Molecular Mechanisms of KSHV Latency Establishment and Maintenance

Abstract

Purpose of Review

This review summarizes results from recent studies that shed light on viral and cellular mechanisms that contribute to latency and persistence of Kaposi sarcoma-associated herpesvirus (KSHV). We discuss the initial molecular events of latency establishment starting from entry of the viral genome into the nucleus to how viral genomes are properly segregated when latently infected cells undergo mitosis. Finally, we discuss the critical role of the latency-associated nuclear antigen (LANA) in orchestrating these processes.

Recent Findings

Upon entering the host-cell nucleus, naked viral DNA circularizes and rapidly associates with cellular histones and the transcription machinery. This permits a burst of viral gene expression including LANA which quickly recruits chromatin remodeling complexes and modifiers to organize the viral chromatin architecture. These events restrict transcription to the latency locus and switch the lytic genes to a repressed state that is poised for reactivation. Structural studies demonstrate that LANA assembly on the viral terminal repeats leads to the formation of higher order structures that stabilize the latent viral chromatin and impact how the viral episome is segregated during cell division.

Summary

Understanding molecular mechanisms of latency establishment and maintenance provide insights into gammaherpesvirus biology and thus may reveal new strategies to prevent or treat KSHV-associated malignancies.

Introduction

Kaposi sarcoma-associated herpesvirus (KSHV) is the causative agent of several malignancies and lymphoproliferative diseases including Kaposi sarcoma (KS), KSHV inflammatory cytokine syndrome (KICS), primary effusion lymphoma (PEL) and multicentric Castleman disease (MCD) [1]. KSHV infections are found across the world but certain regions like sub-Saharan Africa have high seroprevalence in the general population [2]. While infection does not usually lead to disease, immunosuppression is a major risk factor for KSHV malignancies. This strong association first became evident during the AIDS epidemic where AIDS patients exhibited prominent KS lesions [3, 4]. Although the introduction of combination anti-retroviral therapy (cART) significantly decreased the over-all incidence of KSHV-associated cancers in people living with HIV, the risk of KS remains elevated in patients whose HIV infections are effectively controlled by cART [5–7].

KSHV is a gammaherpesvirus that has a large ~140 kilobase (kb) linear double-stranded DNA (dsDNA) genome [8]. KSHV infects many different cell types where, characteristic of the herpesvirus productive or “lytic” replication cycle, the virus initiates a highly coordinated and temporally regulated gene expression cascade of three lytic gene classes (immediate early [IE], early [E], and late [L] lytic genes) that culminates in virion production. However, like all herpesviruses KSHV also establishes life-long latent infections, and like the related human pathogen Epstein-Barr virus (EBV), KSHV latently infects B cells. KSHV can also establish latency in vitro in lymphatic endothelial cells [9–12] and mesenchymal stem cells [13–16], both of which are proposed as the potential cell-types from which KS originates. During latency, viral gene expression is restricted primarily to the ~12 kb latency locus which encodes the latency-associated nuclear antigen (LANA, encoded by ORF73), viral cyclin (vCYC, encoded by ORF72), viral FLICE inhibitory protein (vFLIP, encoded by ORF71), at least 20 mature viral microRNAs, and the circular viral interferon regulatory factor 4 RNA (circ-vIRF4) [17–20]. KSHV also expresses the B cell-specific latent protein vIRF3 in PEL cell lines and PEL tumor samples, but not in KS tumors [21].

Latency is distinct from chronic or subclinical infections as no viral particles are produced. True latency is characterized by two properties: persistence of the viral genome and the capacity to reinitiate full viral gene expression [22]. During latency, viral gene expression is restricted to the latency genes that function to allow the viral genome to be maintained and persist as an extrachromosomal plasmid, even during host-cell division. The minimalistic latency transcription program results in a reduced number of viral antigens being expressed which thereby facilitates evasion of host immune surveillance. A second property of latency is its reversibility, or the capacity of the virus to reinitiate full viral gene expression and productive virion generation in response to environmental cues. Reinitiation of the lytic replication cycle, a process termed reactivation, is triggered by the replication and transcription activator protein (RTA, encoded by ORF50), the master transcription factor that initiates the lytic gene expression cascade.

A critical and essential protein in KSHV latency is LANA. LANA tethers latent KSHV plasmid to host chromosomes during mitosis to ensure that viral genomes are maintained and efficiently partitioned to daughter cells during cell division [23, 24]. In addition to this structural role, LANA also directs silencing of genes involved in lytic replication [25–27]. LANA binds directly to the ORF50 promoter to prevent RTA expression and progression of the lytic gene expression cascade [27]. Moreover, LANA in part facilitates the heterochromatinization of the viral genome during latency establishment [26]. Thus, LANA is integral to KSHV latency via its roles supporting the two properties of true latency: genome persistence and mediating a tightly controlled latency program.

In this review, we discuss new perspectives and insights into molecular mechanisms of KSHV latency. Specifically, we focus on recent studies that investigate the epigenetic reorganization of the viral genome during latency establishment, principles of latent genome partitioning, and how LANA facilitates these processes.

Establishing the Latent Viral Chromatin

Like other herpesviruses, the linear KSHV genome rapidly circularizes and associates with cellular histones upon entry into the nucleus [28, 29]. During this early stage of infection, the viral chromatin gains activating histone modifications (e.g. H3K4me3, H3K27ac) that allow for a burst of latent and lytic gene expression (Figure 1) [29, 30]. However, a critical part of establishing viral latency is the ability to selectively shut off lytic transcription while still allowing for the expression of the latency genes. Transcriptional silencing in eukaryotes is often mediated by the Polycomb group (PcG) proteins, which function in multi-subunit complexes known as the Polycomb repressive complex 1 (PRC1) and PRC2 that deposit repressive post-translational modifications on chromatin [31].

Figure 1. Establishment of the latency program.

1) During de novo infection, the linear KSHV genome circularizes through recombination of its terminal repeats (TR, gray boxes). 2) Cellular histones rapidly associate with the viral genome to form the viral chromatin and begin the transcription of latent and lytic viral genes. 3) Cellular chromatin modifiers are recruited either directly by the viral genome or indirectly through LANA. 4) The orchestrated binding of these chromatin factors remodels the 3D structure of the viral genome to shut off the expression of lytic genes while restricting transcription to latency genes in latency domains.

Clues to how the KSHV genome transitions from a transient burst of lytic gene transcription into a primarily silent latent state emerged from studies performed in SLK cells [32]. Although SLK was later identified to be a contaminant renal epithelial carcinoma cell line Caki-1 [33], it is a well-established cell-culture model in which KSHV infection is tightly latent. A temporal analysis of proteins that bind the viral genome during de novo infection of SLK cells demonstrated that newly synthesized LANA occupies viral promoters as early as 8 hours postinfection (hpi) and precedes the appearance of histone modifications associated with transcriptional repression and chromatin compaction, in particular the trimethylation of lysine 27 on histone H3 (H3K27me3) [26, 29]. LANA mediates transcriptional repression across the viral genome by directly engaging subunits of PRC2, but not PRC1, and recruiting them to viral DNA [26]. The PRC2 subunit EZH2 catalyzes H3K27me3 deposition on the viral genome, which gradually increases over time as viral transcription becomes restricted to the latency genes [26, 34]. Importantly, KSHV mutants that abrogate LANA binding to EZH2, chromatin or DNA, all fail to recruit PRC2 to lytic promoters, which results in aberrantly active lytic gene transcription. Thus, LANA facilitates latency establishment by globally silencing lytic gene expression via the recruitment of PRC2 to the viral DNA.

In contrast to PRC2, the role of PRC1 in silencing the viral genome during latency establishment is more complex. PRC1 can be comprised of canonical (cPRC1) or non-canonical (ncPRC1) complex variants. Both variants share a core Really Interesting New Gene protein (RING1A or RING1B) which acts as an E3 ubiquitin ligase that monoubiquitinates Lys118 and Lys119 of histone H2A (H2AK118Ub and H2AK119Ub), which is indicative of repressed chromatin [35]. cPRC1 differs from ncPRC1 as cPRC1 has a chromodomain protein subunit (CBX) while ncPRC1 instead binds to RING1 And YY1 Binding Protein (RYBP) or YY1 Associated Factor 2 (YAF2) [36]. cPRC1 is recruited to H3K27me3 through the CBX subunit to induce chromatin compaction [37], whereas ncPRC1 is recruited to ubiquitinated chromatin through RYBP/YAF2 to propagate repressive ubiquitination of nearby histones [38].

Systematic knockdown of representative subunits of cPRC1 and ncPRC1 revealed that RYBP and core protein RING1A are required for ncPRC1 silencing of viral lytic genes, but knockdown of cPRC1 components CBX4 or CBX7 had no effect [34]. Furthermore, RYBP and RING1B occupancy of lytic promoters occurred rapidly, within 2 hours of infection, and coincided with H2AK119Ub deposition, which contrasted with the delayed appearance (72 hours) of PRC2-dependent H3K27me3 modifications. Interestingly, RYBP knockdown did not impact the detection of activation mark H3K4me3 nor repressive mark H2AK119Ub on viral lytic promoters. Instead, RYBP knockdown increased transcriptional elongation-associated histone modifications such as H3K36me3, H3K79me2 and H2BK120Ub, and promoted transcriptional readthrough of the distal end of the RTA transcript [34]. This apparent discrepancy suggests that RYBP silences lytic genes at the RNA level during transcriptional elongation rather than functioning at the promoter level.

In addition to the PcG proteins, other epigenetic factors contribute to silencing lytic gene transcription during de novo infection [39] and latency maintenance [40]. Using a siRNA screen against 392 human epigenetic factors in SLK cells, knockdown of 46 host proteins prior to infection resulted in an increase of RTA transcripts without affecting LANA expression [39]. These included subunits of the Nucleosome Remodeling and Deacetylase complex (NuRD; e.g., GATA Zinc Finger Domain-Containing 2B or GATAD2B, Chromodomain Helicase DNA Binding Protein 4 or CHD4, and Methyl-CpG Binding Domain Protein 3 or MBD3) and the Lysine Acetyltransferase 5 (KAT5) co-repressor complex (e.g., KAT5, Histone Deacetylase 7 or HDAC7, HDAC9, and ETS variant transcription factor 6 or ETV6). The identification of multiple histone deacetylases from distinct complexes suggests that deacetylation of chromatin is important in silencing lytic genes. While it remains to be determined whether deacetylases act directly on viral chromatin, it is notable that HDAC inhibitors are commonly used to facilitate KSHV reactivation [41–43]. Interestingly, CHD4 also participates in another chromatin remodeling complex, ChAHP (CHD4-Activity Dependent Neuroprotector Homeobox or ADNP-Heterochromatin Protein 1γ or HP1γ), which occupies the same regions as LANA on latent DNA to inhibit RTA expression [40], thus underscoring the multiple roles of CHD4 at different stages of viral latency.

The same siRNA screen also identified several histone demethylases important for inhibiting RTA transcription, including Lysine Demethylase 2B (KDM2B) [39]. KDM2B primarily catalyzes the removal of activating histone methylations H3K4me3, H3K36me2 and H3K79me2 in mammals [44–46]. It also recruits ncPRC1 through RING1B to unmethylated CpG islands [47, 48]. In de novo KSHV infections, KDM2B binds to the promoters of lytic genes such as RTA and ORF25 to silence these genes during the establishment of latency [39]. Knockdown of KDM2B increased H3K4me3, H3K36me2, and H3K79me2 modifications on the RTA promoter consistent with its expected role as an inhibitor of lytic transcription.

The 3D Organization of Viral Chromatin

Transcription programs during discrete stages of development are orchestrated by higher order structures of chromatin that form local units of coordinated transcription. It is becoming clear that the chromatinized KSHV genome is also organized into specific conformations that enhance the latency transcription program. Pioneering studies by Lieberman and colleagues demonstrated that the chromatin architectural protein CCCTC-binding factor (CTCF) and cohesins segregate a structural loop of the latency locus that allows self-contained transcription during latency (Figure 1) [49–52]. Interestingly, a second loop is formed through CTCF sites on the LANA promoter and the promoter of the master lytic regulator RTA. This loop is thought to instead prevent RTA expression and the aberrant initiation of lytic reactivation. The unique organization of the latency locus and the RTA promoter may act as a multifunctional structure that supports latency, yet remains poised to initiate lytic transcription.

More recently, unbiased genome-wide mapping of long-range intrachromosomal interactions of the KSHV genome was performed in PEL cell lines using chromosome conformation capture followed by high-throughput sequencing (Hi-C) [53, 54]. These experiments revealed that viral genes with similar temporal regulation are found within the same chromatin loops or transcription regulatory domains (TRDs). For instance, although the region containing the B cell-specific latency factor vIRF3 is ~25 kb away from the core latency gene locus, DNA looping puts these two regions near each other to form linked latency TRDs. These TRDs coincide with CTCF binding sites and appear to confine histone modifications within each domain [53]. Interestingly, the vIRF TRD is specific to PEL cell lines and does not form in SLK cells. This is consistent with the B cell-restricted expression of vIRF3 during latency [21], and suggests that cell lineage influences the 3D conformation of the viral genome and thus cell type-specific latency programs. During lytic reactivation, however, the latent TRDs are disrupted and new TRDs involving IE and E lytic genes form and loop back to the RTA promoter. It is postulated that clustering of these lytic TRDs to the RTA locus creates a hub for transcription that facilitates rapid reactivation.

A Super Resolution Model of the Viral Chromatin Tether

Since KSHV establishes latency in dividing cells, efficient segregation of viral episomes into daughter cells during mitosis poses a fundamental problem for viral genome maintenance and persistence. This potential problem is overcome by the functions of latency protein LANA. Similar to the Epstein–Barr virus nuclear antigen 1 (EBNA1), LANA facilitates proper genome segregation during cell division by tethering the viral DNA to mitotic chromosomes [23]. To do so, LANA binds as a dimer to each of three adjacent LANA binding sites (LBS1, LBS2, and LBS3) within each of the 35-45 repeat units present in the terminal repeat (TR) region of the viral DNA [55–60]. This high avidity interaction is mediated by the C-terminal DNA binding domain (DBD) of LANA (aa 951-1162), which is a structural homolog of the DBD of EBNA1 [61].

In contrast to EBNA1 which directly binds to cellular DNA, LANA binds to host chromosomes through chromatin-associated proteins. The DBD interacts with chromatin remodelers and modifiers including Bromodomain-Containing Protein 2 (BRD2) and BRD4 [62–65], which are known to bind to acetylated histones H3 and H4. In addition, the first 22 aa of the LANA N-terminal domain (NTD) (aa 1-330) is required for tethering to mitotic chromosomes [66–68] by binding to histones H2A and H2B [66].

Although several crystal structures of LANA bound to DNA have been solved [60, 69–71], these studies focused on the interactions of LANA to either the viral DNA or host chromatin, but not to both. Only recently was the structure of the full tripartite tether (viral DNA-LANA-host chromatin) elucidated [72]. Kedes, Smith and colleagues used direct stochastic optical reconstruction microscopy (dSTORM) to visualize single molecules of LANA in both SLK cells latently infected by KSHV BAC16 and in the PEL cell line BCBL-1. Comparisons of LANA “dots” within these two cell lines found that the viral tethers have consistent dimensions both within and between cell types. These observations led to the hypothesis that intrinsic properties of the viral DNA (i.e. number of TRs available for LANA binding) drives the size of the LANA tether. To test the model that cis elements in the viral DNA dictate the dimension of the viral tether, the authors visualized LANA dots in the uninfected B cell line BJAB following co-transfection with a plasmid encoding LANA and another plasmid containing increasing numbers of TRs [72]. Consistent with their hypothesis, the LANA tethers increase in size proportionally to the number of TRs. Because binding does not saturate (at least up to 21 TRs tested), each TR is likely independently accessible to LANA binding.

Detailed analyses of these tethers determined that the LANA clusters in adjacent TRs have asymmetrical orientations that form a 46° bend. Modeling these images to published crystal structures and biophysical studies [60, 71, 73] suggests that the asymmetric DNA bend results from the combined effects of the relative positioning of the three LBSs as well as four nucleosomes within each TR. Their proposed model suggests that all three LBSs are simultaneously occupied by LANA dimers (6 molecules per TR). Importantly, the structure of the viral tether orients the LANA NTDs outward in an orientation that should be readily accessible for tethering to host chromatin.

One should note that this model for the viral tether differs from published crystal structures and biochemical studies with LANA. Other groups have observed that LANA can form ring-like complexes composed of tetramers to decamers of LANA dimers [60, 69]. Formation of these structures are induced by binding to TR DNA in vitro [69] but they can also self-assemble without DNA [60, 69]. The discrepancies between these studies likely stem from the limitations of the respective systems, for instance evaluating only the LANA DBD in contrast to full-length LANA protein used by Grant et al. While reducing LANA to the DBD was beneficial for crystal formation and atomic-level structure determination, it is important to note that these structures lack of the first ~1000 amino acids of LANA. In contrast, dSTORM imaging was performed in infected cells, which therefore also include cellular factors that can influence LANA-TR interactions [72]. The authors note, however, that their analysis was restricted to single-molecule structures to aid in establishing a best-fit model. This excludes ambiguous structures that are difficult to resolve including larger LANA “dots”, which may represent higher order structures or clusters. Additional studies should investigate the high-resolution structures of these clusters that form in infected cells. Nevertheless, the interaction of LANA multimers on viral TRs provides intrinsic properties that shape the tethering architecture critical for genome persistence.

Functions of LANA Oligomers

From these structural studies, the oligomerization of LANA appears to be a structural determinant of LANA functions. A recent study to define the necessity of LANA oligomerization in latently infected cells showed that mutations that prevent oligomerization, but still allow for DNA binding by monomers [69], were compromised in their capacity to bind TRs [74]. The oligomerization mutants also failed to recruit the Origin Recognition Complex Subunit 2 (ORC2), a cellular replication factor that is required for latent plasmid replication [75], which may explain the earlier observation that these mutations in LANA abrogated plasmid replication and maintenance [69].

Mutations impacting LANA oligomerization also altered how latent viral chromatin is structurally organized. For instance, looping of the latency locus to the lytic RTA promoter [51] is completely lost for LANA oligomerization viral mutants [74]. Another regulatory loop linking the TR to the LANA locus similarly becomes undetectable. Moreover, loss of oligomerization disrupted the colocalization of LANA with the PRC2 subunit EZH2 and decreased H3K27me3 modifications at the latency locus. These data collectively suggest that LANA oligomerization is also required for the proper control of the latency program by coordinating the structural loops and recruitment of PRC2.

Interestingly, a tiling qPCR analysis of the KSHV genome showed that a large percentage of the viral genome (nucleotides 1-100,000) consistently had lower copy numbers when the LANA oligomerization residues were mutated, which suggests that LANA oligomerization is important for maintaining genome integrity. This likely contributes to the known defects of LANA mutants in long-term plasmid maintenance [69] and is consistent with a recent report describing a protective role for LANA in preventing viral genome degradation [76].

LANA Nuclear Bodies

Perhaps not surprisingly, the oligomerization mutants fail to form LANA speckles or nuclear bodies (LNBs) [74]. LNBs are large aggregates that are easily visualized using epifluorescence microscopy [23]. These have long been considered a hallmark of LANA binding to KSHV genomes, because they colocalize with the viral DNA during interphase and metaphase. The size and speckling of LNBs are reminiscent of nuclear membraneless organelles (MLOs). MLOs are biomolecular condensates that form through self-assembly of proteins and/or nucleic acids [77]. The increased local concentration of macromolecules induces the biophysical liquid-liquid phase separation (LLPS) of these condensates as liquid droplets. Several functions are attributable to MLOs including acting as a reaction crucible, sequestering biomolecules to negatively regulate processes, and serving as an organizational hub for chromatin [78]. For instance, promyelocytic leukemia protein (PML) NBs facilitate the epigenetic silencing of DNA viruses as an antiviral mechanism to prevent viral replication [79]. PML NBs are multi-protein complexes nucleated by the PML protein, which recruits other proteins such as Death Domain Associated Protein (DAXX), ATP-Dependent Helicase, X-linked (ATRX), Sp100, and Small Ubiquitin Like Modifier 1 (SUMO) [80]. In herpes simplex virus 1 (HSV-1) infections, PML NBs rapidly associate with the incoming HSV-1 genome where the DAXX-ATRX histone chaperone complex deposits the histone variant H3.3 bearing the repressive H3K9me3 mark to repress viral transcription [81]. Similarly, DAXX also colocalizes with LNBs during de novo KSHV infection [82]. However, the PML protein does not colocalize with LNBs suggesting that LNBs are distinct from PML NBs.

Recent biophysical characterization of LNBs suggests that LNB formation is dependent on LLPS much like other MLOs [83]. 1,6-hexanediol (1,6-HD) disrupts weak hydrophobic interactions and dissolves condensates, thus impairing the formation of LLPS-dependent structures [84]. Treatment of BCBL-1 and infected iSLK cells with 1,6-HD dramatically decreases the number of cells with LNBs and colocalized ATRX without affecting LANA expression or LANA binding to TRs [83]. Moreover, chromatin conformation capture (3C) analysis revealed that 1,6-HD-treatment modestly disrupted the LANA-dependent loop between the LANA promoter and the TR, but not the latency locus-RTA loop, suggesting that LLPS may impact KSHV genome conformation. Despite the impact on latent chromatin structure, 1,6-HD-treatment does not induce aberrant lytic gene transcription.

Interestingly, LNBs form different patterns of speckling during latency and chemically-induced reactivation in iSLK cells [83]. LNBs in lytic cells increase in size and fluorescence intensity, and rearrange to form larger, ring-like superstructures. Confocal imaging analyses also showed that the spatial interactions of the viral DNA with LANA transitions from compact colocalized structures into larger LNBs with viral DNA intertwined during the latency-to-lytic switch. Surprisingly, DAXX is evicted from LNBs during reactivation, suggesting that DAXX contributes to the epigenetic control of the latent chromatin. In line with this, knockdown of DAXX in latent iSLK cells decreases latency-gene transcription, however it is important to note that DAXX depletion does not affect latency or lytic gene expression in BCBL-1 cells [82]. With this discrepancy in mind, the major role of DAXX recruitment to LNBs in KSHV latency remains unresolved.

The reorganization of LNBs during the latency-to-lytic switch suggests that the LNB ultrastructure is flexible and mediated by LLPS. These larger lytic LNBs may function to create replication compartments that concentrate high amounts of DNA replication and transcription factors required to support reactivation. Such replication compartments were previously observed that included both viral proteins, RNAs, and cellular factors [85–89]. It is therefore reasonable to suggest that the dynamic nature of the LNBs allows the viral genome to rapidly adapt to environmental cues to enable switching between the latency and lytic replication programs.

Genome Partitioning

Because the KSHV genome does not integrate and is retained within an infected cells as an extrachromosomal plasmid during latency, a critical issue for the virus is how to ensure that viral plasmids are properly segregated in dividing cells and not lost. Innovative work by Sugden and colleagues to enable real-time imaging of live cells provides insight to this process [90]. This was accomplished using a KSHV BAC36 (KLacO) derivative and a smaller plasmid replicon system (miniKLacO) that contained 250 tandem repeats of the lactose operator lacO [91] to enable plasmid visualization when introduced in cells expressing a Lac repressor protein (LacI) fused to a tdTomato red fluorescent protein [92]. Imaging of the individual TdTomato puncta from KLacO or miniKLacO showed a wide range of signal intensities that vary up to 100-fold. This contrasts to puncta observed using a similar EBV LacO system in HEK293 cells, which exhibited uniform signal intensities [93]. Superresolution structured illumination microscopy (SIM) was not able to resolve the KSHV clusters to individual components. However, absolute quantification of viral copy numbers in PEL cell lines showed 30% more genomes per cell than total KSHV foci as detected by fluorescence in situ hybridization. EBV plasmids, on the other hand, had the same number of genome copies and visible foci suggesting that each punctate is a single plasmid. These data suggest that the KSHV puncta are comprised of clustered plasmids while EBV puncta contain single genomes.

The difference in clustering observed for KSHV and EBV DNAs is dependent on how their respective viral protein tethers LANA and EBNA1 attach to chromosomes. EBNA1 uses AT-hooks at its NTD to directly tether to host DNA while the LANA NTD binds to histones and other chromatin-binding proteins. By transfecting miniKLacO in SLK cells expressing a chimeric LANA with the EBNA1 AT-hook instead of its histone binding domain, the intensities of the miniKLacO puncta became more uniform and like those of EBV DNA, which strongly suggests that they did not cluster. A model was proposed in which clustering of KSHV DNA is mediated by LANA acting as a molecular “zipper” that brings together two viral plasmids at a single TR – one bound by the LANA NTD through interactions with the viral DNA-associated histones, and another through the LANA DBD. This end-to-end joining together with LANA oligomerization may contribute to the formation of LNBs.

While KSHV and EBV are related gammaherpesviruses that share much common biology, they employ distinct strategies for plasmid segregation. EBV follows a quasi-faithful partitioning of the viral genomes ensuring that almost all daughter cells receive plasmids. KSHV DNA form different sizes of clusters due to the interactions with LANA; and each cluster segregates as a unit leading to uneven distribution of DNA copies (Figure 2). This asymmetrical segregation may appear, on the first glance, a counterintuitive mechanism for viral persistence. However, Sugden and colleagues propose that this can be advantageous in establishing latency faster due to more viral DNA accumulating in some infected cells, with the expense being that viral DNA is lost in some cells [90]. Despite these trade-offs and differences, the unique mechanisms employed by both EBV and KSHV to maintain latency are clearly effective given the large number of humans harboring these viruses.

Figure 2.

In latently infected cells, LANA functions to tether the KSHV genome to host chromosomes to facilitate genome segregation. The binding of LANA to the viral terminal repeats (gray box) induces the oligomerization and formation of LANA aggregates often referred to as LANA nuclear bodies (LNBs). Formation of these higher-order structures lead to the uneven clustering of the viral plasmids. During cell division of latently infected cells, the viral genomes segregate as a cluster. Thus, daughter cells receive unequal copies of the KSHV genome.

Conclusions

Central to the persistence of KSHV is the multifunctional roles played by the major latency protein LANA at different stages of latency. At early stages of infection, LANA is required to recruit cellular chromatin modifiers and remodelers to organize the latency TRDs and prevent lytic gene transcription. Upon establishment of latency, LANA maintains viral chromatin structure and protects the integrity of the viral genome. Finally, LANA is critical for latent replication and plasmid segregation. In all of these stages, the formation of higher order structures is integral to the function of LANA. In this review, we discussed recent studies that shed light on the molecular mechanisms of KSHV latency. However, major questions remain that warrant future investigation to better understand KSHV biology.

First, structural studies on the viral tether have resulted in different models for how LANA functions in this role. While the LANA-DNA structures from crystallography and superresolution microscopy agree that LANA forms complex higher order structures with the viral DNA and chromatin, specific details differ including the number of bound LANA molecules, their correct orientation, and occupancy of the LBSs. Such details directly impact how we understand the functions of LANA in chromatin remodeling and genome partitioning. Future studies will need to overcome several complicating factors, including the presence of intrinsically disordered regions in LANA, cellular proteins that may participate in stabilizing these structures, and the heterogeneity of LANA oligomers. It is possible that these challenges can be addressed using cryo-electron microscopy approaches that allow near-native structure determination of biomolecular complexes in infected cells, as well as facilitate observation of intermediate structures.

A second major question raised stems from the differences in the structures of the latent KSHV chromatin in PEL cells and SLK cells – how does infecting different cell types for latency lead to unique regulation of the latency transcription program? While there are common features of latency used across multiple cell types, it is clear that additional mechanisms specific to discrete cell lineages are used by KSHV. Understanding what these lineage-specific factors are and how they contribute to the organization of the viral chromatin may offer insight into how KSHV establishes latency in such a wide range of cellular niches, eventually influencing the development of multiple diseases that reflect the specific cell types involved.

Finally, the influence of cell-extrinsic factors that influence latency should be considered. Because KSHV intimately evolved with the human host, the microenvironment of the site of the infection likely contributes to viral processes, especially latency. It is interesting to note that the anatomical site of tumor implantation alters both cellular and viral gene expression in PEL cell lines when xenografted in mice [94]. Kaye and colleagues recently demonstrated that mononuclear phagocytes promote the survival, proliferation and differentiation of KSHV-infected primary B cells [95]. This study suggests a mechanism by which KSHV reprograms infected monocytes to secrete chemoattractants and pro-survival cytokines to promote the infection of recruited B cells. Nevertheless, more work is needed to tease out this proposed mechanism. Advances in in vivo models of KSHV infection and pathogenesis will serve as useful tools to understand cell-extrinsic factors that promote latency.