A Region of the Epstein-Barr Virus (EBV) mRNA Export Factor EB2 Containing an Arginine-rich Motif Mediates Direct Binding to RNA*

The Epstein-Barr virus (EBV) protein EB2 (also called Mta, SM, or BMLF1) has properties in common with mRNA export factors and is essential for the production of EBV infectious virions. However, to date no RNA-binding motif essential for EB2-mediated mRNA export has been located in the protein. We show here by Northwestern blot analysis that the EB2 protein purified from mammalian cells binds directly to RNA. Furthermore, using overlapping glutathione S-transferase (GST)-EB2 peptides, we have, by RNA electrophoretic mobility shift assays (REMSAs) and Northwestern blotting, located an RNA-binding motif in a 33-amino acid segment of EB2 that has structural features of the arginine-rich RNA-binding motifs (ARMs) also found in many RNA-binding proteins. A synthetic peptide (called Da), which contains this EB2 ARM, bound RNA in REMSA. A GST-Da fusion protein also bound RNA in REMSA without apparent RNA sequence specificity, because ∼10 GST-Da molecules bound at multiple sites on a 180-nucleotide RNA fragment. Importantly, a short deletion in the ARM region impaired both EB2 binding to RNA in vivo and in vitro and EB2-mediated mRNA export without affecting the shuttling of EB2 between the nucleus and the cytoplasm. Moreover, ectopic expression of ARM-deleted EB2 did not rescue the production of infectious virions by 293 cells carrying an EBVΔEB2 genome, which suggests that the binding of EB2 to RNA plays an essential role in the EBV productive cycle.

Several herpes viruses contain a gene whose product has the characteristics of an mRNA export factor, e.g. herpes simplex virus type 1 (HSV-1) 1 protein ICP27, Epstein-Barr virus (EBV) protein EB2, human herpes virus type 8 (HHV8) protein ORF57, and herpes virus saimiri (HVS) protein ORF57 (for references, see Ref. 1). At least for HSV-1 and EBV, the respective deletion of the ICP27 (2) and EB2 (3) genes abolished both the cytoplasmic accumulation of specific viral mRNAs and the production of infectious viral particles, demonstrating that ICP27 and EB2 are essential viral factors whose function cannot be transcomplemented by cellular factors. The specific function(s) of these viral factors in mRNA export is, however, not yet understood.
Most of the early and late HSV-1 and EBV mRNAs are transcribed from intronless genes. However, transcription, splicing, and mRNA export are linked in eukaryotic cells (4 -10). For example, one model for mRNA export suggests that intron removal in metazoans is associated with the deposition, 20 -24 nucleotides upstream of the exon-exon junction, of a multiprotein complex (EJC) including REF/Aly, Y14, RNPS1, SRm160, and magoh (11)(12)(13)(14). In such a complex, REF/Aly recruits TAP/NXF1 to cellular mRNPs (15)(16)(17)(18). TAP/NXF1 interaction with nuclear pore components then targets mRNPs to the nuclear pore complex and promotes their translocation into the cytoplasm (19). However, mRNAs are still detected in the cytoplasm of Drosophila cells depleted of all EJC proteins, suggesting that additional adaptor protein(s) may mediate the interaction between TAP and cellular mRNAs (20). Cellular mRNAs generated from intronless genes are likely to be exported out of the nucleus through nonspecific interactions with mRNA-bound adaptors like REF/Aly (21) or through sequencespecific interactions with SRp20, 9G8 (22), or U2AF (23). It must be emphasized that SRp20 and 9G8, like REF, interact directly with TAP/NXF1 (24). Interestingly, it has been demonstrated that ICP27 recruits the cellular mRNA export factors REF/Aly and TAP/NXF1 to viral mRNAs generated from viral intronless genes (25), thus allowing viral mRNAs to access the cellular mRNA nuclear export pathway. ICP27 also interacts with SRp20 (26), making it likely that ICP27-mediated mRNA export occurs by recruitment of TAP by REF and/or SRp20. The above results strongly suggest that several adapters may recruit TAP for efficient and perhaps cooperative nuclear export of mRNAs generated from both intron-containing and intronless genes (24). The EBV EB2 protein also exports RNAs generated from both intronless and intron-containing genes (27)(28)(29) and binds REF/TAP-containing mRNPs in vivo (30). EB2 and its HSV-1 functional homologue, ICP27, are likely to be factors involved in the specific export of intronless viral mRNAs, probably via direct interaction.
ICP27-mediated, intronless viral mRNA export is thought to occur through direct RNA-binding via an "RGG box" located within the N terminus of the protein (1,31). However, three C-terminal domains that display homology to the KH RNA interaction motifs of hnRNP K could also be involved in RNA-binding affinity and specificity (32). EB2 expressed as a GST fusion protein has also been shown to interact directly with RNA in vitro (27,28,33,34). However, Northwestern blot experiments labeled a discrete band of 55 kDa but not the full-length 90-kDa GST-EB2 protein, indicating that RNA interaction was restricted to GST-EB2 degradation products in which the EB2 RXP repeats were unmasked and C-terminal (34). This finding confirmed our previously published results showing that a GST fusion protein in which the RXP repeats are C-terminal bound RNA in vitro (27), and this is also where the RXP repeats are located in the HSV-1 US11 RNA-binding factor (35). However, deletion of the RXP domain did not affect EB2-mediated mRNA export, which demonstrates that the RXP repeat is not an RNA-binding motif essential to this function (27). It is therefore not known whether EB2 binds RNA and, if so, by which motif(s).
In this study we have used EB2 and EB2 mutants affinitypurified from human cells as well as GST-EB2 fusion peptides purified from Escherichia coli to locate, by Northwestern analysis and RNA electrophoretic mobility shift assay (REMSA), an RNA-binding motif present in a 33-amino acid domain of EB2 (peptide Da). This motif is not the RXP repeat but is rich in arginines and has structural features of the arginine-rich RNAbinding motif (ARM) also found in several RNA-binding proteins (36,37). The motif has an arginine-rich core and is predicted to fold into an ␣-helix, and, as a synthetic peptide, it binds RNA in the M range in vitro. At least 10 GST-Da molecules bind onto a 180-nt-long, highly structured RNA probe in vitro, suggesting that the EB2 ARM is not sequence specific. A deletion in the ARM disrupts the capacity of EB2 to bind RNA in vitro and in vivo and export mRNA, but without affecting its nucleocytoplasmic shuttling. Using a transcomplementation assay, we also show here that EB2-RNA binding is essential for the production of EBV-infectious particles.
Transfections and CAT Assays-Plasmids used for transfection were prepared by the alkaline lysis method and purified through two CsCl gradients. HeLa cells or 293T cells were grown at 37°C in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum and were seeded at 8 ϫ 10 5 cells per 100-mm-diameter Petri dish 10 h prior to transfection. Transfections were performed by the calcium precipitate method as described previously (29). To evaluate CAT protein expression, we used the Roche Applied Science CAT enzyme-linked immunosorbent assay kit. 48 h after transfection, cells were collected in phosphate buffered saline (PBS). Half of the cells were used for CAT assays according to the manufacturer's instructions. The other half was used to monitor protein expression by Western blots using the anti-FLAG M2 antibody.
Immunoprecipitation of EB2-bound RNAs-To cross-link proteins to RNA, UV irradiation of transfected cells was made at 254 nm for 12 s in a Stratalinker 1800 (Stratagene). Total cell extracts were obtained after incubation of the transfected cells (10 6 ) in 300 l of TNE buffer (10 mM Tris, pH8, 100 mM NaCl, and 1 mM EDTA) containing 1 mM Pefabloc (Uptima) and 500 units of RNasin (Promega) and several passages of the cells through a syringe needle (gauge 26). After centrifugation for 10 min at 12,000 ϫ g, supernatants were incubated for 2 h at 4°C with 7 l of anti-EB2 antibody raised against peptide B (Fig. 1) and for 2 additional hours with 30 l of protein A/G-agarose beads pre-equilibrated in TNE buffer. Beads were successively washed three times in high salt extraction buffer (PBS containing 0.5 M NaCl, 0.5% Nonidet P-40, 1 mM Pefabloc, and 500 units of RNasin), twice in PBS containing 0.5% Nonidet P-40, and once in PBS. They were finally treated for 1 h with proteinase K (500 g/ml) in 1ϫ PBS in the presence of 0.5% SDS. RNA recovered by phenol/chloroform extraction and ethanol precipitation was analyzed by RT-PCR essentially as described previously (29). Briefly, primers located in the pDM128 CAT coding sequence were used both for cDNA synthesis and PCR amplification (Fig. 5A).
Heterokaryons Assays-24 h post transfection, 2 ϫ 10 5 HeLa cells were seeded on glass coverslips with an equal number of NIH3T3 cells in 35-mm dishes and left to grow overnight at 37°C. Cells were then treated for 2 h with 100 g/ml cycloheximide to inhibit protein synthesis. Subsequently, cells were washed in PBS, and heterokaryon formation was performed by incubating the coverslips for 2 min in 50% polyethylene glycol (PEG) 3000 -3700 (Sigma) in PBS. Following cell fusion, coverslips were washed extensively in PBS and returned to fresh medium containing 100 g/ml cycloheximide. After 0.5-2 h at 37°C, cells were fixed with 4% paraformaldehyde, and indirect immunofluorescence was performed essentially as described previously (29), except that either EB2 polyclonal antibody or monoclonal anti-FLAG or anti-hnRNPC (4F4, kindly provided by Dr. G. Dreyfuss) antibodies were used, and Hoechst 33258 (Sigma) was added at 5 g/ml during the secondary antibody incubation.
Western and Northwestern Blots-Western blots were performed as described previously with monoclonal anti-FLAG M2 antibody (Sigma) and rabbit anti-EB2 antibody (29). Northwestern blots were performed essentially as described previously (40), except that the 180-nt 32 Plabeled RNA was incubated with the membrane.
REMSA-RNA-protein interactions were performed with GST and GST fusion proteins purified from E. coli. The 33-amino acid Da peptide N-RSTRKQARQERSQRPLPNKPWFDMSLVKPVSKIT-C was purchased from Genosphere Biotechnologies (France). Peptide Dn N-PVS-KITFVTLPSPLASLTL-C, which partially overlaps with the Da peptide, was purchased from EPYTOP (France). Human ␤-globin 32 P-labeled RNAs were used as probes and prepared as follows. A 180-base pair sequence corresponding to the second exon of the human ␤-globin gene was PCR-amplified using primers 5Ј-TGAATTGTAATACGACTCAC-TATAGGGGTCTACCCTTGGACCCAGA-3Ј and 5Ј-GTGCAGCTCACT-CAGTGTGGC-3Ј, thereby introducing a T7 RNA polymerase promoter at the 5Ј-end of the DNA fragment. A 52-base pair sequence corresponding to the 5Ј part of the second exon of the human ␤-globin gene was PCR-amplified using primers 5Ј-TGAATTGTAATACGACTCACTATA-GGGGTCTACCCTTGGACCCAGA-3Ј and 5Ј-TGGACAGATCCCCCA-AAGGAC-3Ј. A 72-base pair sequence corresponding to the first intron of the human ␤-globin gene was PCR-amplified using primers 5Ј-TGA-ATTGTAATACGACTCACTATAGGGTTGGTATCAAGGTTACAAG-3Ј and 5Ј-GTCTTCTCTGTCTCCACATGC-3Ј. Radiolabeled RNA probes were transcribed from these DNA fragments with T7 RNA polymerase and [␣-32 P]UTP using standard in vitro transcription reactions. Binding reactions were performed with proteins and RNA probes as indicated in the legends to Figs. 2-6 for 15 min at 4°C in binding buffer (10 mM Tris, pH 7.4, 100 mM KCl, 2 mM dithiothreitol, and 1% Triton X-100). Binding reactions were analyzed by polyacrylamide gel electrophoresis in 0.5ϫ TBE buffer. To estimate binding constants, the fraction of the bound RNA was determined by measuring the disappearance of the unbound RNA on gels exposed to a phosphorimaging screen and scanned with an Amersham Biosciences PhosphorImager. Quantitation of free and bound RNA was made using ImageQuant software.
Transcomplementation Assay-The EBV genome was inserted into a mini-F-factor replicon, called p2089 (41,42). p2089 carries the F factor origin of DNA replication, the chloramphenicol resistance gene, and the EBV genome together with the gene for the green fluorescent protein (GFP) under the control of the cytomegalovirus immediate early promoter/enhancer plus the hygromycin resistance gene as a selectable marker in eukaryotic cells. From this plasmid was derived an EBV mutant deleted of the EB2 gene (3). This plasmid was transfected into 293T cells, and a clone called 293 BMLF1-KO was selected (3). The 293 BMLF1-KO cells were transiently transfected with either an empty vector or an EB1-expressing vector or with both the EB1-and EB2expressing vectors as described previously (3). 2-3 days later, the cellfree medium was collected and filtered through a 0.45-m filter. Infection of Raji cells with the filtered medium was carried out as described previously (43,44). Because of constitutive expression of the GFP protein from the EBV genome, infected Raji cells were bright green under UV light and were quantified by fluorescence-activated cell sorter analysis (FACSCalibur).
Prediction of RNA Secondary Structure-RNA secondary structure was determined using a program available on the MFOLD web site (www.bioinfo.rpi.edu/applications/mfold/old/rna) (45,46).
Protein Sequence Analysis-Secondary structure predictions were performed using the IBCP website facilities (npsa-pbil.ibcp.fr) (47). The secondary structure of EB2 peptides Da and Dn was predicted using a large set of methods available including DSC, HNNC, SIMPA96, SOPM, GOR4, PHD, and Predator (for references, see npsa-pbil.ibcp.fr/ NPSA). All methods yielded closely related results.

RESULTS
EB2 Amino Acids 190 -223 Mediate RNA Binding in Vitro-To determine whether EB2 binds directly to RNA, we used Northwestern technology as it does not require extensive biochemical purification of proteins and reliably demonstrates direct binding of proteins to RNA. Because it has not as yet been possible to produce and isolate full-length EB2 from prokaryotic cells, we immunopurified F.EB2 and an EB2 mutant with the N-terminal 184 amino acids deleted, F.NLS.EB2.Cter, from 293T cell nuclear extracts (Dignam extracts; see "Experimental Procedures") ( Fig. 1). Although the deletion of the 184 N-terminal amino acids removes not only the RXP repeat but also the two nuclear localization signals (NLS) of EB2, F.NLS.EB2.Cter is nuclear because of the introduction of the SV40 NLS domain at its N terminus (30). As shown in Fig. 2A, increasing amounts of F.EB2 and F.NLS.EB2.Cter (panels W) bound a 32 P-labeled RNA fragment corresponding to the human ␤-globin second exon RNA (panels N). This demonstrates that EB2 binds directly to RNA in vitro independently of the RXP repeat and that there is an RNA-binding motif in the C-terminal domain of EB2.
To locate precisely a motif(s) mediating RNA-binding in EB2, we produced and purified from E. coli various GST chimeric proteins fused to the overlapping EB2 peptides B to H, depicted in Fig. 1. As a first qualitative screen (Fig. 2B), equal amounts of GST fusion peptides (170 nM) were incubated with 72-nt 32 P-labeled RNA fragment (830 nM) corresponding to the first intron of human ␤-globin. The RNA-protein complexes were resolved by REMSA. As shown in Fig. 2B, a retarded RNAprotein complex was clearly visible only with GST-peptide D (lane 7). However, GST-peptide B also seemed to bind to the RNA probe (Fig. 2B, lane 3), but no individual RNA-protein complex could be detected. Moreover and surprisingly, GST-C, in which the RXP repeats are embedded, did not bind RNA in this assay (Fig. 2B, lane 6). To further locate the RNA-binding motifs in GST-B and GST-D, we expressed and purified GST chimeras fused to shorter peptides (GST-Ba, GST-Bb, GST-Da, and GST-Db; Fig. 1), and used them in a REMSA assay. As shown in Fig. 2B, GST-Ba (lane 4), GST-Bb (lane 5), and GST-Db (lane 9) did not bind RNA, whereas a GST-Da⅐RNA complex was clearly detected (lane 8). As shown in Fig. 2C, we also conducted Northwestern blot assays with increasing amounts of the GST and the GST fusion proteins, using as a probe the 32 P-labeled RNA fragment corresponding to the second exon of human ␤-globin. The results obtained and presented in Fig. 2C corroborate those presented in Fig. 2B. Taken together, these results suggest that the peptide Da contains an RNA-binding motif.
GST-Da Occupies at Least 10 Sites on a 180-Nucleotide-long RNA-Because GST-Da bound to RNA in an in vitro analytical gel shift assay, we further studied its relative RNA-binding affinity. Purified GST-Da (Fig. 3, panel a), GST-Db (Fig. 3,  panel b), or GST (Fig. 3, panel c) were incubated at different concentrations (0 -750 nM) with a 32 P-labeled RNA fragment (100 nM) corresponding to the second exon of human ␤-globin. The number of GST-Da⅐RNA complexes increased with the increase in GST-Da concentration (Fig. 3, panel a), strongly suggesting that peptide Da fused to GST bound several times to the same RNA molecule. As expected, neither GST-Db (Fig. 3,  panel b) nor GST (Fig. 3, panel c) bound detectably to RNA in this experiment. To further document the repetitive binding of GST-Da to the same RNA molecule and estimate the apparent K d , we incubated increasing concentrations of GST-Da with a 52-nt-long, 32 P-labeled RNA fragment (25 nM) generated from the second exon of human ␤-globin. As shown in Fig. 3, panel d, only two GST-Da⅐RNA complexes formed on the short RNA fragment with an estimated apparent K d of 0.6 M. Because the GST we used can form dimers, it is likely that GST-Da binds RNA as a dimer, which suggests that one GST-Da dimer binds every 20 nt on the 52-nt-long RNA. This is in accordance with ϳ10 GST-Da molecules bound to the 180-nt-long RNA (Fig. 3,  panel a). However, we cannot rule out the possibility that the slower migrating complexes formed on the 180-nt-long RNA could be the result of protein aggregation at high protein concentrations rather than binding to multiple sites. These results demonstrate that the peptide Da bound several times to the same RNA molecule and that the binding is probably not sequence-specific.
Synthetic Peptide Da Binds RNA in Vitro and Contains an Arginine-rich RNA-binding Motif-Using an in vitro gel shift assay, we also examined binding of the synthetic peptide Da to the 32 P-labeled RNA fragment of 52-nt that is thought to adopt the theoretical secondary structure shown in Fig. 4B (45,46). Peptide Da bound RNA with apparent low affinity (K d ϳ3 M), assuming that peptide Da bound only once to an RNA molecule at the peptide concentration titrating 50% of the RNA probe (Fig. 4C, lane 6). However, it is likely that, at a RNA:peptide ratio Ͼ3.2 ⌴, more than one peptide binds to the 52-nt-long RNA probe (Fig. 4B). Moreover, we do not know if the RNA structure depicted above is the one recognized by peptide Da. A shorter peptide called Dn (see "Experimental Procedures"), partially overlapping with the five C-terminal residues of peptide Da, did not bind RNA detectably in our assay (Fig. 4C). This reinforces the fact that peptide Da contains an RNAbinding motif. The amino acid sequence of peptide Da is rich in arginine and glutamine (Fig. 4A). More precisely, the hexapeptide TRKQAR in peptide Da has similarities to the argininerich motif (TRRRER) originally found in the bacteriophage N protein (36) and in many other RNA-binding proteins including HIV-1 Rev (TRQARR), HIV-1 Tat (TRGKGR), and Jembrana disease virus (JDV) Tat (TRGKGR) (37,48). Taken together, our results strongly suggest that peptide Da binds RNA in vitro through an ARM.
A Deletion in the EB2 RNA-binding Domain Impairs RNA Binding in Vitro and in Vivo-We next evaluated whether the deletion of the putative ARM in the EB2 protein would affect its binding to RNA in vitro. We generated a vector expressing the EB2 mutant F.EB2.⌬D1 in which the N-terminal part of peptide Da containing the ARM is deleted (Fig. 5A). The F.EB2.⌬D1 protein was expressed in 293T cells and immunopurified from nuclear extracts. Northwestern assays were performed using the 32 P-labeled human ␤-globin second exon RNA fragment. As shown in Fig. 2A EB2 is known to co-immunoprecipitate mRNAs from transfected cells (34). To test whether F.EB2.⌬D1 would bind a target RNA in vivo as compared with F.EB2, HeLa cells were transiently transfected with the pDM128 reporter construct either alone or together with expression vectors for the F.EB2 protein or the F.EB2.⌬D1 protein (Fig. 5A). The pDM128 reporter plasmid expresses a two-exon, one-intron pre-mRNA with the CAT gene located in the intron (38). We have demonstrated previously that EB2 can mediate the nuclear export of unspliced polyadenylated RNA generated from plasmid pDM128 (29). If EB2 binding to pDM128 unspliced RNAs through the ARM is essential for their nuclear export, such RNAs should co-immunoprecipitate with the F.EB2 protein but not with the F.EB2.⌬D1 protein. Protein extracts were made from whole cells after UV irradiation of the transfected cells to cross-link proteins to RNA, as described in Fig. 5B. When whole cell extracts were incubated with an anti-EB2 antibody and the immune complexes pulled down by protein A/G beads, CAT cDNA was effectively amplified by RT-PCR when HeLa cells were expressing F.EB2 (Fig. 5C, lane 4). This suggests that, within cells, a fraction of the unspliced CAT RNA was complexed with EB2. On the contrary, we failed to detectably amplify the CAT cDNA when HeLa cells expressed F.EB2.⌬D1 (Fig. 5C, lane 6), strongly indicating that the EB2 ARM is essential for binding to RNAs within living cells. As expected, no pDM128 unspliced RNA was pulled down by protein A/G beads incubated alone with whole cell extracts, because no CAT cDNAs could be amplified by RT-PCR from the immunoselected material (Fig. 5B, lanes 1, 3, and 5). Finally, the differential RT-PCR amplification of CAT cDNA was not due to individual variations in the amount of unspliced pDM128 RNA in the extracts. Indeed, RNAs were purified from a fraction of the whole cell extracts, and comparable amounts of CAT cDNAs could be amplified by RT-PCR (Fig. 5C, lanes 7-9).
ARM Deletion Affects EB2-mediated mRNA Export without Affecting Its Nucleocytoplasmic Shuttling-As F.EB2.⌬D1 did not bind unspliced pDM128 RNAs in vivo, it should also not export unspliced pDM128 RNAs. We therefore co-transfected HeLa cells with F.EB2 or F.EB2.⌬D1 together with pDM128. In this assay, unspliced CAT-containing RNAs transcribed from pDM128 are rarely found in the cytoplasm, but EB2 strongly increases their nuclear export, and this can be quantified at the level of both CAT-expressed protein and unspliced CAT mRNA detected in the cytoplasm (29). As shown in Fig.  6A, F.EB2 expression strongly increased the cytoplasmic accumulation of unspliced pDM128 mRNAs (lane 2), whereas F.EB2.⌬D1 failed to do so (lane 3). As both proteins were expressed at a comparable level (Fig. 6B), these results demonstrate that deletion of the EB2 ARM specifically impairs EB2-mediated export of pDM128 unspliced mRNAs.
To ascertain that no other function than RNA binding was affected by the deletion in F.EB2.⌬D1, we also examined the nucleocytoplasmic shuttling of F.EB2.⌬D1 by performing a hu- , and GST (panel c) were incubated with a 32 P-labeled 180-nt RNA fragment (100 nM) of the human ␤-globin second exon. Increasing amounts (nM) of affinity-purified GST-Da were also incubated with a 32 P-labeled 52-nt RNA fragment (25 nM) of the human ␤-globin second exon depicted in Fig. 4A (panel d). The RNA-protein complexes were separated by electrophoresis on a 4.5% acrylamide gel in 0.5ϫ TBE.
man-mouse heterokaryon assay. In this assay, HeLa cells were transfected with a vector expressing either F.EB2 or F.EB2.⌬D1. After 24 h, HeLa cells were co-cultivated overnight with mouse NIH3T3 cells. The cells were then fused by incubation with polyethylene glycol in the presence of cycloheximide to suppress de novo protein synthesis and then incubated a further 2 h in the presence of cycloheximide. Indirect immunofluorescence was then performed to detect HeLa-NIH3T3 heterokaryons and evaluate in how many heterokaryons the protein expressed in the HeLa cell nucleus has been transported to the mouse nucleus, thereby demonstrating that both the NLS and the nuclear export signal (NES) are functional. As shown in Fig. 6C, the F.EB2.⌬D1 mutant was found to shuttle as efficiently as the wild-type F.EB2 protein. The endogenous non-shuttling hnRNPC protein was used as control, and, as expected, was not transported from the human to the mouse nucleus (Fig. 6C).

ARM-deleted EB2 Did Not Rescue Infectious Virus Production by 293 Cells Carrying an EBV Mutant Lacking the EB2
Gene-We recently reported the generation of an EBV mutant in which the gene encoding the EB2 protein has been deleted and also the introduction of the mutated viral genome into 293 cells (3). To induce the productive cycle of EBV, it is necessary and sufficient to transiently express, in EBV-infected cells, the EBV transcription factor EB1/Zta, which is the activator of all the EBV early genes including the EB2 gene (41, 49 -51). Our recent studies showed that 293 cells harboring the EB2-defective EBV genome only produced infectious EBV particles when both EB1 and EB2 were transiently expressed, demonstrating that EB2 is essential for the production of infectious virions (3). We also showed that there is a clear correlation between the expression of EB2 in trans and an increase in the nuclear export of a specific subset of EBV early and late mRNAs (3). In such an assay and in contrast to F.EB2, the EB2 mutant F.EB2.⌬D1, did not rescue the production of infectious viral particles when co-expressed with EB1 (Fig. 7A) under conditions where the amounts of EB1 and EB2 protein detected by Western blotting were comparable (Fig. 7B). The above results strongly suggest that binding of EB2 to RNA and, possibly, EB2-mediated mRNA export are required for the production of EBV-infectious virions.

DISCUSSION
In this study we show that the full-length EB2 protein isolated from human cells interacts directly with RNA in vitro and that this interaction is likely to be mediated by a putative arginine-rich RNA-binding motif located in the Da peptide. Moreover, deletion of the arginine-rich motif (mutant F.EB2.⌬D1) severely impaired EB2 binding to RNA in vivo, EB2-mediated mRNA export, and transcomplementation of EB2 function in infectious virus production. It is noteworthy that peptide B also bound RNA in our in vitro assays. However, binding of peptide B to RNA was very weak, and none of the overlapping subfragments of peptide B tested was found to bind RNA either by REMSA or Northwestern blot analysis. We therefore did not study peptide B further, although we cannot rule out that it might contribute to RNA-binding in the context of the full-length EB2 protein. This would explain why, with the F.EB2.⌬D1 mutant, there is clearly some CAT expression in the in vivo assay with pDM128 (Fig. 6A) and some virus replication over background in the transcomplementation as-say (Fig. 7), albeit highly reduced. Further detailed investigations are required to clarify these points.
A computer-assisted search allowed us to characterize a known RNA-binding motif in the Da peptide; peptide Da contains a short sequence rich in arginine/lysine residues, TRKQAR, which is characteristic of ARMs. This sequence is similar, at the primary sequence level, to those found in HIV-1 Rev and Tat, BIV Tat, as well as viral coat proteins and ribosomal proteins (36,37,48,52). Peptide Da also bound RNA as a synthetic peptide, another property of arginine-rich RNAbinding motifs. However, the TRKQAR motif is obviously not sufficient by itself to ensure RNA binding, because the GST-C fusion protein, which also contains this motif, does not bind to RNA (Fig. 2, B and C). Additional amino acids that are present only in peptide Da are thus required for specific contacts with RNA. Examination of the Da peptide sequence around the arginine-rich hexapeptide revealed the presence of additional Arg residues on both extremities, suggesting a more extensive RNA-binding motif. Binding to BIV TAR of the arginine-rich Tat peptide also depends on one isoleucine, one arginine, and one glycine located outside the TRGKGR hexapeptide core (48). In addition, all secondary structure prediction methods tested predicted an ␣-helix that could extend from residue 190 to FIG. 6. Deletion in the EB2 RNA-binding motif impairs EB2mediated mRNA export but not nucleocytoplasmic shuttling. A, HeLa cells were transfected with plasmid pDM128 and the expression vectors as indicated at the left of the panel. Unspliced CAT mRNA export was quantified as the relative amount of CAT protein detected by enzyme-linked immunosorbent assay, with a fixed value of 100 given to the amount of CAT protein expressed in the presence of the wild-type F.EB2. B, Western blot analysis of F.EB2 and F.EB2⌬D1 proteins expressed in transfected HeLa cells. Immunostaining was made using the anti-FLAG antibody (M2, Sigma). C, HeLa cells were transfected with vectors expressing F.EB2 or F.EB2⌬D1. After 48 h, transfected HeLa cells were fused with NIH3T3 cells to form heterokaryons and were further incubated for 2 h in the presence of cycloheximide as described under "Experimental Procedures." Cells were then immunostained (IF) with anti-FLAG (M2, Sigma) or anti-hnRNPC 4F4 (a generous gift from Dr. G. Dreyfuss) primary antibodies and Alexa Fluor goat anti-mouse IgG (HϩL) (Interchim) as a secondary antibody and then stained with Hoechst, which allows HeLa and mouse nuclei to be differentiated in the heterokaryons. Numbers on the right side of the panels indicate the number of heterokaryons with positive mouse cell nuclei immunostaining (indicating protein shuttling)/the total number of heterokaryons with positive HeLa nuclei immunostaining examined. Arrows indicate the mouse nuclei. residue 203 of EB2, i.e. including the TRKQAR sequence (Fig.  4A). The ␣-helix projection of this segment reveals that the arginine residues are essentially located on one side of the helix, forming a basic groove, whereas the other residues form a polar groove on the opposite face of the helix (not shown). Although the existence of this helix remains to be demonstrated experimentally, such a topology suggests that the arginine residues make specific interactions with an RNA partner.
As shown by our in vitro assays, the EB2 peptide Da containing the ARM binds to RNA with no apparent sequence specificity. However, some specificity could reside in the folding of the RNA. Indeed, the RNA probes used in our in vitro binding studies are highly structured. Peptide Da could bind to the RNA phosphate backbone, the major groove, or the minor groove of the RNA probes used. It is also possible that we have missed some sequence specificity by using only heterologous RNAs. Another possibility is that EB2 binding to RNA stabilizes some specific structures in the way that the HIV-1 Rev ARM stabilizes an internal loop upon binding to the Rev response element (RRE) in vitro (53,54). Specific structural studies (e.g. by CD and NMR) will be required to answer such questions.
Despite the apparent lack of EB2 RNA binding specificity in vitro, EB2 selectively exports mRNAs in vivo. Indeed, EB2 has no effect on the export of human ␤-globin mRNAs and, in general, on the export of mRNAs generated by the use of "constitutive" splice sites (27), although the present studies demonstrate that EB2 binds to human ␤-globin RNA exon and intron sequences in vitro. However, EB2 efficiently exports unspliced ␤-thalassemia mRNAs (27) in which a mutation in the first exon 5Ј-splice site has caused the activation of three cryptic 5Ј-splice sites (55). This is reminiscent of the seminal studies made by Chang and Sharp (56) on HIV Rev-mediated mRNA export, which demonstrated that a ␤-globin pre-mRNA containing the Rev response element is exported unspliced by Rev only if either the 5Ј or 3Ј splice sites are mutated. This and other results described elsewhere (27) clearly point to the fact that EB2 exports many different unspliced mRNAs (chloramphenical acetyl transferase, human ␤-globin, pUC18, etc.) but only if they contain introns excised by the use of cryptic 5Јsplice sites. This property was also exploited in this study through the use of the RNA export reporter gene pDM128 (Fig.  6), which is an HIV-derived reporter gene carrying "sub-optimal splice sites," allowing export of unspliced RNAs. EB2 might therefore overcome, by as yet unknown mechanisms, the splicing of pre-mRNAs, which are poor substrates for spliceosome assembly and therefore export unspliced mRNAs. It is noteworthy that ICP27, the HSV-1 mRNA export factor, inhibits splicing by inactivating the SR protein kinase SRPK1 (26). However, ICP27 does not export unspliced pDM128 RNAs. 2 During the productive cycle, EB2 also selectively exports some mRNAs generated from EBV intronless genes (3). Taken together, these data suggest that the EB2 mRNA export specificity could be dictated by the fate of a particular mRNA. Although EB2 might bind to many pre-mRNAs in vivo, those which are poor substrates for splicing or are the products of intronless genes that do not carry constitutive export signals will be the ones exported by EB2. Alternatively, because transcription and mRNA export are coupled in eukaryotic cells (8 -10), EB2 might increase the cytoplasmic accumulation of only a subset of EBV-specific early and late mRNAs as a direct consequence of the differential "loading" of EBV transcription factors EB1 or R and, possibly, cellular factors on EBV early and late promoters.
In conclusion, the EBV mRNA export factor EB2 appears to contain a novel arginine-rich RNA-binding motif with no apparent specificity for defined RNA sequences. Further studies are required to characterize how the EB2 ARM interacts with RNA(s) and what makes an RNA a target for EB2-mediated mRNA export.