Epigenetic Regulation of EBV and KSHV Latency

Abstract

The gammaherpesviruses are unique for their capacity to establish a variety of gene expression programs during latent and lytic infection. This capacity enables the virus to control host-cell proliferation, prevent programmed cell death, elude immune cell detection, and ultimately adapt to a wide range of environmental and developmental changes in the host cell. This remarkable plasticity of gene expression results from the combined functionalities of viral and host factors that biochemically remodel and epigenetically modify the viral chromosome. These epigenetic modifications range from primary DNA methylations, to chromatin protein post-translational modifications, to higher-order chromosome conformations. In addition, gammaherpesviruses have acquired specialized tools to modulate the epigenetic processes that promote viral genome propagation and host-cell survival.

Variations of Gene Expression During Latency

The human gammaherpesviruses, Epstein-Barr Virus (EBV, also HHV4) and Kaposi’s Sarcoma-Associated Herpesvirus (KSHV, also HHV8) can establish latent infections in multiple cell types. During latency, the virus can express a restricted, but variable pattern of viral genes. These different gene expression patterns are essential for viral adaptation to the host and contribute significantly to viral-associated pathogenesis, including carcinogenesis. Epigenetic factors contribute substantially to the formation and heritability of these viral gene expression patterns.

EBV can assume at least four different gene expression programs during latent infection [1]. These “latency types” correlate with the host cell type (e.g. epithelial or lymphoid) or tumor origin (e.g. NPC or BL). In tissue culture, the latency types are metastable, in so far as they can maintain a stable pattern for multiple generations, but can also drift over time or switch in response to environmental signals or pharmacological manipulation. Cells that express all of the latency-associated transcripts (EBNA-LP, 2, 3a, 3b, 3c, 1, LMP-1, 2a, 2b, EBERS, miRNAs, BARTs) are referred to as type III, while those with a more restricted gene expression program are referred to as either Type 0, I, IIa, or IIb. Type 0 refers to the persistence of viral genomes in the absence of viral gene expression, a condition associated with latency in resting memory B-cells. The different latency types have distinct gene expression patterns and corresponding epigenetic modifications, which have been referred to as latency epigenotypes [2].

KSHV also shows variations in gene expression that depend on the cell and tumor type (reviewed in [3]). Most KSHV positive pleural effusion lymphoma (PEL) derived cell lines, express LANA, vCyclin, vFLIP, vmiRNAs, K12, and some variable expression of vIRF-3/LANA-2 [4-8]. In PEL tumor biopsies, rather than established cell lines, KSHV can express additional pathogenic genes, like vGPCR [9]. More sensitive methods using micro-arrays and proteomics have identified additional viral sense and anti-sense transcripts, as well as small peptides that may further refine our understanding of viral gene expression during latency [10]. Similar to PEL cells, KS lesions express the major latency transcripts, but a subset of cells also express some lytic genes, including the multicistronic transcript encoding vGPCR and K14 [11]. Interestingly, in PEL cell lines co-infected with EBV and KSHV, EBV adopts a type II restricted gene expression program [12,13].

Epigenetic Controls of Viral Gene Expression

DNA Methylation

Among the viral genes differentially regulated during EBV latency are those encoded by the long (~100 kb) alternatively spliced transcript initiating at the BamHI C promoter (Cp). These genes include EBNA-LP, EBNA-2, EBNA-3A, EBNA-3B, EBNA-3C, and EBNA1. It is well-established that differential DNA methylation upstream of Cp can lead to stable epigenetic repression [14-19](reviewed in [20]. When Cp DNA is methylated, transcription initiates at Qp to generate a smaller transcript that produces EBNA1, but none of the other EBNA genes. This switch from Cp to Qp is thought to occur during the natural infection as B-cells differentiate from proliferating centroblasts to resting memory B-cells [21]. Concurrently, methylation occurs at the promoter regions for the latency membrane proteins (LMP1 and LPM2), which results in their stable repression. Although BL tumors typically express type I latency pattern, BL derived cell lines can drift into a less restricted type III latency during laboratory cell culture. This shift to type III is thought to be driven by selective advantages of expressing type III genes that promote proliferation and survival, but compensatory changes in c-Myc expression may also be involved [22]. Changes in CpG methylation correlates strongly with the changes in viral promoter selection and latency type [14-16,23,24]. Importantly, treatment with DNA methylation inhibitors, like 5’azacytidine, can reverse repression of Cp transcripts in type I latency, indicating that CpG methylation can be a determinant of latency type [24]. In addition, some regions of the EBV genome are spared DNA methylation. This is particularly true of the regions surrounding Qp and OriP [18,25,26]. Methylation sparing of these regions may be essential for genome persistence during latency since Qp is the default promoter for EBNA1 in type I latency, and OriP is the episome maintenance element [27]. Since EBNA1 binds to both of these regions, it has been proposed that EBNA1 prevents local CpG methylation [28].

Methylation patterns have been examined genome-wide for KSHV [29,30]. Various peaks and valleys of CpG methylation were observed, with remarkable similarity among different cell types, ranging from B-cell derived PELs to in vitro infected SLK cells [29]. In all cases, regions of the viral genome were spared CpG methylation. These include the region immediately upstream of the latency promoter for LANA (LANAp) and several other locations including K9, ORF45/50, K7, and ORF8. The methylation pattern required significant time to establish, since KSHV genomes at 6 days post infection in SLK cells failed to establish any significant patterns of CpG methylation.

Viral-encoded factors can alter DNA methylation machinery by targeting DNA methyltransferases, DNA methyl binding proteins (MBPs), MBP-associated corepressor complexes, and direct binding to methylated DNA [31]. KSHV LANA binds to the de novo methyltransferase DNMT3a [32] and the methyl cytosine binding protein MeCP2 [33]. LANA can stabilize the repression of KSHV lytic cycle gene expression and its interactions with DNA methylation machinery may partly account for this repressing activity. Consistent with this finding is the observation that DNMT3a and 3b mediate CpG methylation and transcription repression of MHV68 ORF50 promoter in latently infected B-lymphocytes in vivo [34]. A remarkable series of findings have revealed a complex and paradoxical role of DNA methylation in the regulation of EBV lytic reactivation (reviewed in more detail in this journal series) [35-40]. The EBV lytic activator Zta has the unusual capacity to bind selectively to methylated DNA, enabling the establishment and reactivation of methylated genomes [35,41-43]. Viral DNA methylation patterns can also be regulated by non-coding RNAs [30]. In one study, deletion of the KSHV miRNA cluster led to a loss of DNA methylation at the ORF50 locus. At least one viral miRNA, miRK12-4-5p, affected viral DNA methylation through the indirect targeting of DNMT3b through the global repressor Rbl2.

Histone Modifications

Viral genomes appear to utilize cellular transcription factors and chromatin regulators indistinguishably from host chromosomes. Most viral promoters share conventional features of cellular epigenetic regulation, with histone acetylation and H3K4me3 methylation occurring at transcribed promoters. Increase histone H3 and H4 acetylation has been demonstrated for the activated Cp in response to EBNA2 activation in type III latency [44], and for activation of the EBV BZLF1/Zta [45,46], and KSHV ORF50/Rta immediate early gene promoters [47]. However, it remains possible that viral episomes are distinguished from cellular chromosomes by some as yet unknown, viral-specific epigenetic features.

Genome-wide studies of histone modifications have revealed several unexpected observations. Initial studies using low resolution PCR arrays, suggested that EBV genomes may be partitioned into relatively large active and inactive chromatin domains, with Cp and Qp representing distinct boundaries for different latency types [48]. More recent studies, using higher resolution tiling arrays or ChIP-Seq methods reveal a more discrete and mottled pattern of epigenetic regulatory marks [29,49,50]. For KSHV genomes, several loci including the immediate early promoter region for ORF50/Rta were found to have a bivalent epigenetic state, similar to that described for embryonic stem cell regulatory genes [51]. The bivalent marks consist of an overlap for H3K4me3, correlating with transcription activity, and H3K27me3, correlating with transcription repression[51]. The EZH2 methylase, a member of the polycomb complex (PcG), was shown to be critical for histone H3K27me3 and repression of KSHV lytic gene reactivation. Interestingly, a pharmacological inhibitor of H3K27 methylation, DZNep, stimulated reactivation of KSHV, but not EBV in dually infected JSC-1 PEL cells, suggesting that KSHV and EBV have different epigenetic controls over lytic reactivation [50].

Mapping of the human ENCODE data base for epigenetic modifications and transcription factor binding sites to the EBV genomes in human LCLs revealed important differences from previous low resolution studies, as well as comparative analysis with KSHV epigenetic marks [49]. Unlike KSHV, levels of H3K27me3 are surprisingly under-represented on the EBV genome, relative to other histone modifications or to the KSHV genome. Examination of the EBV epigenome revealed sharp peaks of histone H3K4me3 and acetylated histones (H3K27ac and H3K9ac) at sites of transcription initiation for EBERs, Cp, LMP1/LMP2, and RPMS1p (the promoter for the BART transcripts and miRNAs). Another feature of the EBV epigenome was the clustering of multiple transcription factors at the sites of active promoter regions, and very few interactions at transcriptionally silent genomic locations.

Chromatin Boundaries

The latent genomes of EBV and KSHV have meta-stable epigenetic modification patterns. The localization of histone H3K4me3 and H3K27ac at sites of transcription initiation are mechanistically explained by the recruitment of RNA polymerase II or III transcription and initiation factors to their cognate DNA recognition sites. High levels of the TATA binding protein (TBP) are found at the EBER promoter, while TBP-associated factors (TAF-1) and histone acetyltransferase co-activator p300 are elevated at the major RNA Pol II initiation sites at Cp, LMP1/LMP2, and RPMS1p [49]. Histone H3K4me3 is elevated at all of these sites, but it is not yet known what histone methylase is responsible for each of these promoter modifications. More cryptic is the mechanism for restricting repressive marks, like H3K9me3 and DNA methylation, to only a few subdomains. As noted above, DNA methylation and H3K9m3 are never found at some protected sites, like the EBV Qp [18] and the KSHV LANA promoter region [29,50]. Chromatin boundary elements may prevent these repressive marks from spreading into protected regions. Among the best-characterized metazoan chromatin boundary factor is the 11-zinc finger DNA binding protein CTCF[52,53]. CTCF binds to a region upstream of the EBV Qp [54] and surrounding the LANA promoter region [55]. Deletion of CTCF from the Qp region leads to an increase in histone H3K9m3, followed by encroaching CpG methylation. Deletion of CTCF from the LANA promoter region results in a loss of episome stability, but there is no evidence that this loss corresponds to an increase in H3K9m3 or DNA methylation. Thus, CTCF appears to function as a chromatin insulator that prevent histone H3K9m3 and DNA methylation at Qp, while it may serve a more complex function at the LANA promoter region. Chromatin boundaries may also be generated by other factors, including the viral encoded maintenance proteins, EBNA1 and LANA, that control nucleosome positioning [56,57] and DNA replication fork stalling [58,59], which may regulate the processive spreading of histone modifications.

Genome Conformation

Higher order DNA structures, like those involving DNA looping that mediate enhancer-promoter interactions, have been identified on both EBV and KSHV chromosomes [60,61]. For EBV, these DNA loop structures correlate with promoter selection, and therefore are likely to be coordinated with epigenetic patterning associated with each latency type. Therefore, higher-order chromatin structure is likely to be a heritable epigenetic regulatory feature. The DNA loops for EBV were found between OriP and either Cp in type III latency or Qp in type I latency [60]. CTCF binding sites located upstream of Qp and Cp were found to be important for mediating the long-distance looping between OriP-Qp in type I latency. These loops may also be mediated by EBNA1, which binds to both OriP and Qp. In KSHV, a DNA loop was identified for the CTCF binding sites in the LANA promoter region. In this case, a loop was formed with the 3’ end of the KSHV latency transcripts ending at K12, suggesting that the entire latency transcription area is contained within a DNA loop mediated in part by CTCF binding. Additionally, a longer DNA loop was also found between the LANA promoter region and the ORF50/Rta promoter control region. DNA loops are stabilized by the cellular factors involved in sister-chromatid cohesion, namely cohesin subunits SMC1, SMC3, RAD21, and SCC2. The cohesins form a ring like structure that can either embrace or hand-cuff two separate DNA molecules to stabilize a non-contiguous DNA-DNA interaction [62-65]. Deletion of the Rad21 subunit leads to a loss of the DNA loop between the latent and lytic control regions, and the reactivation of lytic cycle transcription [66]. These findings suggest that DNA loops are important for maintaining a stable gene expression program during latent infection, but that EBV and KSHV have very different conformations and control mechanisms.

Viral-Encoded Epigenomic Regulators

Viral-encoded factors can modulate the viral and host epigenomes through various mechanisms, including direct interactions with host epigenetic regulatory factors [67]. Here, we consider the contribution of the viral encoded episome maintenance proteins that bind directly to the viral genome and directly influence viral chromosome organization.

EBNA1 binds with high affinity to the FR, DS, and Qp regions of EBV. It can also bind with lower affinity to Rep*[68], and to several sites in the host chromosome [69,70]. At Qp, EBNA1 can function as a repressor of transcription [71], while at OriP it functions as an essential enhancer of Cp and LMP1/LMP2 promoters [72]. At cellular chromosomal sites, EBNA1 binding has no clear effect on transcription of neighboring cellular genes [69,70]. EBNA1 can activate some cellular genes, including NOX2 [73], but its mechanism of activation appears less direct than many well-characterized transcriptional activators and repressors. EBNA1 does not interact with any well-characterized co-activators or co-repressors of transcription. However, EBNA1 does interact with nucleosome assembly proteins and bind to chromatinized DNA [57,74], suggesting that it may play a role in organizing nucleosome position or higher-order chromatin structure. EBNA1 mediates a loop between the FR and DS in vitro, and also mediates transcription activation of Cp during primary infection [72]. Thus, EBNA1 may function as a chromosome architectural protein that organizes nucleosome position and long-distance DNA interactions. EBNA1 is subject to many post-translational modifications, including phosphorylation and poly-ADP ribosylation (PARylation) [75,76]. PARylation of CTCF has been implicated in DNA-looping, and it has been suggested that PARylation of EBNA1 may also contribute to the chromosome architectural functions of EBNA1.

LANA is a functional orthologue of EBNA1, and shares several important features, including the ability to bind to its respective viral genome with high affinity, and attach to metaphase chromosomes [77]. LANA binds the KSHV TR with high affinity, but does not show similar levels of site-specific enrichment at host chromosome sites as does EBNA1[78]. LANA can bind to several sequence specific DNA binding proteins, the best-characterized interaction being with RBP-jK[79]. RBP-jK (also referred to as CBF1 or CSL) is the major target of intracellular Notch, and a common interactor of gammaherpesvirus regulators, including KSHV Rta and LANA, and EBV EBNA2 and EBNA3C. LANA can prevent Rta activation of lytic replication through its competitive interactions with RBP-jK [79-82].

LANA interacts directly with numerous chromatin-associated and epigenetic regulatory factors, including histone H2A/H2B [83], BRD2/4[84], DEK [85], MeCP2 [85], nucleophosmin [86], CENP-F [87], DNMT3a [32], TRF1 [88], p53 [89], TopoII beta [90], and TIP60[88]. Interaction with TIP60, a histone acetyltransferase, may also mediate DNA damage response through acetylation of ATM and other DNA damage response proteins [88]. Interaction with MeCP2 occurs independently of its binding to nucleosomes and functions to localize LANA to chromocenters where it facilitates the transcriptional repression and heterochromatinization of the TRs [33]. However, MeCP2 can also cooperate with LANA to activate some promoters, suggesting that LANA has complex functionality at different chromosomal locations.

LANA is subject to several post-translation modifications that may confer epigenetic regulation of the viral chromosome. These include lysine acetylation[91], arginine methylation by PRMT1 [92], and phosphorylation by PIM1[93,94], CK1, and RSK3 [95]. EBNA1 is also subject to phosphorylation by CK2[96], and CDKs[97], which control important replication and maintenance functions. LANA and EBNA1 can also interact with HAUSP7, a ubiquitin protease that can stabilize p53 and modulate transcription through regulation of histone ubiquitylation cycles [98-100].

Metaphase Attachment Functions

Both LANA and EBNA1 tether the viral episome to the metaphase chromosome, a feature that has correlated strongly with episome maintenance. A potent metaphase attachment motif can be mapped to a stretch of basic amino acids in the amino terminal domain of LANA. This domain interacts with histone H2A/H2B and the interaction has been solved by X-ray crystallography[83]. EBNA1 interacts with metaphase chromosomes through two distinct motifs, an AT-hook motif that can interact with AT-rich DNA [101], and a separate domain that interacts with EBP2, an RNA binding protein that associates with metaphase chromosomes [102]. Both EBNA1 and LANA can interact with the BET family members, BRD4 for EBNA1[103], and BRD2 and 4 for LANA [84,104]. BET family proteins have bromodomains implicated in recognizing acetylated lysine on histone tails [105]. BRD family members have been implicated in the metaphase attachment of HPV E2 proteins [106,107], although they may also play a significant role in transcription regulation [108,109]. BET family proteins can interact with and regulate RNA polymerase II elongation factor pTEFb [110]. BET proteins have also been implicated in mitotic attachment and cell cycle bookmarking of transcriptionally active cellular genes [111,112]. How these viral factors integrate mitotic attachment with episome stability and transcriptional elongation will be an area of important future research.

Conclusions

The epigenetic control of gammaherpesvirus latency is an integral feature of chromosome biology. Epigenetic factors are essential for the complex orchestration of transcription, replication, chromosome mobilization and segregation at the appropriate time and place. Epigenetic features coordinate regulatory events between different genetic locations, through common protein complexes and higher order chromatin structures, including long-intervening DNA loops. These epigenetic regulators play important and significant roles in determining gammaherpesvirus gene regulation during latent infections. More importantly, these epigenetic controls provide both stability and plasticity to the viral genome that enables a programmed response to cellular and environmental changes. Much remains to be discovered for understanding gammaherpesvirus epigenetic controls and how these distinguish latency types. More urgently, will be to determine how epigenetic controls contribute to the formation of aberrant latent epigenotypes that increase risk of virus-driven carcinogenesis [113].

Figure 1. Progressive Epigenetic Modifications Regulate Gammaherpesvirus Latency.

Early stage histone modifications lead to metastable gene expression programs, which are reinforced by secondary higher-order chromosome conformations, including DNA looping. DNA methylation occurs more gradually and may take the place of repressive histone modifications to increase epigenetic stability.

Figure 2. Viral Encoded Epigenetic Modulators.

(A) Cellular epigenetic regulators that interact with gammaherpesvirus chromosome organizing/episome maintenance proteins LANA and EBNA1. (B) Schematic of promoter regulatory interactions mediated by viral chromosome organizing factors for KSHV, LANA, and CTCF. (C) Promoter regulatory interactions for EBV in type I or type III latency conformations. OriP, EBNA1, and CTCF are highlighted.

Highlights.

• Gammaherpesviruses can adopt a variety of gene expression patterns during latent infection.

• Gene expression patterns correlate with changes in epigenetic modifications, including DNA methylation, histone modifications, and chromosome conformations.

• Viral episome maintenance proteins play a central role in epigenetic regulation and chromosome organization.