Immune evasion by gammaherpesvirus genome maintenance proteins

EBV: Epstein–Barr nuclear antigen 1 (EBNA1).

The GMP of EBV is encoded by the latent antigen EBNA1, which is expressed along with seven other viral antigens during EBV latent infection (Kieff & Rickinson, 2007). The initial productive infection with EBV occurs within the oropharynx, involving either epithelial cells or B lymphocytes. EBV subsequently establishes a latent infection within the B-lymphocyte population. Expression of the eight viral latent antigens, six nuclear antigens (EBNA-1, -2, -3A, -3B, -3C and -LP), along with two latent membrane proteins (LMP1 and LMP2), leads to the transformation and proliferation of the infected B cell. This can be mimicked in vitro and such transformed B lymphocytes are termed B-lymphoblastoid cell lines (B-LCLs). This pattern of EBV latent infection has been termed ‘latency III’ (or the growth programme) (Thorley-Lawson & Gross, 2004; Rickinson & Kieff, 2007). As this initial proliferative infection is brought under control by the host immune response, some of the EBV-infected B cells survive to become long-lived peripheral blood memory B cells, carrying the EBV genome as an episomal element. In this form of latency, termed ‘latency 0’ (or the latency programme), all viral gene expression is switched off (Thorley-Lawson & Gross, 2004; Rickinson & Kieff, 2007). However, it is hypothesized that EBNA1 will be present within these cells to ensure binding of the viral episome to host chromosomes. Indeed, it has been shown that EBNA1 can bind to host chromosomes within cells in interphase (Nayyar et al., 2009). When these EBV-positive memory B cells undergo cell division, EBNA1 transcription is initiated and EBNA1 protein is expressed, so that when the viral episome is replicated and partitioned to daughter cells, sufficient EBNA1 protein is available to ensure that new viral episomes are tethered to the host-cell chromosomes within these cells (Hochberg et al., 2004). This form of latency is termed ‘latency I’ (or the EBNA1-only programme) (Thorley-Lawson & Gross, 2004; Rickinson & Kieff, 2007). Differential expression of EBNA1 within each form of EBV latency is determined by epigenetic events controlling alternative promoter usage (Schaefer et al., 1997; Hutchings et al., 2006; Day et al., 2007).

The genome-maintenance functions of EBNA1 have been well-documented, and EBNA1 could be considered the prototypic GMP. For the B95.8 strain of EBV, EBNA1 is a 641 aa protein consisting of small N-terminal and larger C-terminal domains, separated by a 239 aa repeat domain consisting of glycine (G) and alanine (A) (GAr domain) (Fig. 1⇑) (Baer et al., 1984). Within different viral isolates, the GAr domain is conserved but varies in size (Falk et al., 1995). Its role in maintaining the viral episome was first deciphered in the early 1980s (Yates et al., 1985). EBNA1 binds via its C-terminal domain to a region in the viral genome termed OriP, the viral origin of replication. This region contains two elements, a family of repeats (FR element) consisting of 20 high-affinity 30 bp EBNA1-binding sites linked to four low-affinity binding sites within a region termed the dyad symmetry (DS) element (Rawlins et al., 1985; Reisman et al., 1985). This interaction is important for initiating replication of the viral episome and is vital for episomal maintenance. EBNA1, while bound to the viral genome, also attaches to host-cell chromosomes. This attachment is mediated by interaction of EBNA1 with host-cell chromosome-associated proteins, including histone H1 and EBNA1-binding protein 2 (Marechal et al., 1999; Kapoor et al., 2005). However, an alternative model where there is direct interaction between the GMP and chromosomal DNA via an AT-hook motif has also been proposed (Sears et al., 2004); see Fig. 1⇑ for a schematic of the structural features of EBNA1 that are important for genome maintenance.

The immune response to EBV latent infection has been studied intensively, in particular the determination of how latently infected cells are able to avoid detection and elimination by host CD8⁺ CTLs. Control of the proliferative latent infection (latency III) is the result of a vigorous CTL response targeting mainly the EBNA3 family of proteins (Rickinson & Moss, 1997). The downregulation of these antigens in other forms of latency allows EBV-infected cells to escape this aspect of the EBV-specific CTL response. However, given the critical role of EBNA1 during latency, recognition of EBNA1 by CD8⁺ CTLs has been of longstanding interest. In a mouse tumour-rejection model, introduction of EBNA1 into a non-immunogenic carcinoma cell line failed to confer immunogenicity, whereas the equivalent experiment with LMP1-transfected carcinoma cells successfully induced a rejection response (Trivedi et al., 1991), suggesting that EBNA1 may not be a strong immunogen for CD8⁺ CTLs. These observations were reinforced by studies in humans. Independent reports analysing the fine specificity of the CD8⁺ T-cell response to EBV latent antigens failed to detect any responses to EBNA1 (Khanna et al., 1992; Murray et al., 1992). However, at the same time, using computer algorithms to predict potential human leukocyte antigen (HLA)-binding peptides and in vitro HLA peptide-binding assays, it was shown that EBNA1 contained a number of peptides capable of binding to HLA class I molecules (Stuber et al., 1995). This led to the hypothesis that the EBNA1 protein was protected from recognition by CD8⁺ CTLs and, amongst the potential mechanisms suggested, it was postulated that the unique GAr domain might in some way prevent the generation of antigenic peptides (Fig. 2⇑). In seminal work from the Masucci group (Levitskaya et al., 1995), this was shown to be the case. Lacking a natural EBNA1 CD8⁺ T-cell epitope, they introduced a model T-cell epitope into the backbone of EBNA1, with and without the GAr domain, and looked at presentation to epitope-specific CTLs in standard 5 h cytotoxicity assays. The GAr domain was shown to act in cis to prevent presentation of antigenic peptide via the MHC class I pathway to CD8⁺ CTLs (Levitskaya et al., 1995). It was subsequently shown that the GAr domain conferred enhanced stability to the EBNA1 protein by preventing proteasome-mediated degradation (Levitskaya et al., 1997). Thus, it was proposed that, by inhibiting proteasome function, the GAr domain prevented the generation of antigenic peptides (Fig. 3⇓). The ability of the GAr domain to inhibit proteasome-mediated degradation has been well-characterized (Sharipo et al., 1998; Heessen et al., 2002, 2003; Hoyt et al., 2006). However, a more recent report by Daskalogianni et al. (2008), comparing the turnover of EBNA1 protein with that of EBNA1 deleted of the GAr domain in cycloheximide pulse–chase experiments, suggested that the stability of EBNA1 was not related to the presence of the GAr domain. However, this requires further experimental support.

Despite the apparent protection of EBNA1 from processing and presentation by the MHC class I pathway, it was subsequently found that EBNA1-specific CD8⁺ CTLs were present in EBV-seropositive individuals. By conventional in vitro stimulation of peripheral blood mononuclear cells with autologous EBV-transformed B-LCLs, EBNA1-specific T-cell clones were identified by screening against EBNA1 deleted of the GAr domain (Blake et al., 1997). The EBNA1 peptide epitopes and HLA-restricting elements were mapped, confirming that these were MHC class I-restricted responses. However, in standard cytotoxicity assays, these clones failed to recognize either the autologous B-LCL or full-length EBNA1 when overexpressed in HLA-matched target cells, supporting the role of the GAr domain in protecting EBNA1 from recognition by T cells. It was suggested that these EBNA1-specific clones were primed in vivo by the MHC class I cross-priming pathway, which could process and present GAr-positive EBNA1 efficiently (Blake et al., 1997). It was proposed that these cross-primed EBNA1-specific T cells would not recognize endogenously expressed EBNA1, allowing EBNA1-positive cells to remain hidden from immune surveillance. However, although this may indeed be the mechanism of induction of such responses in vivo, it turns out that these T cells are able to recognize EBNA1-positive cells, albeit inefficiently.

Based on the premise that EBNA1-specific responses were present in EBV-seropositive individuals, ELISPOT assays were used to re-examine the CD8⁺ T-cell response to EBNA1. Responses were found in a higher number of individuals than had previously been thought and, in some cases, these responses were found to be the immunodominant response (Subklewe et al., 1999; Blake et al., 2000). The identification of strong EBNA1-specific CD8⁺ T-cell responses led to the re-evaluation of EBNA1 as a target for CTLs. Three independent studies have subsequently shown that CD8⁺ T-cell epitopes present in EBNA1 can be presented via the endogenous MHC class I pathway. Using gamma interferon (IFN-γ) release and extended cytotoxicity assays, Voo et al. (2004) were the first to report recognition of endogenously expressed EBNA1. Significantly, they reported a small increase in recognition if the GAr domain was removed. They made a second important observation when they showed that new protein synthesis was required for effective recognition of EBNA1. This was the first evidence that CD8⁺ T-cell epitopes from EBNA1 were derived from DRiPs and was in support of work from the Fahraeus group, which had initially implicated the importance of DRiPS for EBNA1 by biochemical studies (Yin et al., 2003; see later). This was followed by two further reports supporting these observations. In a study re-examining recognition of B-LCLs by EBNA1-specific T cells, Lee et al. (2004) used paired cell lines carrying either EBNA1 or a GAr-deleted EBNA1 as target cells. These experiments measuring IFN-γ release showed that EBNA1 was naturally processed within a B-LCL and presented to CD8⁺ T cells. Again, recognition was lower than for the matched cell line carrying the GAr-deleted EBNA1. In long-term co-culture assays, this recognition of B-LCLs by EBNA1-specific T cells led to a strong and specific inhibition of B-LCL outgrowth in vitro. This did not appear to be influenced by the GAr domain (Lee et al., 2004). The Khanna group have also shown endogenous processing of EBNA1 and, again, GAr-deleted EBNA1 was presented at a slightly higher level (Tellam et al., 2004). They also provided support for DRiPs as the source of EBNA1 antigen.

Taken together, these studies suggested that the GAr-mediated protection was not as robust as first thought and that EBNA1 could be targeted by CD8⁺ CTLs. However, there was general agreement that deletion of the GAr domain leads to an increase in recognition, implying a protective role for the domain. However, in long-term outgrowth assays, this protection was not sufficient to protect EBNA1-expressing B-LCLs from recognition and destruction by EBNA1-specific CD8⁺ T cells. These data also suggested that DRiPs derived from the translation of EBNA1 mRNA and not the degradation of stable EBNA1 protein were the source of antigenic peptide. These results sit well with the mechanism of long-term EBV latency. In the resting B cell, EBV gene expression is silent (i.e. latency 0), but EBNA1 protein is present, tethering the viral episome to the host chromosome. As mature EBNA1 protein is not processed for presentation to CD8⁺ CTLs and is an extremely stable protein, due to the GAr domain preventing proteasome-mediated degradation, no EBNA1 expression is required and this cell will remain hidden from virus-specific CTLs. However, when these cells are stimulated to divide, EBNA1 gene expression is switched on (i.e. latency I) and the latently infected cells will become a target for CD8⁺ CTLs, due to the production of EBNA1-derived DRiPs. This will prevent uncontrolled expansion of EBNA1-expressing cells, as seen in the outgrowth assay. However, in vivo it is likely that this expansion of memory B cells is very brief and that they soon revert back to resting cells, in which EBNA1 expression is turned off. Thus, there is a small period where EBNA1 is being expressed and the cell becomes susceptible to EBNA1-specific CTLs. However, EBV has another trick up its sleeve that allows these cells to evade efficient detection. The Fahraeus group (Yin et al., 2003) showed that the GAr domain also functions to delay the generation of DRiPs by reducing the translation rate of EBNA1 mRNA. Starting from the hypothesis that the GAr-mediated inhibition of proteasomal degradation would mean that, over time, EBNA1 protein would accumulate in cells, they tested this by simply transfecting target cells with appropriate plasmid constructs and analysing steady-state protein levels. The observed result disproved their initial hypothesis, as steady-state EBNA1 protein levels were significantly lower than those of EBNA1 deleted of the GAr domain. This was not an effect at an RNA stage, as mRNA levels for each transcript were equivalent. After further biochemical analysis, they determined that the GAr domain functions in cis to inhibit self-synthesis of EBNA1 via a reduction in the mRNA translation rate (Yin et al., 2003). This correlates with a reduction in recognition of EBNA1 by CD8⁺ CTLs (Fig. 3⇑) (Yin et al., 2003; Tellam et al., 2007a). This evasion strategy will allow EBV to express the GMP as cells divide, without the CD8⁺ T-cell response being able to eliminate all EBNA1-expressing cells. When some cells revert back to the resting stage, EBNA1 expression will be switched off and the cells will again be invisible to the EBV-specific immune response, thus ensuring a relatively constant population of long-term latently infected cells.

The mechanism by which the GAr domain reduces the rate of translation is unclear. A study by the Khanna group has implicated the EBNA1 mRNA as the mediator of this effect (Tellam et al., 2008). The EBNA1 mRNA sequence has an extensive purine bias within the GAr domain-encoding region, in contrast to the normal mammalian codon usage for these amino acids. Analysis of the secondary structure of the GAr-encoding sequence revealed a lack of any real consensus structure. However, replacing the third base position purines with pyrimidine nucleotides increased the stem–loop structure. This also greatly increased the rate of EBNA1 translation both in vitro and in vivo, and resulted in enhanced presentation of EBNA1 to CD8⁺ CTLs. Such an immune-evasion strategy based on codon usage has been proposed previously (Cristillo et al., 2001). Their predicted model is that the lack of any real secondary structure across the GAr domain causes an inhibition of elongation rate. This model has been questioned by a recent study claiming that the nascent GAr peptide delays the assembly of the initiation complex on its own mRNA (Apcher et al., 2009).

The final piece of weaponry that EBV has to protect EBNA1 from the immune response is at the level of transcription. During latency I, EBNA1 autoregulates its own transcription (Sample et al., 1992). In a more recent study, this autorepression of EBNA1 expression is shown to be the result of an inhibition of pre-mRNA processing: EBNA1 acts either post- or co-transcriptionally to prevent processing of the primary EBNA1 transcript (Yoshioka et al., 2008). It is proposed that this autoregulation would control unnecessary new synthesis of EBNA1 protein. In turn, this would subsequently limit the generation of EBNA1-derived DRiPs, so contributing to the overall immune-evasion strategy of EBNA1 (Fig. 3⇑).

CD4⁺ T-cell recognition of EBNA1 has also been investigated. It has been shown that EBNA1 protein can be processed and presented to CD4⁺ T lymphocytes (Leen et al., 2001). Importantly, recent studies show that endogenous EBNA1 can be loaded onto MHC class II molecules via an intracellular mechanism, perhaps autophagy (Paludan et al., 2005; Mackay et al., 2009). The GAr domain appears to have little influence on the processing of CD4⁺ T-cell peptides from EBNA1. However, the recognition of EBNA1 by CD4⁺ T cells was found to be an inefficient process, requiring a long period for processing and presentation of EBNA1 protein (in comparison to CD8⁺ recognition) (Mackay et al., 2009). Nevertheless, these data may have important implications for T-cell recognition of the latently infected cell, which has switched off gene expression but remains GMP-positive. These studies are in their infancy and require further detailed experimental analysis.

Other lymphocryptoviruses.

LCVs closely related to EBV have been identified in both Old and New World primates. The establishment of LCV-infected B-LCLs has allowed complete or partial sequencing of the genomes of these viruses, and the LCV genes have been compared with their EBV counterparts (reviewed by Wang et al., 2001). The most extensively studied are the LCVs of baboons and rhesus monkeys: Papiine herpesvirus 1 (herpesvirus papio; baLCV) and Macacine herpesvirus 4 (rhesus lymphocryptovirus; rhLCV), respectively. Gene homologues of EBV EBNA1 have been identified in both viruses (Table 1⇑). Baboon EBNA1 (baEBNA1) and rhesus EBNA1 (rhEBNA1), although being slightly smaller than EBV EBNA1 at 511 and 476 aa, respectively, compared with 641 aa for EBV EBNA1, overall each have a high degree of amino acid identity to EBV EBNA1 (Yates et al., 1996; Blake et al., 1999). Both baEBNA1 and rhEBNA1 contain internal amino acid repeats differing slightly from that present in EBV EBNA1, as in addition to having glycine and alanine, they also contain serine (Fig. 2⇑). baEBNA1 has a repeat domain consisting of seven perfect repeats of a GAGAGGS motif. rhEBNA1 contains four perfect repeats of a GAGGS motif preceded by three GAGGS repeats interspersed with 12 additional amino acids, forming seven glycine/alanine-rich repeats within a 47 aa stretch. In this review, these repeats will be termed the GSAr domains to depict the presence of serine amino acids within the repeat. The highest region of conservation is within the C terminus, and the functional properties important for genome maintenance appear well-conserved in both of these predicted LCV GMPs; both proteins have been shown to support EBV OriP-dependent plasmid maintenance and replication (Yates et al., 1996; Blake et al., 1999).

Their predicted role as GMPs and the conservation of GAr-like internal repeats suggested that the ba- and rhEBNA1 proteins would, similarly to EBV EBNA1, be protected from recognition by CD8⁺ CTLs. This was initially investigated by using a series of chimeric constructs in a manner analogous to that used to first describe the EBV EBNA1 GAr-mediated inhibition of antigen processing (Blake et al., 1999). The ba- and rhGSAr domains were engineered into the EBV EBNA1 sequence in place of the endogenous GAr. These constructs were introduced into both human and simian B-LCLs, and presentation of the endogenous EBNA1 CD8⁺ T-cell epitopes was investigated by conventional cytotoxicity assay. These experiments suggested that the ba- and rhGSAr domains did not prevent processing and presentation of these chimeric constructs to CD8⁺ CTLs. Similar data were generated in a parallel experiment using all simian-derived components (Blake et al., 1999), suggesting that, in contrast to the human LCV EBNA1, these simian LCVs did not employ a similar immune-evasion strategy for their GMP. Surprisingly, a follow-up study looking at immune response to rhLCV latent antigens generated contradictory data (Fogg et al., 2005). Ex vivo analysis of T lymphocytes from naturally or experimentally infected rhesus macaques at the New England Primate Research Center (Southborough, MA, USA) identified a strong response to rhEBNA1 in 11 of 23 animals. These responses were detected by using an IFN-γ ELISPOT assay after stimulation with full-length rhEBNA1. However, in cytotoxicity assays, polyclonal T cells expanded in vitro using autologous rhLCV B-LCL failed to recognize endogenously expressed rhEBNA1. This was despite the fact that putative rhEBNA1 peptide epitopes were mapped and used to confirm the presence of rhEBNA1-specific T cells within the polyclonal populations. Even after the isolation of rhEBNA1-specific T-cell clones by limiting-dilution analysis, there was no recognition of either the rhLCV B-LCL- or recombinant vector-expressed full-length rhEBNA1 in either cytotoxicity or IFN-γ-release assays. However, in the authors' words, there was ‘modest’ recognition of the GSAr-deleted rhEBNA1, suggesting that, in this system, the internal repeat was functioning to inhibit processing and presentation of rhEBNA1.

A more recent series of experiments has also addressed this issue and provides data in agreement with the first report by Blake et al. (1999). Expanding on work probing the effect of the rate of translation on T-cell recognition of EBNA1, Tellam et al. (2007b) used a biochemical approach to compare translation of simian and EBV EBNA1 proteins. Both rh- and baEBNA1 were translated at a higher rate than EBV EBNA1 and, unlike EBV EBNA1, deletion of the GSAr domains had no effect on translation rates. The authors inferred from these data that the rh- and baGSAr domains had no effect on self-protein synthesis and were unlikely to affect the generation of DRiPs. In support of this, the rhGSAr domain had no significant effect on the presentation of a model epitope from a chimeric rhEBNA1 in in vitro antigen-presentation assays and ex vivo T-cell stimulation assays (Tellam et al., 2007b). Studies with cycloheximide suggested that the source of antigenic peptide in these assays was newly synthesized protein.

It is difficult to reconcile the different results from these studies. It may simply be that the rhEBNA1-specific clones isolated were of low affinity such that, in both IFN-γ-release and cytotoxicity assays, they failed to recognize endogenously expressed rhEBNA1 and only marginally recognized the GSAr-deleted form. This contrasts directly with the EBV EBNA1 studies (see earlier). In the study by Fogg et al. (2005), no mechanism was put forward for the proposed inhibitory action of the rhGSAr domain and, based on the work of Tellam et al. (2007b), any mechanism is unlikely to be similar to that of the EBV GAr. The mechanism(s) for controlling transcription of simian EBNA1 genes appear analogous to those of EBV EBNA1 (Ruf et al., 1999). Thus, it would appear surprising that the simian EBNA1 proteins are not protected from CTL recognition, but this remains an open question that can only be resolved by further studies.

The first LCV from a New World monkey was identified when an EBV-related virus was isolated from spontaneous B-cell lymphomas of common marmosets; it was termed marmoset lymphocryptovirus or, more formally, Callitrichine herpesvirus 3 (CvHV3; Cho et al., 2001). Sequencing of the viral genome revealed a high degree of similarity to EBV and other Old World primate LCVs, with a large number of gene homologues (Cho et al., 2001; Rivailler et al., 2002). Open reading frame (ORF) 39 was found to have 39 % overall amino acid identity with EBV EBNA1, and 42 % identity within the C-terminal regions. Despite this high degree of sequence similarity, no significant GAr-like internal repeat was found in this protein. No functional studies have yet been carried out on ORF39, so it remains to be determined whether this is the CvHV3 GMP (Table 1⇑). However, based on the sequence similarity, particularly within the C terminus, this would seem likely. It also remains to be determined whether ORF39 has immune-evasion properties.

KSHV: LANA1.

In asymptomatic KSHV carriers, the virus establishes a latent infection in B lymphocytes and the viral genome persists within cells as a nuclear episome (Mesri et al., 1996; Renne et al., 1996). Understanding of KSHV latency is less advanced than that of EBV latency, although it is assumed that there will be a number of parallels between the two. KSHV gene expression is restricted within latently infected cells. In vitro studies using KSHV-infected PEL cells show that five genes are expressed in this model of latent infection. The major latent transcript is a tricistronic message encoding three genes: ORFs 71, 72 and 73 (Dittmer et al., 1998; Talbot et al., 1999). Two other transcripts have been identified, corresponding to the kaposin locus and ORF K10.5 (Sadler et al., 1999; Rivas et al., 2001). On the major latency transcript, ORFs 71 and 72 encode a viral homologue of cellular cyclin D (v-cyclin) and a viral inhibitor of FLICE (v-FLIP), respectively (Chang et al., 1996; Thome et al., 1997), whilst ORF73 encodes a large nuclear protein termed LANA1 (Kedes et al., 1997; Kellam et al., 1997; Rainbow et al., 1997). LANA1 derived from the PEL line BC-1 comprises 1162 aa (Russo et al., 1996) and consists of three distinct regions: N- and C-terminal domains of 337 and 240 aa, respectively, are separated by a large (585 aa) central repeat (CR) domain consisting mainly of the acidic amino acids glutamine (Q), glutamic acid (E) and aspartic acid (D). The CR domain comprises imperfect repeats of these amino acids and is subdivided further into three subdomains based on the amino acid content: CR1 (aa 30–442) consists of a DE-rich region, CR2 (aa 442–768) consists of a DEQ-rich region and CR3 (aa 769–935) consists of an EQL-rich region (Fig. 2⇑). Variation in the size of the central acidic domain contributes to the considerable variation in the sizes of LANA1 seen in different KSHV isolates (Gao et al., 1999).

Although LANA1 has little similarity to EBNA1, it has been identified as the GMP. Its role in episomal maintenance has been well-characterized. LANA1 associates with the viral genome in infected cells and co-localizes with the viral genome in interphase nuclei and on mitotic chromosomes (Ballestas et al., 1999; Cotter & Robertson, 1999). Studies using both KSHV-derived cosmids and a recombinant KSHV cloned in a bacterial artificial chromosome (BAC) have shown that LANA1 is essential for long-term maintenance (Ballestas et al., 1999; Cotter & Robertson, 1999; Ye et al., 2004). LANA1 binds preferentially to the terminal repeat (TR) region of the KSHV episome via its C-terminal domain (Ballestas & Kaye, 2001; Cotter et al., 2001) and, via its N-terminal domain, interacts with the host chromosome by association with histone H2A and H2B (Barbera et al., 2004, 2006). Although not yet formally proven, by analogy with EBV, it is predicted that similar forms of KSHV latency occur in asymptomatic individuals, where LANA1 is likely to be present within resting memory lymphocytes but not expressed, and where LANA1 alone is expressed as the resting latently infected cells divide. Thus, it was predicted that LANA1 would be protected from recognition by CD8⁺ CTLs and that the acidic CR domain would play a functional role. A number of studies using ELISPOT analysis and T-cell proliferation studies have identified KSHV-specific CD8⁺ T-cell responses targeting LANA1 (Brander et al., 2002; Woodberry et al., 2005; Bihl et al., 2007). So, like EBNA1, LANA1 is recognized by CD8⁺ T cells in vivo. However, no-one has yet isolated these LANA1-specific T cells in vitro and tested them for recognition of endogenously expressed LANA1. Presentation of endogenously expressed LANA1 to CD8⁺ CTLs, and the effect of the acidic CR domain, have been investigated by using chimeric LANA1 constructs containing a model T-cell epitope (Zaldumbide et al., 2007). This revealed that the acidic repeat domain of LANA1 did block presentation of CD8⁺ T-cell epitopes. The LANA1 repeat functioned in cis and also contributed to the stability of LANA1 (Fig. 3⇑) (Zaldumbide et al., 2007). A more biochemical approach to investigate the properties of the acidic CR domain was used by the Moore group (Kwun et al., 2007). As with the EBNA1 GAr domain, the acidic repeat was shown to reduce the translation rate of LANA1. The effector region was mapped to a peptide sequence spanning the junction of the CR2 and CR3 subdomains (Fig. 2⇑). The CR2CR3 region was also shown to stabilize proteins by inhibiting proteasomal processing (Kwun et al., 2007). Thus, it was proposed that the acidic CR domain has similar immune-evasion functions to the GAr domain of EBNA1 (Fig. 3⇑). Retardation of translation will interfere with production of DRiPs, and stabilization of LANA1 will maintain the presence of the protein in resting latently infected cells.

Although there were clear functional similarities to the EBNA1 GAr domain, intriguingly there were also a number of areas in which the LANA1 acidic domain behaved differently. When looking at the ability of the domain to act in cis or in trans, surprisingly Kwun et al. (2007) found that it could also block translation in trans and, in their hands, this was also the case for the EBNA1 GAr domain. However, this required high levels of expression for both LANA1 and EBNA1 in these in vitro experiments and so may not reflect the in vivo mechanisms accurately. The CR2CR3 domain was also found to retard translation when fused to a heterologous protein at both the N- and C-terminal regions, contrasting with the EBNA1 GAr domain, which was shown to retard translation only when located towards the N terminus (Yin et al., 2003; Kwun et al., 2007). The mechanism by which the central acidic domain influences the translation rate remains to be determined. However, the effect is mediated by the amino acid sequence rather than at the nucleotide level, as introduction of a stop codon between the CR2 and CR3 subdomains negates the inhibitory effect (Kwun et al., 2007). This is a significant observation, as although there is little amino acid similarity between the repeat domains of LANA1 and EBNA1, it is interesting to note that the mRNA sequences are very similar. Interestingly, a frameshift in the EBNA1-encoding sequence switches from a repeat domain consisting of the amino acids G and A to a repeat of amino acids G, Q and E that shares approximately 65 % similarity with the LANA1 repeat, and can inhibit the presentation of CD8⁺ T-cell peptides (Ossevoort et al., 2007). This is suggestive of a common origin for the individual nucleotide repeat sequences, although they may have diverged translationally and mechanistically.

HVS: ORF73.

HVS naturally infects squirrel monkeys and establishes an asymptomatic latent infection in T lymphocytes, although infection of other New World primates can lead to acute T-cell lymphomas (reviewed by Fickenscher & Fleckenstein, 2001). HVS is also capable of transforming T lymphocytes of humans and rabbits (Medveczky et al., 1989; Biesinger et al., 1992). HVS strains are classified based on their pathogenic potential and sequence divergence within the left-hand TR region of the genome into three subgroups: A, B and C (Medveczky et al., 1984). The transforming properties of HVS subgroup A are due to a single protein, the saimiri transforming protein A (stp-A), whereas subgroup C viruses encode a homologue of this protein termed stp-C and a second transforming protein termed TIP (Biesinger et al., 1990; Jung & Desrosiers, 1991; Jung et al., 1991). Immortalized T cells contain the viral genome as a multi-copy episome, without production of virus particles (Werner et al., 1977; Kaschka-Dierich et al., 1982). Initially latently infected T cells were shown to express the transforming genes, although subsequent analysis revealed that they were dispensable for long-term episomal maintenance (Fickenscher et al., 1996, 1997; Duboise et al., 1998). Using a lung carcinoma cell line stably maintaining HVS DNA as a non-integrating episome as an in vitro model of HVS latency, gene expression was shown to be limited to ORFs 71, 72 and 73 (Hall et al., 2000). These genes are expressed as a polycistronic message and, as with KSHV, ORF71 encodes a v-cyclin D (Chang et al., 1996), ORF72 encodes a vFLIP (Thome et al., 1997) and ORF73 encodes the GMP (Collins et al., 2002; Verma & Robertson, 2003; Calderwood et al., 2004). The GMP from the prototype HVS strain A11 encodes a 407 aa protein. Like most GMPs, it comprises three distinct domains, with a large repeat region flanked by small N-terminal and larger C-terminal domains (Fig. 2⇑). With the exception of the C-terminal domain, there is little similarity to KSHV LANA1 or other GMPs. However, the HVS ORF73 internal repeat domain is analogous to that of KSHV LANA1, being composed of acidic amino acids. In this case, it consists of two distinct elements, a glutamic acid- and glycine-rich (EG-rich) region linked to a glutamic acid- and alanine-rich (EA-rich) region (Fig. 2⇑). In the A11 and C488 strains, respectively, the EG-rich region is 15 and 111 aa, whilst the EA-rich region is 147 and 132 aa (Albrecht et al., 1992; Ensser et al., 2003; Verma & Robertson, 2003). HVS ORF73 has been shown to have similar genome-maintenance properties to KSHV LANA1 and EBV EBNA1. A recombinant HVS BAC lacking ORFs 71–73 failed to persist within dividing cell populations, and episomal maintenance could be rescued by reintroduction of ORF73 (Calderwood et al., 2005). HVS ORF73 binds to cis-acting DNA elements in the TR region of the virus genome via its C-terminal domain and this association is essential for episomal persistence (Collins et al., 2002; Verma & Robertson, 2003; Calderwood et al., 2004). HVS ORF73 also interacts with host-cell mitotic chromosomes, mediated by interaction with the host-cell proteins methyl-CpG-binding protein 2 and histone H1 (Griffiths & Whitehouse, 2007; Griffiths et al., 2008).

The important role played by HVS ORF73 suggested that it too would be protected from recognition by CD8⁺ CTLs. Perhaps not unexpectedly, the large acidic repeat of HVS ORF73 was shown to have immune-evasion properties. By using chimeric HVS ORF73 constructs containing a model T-cell epitope, it was shown that the presence of the acidic repeat reduced presentation of peptides to CD8⁺ CTLs (Gao et al., 2009). Steady-state levels of ORF73 protein are reduced due to the presence of the acidic repeat; however, analysis of the mechanisms responsible revealed some critical differences with respect to the mechanism of action of the HVS acidic repeat, compared with both KSHV LANA1 and EBV EBNA1 repeats. The stability of HVS ORF73 is not influenced by the acidic repeat and, importantly, neither does the acidic repeat have any inhibitory effect on the rate of ORF73 mRNA translation. In contrast, it was shown that steady-state levels of ORF73 mRNA were reduced due to the presence of the repeat (Fig. 3⇑). Comparison of the ORF73 sequences from the A11 and C488 strains of HVS highlighted the EA region as the most conserved element of the repeat. A single copy of the motif EEAEEAEEE, present multiple times in the EA-rich region of both strains, was sufficient to inhibit antigen presentation and to reduce ORF73 mRNA levels, and an additive effect was seen as the number of motifs was increased to six (Gao et al., 2009). It is proposed that the reduction in mRNA levels will reduce the total amount of ORF73 translated, rather than translation rates per se, and thus will also influence the generation of ORF73-derived DRiPs. It was predicted that the HVS ORF73 acidic repeat functions either by destabilizing ORF73 mRNA or by reducing the transcription rate of the ORF73 gene. Surprisingly, we found that the presence of the acidic domain actually acts to stabilize the mRNA (Gao et al., 2009). To maintain steady-state levels, it must also act to control the rate of accumulation, i.e. transcription rate, although this has yet to be formally proven. The inhibitory action of the acidic domain is mediated via the nucleotide sequence rather than the amino acid sequence (E. Callery & N. Blake, unpublished data). However, it is likely to be distinct from that proposed for the EBNA1 GAr domain, which was shown to have no effect on mRNA levels and not to influence the stability of EBNA1 mRNA (Yin et al., 2003; Tellam et al., 2008). It is postulated that the HVS ORF73 nucleotide domain functions at two levels. Firstly, it is proposed that the nucleotide domain acts as a target of an RNA-binding protein that has mRNA-stabilization properties (Garneau et al., 2007). Secondly, it is predicted that the domain acts to repress self-transcription. Clues as to how the repeat-encoding domain may control transcription rates come from work on the inherited disease Friedreich's ataxia (Cossee et al., 1997). This disorder is due to expansion of the trinucleotide repeat GAA within the frataxin gene, which causes low levels of transcription of the gene (Pandolfo, 2002). Mechanistically, this is thought to be due to either transcriptional pausing or induction of heterochromatin regions within the gene (Krasilnikova et al., 2007; Soragni et al., 2008). The long stretches of glutamic acid within the repeat domain of HVS ORF73 are encoded exclusively by the GAA codon and, thus, this sequence may mimic the expanded repeat present in the disease-associated frataxin gene.

MuHV-4: ORF73.

The rhadinovirus MuHV-4 is related more closely to KSHV than to EBV, but has been used widely as a small-animal model of general gammaherpesvirus infection and pathogenesis. In a mouse model, intranasal inoculation with MuHV-4 leads to an acute, self-limiting infection in epithelial cells of the lung, followed by establishment of latent infection in the lungs and lymphoid organs. Long-term latency is established in B cells, although macrophages, dendritic cells and epithelial cells are also recorded as sites of latent infection. Similarly to EBV, latent infection is characterized by expansion of latently infected germinal centre B cells, followed by persistence in the resting memory B-cell population (Flano et al., 2002; Willer & Speck, 2003). As with other gammaherpesviruses, there is a reduction in viral gene transcription during latent infection (Virgin et al., 1999). For MuHV-4, the number of genes expressed varies depending on the cell analysed (Marques et al., 2003). Within the latently infected B-cell population, the profile of gene expression also varies, reminiscent of EBV latency. Amongst the predicted latent genes is MuHV-4 ORF73. Based on co-localization of ORF73 within the MuHV-4, KSHV and HVS genomes, this was predicted to be the GMP of MuHV-4. Within the latently infected B-cell populations, ORF73 is expressed in the germinal centre and other B-cell subsets (Marques et al., 2003). It has also been shown that there is different promoter usage for ORF73 during latent infection (Coleman et al., 2005; Allen et al., 2006). This all suggests that there is tight control of MuHV-4 ORF73 expression during latent infection. The in vitro functional properties of MuHV-4 ORF73 have not yet been investigated; however, in vivo models have shown it to be essential for the establishment and maintenance of latency. A MuHV-4 mutant virus containing a translation-termination codon within the coding region was found to have a severe defect in its ability to establish latency in the spleen of C57Bl/6 mice after intranasal infection (Moorman et al., 2003). A second study using two separate ORF73-deficient viruses, an ORF73-deletion mutant and a frameshift mutant, also demonstrated a severe latency defect after both intranasal and intraperitoneal inoculation of BALB/c mice (Fowler et al., 2003).

The MuHV-4 ORF73 protein encodes a 314 aa protein (Fig. 2⇑) with little overall similarity to KSHV LANA1 or HVS ORF73 (or EBNA1). There is a greater degree of similarity if the C-terminal regions alone are compared. However, significantly, MuHV-4 ORF73 does not encode an internal amino acid repeat sequence. Despite the lack of an internal repeat domain, in light of the importance of ORF73 in establishment and maintenance of latency in the mouse model, Bennett et al. (2005) predicted that there would be similar pressure on ORF73 to evade recognition by CD8⁺ CTLs. Again by using chimeric constructs, this group showed that, when a model epitope was fused to MuHV-4 ORF73, there was limited presentation of the peptide to CD8⁺ CTLs (Bennett et al., 2005). The region responsible for this inhibition was mapped to aa 170–220 (Fig. 2⇑). This sequence lies within the C-terminal half of MuHV-4 ORF73 that has similarity to the C-terminal domains of both KSHV LANA1 and HVS ORF73. These regions in the KSHV and HVS GMPs are important for genome maintenance and have not been implicated in immune-evasion properties. Bennett et al. (2005) showed that, in MuHV-4 ORF73, this amino acid sequence causes a reduction in steady-state levels of ORF73 protein and contributes to the stability of ORF73. Labelling cells at various time points after transfection, these authors showed that aa 170–220 cause a reduction in self-protein synthesis, although the effect on the rate of translation was not analysed (Bennett et al., 2005). However, it is predicted that this region causes a reduction in ORF73-derived DRiPs and thus acts as a cis immune-evasion element (Fig. 3⇑). The mechanism of action has not been deciphered. The authors speculated that it may act at the RNA stage, but this has yet to be investigated.

Perhaps the most important observation from this study came from an in vivo experiment analysing the effect of bypassing the cis-inhibitory domain and expressing an immunogenic peptide from the ORF73 mRNA. A mutant virus was constructed with three tandem CD8⁺ T-cell epitopes linked by an internal ribosome entry site to the C terminus of the ORF73 mRNA. These immunogenic peptides would be translated efficiently from this message, which would be transcribed as for a wild-type virus. Crucially, this led to a severe reduction in virus latency, mediated via an MHC class I-restricted and CTL-dependent mechanism. This was the first clear evidence for the importance of immune evasion by the GMP for establishment of gammaherpesvirus latency in vivo (Bennett et al., 2005).

Other rhadinoviruses.

The genomes of a number of other species of the genus Rhadinovirus have been sequenced completely or in part. ORF73 positional and sequence homologues have been identified in Ateline herpesvirus 3 (herpesvirus ateles strain 73), Bovine herpesvirus 4 (Movar virus) and Macacine herpesvirus 5 (rhesus rhadinovirus) (Table 1⇑). Ateline herpesvirus 3 ORF73 is a 447 aa protein that contains a 169 aa internal repeat consisting of the amino acids glycine, aspartic acid and glutamate (Albrecht, 2000). The ORF73 gene of bovine herpesvirus 4 has been sequenced and found to encode a small protein of 253 aa; no internal repeat motif was identified (Lomonte et al., 1995). The species Macacine herpesvirus 5 represents a number of rhadinovirus strains first found in rhesus macaques, but closely related viruses have also been identified in other macaque species (Damania & Desrosiers, 2001). ORF73 sequence data are available for a number of these viruses: rhesus macaque herpesvirus strains H26-95 and 17577 (Alexander et al., 2000; Searles et al., 1999), Japanese macaque herpesvirus strain JM153 (GenBank accession no. AY528864) and southern pigtail macaque herpesvirus strains M78114 and J97167 (Burnside et al., 2006). As discussed earlier, the rhadinoviruses of Old World primates have been split into two lineages (RV1 and RV2). The southern pigtail macaque herpesvirus strain M78114 is within the RV1 lineage, which includes KSHV. The ORF73 gene from this virus encodes a 1071 aa protein that contains a 521 aa repeat (consisting of glutamate and proline) (Table 1⇑). The other sequenced strains all belong to the RV2 lineage. The ORF73 genes encode proteins of 436–448 aa that do not contain internal amino acid repeat domains. There have been few functional studies on these ORF73 proteins. Rhesus macaque herpesvirus strain H26-95 ORF73 has been shown to bind the viral episome during latency and to be essential for establishment of latency within B cells (DeWire & Damania, 2005; Wen et al., 2009). There have been no studies investigating any immune-evasion properties of these proteins.