, which is often referred to as the lytic switch. BZLF1 expression alone is sufficient to activate downstream lytic genes and complete viral replication in a permissive cell type (10, 46). The immediate-early transcription factor, Zta, is crucial for inducing the expression of early lytic genes such as BALF2, involved in viral DNA replication, and of late lytic genes such as BcLF1 for viral packaging (13).

The regulation of BZLF1 during reactivation is currently best understood from in vitro assays. Reagents such as 12-O-tetradecanoylphorbol-13-acetate (TPA), N-butyric acid, or immunoglobulin-cross-linking antibodies have been shown to activate BZLF1 from latently infected B-cell lines. Investigators have used this approach to identify a number of cellular factors involved in binding and transactivating the BZLF1 promoter (Zp) (44). For example, the myocyte enhancer factor 2D (MEF2D) and Sp1/Sp3 transcription factors bind to the ZI elements of Zp, whereas basic leucine zipper transcription factors such as ATF1, CREB, and ATF2 bind to the ZII element of Zp at the CREB/AP-1 binding sites (Fig. 1) (6, 33, 51). It is of note that the ZII element is absolutely required for transactivation of the BZLF1 promoter since Zp reporter constructs with mutations at the CREB/AP-1 binding site show little activity (44). However, these in vitro systems use acute activation signals that lead ultimately to apoptosis, a mechanism that rapidly induces the expression of BZLF1 and virus production within 2 days, leaving no time for terminal differentiation to occur (22). When latently infected memory B cells are cultured, the virus can undergo reactivation (39); however, this again is an acute response associated with apoptosis, and terminal differentiation is never seen. Whether the mechanism of acute reactivation occurs in vivo is unclear. Thus, there are either two mechanisms by which EBV can reactivate in vivo (acutely or in association with differentiation) or the acute mechanism is an in vitro artifact.

FIG. 1.

Schematic of the BZLF1 promoter Zp. Depicted here are the regulatory elements of the BZLF1 promoter Zp. The nucleotide sequence of the ZID/ZII region, which we designate ZPRO, is shown. ZPRO contains two putative XBP-1 binding motifs designated Zp1 and Zp2. The XBP-1 consensus binding motif is shown below the ZPRO sequence.

Unlike apoptosis, physiological plasma cell differentiation is a process that spans several days (4, 29). This lengthy process is necessary to create the ideal environment for secreting immunoglobulin. It involves a complex interplay of cellular division, plasma cell-specific transcription factors BLIMP-1 and XBP-1, and the unfolded protein response (UPR).

When a memory B-cell is activated by cognate antigen, T-cell help, and cytokines, it begins the differentiation process by undergoing several rounds of cellular division. This is known as the plasmablast stage, where stochastic division is a mechanism to acquire the plasma cell phenotype (17, 47). First, the master regulator of plasma cell differentiation, B-lymphocyte-induced maturation protein 1 (BLIMP-1), is expressed, and its level is increased upon each cellular division (42). Second, BLIMP-1 functions to repress Pax5, a negative regulator of another plasma cell differentiation factor, X-box binding protein-1 (XBP-1) (31). Therefore, BLIMP-1 indirectly induces transcription of XBP-1. The expression of XBP-1 is essential for plasmacytic differentiation because mice without XBP-1 do not make any immunoglobulin-secreting plasma cells despite having normal levels of BLIMP-1 (38).

XBP-1 transcription is also induced by the UPR pathway. This elegant signaling pathway ensures cell survival during endoplasmic reticulum (ER) stress resulting from misfolded proteins, altered glycosylation, or overwhelming protein synthesis (40). Under normal conditions, ER stress proteins ATF6 and inositol-requiring 1α (IRE1α) are sequestered in the ER by Bip chaperone proteins. Upon ER stress, Bip binds misfolded proteins for folding correction, in effect releasing ATF6 and IRE1α to the cytoplasm, where they function to increase and splice XBP-1 transcripts, respectively. Therefore, during unstressed conditions, XBP-1 protein exists in its unspliced, transcriptionally inactive state (XBP-1u), while ER stress results in the generation of a potent transcription factor, XBP-1s (52). XBP-1s is a key component of the UPR since it serves to reduce ER stress by acting on genes that enhance protein folding, promote degradation, and regulate translation and apoptosis (16, 40).

XBP-1 is a basic leucine zipper protein that is part of the CREB/ATF family of transcription factors. It was first found to regulate major histocompatibility complex (MHC) class II DRα expression by binding to the promoter (32) and, subsequently, a consensus binding site for XBP-1 was defined (9). During plasma cell terminal differentiation it acts in two phases (43). The initial phase is to prepare the cell for antibody secretion where XBP-1s functions to expand the ER and increase the secretory apparatus and cell size (16, 25, 41, 50). The second phase of XBP-1s activity is to alleviate stress due to the overwhelming production of immunoglobulin in the ER. XBP-1s serves to upregulate chaperone proteins (such as Bip) that aid in folding, induce the degradation machinery to discard misfolded proteins, enable the secretion of immunoglobulin, allow continued translation, and prevent apoptosis of the plasma cell (16, 30). XBP-1s is thus considered the link between the UPR pathway and terminal differentiation into plasma cells (25).

We sought to determine here whether XBP-1s participates in the transcription of the BZLF1 promoter, thereby potentially providing the signal that leads to EBV replication in plasma cells.

MATERIALS AND METHODS

Cell lines.

MM.1S (kindly provided by Kenneth Anderson) is an EBV-positive human plasma cell line derived from a multiple myeloma. BJAB (a gift from Elliot Keiff) is an EBV-negative B-cell line. RAJI (ATCC), P3HR1, and Rael are EBV-positive Burkitt's lymphoma cell lines. Jurkat (E-6; a gift from Steve Bunnell) is a human T-lymphocyte line. HeLa cells (ATCC) are a human epithelial cervical cancer cell line; HaCaT is a human keratinocyte cell line, and Cos-7 is a monkey kidney epithelial cell line (gift from Ananda Roy). The lymphoid cell lines were grown in RPMI 1640 medium and epithelial cell lines were cultured in Dulbecco modified Eagle medium with high glucose (Gibco) at 37°C with 5% CO₂. Both media were supplemented with 10% fetal calf serum, 2 mM sodium pyruvate, 2 mM l-glutamine, 100 U of penicillin/ml, 100 U of streptomycin/ml, and 5 μg of ciproflaxin/ml. B95-8 is an EBV-positive, marmoset cell line that spontaneously reactivates the virus.

Plasmids.

We obtained the XBP-1s expression plasmid, pcDNA3.1-XBP-1s, from Katsutoshi Mori; it was used in luciferase experiments and electrophoretic mobility shift assays (EMSAs). N-terminal FLAG-XBP-1s was created by using pcDNA 3.1-XBP-1s as a template for generating FLAG-XBP-1s PCR inserts using the following forward and reverse primers carrying BamHI and BspEI restriction sites, respectively: FLAGX2F, 5′-TATGGATCCGCCACCATGGATTACAAGGATGACGACGATAAGGCTATGGTGGTGGTGGCAGCC-3′; and FLAGXR, 5′-CTGTAAGCATCCAGTAGGCAGG-3′. The insert was then cloned back into BamHI- and BspEI-digested pcDNA3.1 XBP-1s plasmid to create pcDNA3.1 N-terminal FLAG-XBP-1s (pcDNA3.1-F-XBP-1s). The pGL2:Zp luciferase construct used in luciferase assays was kindly provided by P. Farrell.

Zp1KO-luc was constructed from pcDNA3.1-XBP-1s by site-directed mutagenesis using the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's directions. For mutagenesis, the primer 5′-GGCACCAGCCTCCTCTGTGATAATTCCCTTTGGGACGTGCTAAATTTAGG-3′ was used.

Transfection.

For each transfection, 4 μg of pGL2 firefly luciferase reporter construct, driven by the EBV lytic promoter, Zp, was cotransfected along with 0.1 μg of Renilla luciferase as an internal control. When indicated, pcDNA3.1 empty vector, pcDNA3.1-XBP-1s or pcDNA3.1F-XBP-1s was included.

All lymphoid cells were washed with phosphate-buffered saline and transfected at concentrations of 2 × 10⁶ cells/100 μl of Opti-MEM I (GIBCO BRL), except for Jurkat cells at 4 × 10⁶ cells/100 μl of RPMI, and epithelial cells at 3 × 10⁵ to 2 × 10⁶ cells/100 μl of Dulbecco modified Eagle medium. Cells were electroporated by using the Squareporator BTX 830 and a multiwell chamber HT-200 (BTX) under the following conditions: for BJAB, Raji, P3HR1, and Jurkat cells, 200 V, 10 ms, and three pulses; for Rael and MM.1S cells, 360 V, 10 ms, and one pulse; for HeLa cells, 150 V, 99 ms, and two pulses; for HaCat cells, 250 V, 50 ms, and two pulses; and for Cos-7 cells, 1.5 kV, 99 μs, and two pulses. After transfection, the cells were recovered in 10 ml of complete RPMI 1640 or high-glucose DMEM, depending on the cell type, for 24 h and then incubated for another 18 h with or without 20 ng of TPA/ml. The activity of the transcription factor on the promoter was measured in a luciferase assay.

For EMSA and chromatin immunoprecipitation (ChIP) assay, MM.1S and P3HR1 cells were transfected with either 50 μg of pcDNA3.1 empty vector or 50 μg of pcDNA3.1FLAG-XBP-1s at a concentration of 1 × 10⁷ to 2 × 10⁷ cells/500 μl of Opti-MEM I (GIBCO BRL) and electroporated by using BTX 830 in a 4-mm single-well cuvette (Bio-Rad) under the conditions described above.

Luciferase reporter assay.

Cells were transfected with the Zp luciferase reporter construct, control, or XBP-1s expression vectors and Renilla luciferase using the conditions described above. Cell extracts were assayed for the firefly and Renilla luciferase activities by using the dual luciferase reporter system according to the manufacturer's directions (Promega). Unfortunately, the activity of XBP-1s interfered with the expression of Renilla luciferase, making it an unsuitable internal control. Therefore, each sample's relative light units was normalized to the total cellular protein and was expressed as the relative light units per microgram of protein for comparison between samples. Each experiment was repeated two to three times in duplicate, and the data are presented as the relative fold induction. Cell extracts containing just Zp-luc, Renilla, and 10 μg of control cytomegalovirus (CMV) plasmid were set as 1.

In vitro translation.

Recombinant XBP-1s or F-XBP-1s were generated from the pcDNA3.1XBP-1s or pcDNA3.1-F-XBP-1s constructs described above by using the in vitro translation kit PROTEINscript II (Ambion) according to manufacturer's instructions. To determine XBP-1s protein binding to putative XBP-1s binding sequences, in vitro-translated XBP-1s (ivTr XBP-1s) or F-XBP-1s (ivTr F-XBP-1s) was subjected to EMSA as described below.

EMSA.

Nuclear extracts from 10 × 10⁶ to 20 × 10⁶ MM.1S or Raji cells were prepared as described previously (45). The protein concentration was determined by using a Bradford assay (Bio-Rad). Complementary oligonucleotides (see Table 2 for a listing) were annealed for use as radiolabeled probes or cold competitors. Probes with phosphorylated 5′ ends were radiolabeled by exchange reaction using the T4 polynucleotide kinase in the presence of [γ-³²P]dATP. Free nucleotides were removed using G25 microspin columns (Amersham Biosciences). For each binding reaction, 7 μg of nuclear extract protein and 5 pmol of radiolabeled probe (∼100,000 cpm) were incubated in binding reaction buffer (20 mM Tris-HCl [pH 7.5], 100 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol) and 2 μg of sheered poly(dI-dC) (Amersham Biosciences). Where indicated, unlabeled competitor oligonucleotides were added at a 50- to 100-fold excess of radiolabeled probe. In supershift assays, anti-FLAGM2 antibody (Sigma) or anti-XBP-1 (courtesy of K. Mori) were added at 4 or 1 μg/reaction, respectively, to binding reactions, followed by incubation at 4°C 1 h prior to the addition of radiolabeled probe. The reaction mixtures (total of 20 μl each) were incubated at room temperature for 20 min and then subjected to vertical electrophoresis in a 5 or 6% Tris-borate-EDTA polyacrylamide gel. The gel was electrophoresed at room temperature in 0.25× Tris-borate-EDTA running buffer, dried on Whatman paper, and autoradiographed on Kodak MS film.

TABLE 2.

Sequences of oligonucleotides used for gel shift assays

Name	Sequence (5′-3′)a	Description	Length (bp)	Reference
HLA-DRα	CCTAGCAACAGATGCGTCATCTCAAAA	Human MHC class II promoter; XBP-1 binding site; positive control	27	32
HLA-DRαM	CCTACCAACAcATaCaTCATCTCAAAA	Mutant of HLA-DRα at XBP-1 binding site	27	32
		Negative control
Aα	CGAACTGTCACGTCATCACAAGA	Mouse MHC class II promoter; XBP-1 binding site; positive control	23	9
AMut	CGAACTGTCAccccATCACAAGA	Mutant of Aα	23
		Negative control
ZPRO	GCCTCCTCTGTGATGTCATGGTTTGGGACGTGCTAAATTTAGG	ZID/ZII region of Zp containing putative XBP-1 binding sites	43	26
Zp1	GCCTCCTCTGTGATGTCATGGTTTGGG	Putative XBP-1 binding site 1	27
Zp1KO	GCCTCCTCTGTGATaattcccTTTGGG	Mutant of Zp1	27
Zp2	CATGGTTTGGGACGTGCTAAATTTAGG	Putative XBP-1 binding site 2	27
Zp2KO	CATGGTTTGGGAcccGCTAAATTTAGG	Mutant of Zp2	27

^aPutative XBP-1 binding motifs GTCAT and ACGT are underlined. Letters in lowercase denote mutations that disrupt the binding site.

ChIP.

To determine whether F-XBP-1s can bind to Zp in vivo, we used the E-Z ChIP (Upstate) protocol with slight modifications. F-XBP-1s or empty vector transfected cells were fixed 5 days posttransfection with formaldehyde (1% final concentration) for 10 min at room temperature. Glycine was added to stop the fixing reaction, and the cells were washed with chilled phosphate-buffered saline. The cells were then lysed with sodium dodecyl sulfate lysis buffer with antiprotease cocktail or APC (Roche) and then sonicated (output 5, duty cycle 40%, 20 pulses, 10 sets on ice) using a Branson Sonifier 250. Sonicated chromatin was diluted 1:9 in ChIP buffer with APC. Approximately 1 million cell equivalents or 1 ml of sonicated chromatin was used in the following preclearing and ChIP steps. For the first preclearing step, 100 μl of 50% protein G-agarose (Pierce) slurry preabsorbed overnight with 500 μg of sonicated salmon sperm DNA (Invitrogen) and 500 μg of purified bovine serum albumin was added per ml of diluted chromatin and then mixed at 4°C for 4 h. The supernatants undergo a second preclearing step, with rocking overnight at 4°C, with normal mouse immunoglobulin G (IgG)-agarose (Sigma) at 5 μg/ml. Precleared chromatin solutions were then used for immunoprecipitation with normal mouse IgG-agarose at 5 μg/ml as a control and mouse anti-FLAG-agarose (Sigma) at 5 μg/ml for 1 h at 4°C. The agarose beads were then washed once with low-salt buffer for 15 min, twice with high-salt buffer for 5 min each time, twice with LiCl buffer for 5 min each time, and finally twice with TE buffer 5 min each time. Chromatin was eluted from the washed beads by using fresh elution buffer, reversed cross-linked at 65°C overnight, and subjected to RNase and proteinase K treatment according to the Upstate protocol. The remaining DNA was purified by ethanol precipitation for assessment by reverse transcription-PCR (RT-PCR). The primers used were as follows: promoter of Grp78, Fwd (5′-GTGAACGTTAGAAACGAATAGCAGCCA-3′) and Rev (5′-GTCGACCTCACCGTCGCCTA-3′); promoter of Zp containing XBP-1 binding site (−220 to +1), Fwd (5′-CTTCAGCAAAGATAGCAAAGGTGG-3′) and Rev (5′-TGGGCTGTCTATTTTTGACACCAG-3′); and promoter of Zp without XBP-1 binding site (−507 to −382), Fwd (5′-TTCATTAAGTTCGGGGGTCA-3′) and Rev (5′-AAGGGAGATGGCTGACACTG-3′). The PCR conditions were 95°C for 5 min, followed by 30 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min.

Immunoblots.

Equal amounts of total cellular extracts (10 to 60 μg/sample) were separated on 10% sodium dodecyl sulfate-polyacrylamide gels and electrotransferred to polyvinylidene difluoride membranes (Millipore). The membranes were blocked in 5% low-fat milk in TBS and then incubated with mouse monoclonal antibody FLAG-M2 (Sigma) at a 1:1,000 dilution for the detection of F-XBP-1s. The immunoreactive proteins were detected by horseradish peroxidase-conjugated goat anti-mouse IgG(H+L) (Zymed) and by enhanced chemiluminescence (ECL System; Millipore and Amersham Biosciences) and then visualized on Kodak X-Omat Blue film.

Real-time quantitative RT-PCR.

Total RNA was used purified by using TRIzol (Invitrogen), and some samples were treated with DNase before cDNA synthesis. mRNA was used as a template for the production of cDNA using the cDNA synthesis kit (Bio-Rad) with random primers. cDNA was used to perform quantitative real-time PCR using the SYBR green method (Bio-Rad) to detect the products of interest as described previously (37).

The primers used were as follows: XBP-1s, Fwd (5′-TGTTCTTCAAATGCCCTTCC-3′) and Rev (5′-GCTGAGAGGTGCTTCCTCGA-3′); B-actin, Fwd (5′-GCGGGAAATCGTGCGTGACATT-3′) and Rev (5′-GATGGAGTTGAAGGTAGTTTCGTG3); BZLF1, Fwd (5′-TTCCACAGCCTGCACCAGTG-3′) and Rev (5′-GGCAGCAGCCACCTCACGGT-3′); and BALF2, Fwd (5′-GTCAGGATGTTCAAGGACGTGG-3′) and Rev (5′-CTCATAGCACATACAGATGGGC-3′).

Generating shRNA against XBP-1s.

The pFRT-H1p (FRT, for RNA targeting) expression vector (a gift from Steve Bunnell) contains the RNA polymerase III-dependent H1 RNA promoter for expression of short hairpin RNA (shRNA) in human cells (49). In order to generate XBP-1-specific targeting shRNA molecules, complementary oligonucleotides were synthesized as follows for cloning into the parental vector pFRT-H1p. Each oligonucleotide pair contains a 5′ BglII and 3′ HindIII overhang and 19 nucleotides (N19) of XBP-1s specific sequence separated by a nine-nucleotide loop 5′-TTC AAG AGA-3′. The sequences of the upper and lower nucleotide strands are 5′-GATCC(N19)TTCAAGAGA(61N)CTTTTTGGA-3′ and 5′-AGCTTCCAAAAAG(N19)TCTCTTGAA(61N)G-3′, respectively. The specific targeting sequences (N19) for XBP-1s were designed with the assistance of the Dharmacon siDESIGN Center, using GenBank accession number NM_005080, against 345 to 364 bp (shXBP-1, 5′-ACAGCAAGTGGTAGATTTA-3′), 434 to 453 bp (shXBP-1.2, 5′-AGAACCAGGAGTTAAGACA-3′), and 898 to 917 bp (shXBP-1.3, 5′-AAGCTAATGTGGTAGTGAA-3′). Negative controls of each shXBP-1 targeting sequence were generated by nucleotide mutations (shControl.1, 5′-ACAGCAtGaGGatGATTTA-3′) or by shuffling the positions of shXBP-1.2 (shControl.2, 5′-AATCACGACGGGAAATAAG-3′) and shXBP-1.3 (shControl.3, 5′-ATTGTAGAGAGGATATACG-3′). The targeting sequence was subjected to BLAST search to confirm gene specificity, while the negative control sequence was confirmed to not target any human genes.

The resultant constructs were electroporated into the XBP-1s-producing cell line, MM.1S, to test for knockdown of the XBP-1s transcript. At 48 h posttransfection, cells were harvested for RNA by the TRIzol method and used for cDNA synthesis. Real-time PCR was subsequently performed on the cDNA to determine the knockdown of XBP-1s. To determine whether shXBP-1 reduced XBP-1s protein, MM.1S cells were transfected with shXBP-1 and F-XBP-1s expression constructs, and cells were harvested for cell lysates at 48 h posttransfection and subjected to immunoblot analysis for the expression of F-XBP-1s.

RESULTS

XBP-1s transactivates Zp in the plasma cell line MM.1S.

We have shown previously that tonsil memory B cells induced to terminally differentiate into plasma cells ex vivo and the plasma cell line MM.1S but not mature B-cell lines contain factors that transactivate the BZLF1 promoter, Zp (29). To determine whether XBP-1 was capable of replicating this effect, a Zp luciferase (Zp-luc) reporter construct and an XBP-1s or control expression plasmid were transfected into the plasma cell line MM.1S (Fig. 2A). XBP-1s expression markedly transactivated Zp-luc (up to 18-fold) in a dose-dependent fashion. As a positive control we also pretreated the cells with the phorbol ester TPA, a known inducer of Zp. As expected, TPA treatment transactivated Zp ∼5-fold; however, we also observed synergistic transactivation of Zp by TPA treatment and XBP-1s expression together, reaching a maximum of 45-fold activation. A similar result was seen with a FLAG-tagged version of XBP-1s (F-XBP-1s) (Fig. 2C). This was important because we have been unable to obtain antibodies to native XBP-1s that are effective to verify protein expression. The equivalent efficacy of the FLAG-tagged version meant that the FLAG epitope did not interfere with the activity of the protein and enabled us to monitor XBP-1s protein expression in this and subsequent experiments using the anti-FLAG-M2 antibody in immunoblot analysis. The dose-dependent transactivation of Zp correlated with F-XBP-1s protein expression, indicating that XBP-1s is capable of transactivating Zp in a plasma cell line.

FIG. 2.

XBP-1s dose dependently transactivates Zp in the plasma cell line MM1.S but not the epithelial cell line HeLa. (A) Dose-dependent activation of a Zp luciferase construct by coexpressed XBP-1s in the presence or absence of TPA in the plasma cell line MM1.S. Note that XBP-1s expression augmented the effect of TPA treatment on Zp by almost ninefold. (B) In contrast, XBP-1s failed to induce Zp in HeLa cells and had no additional effect on control cells where Zp was activated by TPA treatment. (C and D) Same as for panels A and B but with FLAG-tagged XBP-1s (F-XBP-1s). The expression of F-XBP-1s was visualized by immunoblot analysis with anti-FLAG antibody. Zp activity values were expressed as the relative induction over empty vector (CMV) control. The data shown are means of duplicates ± the standard deviation (SD) from two independent experiments.

XBP-1s transactivates Zp in lymphoid but not in nonlymphoid cell lines.

XBP-1s is required for the terminal differentiation of B cells into plasma cells, but this process is dependent on a number of other transcription factors (7, 38). Therefore, it was unclear whether XBP-1s alone was sufficient to activate Zp or if other factors were also required. To address this question, we repeated the studies in Fig. 2A and C in a range of cell types listed in Table 1. We observed that ectopically expressed XBP-1s could transactivate Zp in all lymphoid cell lines tested, including plasma cell, B-cell, and T-cell lines. However, it was unable to do so in any of the nonlymphoid cells, including two epithelial cell lines, HeLa and HaCaT, even though XBP-1s protein was expressed (Fig. 1B and D and Table 1). These results allow us to draw several important conclusions. First, it is apparent that XBP-1s is unable to transactivate Zp by itself but requires cofactors that appear to be lymphoid specific. Second, EBV-encoded gene products were not required for XBP-1s to function since it worked equally well in EBV-negative B-cell lines. Third, although it is known that EBV can replicate in epithelial cells, the necessary cofactors for XBP-1s function are not present in these cells, an indication that reactivation in the epithelial cell background involves a different mechanism. Lastly, treatment of all cell lines, even the African green monkey kidney cell line Cos-7, with TPA resulted in the transactivation of Zp. However, in the nonlymphoid lines, including all of the epithelial cell lines, XBP-1s alone could not transactivate Zp nor could it augment the transactivation driven by TPA. These results indicate that XBP-1s and TPA appear to function through separate pathways to transactivate Zp, and this is dependent upon the cellular context.

TABLE 1.

XBP-1s transactivates Zp in a lymphoid- specific environment

Cell line	Type	EBV status	Induction
			XBP-1s	TPA
MM.1S	Plasma cell	+	+	+
RAJI	B cell	+	+	+
Rael	B cell	+	+	+
BJAB	B cell	-	+	+
Jurkat	T cell	-	+	+
HeLa	Epithelial cell	-	-	+
HaCaT	Epithelial cell	-	-	+
Cos-7	Green monkey cell	-	-	+

MM.1S plasma cell nuclear proteins bind to putative XBP-1 sites on Zp.

Zp contains many regulatory elements that have been defined using agents that induce viral replication in vitro (44) (Fig. 1). We had observed that XBP-1s transactivated Zp; however, it was unclear whether XBP-1s directly or indirectly regulated the promoter. To address this question, we used the transcription factor binding program, MatInspector, to look for putative XBP-1 binding motifs in Zp. The program identified two such sites in the ZII domain that we have termed Zp1 and Zp2 (Fig. 1). To determine whether XBP-1s could bind to either of these sites, we first tested whether MM.1S plasma cell nuclear extracts contain endogenous XBP-1s. We performed DNA binding assays (EMSA) with the nuclear extract and a known XBP-1 binding sequence from the MHC class II promoter, Aα (Table 2) (9). As shown in Fig. 3A, MM.1S nuclear extracts formed a specific complex. The complex was XBP-1 specific because cold competitor sequences containing the XBP-1 binding site (Aα) efficiently competed away the complex, while competitor sequences containing a mutation at the XBP-1 binding site (AMut) failed to do so. We repeated the binding reactions with the human MHC class II promoter, HLA-DRAα probe, and a competing sequence HLA-DRAαM and observed the same results (results not shown) (32).

FIG. 3.

Identification of a putative XBP-1 binding site on ZPRO by EMSA. (A) MM.1S nuclear extracts were incubated with a previously defined XBP-1 binding sequence (Aα). The migration of the putative XBP-1s complex is indicated. Cold competitors containing a mutation at the XBP-1 binding site (AMut) fail to compete for binding. (B) Same as panel A, but with the ZPRO probe demonstrating a complex binding specifically to the Zp1 site. Cold competitors containing an intact Zp1 site (ZPRO and Zp2_KO) efficiently competed for the complex, but competitors with a mutation at the Zp1 site (Zp1_KO) could not. (C) MM.1S nuclear extracts formed specific complexes with a Zp1 probe. The complexes were efficiently competed away by the known XBP-1 binding sequence Aα, the Zp1 sequence itself, containing a putative XBP-1 binding sequence, or ZPRO, which includes both Zp1 and Zp2. The complexes were not competed by Aα containing a mutation in the XBP-1 motif (AMut), or either Zp1 or ZPRO when the putative XBP-1 binding motif was mutated (Zp1_KO and Z-Zp1_KO, respectively). Cold competitors were used at 50- and 100-fold excesses of the radiolabeled probe.

Next, we sought to determine whether MM.1S nuclear extracts could form a specific complex with the viral sequence, ZPRO, which contains the two putative XBP-1 binding sites, Zp1 and Zp2, respectively. Using the ZPRO probe, we observed complexes of similar mobility compared to that found with the Aα probe (Fig. 3B). To determine whether either of the two putative sites was responsible for binding the putative XBP-1s protein, oligonucleotides containing mutations at either Zp1 or Zp2 were used as 100× cold competitors (Fig. 3B). The sequences containing an intact Zp1 site, such as ZPRO or ZPRO containing a mutation at the Zp2 site (Z-Zp2_KO), were able to efficiently compete away the complex, an indication that Zp2 did not participate in complex formation. In contrast, cold sequences of ZPRO that contained a mutation at the Zp1 site (Z-Zp1_KO) failed to efficiently compete for the complex. These results suggest that the Zp1 site but not the Zp2 site was responsible for forming the complex.

To confirm this conclusion, we tested for direct binding of the MM.1S nuclear extracts to a shorter probe containing the Zp1 site (Zp1). This site contained a GTCAT core XBP-1s binding motif. Specific complexes were formed by using the Zp1 probe. The complexes were specific for the XBP-1 binding site because they were eliminated in the presence of an unlabeled Aα competitor oligonucleotide containing an XBP-1 site but not when the XBP-1 site was mutated (AMut). Similarly, the complexes were specific to the GTCAT motif since competitors containing the intact motif, Zp1 and ZPRO, fully competed away the complex, whereas competitors mutated in the motif, Zp1_KO and Z-Zp1_KO, did not (Fig. 3C). Thus, we have shown that a complex present in plasma cell nuclear extracts, which binds to the GTCAT motif in Zp1, can be efficiently competed for with a previously defined XBP-1 binding site. This suggests that XBP-1s protein binds to Zp1.

Recombinant XBP-1s binds to Zp1 site.

To test directly whether XBP-1s binds to the Zp1 site in ZPRO, we first generated recombinant untagged and FLAG-tagged XBP-1s (XBP-1s and F-XBP-1s, respectively) proteins by in vitro translation for the binding reaction. We observed (Fig. 4A, left panel) that untagged XBP-1s and F-XBP-1s both bound the XBP-1s-specific Aα oligonucleotide and produced complexes of the same size, an indication that the FLAG epitope did not affect DNA-binding ability. The FLAG tag on the protein allowed us to perform a supershift assay with the anti-FLAG antibody and confirm that F-XBP-1s participated in complex formation with Aα. In the same way, when we incubated F-XBP-1s with the ZPRO probe (Fig. 4A, right panel), it again produced a specific complex that was supershifted with anti-FLAG antibody. The F-XBP-1s complex was specific to the Zp1 site of ZPRO because the same mobility complex was not seen when the protein was incubated with the Z-Zp1_KO probe.

FIG. 4.

Recombinant XBP-1s binds ZPRO at the Zp1 site. (A) In vitro-translated untagged (ivTr XBP-1s) and FLAG-tagged (ivTr F-XBP-1s) XBP-1s bound to both the Aα and ZPRO probes in an EMSA. Complexes containing F-XBP-1s were supershifted with an anti-FLAG antibody in both cases. F-XBP-1s did not bind to a ZPRO probe where the putative XBP-1s motif, Zp1, was mutated (Z-Zp1_KO). The location of specific complexes is denoted with an arrow, and that of the supershifted complexes is denoted with an asterisk. (B) In vitro-translated untagged XBP-1s (ivTr XBP-1s) binds specifically to the XBP-1 binding site of Aα and the GTCAT motif of Zp1. Cold competitors containing intact XBP-1binding sites (Aα, Zp1) efficiently competed for complex formation in both cases; however, cold competitors with mutated sites (AMut, Zp1_KO) did not. (C) F-XBP-1s protein expressed in MM.1S cells binds to Zp1. Cell lysates from MM.1S cells transfected with an F-XBP-1s expression vector (MM.1S F-XBP-1s) bound a Zp1 probe in an EMSA, and the complex was supershifted with an anti-FLAG antibody. Lysates from cells transfected with the empty vector control (MM.1S CMV) failed to form complexes with Zp1.

To confirm that XBP-1s was binding specifically to Zp1, cold competitors carrying either intact or disrupted putative XBP-1 binding sites were added to the binding reaction. As observed previously with plasma cell nuclear extracts in Fig. 3C, sequences containing a mutation at the XBP-1 binding site (AMut) or the putative XBP-1 binding site (Zp1_KO) failed to compete for XBP-1s protein binding to Zp1. However, sequences containing an intact XBP-1 binding site (Aα) or the intact putative XBP-1 binding site in Zp1 efficiently competed for the complex (Fig. 4B, left panel). The same result was seen in a control experiment (Fig. 4B, right panel) when XBP-1s complexes, bound to the Aα probe, were successfully competed away by the addition of the putative XBP-1 binding probe (Zp1) but not with the mutated probe (Zp1_KO). Competition for the complex by the Zp1 probe was not as efficient as expected; however, this could be due to XBP-1s having a stronger binding affinity for nucleotides flanking the core motif in Aα than for those in Zp1 (9). These competition assays confirm that the putative XBP-1 site (Zp1) identified by MatInspector, which bound complexes from plasma cell nuclear extracts, also bound recombinant XBP-1s.

In vitro-translated XBP-1s bound to Zp1; however, the binding activity of recombinant XBP-1s may not be representative of XBP-1s produced in a cell where posttranslational modifications could occur. Therefore, we repeated the supershift assay with cell lysates from F-XBP-1s-transfected MM.1S plasma cells and the Zp1 probe. As expected, we observed a specific complex corresponding to cellular expressed F-XBP-1s binding to the Zp1 probe, and the complex can be supershifted with the anti-FLAG antibody (Fig. 4C).

In summary, recombinant XBP-1s produced in vitro or in vivo was able to bind specifically, through the consensus XBP-1 binding sequence, to the Zp1 site in the ZII domain of the BZLF1 promoter.

Abolishing the XBP-1s binding site, Zp1, abrogates XBP-1s-mediated Zp transactivation.

Having determined that XBP-1s can bind directly to Zp in vitro, we next tested whther the XBP-1s binding site was required for XBP-1s to transactivate Zp in the luciferase reporter assay. In Fig. 5, increasing amounts of F-XBP-1s were able to transactivate the wild-type Zp reporter (Zp-luc) as observed previously. However, F-XBP-1s was unable to transactivate a Zp promoter construct containing a mutation at the XBP-1 binding site (Zp1_KO-luc). Western blot analysis confirmed that the lack of transactivation activity with Zp1_KO-luc was not because of depressed F-XBP-1s expression.

FIG. 5.

The Zp1 site is required for transactivation of Zp by XBP-1s in MM.1S cells. A luciferase assay was performed as described in Fig. 2 except that cells were transfected with either a wild-type Zp-luc reporter construct or one in which the XBP-1 binding site in Zp1 was mutated (Zp1_KO-luc). As shown in Fig. 2, XBP-1s transactivated the Zp reporter in a dose-dependent manner; however, XBP-1s failed to transactivate the Zp1_KO reporter. Immunoblot analysis with anti-FLAG antibody confirmed that the FLAG-tagged protein was expressed. Mutation of the Zp1 site also abrogated activation driven by TPA treatment. Zp activity is reported as the fold induction over the empty vector control.

It has been shown previously that an intact ZII domain is essential for the induction of Zp by TPA (14). The XBP-1s binding site, Zp1, is located at the ZII domain. Therefore, we tested the effect on TPA-mediated activation of mutating the XBP-1s site in Zp1. We found that when the site was mutated, TPA signaling was unable to activate the promoter (Zp1_KO-luc). Thus, although XBP-1 and TPA function differently in lymphoid and nonlymphoid cells (see above), they seem to utilize the same site for transactivating Zp. TPA is known to induce the binding of certain members of the CREB family of transcription factors such as ATF1, CREB, and ATF2 to the ZII element in epithelial cells (33, 51). This suggests that the virus could use either the TPA signaling pathway or XBP-1s binding at the ZII domain to transactivate Zp depending on the cell it infects. We conclude that the Zp1 site is necessary for XBP-1s transactivation of Zp.

XBP-1s binds to Zp in vivo.

We have identified an XBP-1s binding site on Zp and have demonstrated a direct association of XBP-1s to Zp by using in vitro binding assays. To demonstrate that XBP-1s could bind in vivo to endogenous Zp at the region indicated by our in vitro experiments, we utilized the ChIP assay. We used the EBV-positive MM.1S plasma cell line to determine whether ectopically expressed F-XBP-1s could bind to endogenous Zp. Cells expressing F-XBP-1s were sonicated to obtain sheered DNA that averaged 200 to 600 bp. Immunoprecipitation of F-XBP-1s with anti-FLAG-conjugated agarose beads allowed analysis of the DNA bound to F-XBP-1s by RT-PCR. We observed that F-XBP-1s was recruited to the Grp78 promoter in vivo, as previously described by others (Fig. 6) (12). Similarly, in samples where F-XBP-1s was immunoprecipitated, we were able to detect PCR products using primers against the −220 to +1 positions of the BZLF1 promoter that includes the Zp1 site, indicating that XBP-1s was binding to Zp in vivo. However, when we performed RT-PCR for a different region of Zp (−507 to −382) or an irrelevant promoter (GAPDH [glyceraldehyde-3-phosphate dehydrogenase]) that did not contain XBP-1 binding sites, we detected no PCR products that corresponded to this region of Zp or the GAPDH promoter. Therefore, F-XBP-1s was able to bind the endogenous BZLF1 promoter in vivo, and this binding was localized to the region containing the Zp1 site of Zp.

FIG. 6.

F-XBP-1s binds Zp in vivo as detected in a ChIP assay. MM.1S cells were transfected with control CMV or F-XBP-1s expression plasmids and cultured in the presence or absence of TPA. The chromatin from these cells was analyzed for in vivo binding of F-XBP-1s to the endogenous Zp promoter. All samples were subjected to immunoprecipitation by control IgG-conjugated agarose beads (IgG) or anti-FLAG-conjugated agarose beads (αFLAG) for 1 h, and the chromatin was eluted and purified for assessment by RT-PCR. The figure shows the PCR products from undiluted and twofold-diluted samples (denoted with wedges). F-XBP-1s was observed to bind to the positive control (Grp78) and the Zp sequence (−220 to +1) (highlighted with boxes) that includes Zp1. It did not bind to the negative controls GAPDH and Zp sequences outside of Zp1 (−507 to −382). TPA treatment enhanced endogenous binding to Grp78 and Zp by about twofold. The “+” denotes PCR of total input chromatin prior to immunoprecipitation; the “−” denotes a no-template control for the PCR. *, Empty lanes.

Ectopic expression of XBP-1s upregulates lytic gene expression in MM.1S cells.

We have demonstrated that XBP-1s was recruited to endogenous Zp and that XBP-1s can transactivate the BZLF1 promoter in transient assays. To test whether XBP-1s was able to activate endogenous lytic gene expression, we sought to determine whether expressing F-XBP-1s in MM.1S cells could induce expression of the immediate-early gene BZLF1 as measured by real-time PCR. In Fig. 7, we show that MM.1S cells transfected with F-XBP-1s but not empty vector controls upregulated BZLF1 transcripts, indicating that XBP-1s was capable of triggering expression of the viral lytic gene, BZLF1. It is known that increased expression of BZLF1 is sufficient to activate the lytic cycle therefore, since XBP-1s increased BZLF1 expression, we should observe an increase in downstream lytic genes such as BALF2. Transcription of the early lytic gene BALF2 was also induced by the ectopic expression of F-XBP-1s (Fig. 7). Consistent with the reporter assays above, TPA treatment and F-XBP-1s expression synergistically activated BZLF1 transcription, which also led to increased BALF2 transcription. In summary, the plasma cell-specific transcription factor XBP-1s is capable of activating the lytic cycle by directly binding to Zp and triggering expression of the BZLF1 gene, which in turn activates the transcription of downstream lytic genes.

FIG. 7.

Ectopically expressed F-XBP-1s enhanced lytic gene transcription in MM.1S cells. MM.1S cells were transfected with either control CMV or F-XBP-1s expression vectors and then analyzed for expression of the immediate-early viral gene BZLF1 and early viral gene BALF2. The levels of each RNA were determined by RT-PCR. The data are expressed as relative RNA levels compared to MM.1S cells transfected with the control plasmid (CMV, first bar). B958, a marmoset cell line that undergoes spontaneous EBV reactivation, was used as a positive control for lytic gene expression. Raji, a latently infected B-cell line was used as a negative control. Expression of XBP-1s or XBP-1s with TPA treatment increased BZLF1 and BALF2 transcription. The data shown are means of duplicates ± the SD from two independent experiments.

Knockdown of XBP-1s inhibits BZLF1 expression in MM.1S cells.

We have shown that the expression of XBP-1s induced endogenous BZLF1 transcription. To verify that XBP-1s is essential for BZLF1 expression, we conducted a knockdown experiment to test whether inhibiting XBP-1s expression would cause a parallel reduction of BZLF1. We used the use of a shRNA suppression construct which was targeted against XBP-1s mRNA (shXBP-1). We first sought to determine whether shXBP-1 was capable of knocking down exogenously expressed F-XBP-1s. Immunoblotting for the FLAG-tagged protein (Fig. 8A) revealed that the short hairpin dramatically reduced XBP-1s protein levels. This demonstrates that the short hairpin RNA was effective and could be used against endogenous XBP-1s. Since we were unable to detect endogenous XBP-1s protein due to lack of available antibodies, we next analyzed the effect of shXBP-1 on the levels of endogenous XBP-1 and BZLF1 transcription. As shown in Fig. 8B and D, shXBP-1 was able to dramatically reduce the levels of endogenous XBP-1 expression, and this led to a similar decrease in endogenous BZLF1 transcription. This experiment demonstrates that endogenous XBP-1 is capable of transactivating the endogenous BZLF1 promoter.

FIG. 8.

Knockdown of XBP-1s expression with shRNA decreased lytic gene expression in MM.1S cells. (A) MM.1S cells expressing FLAG-tagged XBP-1s (F-XBP-1s) or a control (CMV) were transfected with an expression vector for a short hairpin targeting XBP-1s RNA (shXBP-1) mixed with various ratios of a control shRNA (shControl). The cells were harvested at 48 h posttransfection and analyzed for expression of F-XBP-1s. The shXBP-1 construct efficiently reduced XBP-1s protein expression, assessed by Western blot analysis, which correlated with a reduction in BZLF1 transcript levels, as assessed by RT-PCR (results not shown). (B) Same as panel A except the cells were subsequently assessed by RT-PCR for expression levels of endogenous XBP-1 transcripts. The data are expressed as the fold change compared to untreated cells. (C) The levels of endogenous BZLF1 gene expression were measured on the same samples shown in panel B. The data shown in panels B and C are means of duplicates ± the SD from two independent experiments.

DISCUSSION

We have shown that the plasma cell-specific transcription factor, XBP-1s, transactivates the gene encoding BZLF1, the first step in the gene expression cascade that eventually leads to virus production. We have further demonstrated that XBP-1s binds to a site, designated Zp1, that contains a consensus XBP-1 binding motif in the ZII element of Zp (the BZLF1 promoter) both in vitro and in vivo and thereby transactivates Zp. This observation provides an explanation as to why EBV-infected plasma cells but not EBV-infected memory B cells in vivo express lytic proteins and provides a possible molecular mechanism for how EBV uses plasma cell differentiation to activate its lytic cycle (1, 29, 35, 36).

XBP-1s links both plasma cell differentiation and the UPR (24). We discuss here how EBV might use XBP-1s within these two contexts to regulate its own reactivation. Circulating latently infected memory B cells occasionally home to the tonsils where they would encounter cognate antigen and bystander T-cell help to undergo terminal differentiation into plasma cells (5). It has been proposed that signaling through the B-cell receptor (BCR) targets the transcription factor, Bcl-6 (required for mature B-cell function), for ubiquitination, which allows for expression of the plasma cell transcription factor, Blimp-1. By a mechanism not completely understood, BCR signaling triggers the first phase of the UPR, enabling the initial expression and splicing of XBP-1 (43). The first phase of XBP-1s activity helps to establish an environment suitable for secreting immunoglobulin (41). By this time, the latently infected memory cell has become a plasmablast, which undergoes active cellular division to gain a more differentiated plasma cell phenotype. Blimp-1 also increases XBP-1 transcription by alleviating Pax5 repression of XBP-1. Since inhibition of Pax5 leads to active transcription and production of immunoglobulin, the ER soon becomes overwhelmed with proteins and triggers the stress response. ER stress proteins ATF6 and IRE1 are released from the ER to generate more XBP-1s (52). This transcription factor is then responsible for activating genes involved in misfolded protein degradation, immunoglobulin translation and secretion, and inhibiting apoptosis in the plasma cell (16, 24, 25, 30). We hypothesize that in an EBV-infected plasma cell, the virus would shunt XBP-1s for the purposes of activating BZLF1, leading the plasma cell to fully devote its machinery to virus production and not immunoglobulin production (Fig. 9).

FIG. 9.

Model of EBV reactivation in a plasma cell. BCR signaling and T-cell help set the latently infected memory B-cell on the course of plasma cell differentiation by downregulating bcl-6, a repressor of the plasma cell specific transcription factor Blimp-1. This results in upregulation of Blimp-1, which in turn suppresses Pax5, which maintains mature B-cell identity and is a suppressor of the plasma cell-specific transcription factor XBP-1s. This allows morphological changes and increased secretory capacity to occur. The UPR, particularly the IRE1-XBP-1s pathway, is essential for plasma cell differentiation. In response to ER stress, ATF6 and IRE are released, which function to increase and splice XBP-1, respectively. Without XBP-1s, the cell could neither acquire the plasma cell phenotype nor secrete antibody. Based on the finding that XBP-1s can bind to and activate the lytic switch BZLF1, it is likely that EBV harnesses the multitasking qualities of XBP-1s, specifically the environment it provides and its involvement in continuing translation and inhibiting apoptosis, for optimal viral replication.

We have shown that XBP-1s transactivates Zp in lymphoid cells, and this activation is synergistically increased by TPA treatment (Fig. 1A and C and Table 1). However, XBP-1s did not transactivate Zp and did not further augment the TPA-driven Zp response in epithelial cells (Fig. 1B and D). This suggests that XBP-1s requires other lymphoid specific factors to transactivate Zp and that XBP-1s and TPA can function through different mechanisms even though both use the same element in the ZII domain. Studies using differentiating SCC12F epithelial cells and EBV-associated epithelial cancer lines have shown that transcription factors involved in cell cycle regulation and cellular differentiation, such as ATF1/CREB/ATF2 and C/EBP, bind and activate Zp through the ZII domain in epithelial cells (21, 33). These observations are consistent with the idea that the virus is reactivated via the ZII domain by mechanisms involving different transcription factors in different cell types (28).

XBP-1s can activate the BZLF1 promoter; however, it is apparent from our experiments to date (not shown) that the presence of XBP-1s alone is not sufficient to efficiently reactivate the virus in all EBV-positive cells. It will be interesting now to investigate what other components of the plasma cell differentiation pathway and the UPR are required to ensure efficient reactivation. Use of chemicals that mimic components of plasma cell differentiation may give insight into which other molecules are involved. Sodium butyrate and valproic acid are histone deacetylase inhibitors that are used to induce the lytic cycle in latently infected cells. They act by globally affecting the chromatin structure and increasing the activity of histone acetylases. These changes in chromatin structure would render the immediate-early genes of EBV accessible for the transcription machinery. Sodium butyrate, for example, has been documented to stimulate B-cell terminal differentiation and viral lytic induction in the Raji Burkitt's lymphoma cell line, giving a strong indication that chromatin changes associated with plasma cell differentiation affect the accessibility of the EBV genome for lytic gene expression (2, 3). Another example is the proteasome inhibitor, Bortezomib, which strongly induces BZLF1 expression in latently infected tumor cells (15). This suggests that decreased proteasomal activity somehow activates BZLF1. Progressive reduction in proteasomal activity has recently been proposed as part of plasma cell differentiation and life span, which results in stabilization of XBP-1s, UPR proteins, and proapoptotic proteins that contribute to sensitivity to apoptosis (8). Both proapoptotic proteins and stabilized XBP-1s could trigger BZLF1 expression.

We have identified a possible mechanism by which viral lytic replication might be initiated in infected plasma cells in vivo, but how does viral DNA replication proceed in a host cell that is arrested in G₀ as a result of terminal differentiation? Kudoh et al. (27), using inducible expression of BZLF1 in EBV-positive B-cell lines, showed that EBV DNA replication is not completely dependent upon cellular replication proteins. Interestingly, EBV replication becomes favorable under conditions in which cellular DNA replication does not occur. These researchers found that viral lytic replication could activate the S-phase promoting cyclin-dependent kinase complexes cyclin E/A and Cdk2 and promote the accumulation of hyperphosphorylated Rb protein. Thus, even in a growth-arrested cell environment, EBV can promote an S-phase-like environment that is favorable for its replication (27). On the other hand, this S-phase state may also be found in precursor plasma cells or plasmablasts (already expressing XBP-1s) undergoing their last round of cellular division just prior to terminal differentiation at G₀ (17, 47). The final round of cellular division might be where EBV could most efficiently replicate its DNA and also utilize the UPR of the plasma cell for producing its viral proteins.

It is not surprising that EBV would utilize the protein production machinery of the plasma cell via XBP-1s to generate copious amounts of virus. For example, human CMV and hepatitis C virus are known to activate and modify the UPR pathway to carry out their own viral production programs (23, 53). In contrast, EBV seems to adopt the normal course of the differentiating B-cell UPR pathway to initiate its replication cycle. Indeed, changes in the UPR pathway would lead EBV to undergo the lytic cycle before the cell dies by apoptosis since terminal differentiation involves a gradual proteosome impairment as a built in life span control for plasma cells (8).

In conclusion, our data suggest that when latently infected B cells differentiate into plasma cells, EBV is able to harness the terminal differentiation pathway via XBP-1s as a physiological signal to reactivate and begin viral replication.