An Epigenetic Journey: Epstein-Barr Virus Transcribes Chromatinized and Subsequently Unchromatinized Templates during Its Lytic Cycle

True late genes are expressed after lytic replication of viral DNA.

Like those of all herpesviruses, EBV’s life cycle includes a latent and a lytic phase. A ubiquitous human pathogen, EBV infects up to 95% of the world’s adult population, in whom the virus persists for life in memory B cells (12, 13). Infection with EBV is usually asymptomatic, although primary infection during adolescence and early adulthood can cause infectious mononucleosis (14, 15). EBV is also associated with several human malignancies, including lymphomas and carcinomas (16). In both memory B cells and tumor cells, EBV exists mostly in a latent state, and viral DNAs exist as circularized, extrachromosomal plasmids. During latency, only a few viral genes are expressed, and infectious virions are not produced, limiting viral exposure to the host immune response. In vitro infection of primary B cells with EBV transforms them into actively proliferating lymphoblastoid cells, in which the virus is also mostly latent.

Occasionally, EBV in latently infected cells shifts to a lytic state, during which a temporally ordered cascade of viral gene expression converts host cells into factories amplifying viral DNA and producing infectious virions (17). It is increasingly appreciated that a fraction of cells in EBV-positive tumors enter the lytic phase, thus playing a contributing role in human malignancies (18, 19).

In vivo, differentiation of EBV-positive cells can trigger the viral lytic cycle (20, 21), and several different agents can induce EBV’s lytic cycle in vitro. During the switch from latency to the lytic phase, cellular transcription factors activate EBV’s immediate early (IE) genes, BZLF1 and BRLF1. These genes encode transcription factors—called Z, Zta, or ZEBRA and R or Rta, respectively—whose activities are key for the latent-lytic switch and have been reviewed extensively (6, 17, 22, 23). BZLF1 and BRLF1 upregulate their own and each other’s expression and initiate the transcription of EBV’s early genes, many of which encode proteins involved in viral lytic DNA replication.

Viral DNA replication is followed by expression of EBV’s late genes, which encode primarily structural proteins (24, 25). The first hints that the expression of a subset of herpesvirus genes—including those of EBV—is inextricably linked to viral DNA replication were discovered decades ago (26–29). More recently, Yuan et al. (30) used phosphonoacetic acid (PAA), an inhibitor of EBV lytic DNA synthesis, to determine which EBV genes depend on viral DNA replication for their expression. Although this study identified several putative late genes, there were two complications in interpreting the data. First, the use of PAA to inhibit viral DNA synthesis seemed to have some level of general cytotoxicity and also led to the downregulation of several well-characterized early genes. Second, multiple EBV transcripts overlap extensively, which makes it challenging to unambiguously quantify their levels with oligonucleotide arrays, as employed in this study, or even with transcriptome sequencing (RNA-seq) or reverse transcription-quantitative PCR (qRT-PCR). For example, members of the BFRF family of genes—BFRF0.5, BFRF1, BFRF2, and BFRF3—share the same poly(A) signal. BFRF3 is considered a true late gene, but Djavadian et al. (24) could detect putative BFRF3 mRNA in EBV-positive cells induced to enter the lytic cycle even in the absence of viral DNA amplification. However, they could detect BFRF3 protein only when viral DNA amplification occurred. Assays using traditional RNA-seq or qRT-PCR would have resulted in the erroneous classification of BFRF3 as an early gene.

To circumvent the aforementioned issues with quantifying overlapping viral transcripts, Djavadian et al. (24) used nonamplified, nontagging Illumina cap analysis of gene expression (31), or CAGE-Seq, to identify and quantify transcription start sites of EBV genes. CAGE-Seq involves capturing the 5′ cap of mature mRNAs and allows precise mapping of transcription start sites. Djavadian et al. also generated mutant versions of EBV that lack BALF2, the viral single-stranded DNA-binding protein, which is required for lytic DNA replication (32). The use of the BALF2-knockout EBV, and trans-complementation with BALF2 when desired, allowed the study of late gene expression in the absence or presence, respectively, of viral DNA replication without the need for chemical inhibitors, such as PAA. Two distinct classes of EBV late genes were identified on the basis of their dependence on viral DNA replication: true late genes—hereafter referred to below as late genes—which are expressed only after, and if, viral DNAs are replicated, and “leaky-late” genes, which can be transcribed by both replication-dependent and replication-independent mechanisms (Fig. 1). Analysis of the transcription start sites of the leaky-late gene BLRF1 showed that distinct start sites were used in the absence or presence of lytic DNA replication. Expression of BLRF2, another leaky-late gene, could be detected even in the absence of viral lytic replication, but protein levels of BLRF2 increased in the presence of viral DNA replication, indicating that BLRF2 is expressed by both replication-dependent and replication-independent mechanisms. In support of these findings, a recent study by Lyons et al. (33) showed that mutant versions of the cellular protein AP-1, which can assume some functions of BZLF1, can activate the transcription of some EBV genes that were previously classified as late genes even in the absence of viral DNA lytic replication. Using CAGE-Seq data, Djavadian et al. (24) have newly classified these viral genes activated by the mutant AP-1 proteins as leaky-late genes or, in some cases, as early genes.

The changing epigenetics of EBV DNA influences viral gene expression.

EBV genomes that initially enter the nuclei of infected cells serve as templates for the expression of several latent genes, including the EBV nuclear antigens (EBNAs) and latent membrane proteins (LMPs), as well as viral noncoding RNAs and microRNAs (34). In 2007, Wen et al. (35) showed that mRNAs of the viral transactivator BZLF1, which mediates the switching of latent infections to the lytic phase, could be detected as early as 1.5 h after EBV infection and that the BZLF1 protein could be detected 6 h postinfection. BZLF1 expression soon after EBV infection was confirmed by other studies as well (36, 37). Paradoxically, two of the three studies looked for but did not detect lytic reactivation of EBV or virion production (within 2 weeks of infection), even though BZLF1 was expressed (35, 37).

Why can BZLF1 serve as a switch from latency to the lytic phase at some points during EBV infection and not at others? Clues needed to solve this mystery came from experiments showing that the binding of BZLF1 to viral promoters is enhanced when those promoters have high levels of CpG methylation (38–40). As it turns out, EBV DNA in virus particles is not methylated; it becomes methylated after infection as the infected cells proliferate. The level of methylation increases with time, at least in B cells, such that higher levels are detected between 50 and 100 cell generations after infection (37, 41). So, when BZLF1 is expressed soon after EBV infection, it cannot bind efficiently to the unmethylated viral DNA and therefore cannot activate BRLF1 or many lytic genes, and the lytic cascade is not initiated. On the other hand, in B cells with established latent infections, EBV DNA is heavily methylated, which allows BZLF1 to bind to the promoters of several lytic genes, turn on their expression, and start the lytic cascade (42).

The epigenetics of EBV genomes changes not only during initial infection but also throughout latency and during lytic amplification and virion production (43–47). In fact, because EBV DNA in virions is unmethylated and is packaged without histones (48), the epigenetic state of the viral genome is reset with each infection. The temporal evolution of viral DNA epigenetics influences viral gene expression—as shown for genes whose expression is driven by BZLF1—and we propose that it also causes the dependence of late gene expression on lytic DNA replication.

DNA amplification during EBV’s lytic phase causes yet another switch in the epigenetics of viral DNA. Kalla et al. (42) showed that DNA amplification during the lytic phase occurs without CpG methylation of viral DNA. Live-cell experiments following cells with EBV genomes from latency through the progression of the lytic phase indicated that lytic replication of EBV DNA yields newly synthesized DNA that lacks histones. EBV DNA replication takes place in replication compartments within the cell nuclei. When EBV DNAs are amplified sufficiently to be detected in the replication compartments, they are not bound by visible levels of histone H2B, H3.1, or H3.3, each fused to green fluorescent protein (GFP) (46). Also, analyses of EBV DNA early in the lytic phase showed decreased supercoiling relative to that of EBV DNA in latently infected cells, a finding consistent with viral DNA being bound by fewer nucleosomes (49). Collectively, these findings show that as viral DNA is replicated during the lytic phase, it becomes unmethylated and loses nucleosomes, culminating in the packaging of unmethylated, nucleosome-free genomes into viral capsids.

Do the unchromatinized EBV genomes produced during lytic replication link EBV late gene expression and DNA replication? If so, only those viral DNAs that include an origin of lytic replication in cis—and therefore can be replicated during the lytic phase—should serve as templates for late gene expression.

EBV DNAs must be replicated to serve as templates for late gene expression.

It has been a matter of debate whether late gene expression requires viral templates with an origin of lytic replication in cis or if DNA replication in trans is sufficient. Experiments with HSV-1 indicated that late genes could not be expressed from an unreplicated viral template (50). However, Serio et al. found that EBV late gene expression from transiently transfected plasmids did not need the viral origin of lytic replication, OriLyt, in cis (51). In contrast, Amon et al. showed that EBV-derived plasmids that were established in Akata cells required OriLyt in cis to express late genes during lytic reactivation (52). These disparate results could be explained by the fact that transfected plasmids do not always recapitulate the results seen when established viral genomes are used. For instance, experiments using plasmids derived from Kaposi’s sarcoma-associated herpesvirus (KSHV), another human gammaherpesvirus, did not recapitulate a dependence of late gene expression on viral DNA replication (53). Similarly, Aubry et al. (9) found that if the required trans factors were provided, transiently transfected plasmids without OriLyt could express a luciferase reporter gene driven by a late gene promoter even in cells that did not harbor EBV. Moreover, while PAA—which inhibits the lytic replication of EBV DNA—prevented the expression of late genes from established EBV genomes in 293 cells, it reduced luciferase expression from transfected reporter plasmids only by ∼50% (9).

One reason why transfected plasmids do not always mirror the transcription dynamics seen with established viral DNA may be the tight link between replication compartments and late gene expression. Studies have detected late gene transcripts within replication compartments, while mRNAs of early genes do not show this colocalization (10, 54). In a recent study, Li et al. (55) used cells in which the EBV and KSHV lytic cycles could be induced by the addition of doxycycline in order to test whether continuous viral DNA replication is required for late gene expression. The addition of PAA to inhibit viral DNA amplification at distinct time points after the induction of the lytic cycle seemed to break apart replication factories. Inhibition of continuous DNA replication also seemed to reduce late gene expression in this study, but interpretation of the results is complicated by two factors. First, it is not clear what percentage of the induced cells actually enter the lytic phase; a different study (56) showed that a maximum of 50% of the inducible KSHV-positive cells entered the lytic phase 6 days after the addition of doxycycline. Second, EBV-positive cells enter the lytic phase asynchronously (46). The combination of these two factors makes it challenging to quantify precisely lytic DNA replication or transcript levels in a population of induced cells.

Recently, Djavadian et al. (10) used EBV genomes with (+OriLyt) or without (ΔOriLyt) OriLyt to reexamine the role of OriLyt in the context of EBV’s lytic phase. They generated plasmids expressing a late protein, VCAp18, driven by its native promoter and fused to GFP. These plasmids were also constructed with or without OriLyt. Fusing VCAp18 to GFP allowed the researchers to differentiate between VCAp18 expressed from the viral genome (without GFP) and VCAp18 expressed from the plasmids (with GFP). The EBV genomes, with and without OriLyt, were established in 293 cells. The plasmids, with or without OriLyt, were transfected into the 293 cells with established EBV genomes (Fig. 2). When the EBV lytic cycle was induced in these cells, the native VCAp18 protein—encoded by the established EBV genomes—could be detected only when OriLyt was present in cis, i.e., the presence of OriLyt in trans on the transfected plasmids did not allow the expression of VCAp18 from the established EBV genomes if they lacked OriLyt. Similarly, VCAp18-GFP, encoded by the transfected plasmids, could be detected only when OriLyt was present in cis. These experiments show that plasmids encoding EBV late genes must be replicated before those genes can be expressed during the lytic phase.

There must be something about EBV genomes that are replicated during the lytic phase that allows these DNAs to serve as templates for late gene expression. Although several questions, such as whether all newly synthesized viral DNAs serve as templates for late gene transcription, have yet to be answered, we know quite a bit about the mechanisms of EBV lytic replication and can formulate hypotheses explaining how lytic replication yields at least some viral templates that allow late gene expression to proceed.

DNA replication and nuclear reorganization during EBV’s lytic phase are key for late gene expression.

During latency, EBV genomes exist as extrachromosomal plasmids in latently infected cells (57, 58). These viral DNAs are replicated by the cellular DNA replication machinery in a licensed manner and are partitioned to daughter cells during mitosis. Lytic replication is very different: EBV encodes its own DNA polymerase, BALF5, as well as other proteins for DNA replication (59). During EBV’s lytic phase, viral DNA is amplified several hundred-fold within days. The viral origin of lytic replication, OriLyt, is required in cis for DNA amplification during the lytic phase (60). There are also several proteins that are necessary trans factors for the replication of EBV DNA during the lytic phase (32). Live-cell experiments using D98/HR1 cells with visible derivatives of EBV have shown that when the lytic phase was induced by the nuclear translocation of the IE protein BZLF1, a subset of cells in the population displayed EBV DNA amplification within a few hours (46). In other cells in the same population, viral DNA amplification was delayed by as much as 36 h. Assays using a Cdt1-GFP fusion protein showed that viral DNA began to be amplified 1 to 3 h after Cdt1-GFP reached 90% of peak intensity. Cdt1 levels reach a maximum in late G₁, and the protein is rapidly degraded during S phase (61). Thus, EBV waits for host cells to reach S phase before initiating its lytic cycle.

Lytic amplification of EBV DNA leads to extensive reorganization of the nucleus and relocation of cellular DNA. Live-cell experiments have shown that the nuclear architecture is reprogrammed early in the lytic phase, and the nucleus takes on a honeycombed appearance. Viral DNA replication takes place in distinct locations, referred to as replication compartments, within the nuclei (46). Among several EBV proteins that localize to replication compartments is BMRF1, a DNA processivity factor that is required for viral DNA amplification during the lytic phase (62). Immunofluorescence and fluorescent in situ hybridization (FISH) studies with B95.8 cells induced to enter the lytic cycle showed BMRF1 cores surrounding amplified viral DNA within nuclear replication compartments (63). Subsequent RNA in situ hybridization assays showed that early mRNAs localized mostly outside BMRF1 cores while late mRNAs were found inside BMRF1 cores in B95.8 cells induced to enter the lytic cycle. In agreement with these findings, Djavadian et al. (10) found that mRNAs of the early gene BALF2 were distributed across the nucleus and cytoplasm but that transcripts of the late gene BcLF1 colocalized with viral replication factories.

The colocalization of EBV replication factories and late mRNAs lends further support to the hypothesis that only replicated DNAs can serve as templates for viral late gene expression. Replication compartments may be supporting late gene expression by additional means. For example, there are indications that levels of some vPIC proteins may be quite low in cells during the lytic phase ((25). Replication compartments may create areas in which concentrations of vPIC proteins are sufficiently high to promote complex formation (discussed in more detail below) and thereby foster late gene expression.

Interestingly, the link between herpesvirus lytic DNA replication and late gene expression seems to be mediated differently in beta- and gammaherpesviruses than in alphaherpesviruses. While all herpesviruses require DNA to be replicated during the lytic phase in order to express their late genes, a set of orthologous genes found in beta- and gammaherpesviruses, but not in alphaherpesviruses, encodes proteins that are all necessary for late gene expression (9, 10).

The beta- and gammaherpesviruses have a different strategy to express late genes.

Initial clues that late gene expression may be regulated differently in beta- and gammaherpesviruses than in alphaherpesviruses came from experiments conducted with murine gammaherpesvirus 68 (MHV68). A series of studies (64–66) with MHV68 mutants lacking the ORF18, ORF24, ORF30, or ORF34 gene showed that these proteins were necessary for late gene expression but not for lytic DNA replication or the expression of early genes. These genes have orthologs in other beta- and gammaherpesviruses but not in alphaherpesviruses. Experiments with human cytomegalovirus (HCMV), a betaherpesvirus, showed that HCMV proteins encoded by the UL79, UL87, and UL95 genes—which are orthologous to the MHV68 ORF18, ORF24, and ORF34 genes, respectively—are required for the expression of viral late genes and virion production (67).

Early, seminal in silico studies identified HCMV UL87 and its EBV homolog, BcRF1, as having structural features and amino acid residue arrangements similar to those of the TATA box-binding protein (TBP) (68). In eukaryotes, TBP is part of the general transcription factor TFIID, which itself is part of the cellular preinitiation complex that recruits RNA Pol II to promoters, initiates transcription, and mediates the release of RNA Pol II from the template (reviewed in reference 69). According to the canonical model of transcription initiation, TBP binds sequence-specifically to a DNA motif called the TATA box. Experiments with Saccharomyces cerevisiae yielded a consensus TATA box sequence of TATA(A/T)A(A/T)(A/G) (70). Not all eukaryotic genes have a TATA box in their promoter region, and in fact, only about 20 to 46% of genes in yeast and as many as 35% of genes in humans have an upstream TATA box-like feature (70–72). However, TBP seems to be important for the transcription of genes with or without a TATA box (69, 73).

Serio et al. (74) mapped a distinct TATA box-like feature in promoter regions of EBV’s late genes. Deletion analysis indicated that a variant of the TATA box motif, with the sequence TATTAAA, and its 3′ flanking region in the promoter of the late gene BcLF1 were required for the temporal regulation of BcLF1 expression. More recently, Djavadian et al. (24) found a TATTWAA element to be the only one enriched in EBV late gene promoters. A similar element containing TATT was also shown to be required to activate the KSHV K8.1 late promoter (75). A link between the viral TATT elements and EBV’s TBP homolog, BcRF1, was discovered when Gruffat et al. (76) showed that BcRF1 can bind specifically to TATT sequences present in the promoter regions of EBV’s late genes. A mutant version of EBV lacking BcRF1 was deficient for late gene expression and virion production but not for lytic DNA replication. Furthermore, mutated versions of BcRF1 unable to bind TATT sequences could not rescue late gene expression or virion production.

However, BcRF1 alone is insufficient to induce the expression of viral late genes (76). Recent research has shown that BcRF1 is but one component of a viral preinitiation complex (vPIC), along with the EBV proteins BDLF3.5, BDLF4, BFRF2, BGLF3, and BVLF1 (9, 10, 77). Each component of the vPIC is required for late gene expression in the context of the entire EBV genome, but a lack of vPIC components does not hinder early gene expression or lytic DNA replication (10). Immunoprecipitation experiments support a direct interaction between BcRF1 and cellular RNA Pol II, and this interaction was enhanced when other vPIC components were coexpressed (9). The KSHV homolog of BcRF1, ORF24, can also bind cellular RNA Pol II when ORF24 is expressed as a Strep-tagged protein outside the context of viral infection in 293 cells or as a FLAG-tagged protein in KSHV-infected SLK cells. Chromatin immunoprecipitation (ChIP) experiments showed that ORF24 is also able to bind to the late K8.1 promoter, which has a TATT sequence, but not to an early promoter with a more canonical TATA sequence (78). KSHV ORF24 can thus serve as a bridge linking RNA Pol II and viral late promoters. Other components of the KSHV vPIC are required for late gene expression as well. For example, point mutations that eliminated interactions between KSHV ORF24 and ORF34, another vPIC component, also prevented late gene expression (79), and KSHV mutants lacking either ORF18 or ORF30 were deficient for late gene expression but not for early gene expression or viral DNA replication (80). Similarities in the vPICs of gamma- and betaherpesviruses are highlighted by the finding that the entire HCMV vPIC can complement defective EBV vPIC, although individual components of the vPIC of one virus cannot substitute for corresponding components of another (9). A few EBV late genes seem to be transcribed in a vPIC-independent manner (24, 81), perhaps through mechanisms similar to those in alphaherpesviruses, but the vPIC is crucial for the expression of the vast majority of EBV late genes.

A new paradigm of EBV transcription: vPIC mediates the transcription of late genes from unmethylated, unchromatinized templates.

Each stage of EBV’s lytic gene expression program—immediate early, early, and late—is marked by distinct processes of transcriptional regulation and activation. The immediate early genes, BZLF1 and BRLF1, have complex promoters that can be bound by several cellular activators and repressors, which regulate their expression tightly (6, 59). Early gene promoters have a simpler structure, consisting primarily of binding sites for BZLF1 or BRLF1, called Z response elements (ZREs) or R response elements (RREs), respectively. Late genes seem to have the simplest promoter structures, consisting of only a TATTWAA element needed for vPIC binding. The simplicity of EBV’s late gene promoters mirrors the epigenetic simplicity of late gene templates, as described below. While BZLF1 is expressed from relatively unmethylated templates early during viral infection, expression of BZLF1 during the latent-lytic switch occurs from templates that are heavily methylated and bound with nucleosomes. Subsequently, BZLF1 activates the transcription of early genes from methylated viral DNA.

We propose that newly synthesized EBV DNA produced during the lytic phase is epigenetically distinct from the viral DNA that is present during latency. The viral preinitiation complex binds only the newly synthesized, unmethylated viral DNA that is also free of histones and recruits RNA polymerase II and other factors necessary for the expression of late genes from these templates (Fig. 3). Given that late gene expression in beta- and gammaherpesviruses requires vPIC proteins, it is likely that unchromatinized viral DNA serves as the template for late gene expression in all these viruses. In support of this hypothesis, Toth et al. (82) found that late gene expression during KSHV’s lytic phase was accompanied by a reduction of histone H3 association with viral DNA, as determined by ChIP assays. The epigenetic status of betaherpesvirus DNA during the lytic phase is less clear. Some regions of HCMV DNA appear to be associated with histones even after lytic replication, and histone molecules appear to be present in viral replication compartments, in contrast to what is seen during EBV’s lytic replication (83). However, other research indicates that only about 10% of HCMV DNA is associated with histone H3 late in the lytic phase (84).

Recent studies have also shown that the expression of EBV’s lytic genes has many layers of temporal and promoter complexities. Leaky-late genes have promoters that support transcription by both early and late mechanisms, potentially from distinct transcription start sites (24). While several studies have looked at the cascade of lytic gene expression in EBV (25, 30), one major experimental challenge is inefficient lytic cycle induction in EBV-positive cells. The inefficient induction of the lytic cycle yields a population where only some fraction of cells is lytic while others remain latent (46). In addition, even when lytic induction is relatively efficient, cells enter the lytic phase asynchronously, depending on which phase of the cell cycle they are in at the time of induction. It would be useful to engineer EBV-positive cells that could be selected based on whether they are in the early or late lytic phase. One potential technique is generating EBV-positive cells that harbor a truncated version of the human low-affinity nerve growth factor receptor (LNGFR) molecule as a cell surface marker driven by either an early or a late promoter. For example, EBV-positive cells, in which LNGFR expression is under the control of a tetracycline-dependent promoter that also drives the expression of BZLF1, were selected using anti-LNGFR magnetically activated cell sorting (MACS) beads (85, 86). Similarly, the rat CD-2 protein driven by an EBV early promoter allowed the selection of cells—by affinity purification—in which EBV was entering its lytic cycle (87). Either of these approaches would allow the selection of cells labeled with 5-ethynyl uridine (EU) in the early and late lytic phases, followed by enrichment for EU-labeled RNAs and RNA-seq to assay for viral and cellular genes that are being actively transcribed at different times during the lytic cycle. Similar experiments using ribosome profiling may yield valuable information about which mRNAs are being actively translated during the lytic phase. For example, Bencun et al. (88) used ribosome profiling on EBV-positive cells to show highly variable translation of EBV transcripts. However, only about 4 to 6% of the cells used in the assay were in the lytic phase. The use of a system similar to LNGFR-based selection would allow enrichment for cells in the lytic phase prior to ribosome profiling and a greater understanding of the dynamics of EBV lytic gene expression. Also, we now know that the beta- and gammaherpesviruses encode their own preinitiation complexes but use cellular RNA Pol II to transcribe late genes—a marvelous example of how viruses can mimic and coopt cellular processes at the same time. But it remains unclear whether other cellular factors or TBP-associated factors (TAFs) are part of the vPIC–RNA Pol II complex. The ability to select cells expressing late genes, using the LNGFR system, may facilitate the immunoprecipitation of RNA Pol II and the subsequent use of mass spectrometry to assay for other proteins associated with it during late gene expression.

What advantages might EBV derive from a strict temporal control of its late gene expression? One possibility is a competition between early and late gene expression for limited resources needed for transcription. In support of this possibility, Djavadian et al. (24) found that that EBV mutants defective for late gene expression exhibited reduced early gene transcription when late gene transcription was restored by trans-complementation. A second possible reason is that late promoters appear to be extremely compact compared with early (and immediate early) promoters. Sequences ∼30 bp upstream of the transcription start site appear to be sufficient to mediate late transcription kinetics, and the TATTWAA element may be the only essential cis-acting sequence. This simplicity may be particularly advantageous considering that most herpesvirus lytic transcription units are “nested”; i.e., multiple genes share a single polyadenylation site, and except for the most 5′ gene, the promoter is located within the coding sequence of another gene.

Conclusions.

Recent research has elucidated cis and trans factors linking EBV lytic replication and late gene transcription and has also highlighted differences in how late gene transcription is facilitated in different herpesviruses. We have learned that three factors are necessary for late gene expression during the lytic phase of EBV infection: the presence of OriLyt in cis, all components of the viral lytic DNA replication machinery, and the proteins that constitute the viral preinitiation complex. These discoveries are the foundation of a model of late gene expression in EBV in which late genes are expressed from unchromatinized viral templates by cellular RNA Pol II, recruited by the vPIC (Fig. 3). While vPIC-mediated late gene expression seems to be a predominant pathway in EBV, studies have indicated that other viral genes, including those encoding SM (89) and the EBV kinase BGLF4 (81, 90), may also play roles in late gene expression.

The dependence of beta- and gammaherpesvirus late gene expression on both viral DNA replication and the vPIC suggests that these viruses have evolved to avoid control of their late gene expression by the histone code. The regulation of transcription of host genes requires both the writers and the readers of this code. The writers, such as histone methyltransferases and histone deacetylases, are complexes of enzymes and scaffolds that add modifications to the histones in the nucleosomes at promoters. The readers are components of the transcription machinery that recognize these modifications, and they may be cellular or viral proteins. Together, they can either inhibit or activate transcription (reviewed in references 91 and 92). By using DNA templates that lack histones and by encoding their own vPICs, the beta- and gammaherpesviruses escape the regulatory control imposed on most cellular genes and instead substitute their own control, while still using some elements of the cellular transcription machinery. The alphaherpesviruses have not evolved the same control, perhaps suggesting different evolutionary constraints encountered during reactivation in neurons (93).