In eukaryotes, many latent viruses replicate as extrachromosomal molecules, called episomes, and efficiently segregate to daughter cells by noncovalently attaching to mitotic chromosomes. To understand the mechanism governing the processes, we analyzed the detailed subcellular localization of Epstein-Barr virus (EBV) genomes and a viral protein EBNA1, a bridging molecule between viral genomes and cellular chromatin. In the cells that were infected with a recombinant EBV expressing epitope-tagged EBNA1, EBNA1 localized to intranuclear punctate dots, which coincided with the localization of EBV genomes as revealed by fluorescence in situ hybridization (FISH). A significant number of EBNA1 dots were found to localize symmetrically on sister chromatids of mitotic chromosomes. Such symmetrical localization of EBNA1 dots was observed in prematurely condensed G2 chromosomes as well, correlating with the presence of closely spaced double dots of EBNA1 in G2-phase-enriched cells. The EBNA1 double dots were occasionally interconnected by the FISH signals of EBV episomes, exhibiting a dumbbell-like appearance. Thus, we propose that the partitioning of EBNA1 molecules onto sister chromatids during cellular DNA replication underlies the non-stochastic segregation of extrachromosomally replicating viral genomes.
Accurate and complete replication and subsequent faithful segregation of cellular chromosomes are strictly controlled processes. During the DNA synthesis (S) phase of the cell cycle, all of the chromosomes replicate only once. Pairs of duplicated chromosomes, designated as sister chromatids, remain attached side-by-side until the cell enters mitosis (reviewed by Losada and Hirano, 2005). Sister chromatids start to separate at the metaphase-anaphase transition during mitosis and, once separated, they move to opposite ends of the cell to be incorporated into the nucleus of each daughter cell.
Some extrachromosomally replicating viruses are known to piggyback onto cellular chromosomes in order to stably transmit their genomes to the daughter cells (reviewed in Botchan, 2004; Calos, 1998). Among these, Epstein-Barr virus (EBV) is the best-characterized. EBV genomes exist as 170-kb double-stranded circular molecules, called episomes, in latently infected human B cells (Rickinson and Kieff, 2001). Engineered mini-plasmids that harbor the cis-acting latent origin-of-replication (oriP) and the gene encoding the transacting viral protein EBNA1, but which lack other viral sequences, are stably maintained in dividing human cells (Yates et al., 1985). The result indicates that oriP and EBNA1 are minimal essential elements for episomal maintenance.
EBNA1 binds to the oriP region of the EBV genome, which includes multiple copies of the EBNA1-binding sequences (Rawlins et al., 1985; Reisman et al., 1985). EBNA1 binding to the oriP sequences has multiple functions. As a sequence-specific DNA-binding protein, it stimulates transcription of viral latent genes; it also mediates replication and nuclear retention of oriP-containing plasmids (reviewed in Leight and Sugden, 2000; Mackey and Sugden, 1999). Biochemical analysis has revealed that it binds constitutively to the oriP region throughout the cell cycle (Hsieh et al., 1993). Recent studies have established the importance of EBNA1 for holding viral episomes on host chromosomes. The centrally located glycine/arginine-rich region of EBNA1 has the ability to bind to host mitotic chromosomes (Hung et al., 2001; Kanda et al., 2001a; Marechal et al., 1999; Sears et al., 2004; Wu et al., 2000; Wu et al., 2002), and an N-terminally located one may also contribute to the ability (Hung et al., 2001; Kanda et al., 2001a; Marechal et al., 1999; Sears et al., 2004). However, the C-terminal region of EBNA1 is responsible for oriP binding (Ambinder et al., 1991). As a result, EBNA1 can recruit EBV episomes onto host mitotic chromosomes. Curiously, although EBNA1 is indispensable for oriP plasmid replication, EBNA1 has no reported catalytic functions that are required for DNA replication (Frappier and O'Donnell, 1991). Human replication complexes, such as origin recognition complex (ORC) and mini-chromosome maintenance (MCM) proteins, have been shown to reside in the oriP region of EBV plasmids (Chaudhuri et al., 2001; Dhar et al., 2001; Schepers et al., 2001). Thus, it appears that EBV latent replication is mediated entirely by the cellular replication machinery (reviewed in Wang and Sugden, 2005), which also explains why EBV episomes replicate only once per host-cell cycle (Yates and Guan, 1991).
The once-per-cell-cycle mode of replication of EBV latent infection implies that the copy number of EBV episomes doubles only once per replication cycle. In order to keep the copy number of EBV episomes stable in dividing cells, duplicated EBV episomes need to be partitioned as equally as possible between two daughter cells. In other words, it is advantageous for EBV to have a mechanism for evenly distributing its duplicated genomes to daughter cells. There are several lines of evidence that non-random segregation of EBV episomes occurs in dividing cells. FISH analysis revealed that significant numbers of EBV episomes are present as symmetrical doublet signals on sister chromatids (Delecluse et al., 1993). Also, an immunofluorescence analysis revealed that the signals of EBNA1 are symmetrically localized on sister chromatids in 293 cells (Sears et al., 2004). These cytological data fit well with the hypothesis that viral EBNA1 protein contributes to non-stochastic segregation of EBV episomes (Kanda et al., 2001a; Leight and Sugden, 2000; Mackey and Sugden, 1999; Sears et al., 2004). However, simultaneous visualization of EBNA1 and EBV genomes has not been reported; thus, evidence that EBNA1 directly mediates non-random segregation of EBV episomes is lacking.
Thus, we set out to examine how EBNA1 mediates the segregation of EBV episomes in latently infected cells. Because antibodies raised against EBNA1 protein lacked sufficient sensitivity and specificity to determine the precise intranuclear localization of EBNA1, we generated a recombinant EBV expressing epitope-tagged EBNA1. This method facilitated the detection of EBNA1, which is expressed at low levels in latently infected cells. Furthermore, it enabled us to visualize EBNA1 and EBV genomes simultaneously with immunofluorescence and FISH, respectively. In the current study, we present evidence for EBNA1-mediated, non-random segregation of EBV episomes in latently infected cells.
Generation of cells infected with recombinant EBV expressing epitope-tagged EBNA1
We generated a recombinant EBV to facilitate the detection of EBNA1 in latently infected cells. A BAC clone of the genome of Akata strain EBV, carrying the genes for green fluorescent protein (GFP) and hygromycin resistance, was used as the starting material. A recombinant EBV genome that encoded hemagglutinin (HA)-tagged EBNA1 was constructed by means of the BAC system (Ahsan et al., 2005; Kanda et al., 2004) (see supplementary material Fig. S1A). The resultant BAC clone was first introduced into EBV-positive Akata cells through transfection and then transferred to EBV-negative Akata cells through infection, according to a previously established protocol (Ahsan et al., 2005; Kanda et al., 2004). We obtained many hygromycin-resistant cell clones that were strongly positive for GFP. Western blot analysis, using either EBV-immune human serum or anti-HA antibody, revealed that the established cell clones expressed HA-tagged EBNA1, but not wild-type EBNA1 (Fig. 1A).
Episomal maintenance of the recombinant virus expressing HA-tagged EBNA1
We used several strategies to determine whether the recombinant EBV genomes encoding HA-tagged EBNA1 (referred to as recombinant EBV hereafter) were maintained as episomes in the established cells. When cellular fractions enriched with extrachromosomal DNA molecules were used to transform Escherichia coli, we recovered intact BAC clones, indicating the presence of episomal molecules in the established cells (see supplementary material Fig. S2A). We performed FISH analyses to directly visualize recombinant EBV genomes at the single cell level. We observed a heterogeneous number of FISH signals in cells harboring the recombinant EBV, resembling the appearance of Akata cells harboring naturally infected EBV (referred to as wild-type EBV hereafter) (Fig. 1B). Such heterogeneity in the number of FISH signals is typical of episomally maintained EBV genomes; the appearance of integrated copies of the EBV genome is more consistent and signals are fewer in number in every cell (Hurley et al., 1991b) (see supplementary material Fig. S2B for a typical pattern of FISH signals of integrated EBV genomes). When we treated the cells harboring either wild-type or recombinant EBV genomes with hydroxyurea, which accelerates the loss of EBV episomes from latently infected cells (Chodosh et al., 1998), we observed substantial loss of EBV genomes. Many cells became free of FISH signals after 5 weeks of hydroxyurea treatment (Fig. 1B), arguing against the covalent integration of EBV genomes into chromosomes. Substantial loss of EBV genomes in hydroxyurea-treated cells was also confirmed by Southern blot analysis (Fig. 1C). Of note, in hydroxyurea-treated cells harboring the recombinant EBV, the remaining FISH signals exhibited a pattern that was consistent with episomal maintenance of EBV genomes (see supplementary material Fig. S2C). Therefore, although we cannot completely rule out the possibility that a few integrated copies of the EBV genomes coexist with extrachromosomal ones in the established cells, as has been reported earlier (Delecluse et al., 1993), these results indicated that most, if not all, recombinant EBV genomes were extrachromosomal.
Punctate localization of EBNA1 coincides with the FISH signals of EBV genomes
We next examined the localization of HA-tagged EBNA1 by immunofluorescence analysis using an anti-HA antibody. We observed discrete punctate signals in the nuclei of cells in all phases of the cell cycle (Fig. 2A). The number of punctate signals varied from cell to cell. Furthermore, the fluorescent intensities of the dots were heterogeneous. Such heterogeneity is most likely because of heterogeneity in the numbers of EBNA1 molecules (and EBV genomes) constituting the dots. Comparison of anti-HA antibody staining and anti-EBNA1 polyclonal antibody staining revealed that the signals obtained by anti-HA antibody were more distinct (see supplementary material Fig. S3).
Because there are multiple EBNA1-binding sites clustered within the oriP region of the EBV genome (Rawlins et al., 1985; Reisman et al., 1985), we hypothesized that this staining pattern reflected the accumulation of HA-tagged EBNA1 molecules at the sites of EBV genomes. To test this idea, we established a method that allowed us to detect the localization of EBV genomes (through FISH) and HA-tagged EBNA1 protein (through immunofluorescence) simultaneously. We found that essentially all of the anti-HA staining signals overlapped with FISH signals (Fig. 2B, top panels). By contrast, when cells harboring wild-type EBV genomes were subjected to the same type of analysis, we observed FISH signals but no anti-HA staining (Fig. 2B, bottom panels), indicating that the anti-HA staining was specific.
At the same time, we noticed that anti-HA staining and FISH signals did not fully overlap with each other in merged images. Rather, they partially overlapped with each other. This is reasonable because the FISH probe encompasses the entire EBV genome, whereas EBNA1 specifically binds to the oriP region. The signal intensities of EBNA1 dots were as strong as the FISH signals of EBV genomes. This is most likely because of multiple EBNA1 molecules binding to multiple EBNA1-binding sites within the oriP region of the EBV genomes, and the presence of three copies of HA tags on each molecule of EBNA1 (see supplementary material Fig. S1A).
These results clearly indicated that most, if not all, EBNA1 molecules accumulate at EBV genome loci in latently infected cells.
Symmetrical distribution of EBNA1 molecules on sister chromatids of mitotic chromosomes
We examined the localization of EBNA1 dots in mitotic cells by immunofluorescence analysis. EBNA1 dots were found to associate with prometaphase chromosomes and segregating anaphase chromosomes (Fig. 3A), corresponding well with the fact that EBV episomes piggyback onto mitotic chromosomes for efficient segregation in dividing cells.
We next examined whether EBV genomes target specific chromosomal loci by examining the distribution of EBNA1 dots on mitotic chromosome spreads. An antibody recognizing CENP-C, a centromeric protein, was used for dual-color immunofluorescence analysis together with the anti-HA antibody. CENP-C is known to localize as double dots to the inner kinetochore plates of mitotic chromosomes (Warburton et al., 1997). Thus, CENP-C can serve as a marker for the identification of sister chromatids of mitotic chromosomes.
There was no apparent colocalization of EBNA1 and CENP-C, indicating that EBV genomes do not target centromeric regions of mitotic chromosomes (Fig. 3B). Notably, dual-color immunofluorescence analysis clearly indicated that significant numbers of EBNA1 dots localized symmetrically on sister chromatids, similar to CENP-C dots (Fig. 3B). We also used FISH and immunofluorescence to detect the locations of HA-tagged EBNA1 and EBV genomes on mitotic chromosomes. As expected, HA-tagged EBNA1 and EBV genomes colocalized each other on mitotic chromosomes (Fig. 3C), just as in interphase nuclei. Again, significant numbers of symmetrical, overlapping signals of EBNA1 and EBV genomes were observed on mitotic chromosomes (Fig. 3C). This result was reminiscent of a previous report that episomal EBV molecules could localize symmetrically on sister chromatids (Delecluse et al., 1993).
To exclude the possibility that the symmetrically localized EBV genomes we observed were peculiar to the cells harboring the recombinant EBV, Akata cells with wild-type (i.e. naturally infected) EBV genomes were subjected to conventional FISH analysis. EBV genomes in naturally infected Akata cells are known to be strictly episomal, and clonal cell lines that are free of EBV infection can be derived from Akata cells through extended in vitro cultivation (Shimizu et al., 1994). Two representative metaphase spreads with FISH signals of EBV genomes are shown in Fig. 3D. In both metaphase spreads, we found that some of the FISH signals localized symmetrically on sister chromatids (indicated by arrowheads). The symmetrically localized FISH signals constituted 15% (16/106) of total FISH signals in the left panel and 19% (16/84) in the right panel (Fig. 3D). These results are compatible with the observation of symmetrically localized EBNA1 dots on sister chromatids in the recombinant EBV-infected cells (Fig. 3B,C). Thus, symmetrical localization of extrachromosomally maintained EBV genomes on sister chromatids are not peculiar to the cells harboring the recombinant EBV, and the symmetrically localized EBV genomes constitute up to 20% of EBV genomes on mitotic chromosome spreads.
Symmetrically localized EBNA1 dots on sister chromatids of prematurely condensed G2 chromosomes
We hypothesized that segregation of EBV episomes to sister chromatids takes place while EBV episomes are replicating synchronously with cellular chromatin, and that newly replicated sister EBV episomes should be readily loaded onto newly replicated sister chromatids. If this were the case, symmetrical localization of EBNA1 dots would be most prominent shortly after replication. Therefore, we examined the localization of EBNA1 dots in G2-phase cells.
To enrich G2-phase cells, cells harboring the recombinant EBV were synchronized at the G1-S boundary and then released into S-phase. Fluorescence-activated cell sorting (FACS) analyses revealed that, although the synchronization at the G1-S boundary was imperfect, late S- to early G2-phase cells were significantly enriched 6 hours after release (Fig. 4A). Microscopic observation revealed that most of the cells were in interphase, and that mitotic cells were barely detectable.
Synchronized cells were hypotonically swollen and subjected to immunofluorescence analysis with the anti-HA antibody. We readily noticed that, in interphase nuclei of G2-enriched cells, EBNA1 dots were frequently observed as closely spaced doublets (Fig. 4B). The majority of nuclei were found to contain at least a pair of doublet signals of EBNA1. This observation was reminiscent of the well-known FISH result that replicated genomic (chromosomal) loci appear as doublets in interphase nuclei (Selig et al., 1992).
To determine whether EBNA1 doublets represented replicated EBV genomes, we examined the distribution of EBNA1 doublet signals on G2 chromosomes by using a technique to induce premature chromosome condensation (PCC), which was previously used to analyze the distribution of EBNA1 in living cells (Ito et al., 2002). Calyculin A, a phosphatase inhibitor, was added to G2-phase-enriched cells 1 hour prior to harvesting to induce PCC. With calyculin A treatment, we saw many cells with condensed chromosomes, indicating that calyculin A efficiently induced PCC, at a frequency of up to 30%. The morphological characteristics of PCC allowed us to determine the cell-cycle phase of individual PCC cells (Bezrookove et al., 2003). We focused on prematurely condensed chromosomes in G2-phase cells (G2-PCCs), which were bivalent with their sister chromatids closely attached to each other. Morphologically well-preserved G2-PCCs with clear EBNA1 signals were chosen for the analysis. Out of more than 100 acquired images of G2-PCCs, 40 of them obtained by two independent experiments were examined in detail. We found that all the examined G2-PCCs harbored at least a pair of EBNA1 doublets that were symmetrical on sister chromatids (Fig. 4C and data not shown). Importantly, although the EBNA1 signal intensities were heterogeneous, the signal intensities of paired EBNA1 dots on sister chromatids were similar, supporting the idea that they were indeed sister molecules.
To quantitate the frequency of such symmetrical EBNA1 doublets in individual cells, we again used CENP-C staining to visualize sister chromatid alignment and counted the number of paired EBNA1 dots that were at right-angles to the longitudinal axes of the chromosomal arms. A typical example of a G2-PCC with anti-HA and CENP-C staining is shown in Fig. 4D. In this example, we identified 22 symmetrically localized EBNA1 dots that were accompanied by symmetrically localized CENP-C dots on sister chromatids. The symmetrically localized EBNA1 double dots constituted 39% (22/56) of total EBNA1 dots. Furthermore, we observed eight additional EBNA1 double dots that were not accompanied by CENP-C double dots. In total, 53% (30/56) of the EBNA1 dots existed as closely spaced double dots in this G2-PCC.
Symmetrically localized EBNA1 double dots were observed in three other G2-PCCs at frequencies of 23% (10/43), 38% (10/26) and 34% (14/41), respectively (see supplementary material Fig. S4). In these G2-PCCs, closely spaced EBNA1 double dots constituted 51% (22/43), 61% (10/26) and 49% (20/41), respectively, of the total EBNA1 dots. Our attempt to determine the average frequency of symmetrical distribution of EBNA1 dots on sister chromatids was hampered by the enormous heterogeneity in the number of EBNA1 dots. Nevertheless, based on these four examples, we conclude that the distribution of EBNA1 on G2-PCCs is not random.
These results suggest that EBNA1 doublets in the nuclei of G2-phase-enriched cells most likely represent replicated sister episomes, and thus correspond well with the doublet signals on G2-PCCs.
Simultaneous visualization of EBNA1 and EBV episomes in G2-phase-enriched cells
We took advantage of our ability to perform simultaneous FISH and immunofluorescence analyses, and focused our attention on the closely spaced EBNA1 double dots observed in G2-phase-enriched cells. Based on the results of FACS analysis (Fig. 4A), most of the cells were in either late S-phase or G2-phase.
As expected, EBNA1 double dots, which were located approximately 2 μm apart, frequently overlapped with the double dots of EBV genomes (Fig. 5A, sub-panel 6). Notably, we observed significant numbers of EBNA1 double dots interconnected by EBV genome signals, such that the structures exhibited a dumbbell-like appearance (Fig. 5A, sub-panels 1, 2, 4, 5). Because not all of the cells on slides were consistently double-stained by FISH and immunofluorescence, the frequency of cells harboring such dumbbell-like structures could not be determined. Still, we repeatedly observed such dumbbell-like structures in three independent experiments. In some cases, FISH signals of the EBV genome overlapped the EBNA1 double dots asymmetrically (Fig. 5A, sub-panel 3). Importantly, we rarely observed double dots of EBV genomes interconnected by a single EBNA1 signal.
We also examined whether such dumbbell-like structures were also present in G2-PCCs. In many cases, EBNA1 and EBV genome signals overlapped as symmetrically localized dots on sister chromatids (Fig. 5B, left). We also observed symmetrically localized EBNA1 double dots interconnected by the FISH signals of EBV genomes (Fig. 5B, right). These structures appeared similar to the dumbbell-like structures observed in interphase nuclei (likely to be S- or G2-nuclei).
The result is compatible with the idea that the partitioning of EBNA1 onto sister chromatids precedes the segregation of EBV episomes.
Dumbbell-like structures in G2-arrested cells treated with a topoisomerase II inhibitor
During replication of circular duplex DNA molecules, interlocked rings, often referred to as catenated DNA molecules, are formed and are subsequently unlinked by a topoisomerase. Previously, electron-microscopic analysis demonstrated the presence of catenated dimers of EBV genomes in S-phase-enriched fractions of EBV-positive Raji cells (Gussander and Adams, 1984). To determine whether the dumbbell-like structures observed in this study consist of partitioned EBNA1 molecules interconnected by a catenated dimer of EBV episomes, we used the topoisomerase II inhibitor ICRF-193, which inhibits the decatenation of replicated DNA molecules.
Cells harboring recombinant EBV were released from the G1-S boundary in the absence or presence of ICRF-193, and cell cycle progression was monitored by FACS. In the absence of ICRF-193 treatment, cells progressed through G2/M phase and entered G1-phase 24 hours after release (Fig. 6A). By contrast, in the presence of ICRF-193, cells did not enter G1, even at 24 hours after release. Microscopic analysis revealed that the mitotic index of ICRF-193-teated cells 24 hours after release was nearly zero (data not shown), indicating that the majority of cells arrested in G2. This was probably because of the activation of a DNA decatenation checkpoint by ICRF-193 (Deming et al., 2001).
In ICRF-193-treated cells arrested in G2, in three independent experiments, we repeatedly found cells with multiple dumbbell-like structures (Fig. 6B). These prominent dumbbell-like structures were barely observed in control cells that were not treated with ICRF-193. Although we were unable to accurately determine the frequency of these dumbbell-like structures in the presence or absence of ICRF-193 treatment because of technical difficulties, the dumbbell-like structures in topoisomerase inhibitor-treated cells most likely represent replicated EBV episomes that failed to undergo decatenation.
These results suggest that the dumbbell-like structures, which are frequently observed in late S- to early G2-phase cells (Fig. 5), consist of partitioned EBNA1 molecules interconnected by a catenated dimer of EBV episomes.
Previous studies have shown that autonomously replicating episomes, such as latent viruses (reviewed in Botchan, 2004; Calos, 1998) and cellular acentric chromosomes (Kanda et al., 2001b) (reviewed in Kanda and Wahl, 2000), attach to segregating chromosomes for efficient segregation in dividing cells. However, little is known on whether episomes are randomly partitioned between daughter cells or evenly partitioned in a controlled fashion.
To address this issue, we established cells that were infected with a recombinant EBV expressing HA-tagged EBNA1. Immunofluorescence analysis with an anti-HA antibody revealed that HA-tagged EBNA1 localized to discrete, intranuclear punctate dots, which colocalized with FISH signals of EBV genomes. This punctate localization pattern of EBNA1 corresponded well with the localization of EBNA1 observed in latently infected B95-8 cells (Daikoku et al., 2004). Others have observed a diffuse pattern of EBNA1 localization (in interphase nuclei or on mitotic chromosomes) in other latently infected cells (Petti et al., 1990), and in cells overexpressing EBNA1 (Hung et al., 2001; Ito et al., 2002; Kanda et al., 2001a; Marechal et al., 1999). We argue that the punctuate staining pattern of EBNA1 can only be observed in latently infected cells in which EBNA1 is expressed at low levels. When expressed at low levels, the majority of EBNA1 molecules should form clusters at the oriP region of the EBV genomes, and little EBNA1 would be available to diffusely localize throughout interphase nuclei or on mitotic chromosomes.
Our data indicate that a significant number of EBNA1 dots in latently infected cells localize symmetrically on mitotic chromosomes (Fig. 3) and on calyculin A-induced G2-PCCs (Fig. 4 and see supplementary material Fig. S4). It has generally been assumed that symmetrical FISH signals on sister chromatids represent chromosomally integrated copies of EBV genomes (Hurley et al., 1991a; Reisinger et al., 2006; Trescol-Biemont et al., 1987). However, integration of the EBV genomes usually occurs at one or a few chromosomal loci (Hurley et al., 1991b) (see also supplementary material Fig. S2B); thus, the symmetrical EBNA1 doublets on G2-PCCs, which constitute up to 40% of the total dots in the cell, most likely do not represent integrated copies of the EBV genome. Although calyculin A treatment can produce artifacts, the relative positions of EBV episomes on chromatin should not be affected. A good correlation between the EBNA1 doublets on G2-PCCs and in the nuclei of G2-phase-enriched cells (Fig. 4B) suggests that the doublets are not an artifact caused by calyculin A treatment.
We observed dumbbell-like structures in G2-phase-enriched cells, in which paired EBNA1 dots were connected by the FISH signal of the EBV genome (Fig. 5A). Corresponding to this observation, we observed symmetrically localized EBNA1 doublets that were occasionally interconnected by FISH signals of EBV genomes in prematurely condensed G2 chromosomes (Fig. 5B). Such dumbbell-like structures are most likely resolved to individual dots while cells progress through S- and G2-phases.
We argue that paired EBNA1 dots on sister chromatids are not merely a consequence of the physical tethering of the EBNA1 molecules through the EBV DNA. Rather, the overall data favors an EBNA1-mediated segregation mechanism (Fig. 7). Large clusters of EBNA1 accumulate on multiple EBNA1-binding sequences clustered in the oriP region of the EBV genomes. The clusters of EBNA1 serve as a molecular scaffold to attach EBV episomes to cellular chromatin. When cellular replication takes place, approximately equal amounts of EBNA1 are distributed onto sister chromatids. At the same time, clustered EBNA1 molecules mediate the loading of the cellular replication machinery onto the EBV origin of replication, resulting in the replication of the EBV episome. Replicated dimers of EBV genomes are catenated, as revealed by electron-microscopic observation (Gussander and Adams, 1984), and segregation is not complete until they become decatenated by topoisomerase II. This explains the dumbbell-like structures we frequently observed using combined FISH and immunofluorescence analyses (Figs 5, 6). In the dumbbell-like structures, the distance between the doublets of EBNA1 (2 μm) was far shorter than the expected length of the 170-kb EBV genome when it is stretched (approximately 100 μm) (Reisinger et al., 2006), suggesting that catenated dimers of EBV genomes have nucleosomal and higher-order structures.
Although we observed evidence of missegregation (Fig. 5A, sub-panel 3), we postulate that accurate and equal partitioning of chromatin-associated EBNA1 molecules to sister chromatids occurs in S-phase cells. The percentage of symmetrically localized EBNA1 dots on G2-PCCs was less than 50%, and in mitotic cells the percentage appeared to be even less. This discrepancy could be because of dynamic movement of segregated EBV episomes in S- and G2-phase cells. This movement could be along chromatin, or from one chromatin fiber to another (Fig. 7, bottom). In this way, asymmetrically localized EBV episomes in mitotic cells could also be segregated symmetrically during S phase. It remains to be clarified whether the proposed mechanism contributes to faithful segregation of EBV episomes.
In summary, we present further evidence in support of non-random segregation of extrachromosomally maintained viral episomes. At least two other viruses, Kaposi's sarcoma-associated herpesvirus (KSHV, human herpesvirus 8) and bovine papillomavirus-1, are known to attach to host mitotic chromosomes (reviewed by Botchan, 2004). Among these, KSHV encodes latency-associated nuclear antigen (LANA), which is a functional counterpart of EBNA1 in EBV (Ballestas et al., 1999; Barbera et al., 2006; Shinohara et al., 2002). Previous studies demonstrated that the LANA colocalizes with KSHV genomes on mitotic chromosomes (Ballestas et al., 1999) and occasionally exhibits symmetrical localization on sister chromatids of mitotic chromosomes (Szekely et al., 1999). Therefore, the ability to distribute replicated sister episomes onto sister chromatids appears to be preserved in KSHV and perhaps in other gamma herpesviruses as well. The molecular mechanisms by which EBNA1 and LANA associate with cellular chromatin are still under debate (Barbera et al., 2006; Kapoor et al., 2005; Sears et al., 2004; Shire et al., 1999), and remain to be clarified.
The ability of viral episomes to segregate evenly onto sister chromatids is a fascinating tactic for virus survival in dividing cells. It also raises the intriguing possibility that the stability of episomal vectors could be improved with strategies to ensure the faithfulness of their segregation onto sister chromatids.
Materials and Methods
Akata, a human Burkitt's lymphoma-derived cell line carrying EBV episomes, and an EBV-negative Akata cell line (Shimizu et al., 1994) were grown in RPMI 1640 medium (Sigma, St Louis, MO) supplemented with 10% fetal bovine serum (Gibco, Rockville, MD).
The hygromycin phosphotransferase gene and pBS246-mloxP-zeo (Ahsan et al., 2005) was tandemly connected to make pHyg-mloxP-zeo. The kanamycin resistance marker gene of pUC4K (Pharmacia) was subcloned in the double I-PpoI vector (Kanda et al., 2004) to make pIPpoI-km. The gene encoding EBNA1 protein with three copies of HA tags (aa sequence: YPYDVPDYA-G-YPYDVPDYA-GS-YPYDVPDYA) attached to its C-terminus was generated by standard PCR-based strategy. The resultant EBNA1-HA3 gene was subcloned in the double I-PpoI vector to make pIPpoI-EBNA1-HA3.
Construction of recombinant EBV genome in E. coli
The PCR product of pHyg-mloxP-zeo was used to replace the neomycin-kanamycin resistance gene of AK-BAC-GFP (Kanda et al., 2004) with the hygromycin phosphotransferase gene through GET recombination. The zeocin marker gene was then excised by in vitro Cre recombinase treatment (Ahsan et al., 2005; Kanda et al., 2004) to generate Hyg-GFP-BAC (HGB). A second GET recombination was performed to replace the coding sequence of the EBNA1 gene with the kanamycin resistance gene, which was flanked by two I-PpoI sites, by using a PCR product of pIPpoI-km. Phusion high-fidelity DNA polymerase (Finnzymes) was used in the PCR reaction (98°C for 30 seconds; 25 cycles of 98°C for 20 seconds, 55°C for 30 seconds, 72°C for 60 seconds; 72°C for 10 minutes). The resultant BAC clone, HGB-ΔEBNA1-km, was digested by I-PpoI and self-ligated to make HGB-ΔEBNA1-Δkm. The I-PpoI fragment of pIPpoI-EBNA1-HA3 was subcloned into the unique I-PpoI site of HGB-ΔEBNA1-Δkm to generate HGB-EBNA1-HA3 according to the previously described in vitro ligation protocol (Kanda et al., 2004). Electrotransformation, BAC plasmid (BACmid) preparation and DNA analysis were performed as described previously (Kanda et al., 2004). The details of BACmid construction and the PCR primer sequences for the EBNA1 replacement are provided in the supplementary material Fig. S1.
Transfection of BACmids into Akata cells
EBV-positive Akata cells (5×106) were transfected with 5 μg of BACmid DNA (prepared using Nucleobond BAC100 kit) through electroporation (BioRad Gene Pulser II, 190 V, 950 μF). Transfected cells were resuspended in 5 ml of culture medium and plated into 6-well dishes. At 2 days posttransfection, cells were plated at 104 cells per well in 96-well tissue culture plates in medium containing 150 μg/ml of hygromycin (Calbiochem). Half of the culture medium was replaced with fresh hygromycin-containing medium every 5 days. Hygromycin concentration was decreased to 100 μg/ml (after day 10) in order to propagate the drug-resistant clones. Hygromycin-resistant clones were screened for the presence of episomally maintained BACmids as described previously (Kanda et al., 2004). Cells harboring intact episomes of BACmids were obtained, and the cells were used for virus production.
Virus production and infection
For virus production, dialyzed rabbit anti-human IgG (Dako, Glostrup, Denmark) was added to a cell suspension (2×106 cells/ml) to give a final concentration of 0.5% (v/v), and the cells were incubated at 37°C for 6 hours. The cells were then resuspended in a fresh medium (106 cells/ml) and cultivated for 2 days. The culture supernatant was harvested, filtered through a 0.45-μm filter and used as a virus solution.
For virus infection, EBV-negative Akata cells were resuspended in the diluted (1:5 and 1:10) virus solution (106 cells/ml) and incubated at 37°C for 90 minutes on a rotator. The cells were then washed once with a fresh medium, resuspended in a fresh medium (106 cells/ml) and cultivated for 2 days. Infected cells were plated at 104 cells per well in 96-well tissue culture plates in a medium containing hygromycin (150 μg/ml; Calbiochem), and hygromycin-resistant cells were obtained as described above.
For eliminating EBV episomes from cells, hydroxyurea (1 M stock solution in PRMI medium; Sigma) was added to culture medium at the final concentration of 50 μM. Cells were plated at a density of 5×105 cells/ml and passaged in the hydroxyurea-containing medium for 5 weeks.
Cell cycle synchronization
Cell synchronization was performed as described previously (Shimizu et al., 1998) with minor modifications. Rapidly growing Akata cells harboring the recombinant EBV were first arrested by adding 2 mM thymidine (100 mM stock solution in RPMI medium; Sigma) for 16.5 hours. The cells were then washed with medium, released into fresh medium (no 2′-deoxycytidine added) for 9.5 hours, and then incubated with 2.5 μg/ml aphidicolin (2.5 mg/ml stock solution in DMSO; Sigma) for 17 hours to arrest cells as they entered S phase.
The arrested cells were washed with medium and then released into fresh medium. Cell cycle progression was monitored by FACS. When indicated, at 2 hours after release, the topoisomerase II inhibitor ICRF-193 (10 mM stock solution in DMSO; BIOMOL Research Laboratories) was added to medium at the final concentration of 2 μM. For FACS analyses, samples of cells (5×105 cells) were harvested, resuspended in 600 μl of PBS, and then 1.4 ml of ethanol was added, and the fixed cells were kept on ice for one overnight period. The fixed cells were stained with PBS containing propidium iodide (20 μg/ml) and RNase (200 μg/ml) and analyzed by FACSCalibur (Becton Dickinson). Cells were processed for immunofluorescence and FISH analyses only when good cell synchronization was verified, and all the experiments were repeated more than three times.
Preparation of fixed cells
To obtain the data shown in Fig. 2A and Fig. 3A, cells were smeared on glass slides, dried briefly and fixed with 4% paraformaldehyde for 10 minutes. For immunofluorescence analyses of chromosome spreads (Fig. 3B; Fig. 4) and for combined FISH and immunofluorescence analyses (Fig. 2B; Figs 5, 6), cells were hypotonically swollen in a hypotonic buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 5 mM MgCl2) at a cell density of approximately 1×107 cells/ml. After 5 minutes of incubation at 37°C, hypotonically swollen cells were smeared onto glass slides [10-well slides (Cel-line/Erie Scientific) for immunofluorescence, or slides that were marked with Dako Pen (Dako) for FISH]. Immediately after residual buffer evaporated, cells were fixed in 4% paraformaldehyde at room temperature for 10 minutes and then fixed in 70% ethanol on ice for 10 minutes.
For inducing PCC, calyculin A (40 μM stock solution in ethanol; Sigma) was added to the final concentration of 80 nM at 1 hour prior to harvesting cells. G2-PCCs were identified according to the criteria that were described previously (Bezrookove et al., 2003).
Fixed cells were treated with blocking buffer [2.5% bovine serum albumin (BSA), 0.2 M glycine, 0.1% Triton X-100] at 37°C for 30 minutes. A primary antibody used for detecting HA-tagged EBNA1 protein was anti-HA monoclonal antibody (mAb) 6E2 (1:100 dilution in blocking buffer; Cell Signaling). Following overnight incubation at 4°C, slides were washed with PBS, and then incubated with Cy3-conjugated donkey anti-mouse IgG (1:200 in PBS; Jackson ImmunoResearch) at 37°C for 30 minutes. Following incubation, slides were washed with PBS, and chromosomes were counterstained with 4′,6′-diamidino-2-phenylindole (DAPI) (8 μg/ml) in Vectashield (Vector Laboratories).
For dual-color immunofluorescence to detect centromeres together with EBNA1, anti-CENP-C polyclonal antibody (a gift from D. Cleveland, University of California, San Diego, CA; 1:500 dilution in blocking buffer) was used as a primary antibody and Cy5-conjugated donkey anti-rabbit IgG (1:200 in PBS; Jackson ImmunoResearch) was used as a secondary antibody.
To make a FISH probe, 1 μg of BACmid DNA (HGB-EBNA1-HA3) was labeled with digoxigenin using DIG-Nick Translation Mix (Roche) according to the manufacturer's instructions. Labeled probe was mixed with 24 μg of sheared salmon sperm DNA, ethanol precipitated and redissolved in 60 μl of hybridization solution (50% formamide, 10% dextran sulfate, 2×SSC). Standard FISH protocol was used to visualize EBV episomes in colcemid (100 μg/ml)-induced metaphase chromosome spreads prepared by conventional methanol-acetic acid fixation.
For simultaneous FISH and immunofluorescence, hypotonically swollen cells were fixed as described above, permeabilized with blocking buffer containing RNase (100 μg/ml) for 1 hour, and then incubated in 50% formamide 2×SSC for 30 minutes at room temperature for equilibration. Each spot of cells was overlaid with 5 μl of hybridization mixture and covered by a coverslip. Slides were then sealed in a hybridization bag while aspirating air by an air pump. Sealed slides were then directly immersed in water (85°C) for 5 minutes for denaturation, followed by hybridization at 37°C overnight.
Slides were then washed three times in 50% formamide 2×SSC at 45°C for 3 minutes each, three times in 2×SSC at 45°C for 3 minutes each, and once in 0.1×SSC at 60°C for 10 minutes. Subsequently, slides were incubated in 4×SSC containing 1% BSA and 5% nonfat dry milk at 37°C for 20 minutes, and then in 4×SSC containing 1% BSA and anti-digoxigenin fluorescein (Roche; 1:100 dilution) at 37°C for 30 minutes.
After antibody reaction, slides were washed in 4×SSC, 4×SSC with 0.1% Triton X-100, 4×SSC, and PN buffer [0.1 M phosphate buffer (pH 8.0), 0.5% NP-40] at room temperature for 3 minutes each. Slides were then processed for immunofluorescence using anti-HA antibody as described above. Anti-digoxigenin fluorescein (1:200 dilution) was included in a secondary antibody reaction.
Images were acquired by using the 60× objective of a laser-scanning microscopy system (Fluoview; Olympus). In cases when signals were scattered along the z-axis, multiple Z-series images (0.4 to 0.7 μm intervals) were acquired and extended projection-view images were generated. The acquired digital images were expressed as pseudocolor and then merged using Adobe Photoshop (Adobe Systems, Mountain View, CA).
Whole cell extracts were prepared by lysing cells with SDS lysis buffer (3% SDS, 10% glycerol, 125 mM Tris-Cl pH 6.7, 6% urea) at the cell concentration of 1×107 cells/ml. Whole cell extracts were analyzed by SDS-7% polyacrylamide gel electrophoresis. Immunoblot was performed according to standard protocols. EBV-immune human serum reactive to EBNA1 (1:500 dilution) or anti-HA mAb 6E2 (1:1000 dilution) were used as primary antibodies, and horse radish peroxidase-conjugated anti-human IgG or anti-mouse IgG (1:2000 dilution; Amersham) were used as secondary antibodies.
We thank P. Ioannou for the pGETrec plasmid, T. Daikoku and T. Tsurumi for the anti-EBNA1 antibody, D. Cleveland for the anti-CENP-C antibody, and M. Sato for technical assistance. This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan (T.K. and K.T.).