Mitotic chromosome interactions of Epstein-Barr nuclear antigen 1 (EBNA1) and human EBNA1-binding protein 2 (EBP2)

Epstein-Barr virus (EBV) genomes stably persist in the nuclei of proliferating cells as double-stranded circular DNA plasmids. This persistence, which occurs during latent infection, involves the replication of the EBV genomes once per cell cycle coupled with a mitotic segregation mechanism in which the EBV genomes are tethered to the cellular chromosomes (reviewed by Frappier, 2004; Frappier, 2010). The replication and segregation of EBV DNA requires a single viral protein, Epstein-Barr nuclear antigen 1 (EBNA1), in addition to cellular factors. EBNA1 contributes to the initiation of EBV DNA replication by binding to four recognition sites present in the dyad-symmetry (DS) element of the latent origin of replication, oriP (Koons et al., 2001; Rawlins et al., 1985). This interaction destabilizes nucleosomes at the origin and aids the recruitment of cellular replication factors to the origin (Avolio-Hunter et al., 2001; Dhar et al., 2001; Julien et al., 2004; Norseen et al., 2008; Schepers et al., 2001). The segregation function of EBNA1 involves EBNA1 binding to 20 recognition sites in the family of repeats (FR) element of oriP, as well as the interaction of EBNA1 with the cellular mitotic chromosomes (Krysan et al., 1989; Lupton and Levine, 1985). EBNA1 binds specific sequences in the FR and DS elements through its C-terminal DNA-binding and dimerization domain (Fig. 1), although this domain is not sufficient for the functional activities of EBNA1 (Ambinder et al., 1991; Bochkarev et al., 1996; Summers et al., 1996).

Given their importance in EBV persistence, the mechanisms by which EBNA1 interacts with mitotic chromosomes and mediates the segregation of EBV episomes has been the subject of several studies. EBNA1, EBV episomes and oriP-based plasmids have all been found to be tightly associated with metaphase chromosomes (Grogan et al., 1983; Harris et al., 1985; Petti et al., 1990; Simpson et al., 1996). EBNA1 and the FR element have been shown to be required for the attachment of oriP-containing plasmids to mitotic chromosomes, although EBNA1 interacts with mitotic chromosomes whether or not it is bound to oriP (Kanda et al., 2001; Kapoor et al., 2005; Wu et al., 2000). Functional characterization of EBNA1 mutants has shown that the DNA replication and segregation functions of EBNA1 are distinct in terms of their amino acids requirements. In particular, deletion of the Gly-Arg-rich sequence between residues 325-376 (Fig. 1) abrogates the ability of EBNA1 to stably maintain oriP plasmids in human cells without affecting the replication of these plasmids, a pattern that is consistent with a specific role in plasmid segregation (Wu et al., 2002). This mutation also decreases the association of EBNA1 with metaphase chromosomes, which is consistent with the chromosome-tethering model of segregation (Wu et al., 2002). A similar but more subtle effect on plasmid segregation was observed when residues 8-67 were deleted (Wu et al., 2002). This N-terminal region includes a second Gly-Arg-rich sequence between residues 33 and 53 (Fig. 1), although deletion of this sequence alone had no detectable effect on any EBNA1 function (Wu et al., 2002). Both Gly-Arg-rich sequences of EBNA1 have been shown to have a propensity to bind to mitotic chromosomes when excised from EBNA1 and fused to other proteins (Hung et al., 2001; Marechal et al., 1999). However, the degree to which each is involved in chromosome interactions in the context of the native EBNA1 protein is unclear. The similarity of the sequences in these Gly-Arg regions to AT hook sequences has led to the hypothesis that these sequence might directly contact chromosomal DNA (Sears et al., 2004), although the same sequences are known to interact with at least a few different cellular proteins (Holowaty et al., 2003; Shire et al., 1999; Snudden et al., 1994; Van Scoy et al., 2000; Wang et al., 1997).

To date, one cellular protein, EBNA1-binding protein 2 (EBP2), has been found to have a role in EBNA1-mediated plasmid segregation. This protein was discovered as an EBNA1 binding partner in a two-hybrid screen, and residues 325-376 of EBNA1 are important for this interaction (Shire et al., 1999). It was subsequently shown that EBP2 enables EBNA1 to segregate FR-containing plasmids in yeast by facilitating the interaction of EBNA1 with yeast mitotic chromatin (Kapoor and Frappier, 2003; Kapoor et al., 2001). In keeping with this finding, EBP2 associates with human metaphase chromosomes, and silencing of EBP2 expression results in a large decrease in the amount of EBNA1 associated with metaphase chromosomes (Kapoor et al., 2005; Wu et al., 2000). In addition, the expression of human EBP2 in murine Sp2/0 cells enables EBNA1 to stably maintain oriP-based plasmids in these cells (Habel et al., 2004). However, it is not clear whether EBP2 enables the initial association of EBNA1 with the chromosomes, or whether it is required to maintain or stabilize the EBNA1-chromosome interaction.

EBP2 is highly conserved in eukaryotes (Henning and Valdez, 2001; Shire et al., 1999), where it forms part of the nucleolus (Andersen et al., 2002; Chatterjee et al., 1987). Studies in Saccharomyces cerevisiae show that yeast EBP2 has an essential role in processing 27S pre-rRNA to 25S rRNA (Huber et al., 2000; Tsujii et al., 2000) and, in keeping with this role, EBP2 is also required for cell proliferation in human cells (Kapoor et al., 2005). In mammals, the nucleolus disassembles at the beginning of mitosis and the components disperse, some of which then associate with the cellular chromosomes to form a peripheral layer (Gautier et al., 1992; Hernandez-Verdun and Gautier, 1994). The timing of redistribution of specific nucleolar proteins appears to be highly ordered, although only a small number of nucleolar proteins have been followed through mitosis (Leung et al., 2004; Zatsepina et al., 1997). Our initial observations suggest that EBP2 belongs to the class of nucleolar protein that redistributes to the chromosomes in mitosis (Kapoor et al., 2005; Wu et al., 2000), but nothing is known about the timing of this movement. In addition, controversy on the localization of EBP2 was raised when one group failed to detect EBP2 on mitotic chromosomes (Sears et al., 2004).

Although it is well established that EBNA1 associates with metaphase chromosomes, few studies have examined the timing with which EBNA1 associates with host chromosomes through mitosis. In addition, it remains controversial as to whether or not EBNA1 binds chromosomes in interphase (Daikoku et al., 2004; Ito et al., 2002; Kanda et al., 2007; Kanda et al., 2001; Ritzi et al., 2003). In this study, we address the timing of the association of EBP2 and EBNA1 with host chromosomes through mitosis to try to understand how EBP2 contributes to EBNA1-mediated segregation and how relocalization of EBP2 from the nucleolus compares with that of other nucleolar proteins. We also investigate the contribution of the two Gly-Arg-rich regions of EBNA1 to interactions with EBP2, and to chromosome association at different stages of mitosis.

To date, the subcellular localization of EBP2 has only been examined in interphase- and in metaphase-blocked cells. We wanted to assess the localization of EBP2 throughout mitosis more carefully, in the absence of agents that block progression through the cell cycle. Imaging of mitotic cells is often hampered by their round morphology. To circumvent this problem, we chose to conduct our studies in Rat2 cells, which have a relatively flat morphology throughout mitosis and hence are suitable for mitotic imaging (Bakal et al., 2005). Endogenous EBP2 in log-phase cells was visualized using polyclonal antibody raised against highly purified human EBP2, and cells at various stages of mitosis were identified by the DNA morphology after DAPI staining.

The localization of EBP2 was first examined in prophase cells, which were identified according to their spaghetti-like DNA strands and also, in some cases, by the presence of phosphorylated histone H3 (Fig. 2A). In all prophase cells examined, EBP2 was localized to one or two discreet regions of the nucleus, which is consistent with its known nucleolar localization in interphase (Fig. 2A,B top panel). B23 (also called nucleophosmin) was examined as a marker for a nucleolar protein that is known to redistribute to the cellular chromatin during prophase (Dundr et al., 1997). In prophase cells where B23 was localized to the nucleolus, EBP2 always showed similar localization as B23, confirming that EBP2 is largely nucleolar in prophase (Fig. 2A, middle panel). However, prophase cells were also evident in which B23 was dispersed over the host DNA (Fig. 2A, bottom panel), consistent with previous reports that B23 leaves the nucleolar region during prophase (Dundr et al., 1997; Zatsepina et al., 1997). In these cells, EBP2 remained largely localized to the nucleolar region, although a small amount of staining outside of this region was also detected. These results indicate that EBP2 remains in the nucleolus after the redistribution of B23.

Rat2 cells at various stages of mitosis were also imaged by deconvolution of optical stacks (Fig. 2B, left panels). The majority of EBP2 was observed to be associated with the condensed chromosomes in metaphase, anaphase and telophase, where it localized predominantly to the edges of the chromatin. To verify that this localization is not particular to rodent EBP2, the same experiments were performed with CFPAC cells, a human pancreatic adenocarcinoma cell line also known for its flat morphology in mitosis (Levesque et al., 2003). EBP2 localization in these cells throughout mitosis was found to be identical to that in Rat2 cells (Fig. 2B, right panels). The results indicate that EBP2 relocalizes to the chromosomes very late in prophase and forms part of the chromosomal peripheral layer in metaphase through to telophase.

We next wanted to compare the localization of EBNA1 and EBP2 throughout mitosis, and determine whether EBNA1 expression affected the localization of EBP2. For these experiments, we generated Rat2 cells expressing EBNA1 fused to a dimeric form of the HcRed fluorescent tag. Log-phase fixed cells were visualized for Red-tagged EBNA1 and endogenous EBP2 (using anti-EBP2 antibody) and categorized into mitotic phases based on morphology of the DAPI-stained DNA. Deconvolved images are shown in Fig. 3. In prophase (Fig. 3A), EBNA1 fluorescence largely corresponded to the condensed DNA. By contrast, most of the EBP2 was localized to the nucleolar region, with a small amount exhibiting more dispersed staining, as observed in the absence of EBNA1. Tracings showing a degree of correspondence of EBNA1, EBP2 and DAPI staining are shown for each image (right panels), as determined from a line drawn diagonally through the image (indicated in merged images) using ImageJ software. The results indicate that EBNA1 does not alter the localization of EBP2 and that most of the EBNA1 is associated with host chromatin in prophase.

In metaphase through to telophase, EBNA1 was tightly associated with the host chromosomes, resulting in diffuse staining over the entire chromatin mass (Fig. 3B-D). EBP2 behaved as it did in the absence of EBNA1, with a significant portion of the protein associating with the edges of the chromatin, beginning in metaphase and extending through to telophase. For each stage of mitosis, several images were analysed by ImageJ software and Pearson coefficients of colocalization were determined for six deconvolved images at each stage of mitosis (where a value of `1' indicates perfect colocalization). The average values (shown in Fig. 3E) confirm that, although there is little correspondence of EBNA1 and EBP2 in prophase, the overlap in the localization of these proteins significantly increases, starting in metaphase and continuing into anaphase and telophase.

We also examined the localization of EBNA1 and EBP2 in metaphase in the context of latent EBV infection. EBV establishes latent infection in B cells, in some cases contributing to the development of Burkitt's lymphoma. The round morphology of B cells severely impairs the imaging of these cells; however, localization of protein to mitotic chromosomes can be assessed by imaging spreads of the mitotic chromosomes. Therefore, Akata EBV-positive Burkitt's lymphoma cells were blocked in metaphase, and the chromosomes were spread on slides and stained with antibodies against endogenous EBNA1 and EBP2. Images from several cells were analysed by deconvolution and three layers from the deconvolved Z-stacks of a representative chromosome spread are shown in Fig. 4, starting from the middle of the chromosomes (top row) and moving towards the outer layer (bottom row). Line scans taken diagonally across the images are also present, showing the degree of correspondence of DAPI, EBNA1 and EBP2 staining. EBNA1 exhibited diffuse staining that corresponded well to the DAPI staining, and was most intense in the middle layer of the chromosomes. EBP2 staining also largely corresponded to the DAPI staining, and in many cases with EBNA1 peaks. However, EBP2 was more prevalent in the outer layer of the chromosomes than in the middle layer, which is consistent with the peripheral chromosome staining observed in Rat2 and CFPAC cells. It is important to note, however, that at each layer through the chromosomes, there are yellow dot, which are indicative of an interaction between a proportion of EBNA1 and EBP2. Line scans from three different chromosome spreads were used to calculate Pearson coefficients of colocalization for EBNA1 and EBP2 (Fig. 3E). The average coefficient of 0.38 was similar to that observed in the later stages of mitosis in Rat2 cells, and is consistent with colocalization of a proportion of the two proteins.

Both the N-terminal (residues 34-52) and central (residues 325-376) Gly-Arg-rich regions of EBNA1 have been shown to have some capacity to bind metaphase chromosomes when excised from EBNA1, and the central Gly-Arg region has been shown to be important for association of EBNA1 with chromosomes from metaphase-blocked cells. However, the degree to which the two Gly-Arg regions contribute to chromosome interactions of native EBNA1 at various stages of mitosis has not been examined. We addressed this by expressing EBNA1 deletion mutants (without tags) in human nasopharyngeal carcinoma (CNE2Z) cells, staining log-phase cells for EBNA1 using EBNA1-specific monoclonal antibody and examining cells at various stages of mitosis according to the DNA morphology after DAPI staining (Fig. 5). CNE2Z cells were used because they are derived from an EBV-positive nasopharyngeal carcinoma (a tumour that is highly associated with latent EBV infection), but lose the EBV genomes during growth in culture. The localization of EBNA1 in CNE2Z was indistinguishable from that observed above for Rat2, showing very tight correlation to the condensed DNA at prophase through to telophase (Fig. 5).

Comparison of the localization of EBNA1 mutants (shown in Fig. 1) in prophase cells showed that EBNA1 proteins lacking the N-terminal Gly-Arg region (Δ8-67 or Δ34-52) remained closely associated with the condensed DNA throughout mitosis, although a small amount of staining outside the chromosomes was observed for Δ8-67 at each stage of mitosis (Fig. 5). A more obvious effect on EBNA1 localization was observed when the central Gly-Arg region was deleted in EBNA1 mutant Δ325-376. This mutant exhibited considerable staining both on and outside the chromosomes at all stages of mitosis. Deletion of both of the Gly-Arg regions in Δ8-67Δ325-376 or Δ34-52Δ325-376 resulted in a diffuse staining pattern that did not correspond to the chromosomes at any stage of mitosis (Fig. 5). Note that the intensity of EBNA1 staining was similar for all mutants, indicating that there were no major differences in expression levels. The results indicate that sequences 325-376 are predominantly responsible for chromosome interactions throughout mitosis and that sequences 34-52 make a minor contribution to chromosome attachment that can partially substitute for the 325-376 sequence in its absence.

The contribution of EBNA1 residues 8-67 and 325-376 to metaphase chromosome attachment was also assessed by biochemical fractionation of colcemid-blocked cells (Fig. 6A, right panel). In these experiments, cells expressing the EBNA1 proteins were lysed in isotonic buffer and subjected to centrifugation to separate soluble proteins from insoluble chromatin-bound proteins. Consistent with the microscopy data, this method showed a tight association of EBNA1 with the chromosomal pellet and that deletion of residues 325-376 results in the release of a large proportion of EBNA1 from the chromosomes. EBNA1 Δ8-67, however, remained largely localized to the chromosomal pellet, with a release of a smaller proportion of EBNA1 into the soluble fraction. Note that, because the fractionation method requires that EBNA1 remains associated with the chromosomes during centrifugation, this should result in the release of proteins weakly associated with the chromosomes, and hence show a higher proportion of proteins in the soluble fraction than is evident by microscopy. The fractionation results confirmed that residues 325-376 have a major role in chromosome attachment and that the N-terminal sequences have a more modest role.

There have been mixed reports in the literature as to whether EBNA1 is associated with cellular chromatin in interphase. For example, biochemical fractionation performed by Kanda and co-workers (Kanda et al., 2001) support the conclusion that EBNA1 is bound to chromatin, whereas those performed by Daikoku and colleagues (Daikoku et al., 2004) indicate that EBNA1 is largely soluble. However, these experiments were performed using different buffer conditions (salt and detergent concentrations), which might account for this discrepancy. To further investigate the possible association of EBNA1 with interphase chromatin, we performed biochemical fractionation of log-phase cells using the same conditions that we used to detect EBNA1 binding to mitotic chromosomes (isotonic conditions) and compared this with the same fractionation performed in buffer systems used by Kanda and colleagues [hypotonic sucrose buffer (HSB)] and Daikoku and co-workers [CSK with 0.5% Triton X-100 (mCSK)]. EBNA1 pelleted with the chromosomes in interphase in isotonic and HSB buffers, but not in mCSK buffer. When biochemical fractionation was performed on metaphase-blocked cells, similar results were obtained: the mCSK buffer resulted in the solubilization of EBNA1. Given that the interaction of EBNA1 with metaphase chromosomes is well established, the results indicate that mCSK buffer disrupts the interaction of EBNA1 with chromosomes, and hence does not provide reliable information on whether or not EBNA1 is associated with interphase chromatin. However, the results with both isotonic and mCSK buffers are consistent with EBNA1 binding to interphase chromatin, although association with other insoluble components is also possible.

We also performed biochemical fractionations (using isotonic buffer) with log-phase cells expressing EBNA1 mutants, to assess whether sequences involved in chromosome attachment in mitosis also mediate the putative chromatin interactions in interphase. We found that, as in mitosis, residues 325-376 have a major role in interphase chromosome interactions, whereas deletion of residues 8-67 had a much smaller effect. The results suggest that residues 325-376 have an important role in chromosome interactions throughout the cell cycle.

To better understand the relationship between the mitotic chromosome interactions of EBNA1 and EBP2 binding, we examined the ability of the EBNA1 mutants used in the localization studies to bind to EBP2. We have previously shown that this interaction can be detected by yeast two-hybrid assay and that residues 325-376 are important for this interaction (Shire et al., 1999; Shire et al., 2006). We have now used a more sensitive version of this assay based on LexA fusion proteins, to evaluate whether EBNA1 N-terminal sequences also contribute to EBP2 interactions (Fig. 7A). Deletion of residues 34-52 in EBNA1 was found to have a negligible effect on EBP2 binding, whereas deletion of residues 325-376 severely decreased EBP2 binding. However, some ability to bind EBP2 was still detected with Δ325-376 mutants, and this was further decreased by removal of residues 34-52 (compare Δ325-376 to Δ34-52Δ325-376 in the bottom right panel of Fig. 7A). Differences in EBP2 binding were not due to different levels of expression of the EBNA1 mutants, because all were expressed in yeast at similar levels (Fig. 7B). The observed effects mirror the results for EBNA1 attachment to mitotic chromosomes, raising the possibility that the mitotic chromosome interactions mediated by residues 325-376 or by residues 34-52 (as occurs when 325-376 is deleted) might involve EBP2.

We have examined the timing with which EBP2 and EBNA1 associate with cellular chromosomes in the cell cycle and the involvement of EBNA1 sequences in this process. EBP2 is a nucleolar protein (Andersen et al., 2002; Chatterjee et al., 1987) that has an essential role in rRNA processing (Huber et al., 2000; Tsujii et al., 2000). In mammals, the nucleolus undergoes an ordered disassembly at the beginning of mitosis, then reforms in telophase. The mechanisms by which this complex structure disintegrates and reforms at each cell cycle has been the subject of several studies, but is not well understood. The proteins that normally reside in the nucleolus in interphase have been reported to undergo various fates in mitosis. Proteins involved in rRNA transcription generally remain associated with the nucleolar-organizing regions in the condensed chromosomes. However, protein components involved in pre-rRNA processing have been reported to associate with the periphery of all of the chromosomes and/or form particles in the mitotic cytoplasmic particles, including nucleolus-derived foci (Dundr et al., 1997; Gautier et al., 1992; Hernandez-Verdun and Gautier, 1994; Olson and Dundr, 2005; Shaw and Jordan, 1995). Our observations indicate that EBP2 is one of the nucleolar proteins that associates with the periphery of the chromosomes in mitosis, and this is consistent with previous observations that proteins involved in pre-rRNA processing form part of the peripheral layer of the chromosomes (Leung et al., 2004; Zatsepina et al., 1997).

One of the most studied nucleolar proteins in terms of its localization in mitosis is B23, which is involved in late rRNA processing steps and is part of the granular component of the nucleolus. B23 leaves the nucleolus in prophase to associate with the chromosome periphery and this relocalization has been found to occur later in prophase than for other nucleolar proteins, particularly those from the fibrillar center of the nucleolus (Leung et al., 2004). Here, we have shown that EBP2 leaves the nucleolus even later than B23. However, similarly to B23, EBP2 is largely (but not entirely) associated with the peripheral chromosome layer.

EBNA1 is essential for the long-term maintenance of EBV episomes in replicating cells because of contributions to both the replication and mitotic segregation of the episomes, the latter of which requires EBNA1 attachment to the chromosomes. However, the timing by which EBNA1 associates with chromosomes in the cell cycle has not been clear, owing to conflicting data. For example, Ito and co-workers (Ito et al., 2002) reported that EBNA1 is associated with interphase chromatin that had been artificially condensed by calyculin A treatment; however, this treatment might also artificially induce protein-chromosome interactions. Using biochemical fractionation, Kanda and colleagues (Kanda et al., 2001) observed that EBNA1 from interphase cells sedimented with the chromosomal pellet and was released by micrococcal nuclease, which is consistent with interaction of EBNA1 with chromosomes in interphase. By contrast, biochemical fractionation approaches used by Daikoku, Ritzi and colleagues (Daikoku et al., 2004; Ritzi et al., 2003) found little EBNA1 from interphase cells in the chromosomal pellet. Such divergent results might be due to the different buffer conditions used in each study. Indeed, we showed that seemingly small changes in buffer conditions can have major effects on EBNA1 solubility, and that the association of EBNA1 with mitotic chromosomes is disrupted in some buffers used to demonstrate chromosome interactions of other proteins (such as CSK buffer with 0.5% Triton X-100). Therefore, one cannot draw conclusions on EBNA1-interphase chromatin interactions under these conditions. Using buffers that do not disrupt EBNA1-mitotic chromosome interactions, we found that EBNA1 fractionates with the chromosomal pellet in log-phase cells, which is consistent with a chromatin interaction in interphase. In addition, we have clearly shown by immunofluorescence imaging, that EBNA1 is associated with chromosomes by prophase.

EBNA1 contains two Gly-Arg-rich regions, one between amino acids 33-53 and a larger one spanning residues 325-376. Studies involving excision of these sequences and fusions to other proteins or to the EBNA1 DNA-binding domain have indicated that each region has some ability to interact with metaphase chromosomes, suggesting that they are important in the mitotic segregation function of EBNA1 (Hung et al., 2001; Marechal et al., 1999; Sears et al., 2004). However, functional analysis of EBNA1 deletion mutants indicated that only residues 325-376 are important for segregation. Specifically, deletion of amino acids 325-376 abrogated the ability of EBNA1 to maintain EBV-based plasmids in replicating cells, without affecting the DNA replication function of EBNA1, whereas deletion of residues 34-52 had no effect on either the long-term maintenance or the replication of EBV-based plasmids (Shire et al., 1999; Wu et al., 2002). The fact that these deletion mutants retained at least one functional activity of EBNA1 shows that these deletions did not cause misfolding of EBNA1, a result that is consistent with the fact that these regions are within an extended region of EBNA1 as opposed to being folded into a protease-resistant domain (E. Bochkareva and L. Frappier, unpublished results).

To better understand the above observations, it was necessary to more carefully examine the contributions of residues 33-53 and 325-376 to chromosome attachment throughout mitosis. Here, we have shown that residues 325-376 have a major role in the chromosome interactions of EBNA1 at each stage of mitosis, whereas deletion of residues 34-52 had no obvious effect on chromosome interactions at any stage of mitosis. These results agree well with the functional effects on segregation, and strongly suggest that the segregation defect of Δ325-376 is due to insufficient chromosome attachment. In addition, the ability of EBNA1 N-terminal sequences to mediate chromosome attachment was revealed when residues 325-376 were deleted, because the chromosome interactions that were seen with this mutant were abrogated by removal of amino acids 34-52 or 8-67. This suggests that the N-terminal Gly-Arg-rich sequences can mediate weak chromosome interactions in the absence of residues 325-376 that form the principal chromosome-binding region. It is also interesting that Δ8-67 mutant on its own resulted in detectable effects on chromosome interactions at each stage of mitosis (which was not seen with the Δ34-52 mutant), suggesting that sequences adjacent to the N-terminal Gly-Arg contribute to chromosome interactions. This observation is consistent with previous functional data showing that the Δ8-67 mutant had reduced ability to maintain oriP plasmids (Wu et al., 2002).

In our previous studies, we identified EBP2 as a protein capable of interacting with EBNA1 through its C-terminal domain (Kapoor et al., 2001; Shire et al., 1999). We also showed that this protein interacts with chromosomes in metaphase-blocked cells through its central coiled-coil domain (Kapoor and Frappier, 2003; Kapoor et al., 2005; Wu et al., 2000). Remarkably, expression of human EBP2 in S. cerevisiae enabled EBNA1 to segregate plasmids containing the EBV segregation element (FR) in yeast in a manner that was dependent on both the chromosome-interacting and EBNA1-interacting domains of EBP2, and segregation in this system was shown to be sensitive to the same EBNA1 mutations as human cells (Kapoor and Frappier, 2003; Kapoor et al., 2001). In addition, segregation of FR-containing plasmids in yeast was also achieved by expressing a single protein in which the EBP2 chromosome-binding domain was fused to the EBNA1 DNA-binding domain (Kapoor and Frappier, 2003). The importance of EBP2 for EBNA1-mediated segregation in human cells was suggested by the close correlation of effects of point mutations in residues 325-376 on plasmid-maintenance activity and EBP2 binding (Shire et al., 2006), and further confirmed by the finding that silencing of EBP2 in four different human cells lines (including EBV-positive cells) led to a substantial decrease in the association of EBNA1 with metaphase chromosomes (Kapoor et al., 2005). As expected, this loss of EBNA1 from the mitotic chromosomes was accompanied by a similar loss of oriP plasmids from the chromosomes (Kapoor et al., 2005). The results as a whole indicate that an interaction between EBP2 and EBNA1 is important for proper association of EBNA1 with mitotic chromosomes and hence for the segregation function of EBNA1. However, it is not clear why EBP2 was required for the association of EBNA1 with mitotic chromosomes. Two possibilities seem likely: (1) EBP2 might be required for the initial attachment of EBNA1 to chromosomes by mediating EBNA1 interactions with the chromosomes; or (2) EBP2 might associate with EBNA1 after its initial chromosome interaction and may be required to stabilize the EBNA1-chromosome interaction. The data presented here on the timing of the association of EBNA1 and EBP2 with chromosomes through mitosis strongly support the second possibility.

The cumulative results on EBNA1-chromosome associations suggest a model in which interactions between EBNA1 and chromosomes that govern mitotic segregation involve at least two steps. The first chromosome interaction probably occurs in interphase and does not involve EBP2. At this stage, EBNA1 might interact directly with the cellular DNA, as proposed by Sears and colleagues (Sears et al., 2004), or it might involve interactions with one or more chromatin-associated proteins. The second stage of EBNA1-chromosome interaction occurs during the second half of mitosis, and might require an interaction with EBP2 to keep EBNA1 tightly associated with the chromosomes, either because of stabilization of the initial EBNA1-chromosome contact or the addition of a second contact. This two-step scenario has interesting parallels with the segregation of papillomavirus genomes that is mediated by the E2 protein of bovine papillomavirus (BPV), which, similarly to EBNA1, governs the segregation of BPV episomes by tethering them to the cellular chromosomes in mitosis. Interactions of E2 with mitotic chromosomes have been shown to occur through two different host proteins, Brd4 (Brannon et al., 2005; You et al., 2004) and ChlR1 (Parish et al., 2006), leading to a model in which ChlR1 is required for the initial attachment of E2 to chromosomes before mitosis and/or in early mitosis, and in which Brd4 is involved in E2-chromosome interactions at later stages of mitosis (Parish et al., 2006). Thus several mechanisms of chromosome interaction at different stages of mitosis might be a common theme for viral episomes that segregate by chromosome attachment.

Rat2 thymidine kinase-deficient rat fibroblasts (Reynolds et al., 1987; Topp, 1981) were propagated in DMEM. CFPAC-1 pancreatic adenocarcinoma cells (Schoumacher et al., 1990) were maintained in Iscove's modified Dulbecco's medium (Gibco). Akata cells (Burkitt's lymphoma cell line) were maintained in RPMI 2640 medium (Sigma). CNE2Z EBV-negative nasopharyngeal carcinoma cells were grown in α-MEM. CNE2Z cells were derived from EBV-positive tumour cells that lose the EBV upon growth in culture (Sun et al., 1992). In all cases, medium was supplemented with 10% fetal bovine serum, 5% penicillin and streptomycin and 5% L-glutamine.

To make the diHcRed-tagged EBNA1 construct, EBNA1 (lacking most of the Gly-Ala repeat) was PCR amplified from the pAS2.EBNA1 vector (Shire et al., 1999) with primers that placed HindIII and BamHI sites flanking the cDNA. After digestion with HindIII and BamHI, the cDNA was inserted between the HindIII and BamHI sites of the pdiHcRed-C1 vector [a modification of pHcRed1-C1 from Clontech Laboratories kindly supplied by David Bazett-Jones, The Hospital for Sick Children, Toronto, Canada (Dellaire et al., 2006)] just after the cDNA encoding a dimeric version of HcRed. The cDNA encoding the fusion protein was then excised with NheI and BamHI and inserted between the NheI and BamHI sites of pcDNA3.1/Hygro(–) (Invitrogen). For experiments involving the localization of EBNA1 mutants in human cells, EBNA1 proteins were expressed from pc3oriP and were generated as previously described (Wu et al., 2002). Constructs for yeast two-hybrid assays were generated by subcloning the EBNA1 mutants into pLexA-kan (a gift from Igor Stagljar, University of Toronto, Toronto, Canada) between the NdeI and BamH1 sites, resulting in fusion to the LexA DNA-binding domain.

Rat2 and CFPAC cells were plated on coverslips at a density of 1-2×10⁵ cells/ml at 37°C in complete medium, washed with PBS, and fixed in 3% formaldehyde in PBS. The cells were then permeabilized with 0.1% Triton X-100 for 5-10 minutes, washed with PBS, and blocked in 3% BSA in PBS for 30 minutes before staining with rabbit anti-EBP2 (Wu et al., 2000), goat anti-B23 (C10; Santa Cruz Biotechnology, CA) and mouse anti-phosphohistone H3 (Upstate Laboratories) antibodies. After washing in PBS with 0.01% Tween 20, cells were treated with the following secondary antibodies: fluorescein-isothiocyanate-conjugated donkey anti-rabbit antibody for EBP2 (Santa Cruz Biotechnology), donkey anti-goat Cy3 secondary antibody for B23 (Chemicon) and Texas-red-conjugated goat anti-mouse for phosphohistone H3 (Molecular Probes, Eugene, OR). After washing, cells were mounted on coverslips with Prolong Antifade, containing DAPI (4′,6′-diamidino-2-phenylindole; Invitrogen). To generate Rat2 cells expressing diHcRed-EBNA1, a 10 cm plate of Rat2 cells were transfected with 5 μg pdiHcRed-EBNA1 using lipofectamine then propagated in medium containing 200 μg/ml hygromycin (Gibco) to select for the plasmid. After 3 weeks of selection, cells were fixed and stained for EBP2, as described above.

Microscopy was performed using a Leica DM-IRE2 wide-field inverted fluorescence microscope, a Hamamatsu ORCA-AG CCD camera, appropriate filters and Openlab software (version 4.0.2). Exposure times for DAPI, hEBP2 and diHcRedEBNA1 imaging were kept constant within each set of experiments. Where indicated, digital images were collected at defined intervals (0.2-0.3 μm) along the Z-axis and the images were deconvolved (60 iterations) using the 3D deconvolution software (Image Pro software, MediaCybernetics). Digital images were imported into ImageJ (NIH, Bethesda, MD) for line pixel analysis and calculation of the extent of colocalization between EBNA1 and EBP2. Images were false-coloured using ImageJ software and presented using Adobe PhotoShop 5.5 (Improvision, San Jose, CA). The Pearson product moment correlation coefficient (r), a measure of the correlation of two variables X and Y on the same object or organism, was calculated using ImageJ software.

For chromosome spreads, Akata cells (EBV-positive Burkitt's lymphoma) were blocked in mitosis with colcemid (0.1 μg/ml) for 16 hours, swollen in hypotonic buffer, fixed with methanol:acetic acid (70:30) and dropped onto cooled slides as previously described (Kapoor and Frappier, 2003). The chromosomes were then stained for EBP2 and EBNA1 and counterstained with DAPI, as previously described (Kapoor et al., 2005), using anti-EBP2 rabbit antibody and OT1x monoclonal antibody against EBNA1 (kindly supplied by Jaap Middeldorp, VU University Medical Center, Amsterdam, The Netherlands), followed by fluorescein-isothiocyanate-conjugated donkey anti-rabbit and Texas-red-conjugated goat anti-mouse secondary antibodies. Images were captured and processed as described above.

CNE2Z cells, grown on coverslips in a six-well dish, were transfected with 1 μg pc3oriP expressing EBNA1 or EBNA mutants using Fugene HD (Roche). Two days later, cells were fixed in 3% formaldehyde for 30 minutes and permeabilized with 1% Triton X-100. EBNA1 was detected with OT1x monoclonal antibody followed by Alexa-Fluor-488-conjugated goat anti-mouse antibody (Invitrogen) and DAPI (25 ng/μl) staining. Microscopy was performed as described above.

Biochemical fractionation was performed either on log-phase cells or on cells blocked in mitosis with colcemid treatment (0.1 μg/ml medium) followed by mitotic shakeoff, as previously described (Kapoor et al., 2005). Unless otherwise indicated, cells from a 6-cm dish were lysed in 100 μl of isotonic buffer [20 mM Tris-HCl pH 7.5, 75 mM KCl, 30 mM MgCl₂, 0.5% Nonidet P-40, 1 mM dithiothreitol (DTT), 0.5 mM EDTA, 1 mM PMSF, 1 mM benzamidine]. A 50 μl sample of the cell lysate was removed (sample W) and the remaining 50 μl of lysate was spun for 10 minutes at 4°C in a microcentrifuge. The supernatant (sample S) was removed and the pellet was resuspended in 50 μl isotonic buffer (sample P). Where indicated, isotonic buffer was substituted with modified CSK buffer (mCSK: 10 mM PIPES pH 6.8, 100 mM NaCl, 300 mM sucrose, 1 mM MgCl₂, I mM EGTA, 1 mM DTT, 1 mM PMSF, 0.5% Triton X-100) (Daikoku et al., 2004) or hypotonic sucrose buffer (HSB: 10 mM HEPES pH 7.9, 10 mM KCl, 1.5 mM MgCl₂, 0.34 M sucrose, 10% glycerol, 1 mM DTT, 0.1% Triton X-100, protease inhibitors) (Kanda et al., 2001). Aliquots of 25 μl of each of W, S and P samples were subjected to SDS-PAGE and western blotting using OT1x monoclonal antibody against EBNA1 and goat anti-mouse antibody conjugated with HRP (Santa Cruz Biotechnology). Membranes were developed with the ECL Plus system (Amersham Biosciences).

Two-hybrid assays were performed as described (Shire et al., 2006). Briefly, S. cerevisiae strain L40a (MATa trp1 leu2 his3 LYS2::lexA-HIS3 URA3::lexA-lacZ) was transformed with one of the pLexA.EBNA1 constructs and with pACTII expressing hEBP2 (fused to the GAL4 activation domain) or with empty pACTII (negative control). Yeast were grown overnight in medium selective for both plasmids (lacking Trp and Leu) and tenfold serial dilutions of the cultures were spotted and grown on plates lacking Trp and Leu with and without His.