Interaction between Basic Residues of Epstein-Barr Virus EBNA1 Protein and Cellular Chromatin Mediates Viral Plasmid Maintenance*

Background: Epstein-Barr virus episomes piggyback onto cellular chromosomes in latently infected cells. Results: EBNA1 was functionally sound even when arginines within its chromosome binding domains were replaced with lysines. Conclusion: The basic nature of the domains is critically important for the chromosome binding. Significance: Learning how EBNA1 attaches to chromosomes is crucial for understanding the mechanism of stable maintenance of viral episomes. The Epstein-Barr virus (EBV) genome is episomally maintained in latently infected cells. The viral protein EBNA1 is a bridging molecule that tethers EBV episomes to host mitotic chromosomes as well as to interphase chromatin. EBNA1 localizes to cellular chromosomes (chromatin) via its chromosome binding domains (CBDs), which are rich in glycine and arginine residues. However, the molecular mechanism by which the CBDs of EBNA1 attach to cellular chromatin is still under debate. Mutation analyses revealed that stepwise substitution of arginine residues within the CBD1 (amino acids 40–54) and CBD2 (amino acids 328–377) regions with alanines progressively impaired chromosome binding activity of EBNA1. The complete arginine-to-alanine substitutions within the CBD1 and -2 regions abolished the ability of EBNA1 to stably maintain EBV-derived oriP plasmids in dividing cells. Importantly, replacing the same arginines with lysines had minimal effect, if any, on chromosome binding of EBNA1 as well as on its ability to stably maintain oriP plasmids. Furthermore, a glycine-arginine-rich peptide derived from the CBD1 region bound to reconstituted nucleosome core particles in vitro, as did a glycine-lysine rich peptide, whereas a glycine-alanine rich peptide did not. These results support the idea that the chromosome binding of EBNA1 is mediated by electrostatic interactions between the basic amino acids within the CBDs and negatively charged cellular chromatin.

EBNA1 and the cellular origin recognition complex through the RNA binding activity of EBNA1 (21). These domains also protect EBNA1 from proteasomal degradation (11) and enhance the transfection efficiency of an oriP-containing plasmid (22).
The molecular mechanism by which the CBDs of EBNA1 attach to mitotic chromosomes is still under debate. It has been proposed that EBP2, which interacts with the centrally located glycine-arginine rich domain (aa 328 -377), mediates the interaction between chromosomes and viral episomes (13,19). However, a recent study argued against the idea and identified HMGB2 (high mobility group box 2), a well known chromatin component, as a binding partner for EBNA-1 on chromatin (23). It has also been proposed that RNA molecules mediate the interaction between EBNA1 and mitotic chromosomes (24). Another hypothesis is that EBNA1 directly binds to the AT-rich regions of chromosomal DNAs via the AT hook motifs within the CBDs (10). The AT hooks share the common consensus sequence proline-arginine-glycine-arginine-proline (PRGRP) flanked by other positively charged residues (25,26). Synthetic peptides derived from the putative AT hook motifs of EBNA1 were shown to preferentially bind to a poly(dA-dT) probe in vitro (10). It is noteworthy that the CBDs of EBNA1 can be replaced by HMGA1a or histone H1 without affecting the capacity of EBNA1 to maintain oriP-based plasmids (9), suggesting that EBNA1 and these cellular proteins bind to host chromosomes in a similar fashion.
In this study we focused on clarifying the importance of arginine residues within the EBNA1 CBDs. Our technical advantage is the usage of EBNA1-mCherry fusion proteins and HeLa cells expressing histone H2B-GFP (HeLa-H2B-GFP cells) (27), and thereby EBNA1 mutants (red fluorescence) and host chromosomes (green fluorescence) were readily visualized in living cells. We now present first experimental evidence that arginine residues within the CBDs of EBNA1 are actually important for its chromosome binding in the context of EBNA1 protein. Importantly, the arginines within the CBDs could be replaced with lysines without compromising its episome maintenance ability, indicating that the basic nature of the CBDs rather than arginine residues themselves is a primary determinant of chromosome binding ability of EBNA1.

EXPERIMENTAL PROCEDURES
Recombinant Plasmids-All the EBNA1-mCherry expression plasmids were constructed by using pSG5 vector (Stratagene) as a backbone vector. The genes of EBNA1, EBNA1⌬GA lacking the glycine-alanine repeats (aa 90 -327), and mCherry were cloned into pSG5 vector. Primer sequences and schematic illustrations describing the PCR-based cloning strategies are in the supplemental Methods. The gene encoding the CBD region of EBNA1⌬GA (aa 40 -89 and aa 328 -377 connected in tandem; Fig. 1B) and the mCherry gene were PCR-amplified using primers pairs of CBDup-mCherryCBDdown and CBDmCherryup-mCherrydown, respectively, and the resultant PCR products were used as mixed templates for the second-round PCR using CBDup and mCherrydown primers to obtain the CBD-mCherry fusion gene. The PCR product of the CBD-mCherry was digested with EcoRI and BamHI, and the fragment was cloned into pSG5 vector. Derivatives of CBD-mCherry genes were generated by using an invert PCR method (28) with primers listed in the supplemental Methods. The plasmid CBD1CBD2-mCherry was subjected to a one-step site-directed mutagenesis protocol (29) to obtain plasmids encoding 4RA-mCherry and 4RK-mCherry using primer pairs of RAsub01F-RAsub01R and RKsub01F-RKsub01R, respectively. The plasmid of 4RA-mCherry was further modified to obtain plasmids encoding 12RA-mCherry (using primers of 8RAup-8RAdown), 16RA-mCherry (using primers of 12RAup-12RAdown), and 24RA-mCherry (using primers of 8RAup-8RAdown and a template DNA of 16RA-mCherry) by using the inverted PCR method. The plasmid encoding 24RK-mCherry was constructed by means of the same strategy (using primers of 8RKup, 8RKdown, 12RKup, and 12RKdown). The plasmid CBD1-mCherry was PCR-amplified using CBD1ϫ2up and CBD1ϫ2down primers to obtain the plasmid CBD1 ϫ 2-mCherry. The plasmids expressing EBNA1⌬GA-mCherry were constructed by the inverted PCR method using pSG5-EBNA1⌬GA and pSG5-mCherry as mixed templates. The plasmids expressing 24RAϩTA-mCherry, 24RKϩTA-mCherry, mEBNA1(CBD1-RA)-mCherry, mEBNA1(CBD2-RA)-mCherry, mEBNA1(24RA)-mCherry, and mEBNA1(24RK)-mCherry were constructed by multiple rounds of inverted PCRs using the plasmids encoding 24RA-mCherry and 24RK-mCherry as starting materials. Primer sequences that are not specified in the supplemental Methods are available upon requests. The coding sequences of all the constructed plasmids were verified by DNA sequencing. A retroviral vector expressing CBD-mCherry was made by cloning the AflIII-BamHI fragment of the CBD-mCherry gene between the NcoI and BamHI sites of pCLMFG-MCS (30). Retroviruses pseudotyped with vesicular stomatitis virus G (VSV-G) protein were produced by utilizing HEK 293 cells and pCGCGP (31), a plasmid expressing gag, pol, and VSV-G as an env. pEB-Multi-Neo (Wako Pure Chemical Industries, Ltd, Osaka, Japan) is an improved episomal plasmid having the oriP sequence, an EBNA1 gene that is driven by a thymidine kinase promoter, and a kanamycin resistance gene as a prokaryotic selection marker. The Acc65I-NotI fragment of pEGFPN1 (Clontech) was cloned between the Acc65I and NotI sites of pEB-Multi-Neo to make pEB-Multi-Neo-EGFP. The constructs encoding EBNA1⌬GA, mEBNA1(24RA), or mEBNA1(24RK) were digested by AvrII and SacI to obtain DNA fragments of partial EBNA1 coding sequences. Each AvrII-SacI fragment and the EcoRI-AvrII fragment of pEB-Multi-Neo were cloned into pEB-Multi-Neo-EGFP vector that had been digested with EcoRI and SacI to generate derivatives of pEB-Multi-Neo-EGFP.
Cell Culture-HeLa cells and its derivatives were maintained in DMEM medium (Sigma) supplemented with 10% fetal bovine serum. The establishment of HeLa H2B-GFP cells was described previously (27). HeLa-H2B-GFP cells were infected with a retroviral vector expressing CBD-mCherry, and a derivative cell line expressing both H2B-GFP and CBD-mCherry was obtained by a limited dilution protocol. EBV-negative Akata cells (Akata(Ϫ) cells) (32), derived from EBV-positive Burkitt's lymphoma cells, were maintained in RPMI medium (Sigma) supplemented with 10% fetal bovine serum.
Microscopic Analyses-HeLa-H2B-GFP cells in 12-well dishes were transfected with the expression vectors encoding various CBD-mCherry mutants by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instruction. At 2 days posttransfection, the cells were re-plated onto 35-mm glass base dishes (Asahi Glass Co. Ltd., Tokyo, Japan). The cells were directly subjected to confocal microscopic observation at 3 days posttransfection. Images were captured with a Plan-Apochromat 63ϫ/1.4-numerical aperture oil immersion objective and processed using an LSM510 Meta microscope (Carl Zeiss MicroImaging, Inc.).
Western Analyses-HeLa-H2B-GFP cells were plated in 12-well dishes and transfected with expression plasmids encoding the derivatives of CBD-mCherry by using Lipofectamine 2000. The transfected cells were harvested at 2 days posttransfection, and the cell pellets were resuspended with 30 l of PBS (without magnesium and calcium) and 60 l of 1.5 ϫ lysis buffer (187.5 mM Tris-HCl (pH 6.8), 4.5% SDS, 9% urea, and 15% glycerol) and sonicated to obtain whole cell extracts. The whole cell extracts (10 l each) were subjected to Western analyses by using either anti mCherry monoclonal antibody (Clontech), anti EBNA1 rabbit polyclonal antibody (33), and anti ␣␤-tubulin rabbit polyclonal antibody (Cell Signaling) as primary antibodies.
Plasmid Maintenance Assay-HeLa cells in 12-well dishes were transfected with pEB-Multi-Neo-EGFP and its derivatives by using Lipofectamine 2000. Akata(Ϫ) cells were transfected with the same set of test plasmids by using Neon (Invitrogen) with a parameter setting of 1400 V, 20 ms, 2 times. Transfected cells were harvested at 2 days posttransfection, and the cells were cultivated with culture media containing G418 (800 g/ml for HeLa cells, 700 g/ml for Akata(Ϫ) cells) for 5 days to enrich cells that incorporated plasmid DNAs. The transfected cells were then appropriately diluted and cultivated in the absence of G418 for another 5 days. The aliquots of cells were harvested on day 5 (after G418 selection) and day 10 (after 5 days of cultivation without drug) to be analyzed by FACSCalibur, and the percentages of GFP-positive cells were determined by Cell-Quest software. Episomal plasmids were extracted from aliquots of transfected cells (cultivated in wells of 6-well dishes) at day 10 using the alkaline lysis method as previously described (34), with RNase A (50 l/ml) included in the resuspension buffer. The nucleic acid pellets were dissolved in 15 l of nuclease free water, and 1.5-l aliquots were used to transform 20 l of ElectroMAX DH10B competent cells (Invitrogen) according to the manufacturer's instructions.
Nucleosome Binding Assay-Nucleosome core particles were reconstituted by a salt dialysis method utilizing purified human histone proteins (expressed in Escherichia coli) and an ␣-satellite DNA derivative (146 bp) as described previously (35,36). Synthesized EBNA1-derived peptide (GR peptide) and its derivatives (GA and GK peptides) ( Fig. 5A) (Biosysthesis, Lewisville, TX) were dissolved to dimethyl sulfoxide to make 5 mM stock solutions. Binding assays were performed by mixing increasing molar amounts of each peptide (10 -40 pmol) with reconstituted nucleosomes (2.1 pmol) per each sample in 12 mM Tris-HCl (pH 7.5), 0.6 mM DTT. The mixtures were incubated at 37°C for 60 min and then analyzed by 6% non-denaturing polyacrylamide gel electrophoresis in 0.2 ϫ TBE buffer (18 mM Tris base, 18 mM boric acid, and 0.4 mM EDTA) at 18.75 V/cm for 1 h followed by ethidium bromide staining (36).

RESULTS
The CBD-mCherry Works as an Artificial Chromosome Binding Module-The regions spanning the domain A (GR1) and domain B (GR2) (Fig. 1A) are currently assumed as important subdomains required for the chromosome binding of EBNA1. We first verified the results by making a construct encoding EBNA1 subdomains spanning aa 40 -89 and aa 328 -377 connected in tandem that were fused to mCherry (Fig. 1B). The fusion protein, designated CBD-mCherry, contains an EBNA1 subdomain that is almost identical to that of GR1/2-GFP protein (8). The CBD-mCherry was transiently expressed in HeLa-H2B-GFP cells, and the transduced cells were observed by using live confocal microscopy. Supporting the previous reports, the CBD-mCherry fusion protein associated as strongly and exclusively with mitotic chromosomes as EBNA1⌬GA (an EBNA1 mutant lacking the Gly-Ala repeat) fused to mCherry (Fig. 1C). The domain A (GR1) can be divided into two regions: the first 15 amino acids (aa 40 -54), which are highly enriched with glycine and arginines, and the following 35 amino acids (aa 55-89), which constitute the transactivation domain (TA) of EBNA1 (14,(37)(38)(39). The region of aa 40 -54 is referred to as CBD1 region, whereas the region of aa 328 -377 (domain B, GR2) is referred to as CBD2 region hereafter. The CBD2 region is further divided into CBD2-I and CBD2-II regions. It was then found that a mutant CBD1CBD2-mCherry, in which the transactivation domain of EBNA1 was omitted from the CBD-mCherry, associated with mitotic chromosomes just like the CBD-mCherry (Fig. 1C).
The CBD-mCherry and the CBD1CBD2-mCherry, despite their lack of functional NLS (aa 379 -387), exclusively localized in interphase nuclei ( Fig. 1, C and D), whereas the EBNA1⌬GA lacking the NLS (EBNA1⌬GA⌬NLS-mCherry) exclusively localized to cytoplasm in interphase cells (Fig. 1D). Thus, the CBD-mCherry localizes to interphase nuclei in a way that is independent from the NLS. HeLa cells stably expressing both H2B-GFP and CBD-mCherry proteins ( Fig. 1E) were treated with calyculin A (40), a phosphatase inhibitor, and premature chromosome condensation was induced. The results revealed that CBD-mCherry exclusively associated with prematurely condensed interphase chromatin (Fig. 1F). Thus, the nuclear localization of the CBD-mCherry protein is most likely a manifestation of its binding to interphase chromatin.
Amino Acid Substitution Mutations Affect the Localization of EBNA1 CBD-derived Fusion Proteins-Due to the small size of the CBD1CBD2 domain (65 amino acids, encoded by 195 nucleotides), it was feasible to systematically introduce amino acid substitution mutations into the region. To get direct evidence for the importance arginine residues, arginines of CBD1CBD2-mCherry were incrementally replaced with alanines to generate a series of RA mutants (4RA-, 12RA-, 16RA-mCherry, and 24RA-mCherry) (Fig. 2, A and B). The results revealed that the number of residual arginines in these mutants correlated well with their chromosome binding abilities. Specifically, the 16RA mutant retained very weak chromosome binding. The 24RA mutant, with no residual arginines, completely lost chromosome binding ability, and as a result, mitotic chromosomes were excluded from mCherry staining (Fig. 2C). These results demonstrate that arginine residues are essential for chromosome binding, at least in the context of this artificial chromosome binding module.
In recognition of the importance of arginine residues, a mutant 24RK-mCherry, in which all of the arginine residues within the CBD1 and -2 regions were replaced with lysines (Fig.  2, A and B), was then tested to examine the consequence of mutations that conserved the basic nature of the CBDs. The 24RK-mCherry was predominantly nuclear in interphase cells ( Fig. 2C), in clear contrast to the diffuse intracellular staining of 24RA-mCherry. On the other hand, the mitotic chromosome binding of 24RK-mCherry was significantly impaired compared with CBD1CBD2-mCherry (Fig. 2C).
Importantly, when the TA domain was added back to the 24RK-mCherry, the resultant 24RKϩTA-mCherry exclusively associated with mitotic chromosomes (Fig. 2C). By contrast, the chromosome binding did not recover when the TA domain was added back to 24RA-mCherry (24RAϩTA-mCherry) (Fig. 2C). Thus, in the presence of the TA domain, arginine residues within the CBD1 and -2 regions can be replaced with lysines without impairing chromosome binding ability.
Western analyses revealed that, although mCherry-positive cells were microscopically observed at comparable frequencies, the protein expression levels of the mutants varied significantly; the mutants with chromosome binding abilities (CBD1CBD2-mCherry and 4RA-mCherry) were expressed significantly more than the mutants that bound poorly to chromosomes (24RA-mCherry) (Fig. 2B).
Arginine-to-alanine Substitutions within the CBDs, but Not Arginine-to-lysine Substitutions, Affect Chromosome Binding of EBNA1-The results obtained by the CBD-mCherry fusion protein with arginine-to-alanine or arginine-to-lysine substitutions were then verified in the context of the EBNA1⌬GA protein, which retained the TA, the NLS, and the C-terminally located DNA binding domain. The EBNA1⌬GA was chosen for the mutagenesis because it was technically difficult to introduce mutations while keeping the Gly-Ala repeat of the full-length EBNA1. Arginine-to-alanine substitution mutations were introduced to the CBD1 region (mEBNA1(CBD1-RA)) or to the CBD2 region (mEBNA1(CBD2-RA)), and these mutants were expressed as mCherry fusions in HeLa-H2B-GFP cells (Fig. 3, A  and B). Mutating the arginine residues within the CBD2 region attenuated the chromosome binding ability more significantly than mutating those within the CBD1 region, which was demonstrated by enhanced cytoplasmic staining of mCherry in mitotic cells (Fig. 3B).
Microscopic analyses demonstrated that both EBNA1(24RA)-mCherry and mEBNA1(24RK)-mCherry were localized to the nucleus in interphase cells (Fig. 3C), as both constructs retain the NLS. However, in mitotic cells, whereas mEBNA1(24RA)-mCherry exhibited minimal, if any, chromosome binding, mEBNA1(24RK)-mCherry exhibited chromosome binding ability indistinguishable from that of EBNA1⌬GA-mCherry (Fig. 3C). This correlated very well with the results of 24RAϩTA-mCherry and 24RKϩTA-mCherry (Fig. 2C). Thus, the results obtained by CBD-mCherry-derived mutants had faithfully predicted the behaviors of EBNA1⌬GA-derived mutants, justifying our approach to use the CBD-mCherry as a convenient indicator of EBNA1 chromosome binding. We conclude that arginine residues within the CBD1 and CBD2 regions confer chromosome binding ability in the context of EBNA1⌬GA, and the arginine residues are exchangeable to lysines.

Plasmid Maintenance Abilities of EBNA1 Amino Acid Substitution Mutants Correlate with Their Chromosome
Binding Abilities-The biological functions of EBNA1⌬GA, mEBNA1(24RA), and mEBNA1(24RK) were then examined by plasmid maintenance assay. An EBNA1-oriP plasmid (pEB-Multi-Neo) with an EGFP expression cassette was modified so that its EBNA1 gene was replaced with an EBNA1⌬GA gene, with an mEBNA1(24RA) gene, or with an mEBNA1(24RK) gene, and these test plasmids were introduced into HeLa cells or EBV-negative Akata cells (Akata(Ϫ) cells) (32). No significant differences of GFP expression levels were noticed between the test plasmids shortly after transfection to HeLa and Akata(Ϫ) cells (data not shown). Western analyses revealed that EBNA1⌬GA, mEBNA1(24RA), and mEBNA1(24RK) were expressed in transfected cells (Fig. 4A), although their expression levels were lower than that of the full-length EBNA1 in Akata(Ϫ) cells. This is most likely due to the lack of the Gly-Ala repeats in these proteins. The transfected cells were subjected to G418 selection for 5 days to enrich cells that incorporated the plasmids, and subsequently they were proliferated in the absence of G418 for 5 days. The frequencies of GFP-positive cells were determined by FACS by time course (Fig. 4B). The oriP plasmid encoding EBNA1 was stably maintained in transfected cells, as revealed by sustained GFP expression until day 10 (Ͼ60% GFP-positive both in HeLa and Akata(Ϫ) cells). The oriP plasmid encoding EBNA1⌬GA also exhibited sustained GFP expression until day 10 (ϳ50% in HeLa cells, 40% in Akata(Ϫ) cells), although the efficiency was slightly impaired compared with the plasmid encoding the full-length EBNA1. By contrast, the oriP plasmid encoding mEBNA1(24RA) exhib- ited a rapid decline in GFP expression until day 10 (only ϳ5% in HeLa and Akata(Ϫ) cells). Interestingly, the oriP plasmid encoding mEBNA1(24RK) exhibited sustained GFP expression, just like oriP plasmid encoding EBNA1⌬GA. Thus, mEBNA1(24RA) is functionally impaired, whereas mEBNA1(24RK) is functionally intact.
To demonstrate that the sustained GFP expressions actually represent episomal maintenance of plasmid DNAs, episomal plasmids were extracted from transfected HeLa cells, and they were rescued as bacterial clones. A typical result of transforma-tion is shown in Fig. 4C. Transformation efficiency was highest for the test plasmid encoding wild-type EBNA1 and about 2-fold less for the test plasmids encoding EBNA1⌬GA. Importantly, the test plasmid encoding mEBNA1(24RK) derived as many colonies as the test plasmid encoding EBNA1⌬GA. By contrast, very few colonies were obtained from the sample of mEBNA1(24RA), and its transformation efficiency was ϳ200-fold less compared with those of EBNA1⌬GA and mEBNA1(24RK). These results indicate that mEBNA1(24RK) is functionally equivalent to EBNA1⌬GA, and that the difference of GFP expression levels faithfully represents the difference of episomal maintenance of test plasmid DNAs.
Further Dissection of the CBD-mCherry Demonstrates Relatively Small but Significant Contribution of the CBD1 Region-The excellent correlation between the chromosome binding abilities of CBD-mCherry-derived mutants and the biological activities of the corresponding EBNA1 mutants encouraged us to further dissect the CBD-mCherry in greater details. Various deletion mutants, each fused to mCherry, were then transiently expressed in HeLa-H2B-GFP cells (Fig. 5, A-C). The abilities to bind mitotic chromosomes correlated very well with the abilities to localize in interphase nuclei in all the tested mutants. The CBD2-mCherry exhibited more significant chromosome binding and nuclear localization than the CBD1-mCherry, corresponding well with the mutagenesis data of EBNA1⌬GA-mCherry (Fig. 3B). Importantly, the addition of the CBD1 region could rescue the chromosome binding ability that had been lost by partial deletion of the CBD2 region (CBD1CBD2-I-mCherry and CBD1CBD2-II-mCherry; Fig.  5, A and C). Furthermore, two copies of the CBD1 region conferred chromosome binding and nuclear localization activities comparable to those of the CBD2 region (CBD1 ϫ 2-mCherry; Fig. 5C). These results indicate that the 15-aa polypeptide of the CBD1 region enables mitotic chromosome binding as well as interphase chromatin binding at least in this experimental setting.
Western analyses revealed that, although mCherry-positive cells were microscopically observed at comparable frequencies, the protein expression levels of these mutants varied widely. In analogy to Fig. 2B, it was also found that mutants that bind chromosomes are likely to be more stable than mutants that do not (Fig. 5B).
A CBD1-derived Peptide Binds to Reconstituted Nucleosome Core Particles in Vitro-The CBD1 region, which appears to be a minimal unit with chromosome binding ability, is highly basic and likely binds to negatively charged DNA. However, it remained to be clarified whether a CBD1 peptide can bind to nucleosomes, in which the negative charge of DNA is neutralized by the positive charge of histones. To examine the interaction between the CBD1 peptide and nucleosomes, we took advantage of nucleosome core particles that were reconstituted from purified core histones and 146 bp of DNA (35,36). A chemically synthesized 15-aa peptide of the CBD1 region (GR peptide) and two other control peptides (GK and GA) (Fig. 6A) were examined for their ability to bind to reconstituted nucleosome core particles in vitro by means of electrophoretic mobility shift assays (Fig. 6B). The addition of increasing molar ratios of GR peptide to nucleosome core particles resulted in a clear . Transfected cells were first subjected to 5 days of G418 selection (day 5) and then cultivated without G418 for another 5 days (day 10), and their GFP expression was analyzed by FACS. Representative results obtained by two independent experiments were averaged and presented, and similar results were reproducibly obtained (data not shown). Error bars represent ϮS.D. C, shown is a picture of Petri dishes with bacterial clones that were rescued from transfected HeLa cells. Cellular fractions containing episomally maintained plasmids were extracted from transfected HeLa cells on day 10 (Fig. 4B) and used to transform E. coli, DH10B. EBNA1 derivatives that were encoded by the test plasmids are indicated on the right. upward shift of the nucleosomal DNA (Fig. 6B, lanes 2-4). The GK peptide also caused an upward shift of the nucleosomal DNA, although the shift appeared more smear-like in nature (Fig. 6B, lanes 6 -8). The structural and functional differences between these peptides are suggested by the different band shift patterns. In both cases the band of naked DNA disappeared. By contrast, no band shift was observed when the GA peptide was added to the reaction (Fig. 6B, lanes 10 -12). These data clearly indicate that the GR and GK peptides, but not the GA peptide, bind to reconstituted nucleosomes in vitro. These data fit well with the in vivo data showing that EBNA1⌬GA and mEBNA1(24RK), but not mEBNA1(24RA), bind to chromosomes and stably maintain oriP plasmids (Figs. 3C and 4, B and C).

DISCUSSION
Latently infected EBV genomes are stably maintained in the nuclei of host cells as multi-copy episomes. The chromosome (chromatin) binding ability of EBNA1 is a fundamental necessity of episome maintenance, as EBNA1 is a tethering molecule that recruits EBV episomes onto cellular chromatin. Previous studies identified the glycine-arginine rich regions as the CBDs of EBNA1 (7-9, 11, 12).
We revisited the issue by focusing on the importance of arginine residues within the CBDs for the chromosome binding of EBNA1. Amino acid substitution and deletion analyses revealed that no specific arginine residue is indispensable for chromosome binding (Figs. 2 and 5), at least in the context of FIGURE 5. Further dissection of the CBD-mCherry demonstrates relatively small but significant contribution of the CBD1 region. A, shown is a schematic representation of various deletion mutants of CBD fused to mCherry (mCh) on their C termini. Numbers of residual arginines within the CBD region as well as chromosome binding activities of these mutants are indicated as in the legend of Fig. 2A. B, Western blot data demonstrate the expression of various CBD-mCherry mutants. The results obtained by using anti mCherry antibody (top) and anti-␣␤-tubulin antibody (bottom) are shown. Faint but specific bands of mCherry fusions are highlighted by dots. C, live confocal microscopic images demonstrate the localization of various mCherry fusions. Paired images of interphase cells (top two panels) and mitotic cells (bottom two panels) are shown for each mutant. Note that the abilities to bind mitotic chromosomes correlate with the abilities to localize in interphase nuclei. Scale bar, 5 m.
the CBD-mCherry fusion. Both the results obtained by mutagenesis of CBD-mCherry and EBNA1⌬GA-mCherry led to the same conclusion; the arginine residues within both the CBD1 and -2 regions contribute to the chromosome binding, and contribution of the CBD2 region is predominant. The overall data let us hypothesize that the number of arginine residues within the CBD1 and -2 regions primarily determines chromosome binding ability of EBNA1. The extents of contributions of the CBD1 and -2 regions appear to be proportional to the respective numbers of arginines (6 versus 18) within the regions. The mutants 24RA-mCherry (Fig. 2) and mEBNA1(24RA)-mCherry (Fig. 3C) did not show any chromosome binding activity, just like an EBNA1 mutant lacking nearly the entire N-terminal region (a mutant designated as EBNA1⌬16 -372 in Ref. 12). Thus, mEBNA1(24RA) is expected to behave as a dominant negative mutant of EBNA1 (41).
Intriguingly, the arginines within the CBD1 and -2 regions are exchangeable to lysines in the context of EBNA1⌬GA protein. Functional integrity of EBNA1(24RK) was verified by its ability to maintain oriP mini plasmids, just like EBNA1(⌬GA) (Fig. 4, B and C). The data fit very well with the result of in vitro nucleosome binding assay that the synthetic GR and GK peptide, but not the GA peptide, efficiently binds to reconstituted nucleosome core particles (Fig. 6B). This data raise the possibility that EBNA1 simply targets nucleosomes, either as nucleosomal DNA or histones or both via electrostatic interaction. An important implication of this result is that no other cellular protein(s) or RNA molecules is required for this interaction. The LANA protein of KSHV, a functional counterpart of EBNA1, has been shown to recognize "histone H2A-H2B dimer acidic patch" on the nucleosomal surfaces through its N-termi-nal 23 amino acids, which also contains multiple arginines (42). A cellular protein, RCC1, a guanine nucleotide exchange factor for Ran GTPase (43), also recognizes the H2A-H2B dimer acidic patch, and arginine residues of RCC1 play critical roles to interact not only with histones but also with nucleosomal DNA (44). EBNA1 may utilize a similar strategy to interact with cellular chromatin.
A previous report indicated that the chromosome binding sequences of ␤-papillomavirus E2, EBV EBNA1, and Kaposi's sarcoma-associated herpesvirus (KSHV) LANA all contain common RXXS motifs (45). In EBNA1, four copies of RXXS motifs are found in the CBD2 region (Fig. 1A). Our data indicate that even when all the arginine residues within the CBD2 region were substituted with alanines (mEBNA1(CBD2-RA)-mCherry), chromosome binding ability was not completely abolished (Fig. 3B). Thus, the arginines of the RXXS motifs are not essential for chromosome binding of EBNA1, although they contribute to it to a certain extent.
Our data are not exclusive to the hypothesis that EBNA1 binds to chromosomes via cellular proteins, such as EBP2 (13,19) and HMGB2 (23), or via RNA molecules (24). An alternative model is that the CBDs directly contact chromosomal DNA via AT hooks, as has been proposed in a previous study (10). HMGA1a contains three AT hook consensus motifs (PRGRP), which cooperatively mediate efficient chromosome binding (46). Replacing the second arginines of all three AT hook motifs with glycines abolished the chromosome binding of HMGA1a, whereas two intact AT hook motifs conferred enough chromosome binding ability to form a stable interaction (46). In analogy to this observation, EBNA1 may require at least two AT hooklike motifs to efficiently bind to chromosomes (chromatin). Regarding the observed exchangeability between arginines and lysines, these residues may share similar DNA binding preferences. Some AT hook consensus motifs are known to be flanked by highly conserved lysines (25). Thus, it is possible that lysine substitutions retain AT hook activity.
This study also demonstrated that the CBD-mCherry fusion protein is practically useful as a convenient indicator of EBNA1 chromosome binding. We have shown that the chromosome binding abilities of all the tested CBD-mCherry mutants, with the only exception of 24RK-mCherry, correlate well with their abilities to localize to interphase nuclei. This is because CBD-mCherry does not possess the EBNA1 NLS, and they localize to interphase nuclei via binding to interphase chromatin. This raises the possibility that one can quickly assess the mitotic chromosome binding ability of CBD-mCherry simply by examining whether it localizes to nuclei or not. In this respect, a derivative HeLa cell line that stably expresses both CBD-mCherry and histone H2B-GFP (Fig. 1E) should be extremely useful. This cell line can be subjected to high-content screenings to identify small molecules that specifically inhibit chromosome binding of CBD-mCherry without affecting the chromosomal staining of H2B-GFP. Such small molecules may be novel anti-EBV drugs.
Acknowledgments-We thank Dr. Akihiro Abe for the gift of pCGCGP and Dr. Akimitsu Konishi for helpful discussions. FIGURE 6. A CBD1-derived peptide binds to reconstituted nucleosome core particles in vitro. A, shown are amino acid sequences of CBD1-derived peptide (GR) and its derivative peptides with aa substitutions, GK and GA. B, shown is the nucleosome binding assay. Increasing molar amounts of each peptide (indicated on top) were mixed with reconstituted nucleosome core particles. The mixtures were electrophoresed through a non-denaturing polyacrylamide gel, and the gel was stained with ethidium bromide. The positions of nucleosomal DNA and naked DNA are indicated.