Burkitt lymphoma: the role of Epstein-Barr virus revisited

Authors

  • Sebastian Grömminger,

    1. Institute of Clinical Molecular Biology and Tumour Genetics, German Research Centre for Environmental Health, Helmholtz Zentrum München, Germany
    Search for more papers by this author
  • Josef Mautner,

    1. Clinical Cooperation Group, Paediatric Tumour Immunology, Children’s Hospital, Technische Universität München and Helmholtz-Zentrum München, Germany
    Search for more papers by this author
  • Georg W. Bornkamm

    1. Institute of Clinical Molecular Biology and Tumour Genetics, German Research Centre for Environmental Health, Helmholtz Zentrum München, Germany
    Search for more papers by this author

Professor Georg W. Bornkamm, Institute of Clinical Molecular Biology and Tumour Genetics, German Research Centre for Environmental Health, Helmholtz Zentrum München, Marchioninistr. 25, D-81377 München, Germany.
E-mail: bornkamm@helmholtz-muenchen.de

Summary

The particular epidemiological features of Burkitt lymphoma (BL) in Tropical Africa, first described by Denis Burkitt in 1958, initiated the search for a virus that induces malignant B cell lymphomas in humans and is transmitted by arthropods. The herpes virus (Epstein-Barr virus, EBV) discovered by Epstein and collaborators in cell lines established from BL biopsies fulfilled some of these predictions. It drives primary B cells into unlimited proliferation, induces malignant B cell lymphomas in immunocompromised individuals (post-transplant lympho-proliferative disease, PTLD) in vivo, and footprints of the virus are generally detected in African BL biopsies supporting a causative role of the virus in the pathogenesis of BL. The virus is, however, not transmitted by arthropods and is spread ubiquitously amongst the human population through saliva. Furthermore, BL and EBV-induced PTLD are now recognized as pathogenetically distinct entities: BL involves MYC-immunoglobulin translocations in contrast to PTLD, and different patterns of viral genes are expressed in both diseases. Viral gene products expressed in BL are assumed to contribute to inhibition of apoptosis, although their precise mechanism of action is not fully understood. In the future, next generation sequencing is expected to shed more light on the contribution of EBV to the pathogenesis of BL.

The human herpes virus discovered in Burkitt lymphoma cells is a ubiquitous virus and is the causative agent of infectious mononucleosis

The high incidence of Burkitt lymphoma in Equatorial Africa and the peculiar epidemiological features relating the incidence of the tumour to temperature and humidity (Burkitt, 1958, 1962) prompted Anthony Epstein and collaborators to search for a virus in BL specimens by electron microscopy. This finally led to the discovery of a novel herpes virus in cell suspensions that started to grow out as cell lines (Epstein et al, 1964). Contrary to the herpes viruses known at that time, this novel virus could not be transmitted to other cells and propagated, which severely hampered its further characterization. A major breakthrough came with the discovery by Gertrud and Werner Henle, that cells of BL cell lines, in which a small and variable proportion of cells produces viral particles, can be used as a source for viral antigens enabling the search for antibodies against the virus in individuals or patients in an indirect immunofluorescence assay (Henle & Henle, 1966). This paved the way for large-scale seroepidemiological studies that revealed a number of unexpected results:

  • 1 First of all, the virus (designated ‘Epstein-Barr virus’ [EBV] by the Henles) turned out to be a ubiquitous virus that was not restricted to Central Africa. Primary infection usually occurred in infants in a mostly asymptomatic fashion (Henle & Henle, 1970; Kafuko et al, 1972).
  • 2 In countries with high socioeconomic status, primary infection appeared to be delayed until adolescence or adulthood (Henle & Henle, 1970).
  • 3 The seroconversion of a co-worker, whose serum had served as a negative control, with symptoms of infectious mononucleosis initiated a large seroepidemiological survey among students at Yale University, which revealed that EBV is the causative agent of infectious mononucleosis (Henle et al, 1968).

EBV as a potential human tumour virus

Over the next years and decades, several laboratories made a number of observations that strengthened the notion that EBV is a potential human tumour virus:

  • 1 Patients with BL and with nasopharyngeal carcinoma (NPC), a tumour with high incidence in China, had particularly high antiviral antibody titres, suggesting high viral loads in these patients (Old et al, 1966).
  • 2 It was shown that the virus can infect primary human B cells and drive them into unlimited proliferation, a phenomenon called immortalization or growth transformation (Henle et al, 1967; Pope et al, 1968).
  • 3zur Hausen et al (1970) provided evidence that EBV DNA is present in biopsies of BL and NPC, and even in BL cell lines that failed to produce viral particles, which was reminiscent of Papova viruses in animal model systems.
  • 4 As a functional equivalent to B cell immortalization by EBV in vitro, Shope et al (1973) showed that EBV induces polyclonal B cell lymphoproliferation in marmosets in vivo.
  • 5 EBV causes polyclonal lymphoproliferative disease (PTLD) in immunocompromised patients after solid organ and bone marrow transplantation (Hanto et al, 1981; Schubach et al, 1982).
  • 6 EBV causes polyclonal B cell lymphoproliferation in patients with genetic predispositions (XLP) (Purtilo et al, 1978).
  • 7 EBV DNA was later shown to be present also in a number of other malignancies including a subset of classical Hodgkin lymphoma (Weiss et al, 1989), gastric cancer (Shibata et al, 1991), and some rare Natural Killer (NK)/T and T cell lymphomas (Bornkamm et al, 1976; Fox et al, 2011) (also formerly called angioimmunoblastic lymphadenopathy).

This accumulative evidence, further supported by molecular data on the role of viral gene products expressed in human tumours, led to the classification of EBV as a class I carcinogen by the International Agency for Research on Cancer (World Health Organization [WHO], 1997; WHO, 2011). However, one should keep in mind that this association is, to a large extent, correlative and does not provide a molecular explanation as to the causative role of EBV and its gene products in the various malignancies. This is particularly true for the involvement of EBV in BL as will be outlined in detail below. For a better understanding of EBV′s pathogenic role in different tumour entities, a better understanding of the virus’ life cycle and the complex interplay with the host is required.

The different modes of virus-host interactions in vivo

EBV-induced B cell proliferation is limited by EBV-specific T cells in vivo

B cells are the primary targets of EBV infection in vitro as well as in vivo. Infection of B cells leads to the expression of a limited set of viral gene products, which drive the cells into proliferation. These include six nuclear antigens (Epstein–Barr nuclear antigen [EBNA]1, EBNA-LP, EBNA2, EBNA3A, -3B, -3C), three latent membrane proteins (LMP1, LMP2A and LMP2B), two small noncoding RNAs (EBER1 and 2) and a large set of microRNAs that are transcribed from two independent transcription units (BHRF1 and BART0) (Fig 1, upper panel). This pattern of viral gene expression has first been described in EBV-immortalized cells in vitro and has been denoted latency III (as opposed to latency I viral gene expression in BL cells, see below). The nuclear antigens are translated from one large transcript initiated at the W-promoter (Wp). Promoter activity is shifted to the C-promoter when the viral genome is circularized. EBNA1, EBNA2, EBNA3A and -3C as well as LMP1 are essential for driving B cells into continuous proliferation. Upon infection the viral genome of 172 kb is amplified and circularized. It is replicated episomally and maintained at a fairly constant copy number of 10–50 copies per cell. EBNA1 binds to the plasmid origin of replication oriP and is responsible for proper replication and segregation of the viral episomes into daughter cells during cell division (Yates et al, 1984). It also contributes to transcriptional regulation of the C-promoter. EBNA1 has furthermore been reported to contribute to the regulation of apoptosis (Kennedy et al, 2003) and cellular genomic instability (Gruhne et al, 2009), but these functions are less well established. EBNA2 in conjunction with EBNA-LP is the transcriptional master regulator of EBV in B cells that regulates the C-promoter and thus its own expression, the LMP1 and LMP2 promoter as well as a number of cellular genes, the MYC gene being among the prominent ones (Kaiser et al, 1999; Maier et al, 2006). EBNA2 does not bind to DNA directly, but is recruited to its promoters by the cellular DNA binding protein RBP-Jκ, a repressor complex-recruiting protein, that is physiologically converted into an activator by binding to Notch proteins (Zimber-Strobl & Strobl, 2001). EBNA2 may thus be regarded as a functional Notch equivalent. The ENBA3 proteins bind to chromatin remodelling proteins (Knight et al, 2003; White et al, 2011) and are involved in fine-tuning expression of EBNA2 and thus eventually also in switching off EBNA2 expression at the transition to viral latency in vivo. LMP1 is a membrane protein with transforming activity in fibroblasts (Wang et al, 1985). It operates as a constitutively active CD40 receptor (Rastelli et al, 2008). Likewise, LMP2A acts as a functional B cell receptor equivalent that recruits the tyrosine kinases Syk and Lyn (Merchant et al, 2000). In vivo proliferation of infected B cells is limited by CD8+ and CD4+ T cells (Fig 2). CD8+ T cells primarily recognize peptides derived from the EBNA3 family of proteins expressed in proliferating B cells as well as immediate early proteins of the lytic cycle (Long et al, 2011). CD4+ T cells target viral structural glycoproteins with high efficiency (Adhikary et al, 2006).

Figure 1.

 Different expression patterns of the viral genome define different types of latency. It should be noted that the term ‘latency’ is used for the description of two different phenomena in the EBV field: on one hand it describes the mode of virus persistence in vivo (in vivo latency), on the other hand it describes viral gene expression programmes in vitro or in vivo that are not associated with lytic virus production. The second meaning is used in this figure. The viral genome of 172 kb is shown in linear form as present in virus particles. Note that the viral genome is circularized in latently infected cells and that the LMP2 gene is transcribed across the termini of the linear genome. Transcription in the rightward direction and splicing of the transcripts is depicted above and in the leftward direction below the line. Black flags represent promoters that are active in the respective form of latency. Latency III represents the viral gene expression pattern observed in EBV-immortalized cells and in PTLD. All nuclear antigens are expressed from one large transcription unit initiated at the C- or W-promoter by differential splicing. In lymphoblastoid cells (LCLs) with latency III LMP1, LMP2A, LMP2B, the EBERs and viral microRNAs derived from the BART and from the BHRF1 locus are additionally expressed. The latency I expression programme observed in most Burkitt lymphoma (BL) cells ex vivo and in BL latency I cell lines is characterized by expression of EBNA1 from the Q-promoter, expression of two small non-polyadenylated RNAs (EBER1 and 2) and BART transcripts giving rise to 22 microRNA stem loops. Wp-restricted latency in a subset of BL cases is characterized by a deletion in the EBNA2 open reading frame and transcripts initiated at the W-promoter giving rise to expression of BHRF1, EBNA3A, -3B, -3C and EBNA1. Cells in Wp-restricted latency also express the EBERs and viral microRNAs from the BART and BHRF1 locus. The bottom panel depicts a magnification of the BART locus that gives rise to 22 microRNA stem loops.

Figure 2.

 Virus-host interactions in vivo. Primary infection of B cells leads to B cell proliferation in vivo. The vigorous T cell response elicited by proliferating B cells drives EBV into latency in memory B cells in vivo where the virus is invisible for the immune system (in vivo latency or latency 0). Virus reactivation is initiated by stimulation of memory B cells with their cognate antigen. EBV also infects epithelial cells that are able to replicate the virus. The CD8+ T cell response is directed mainly against proteins of the EBNA3 family in proliferating cells and against immediate early antigens expressed upon reactivation during the lytic cycle. The CD4+ response targets mainly viral glycoproteins present on virus particles and limits viral spreading. EBV-associated malignancies mimic physiological virus-host interactions: Uncontrolled B cell proliferation leads to post-transplant lymphoproliferative disorders (PTLD) in immuncompromised individuals. Nasopharyngeal carcinoma (NPC) and gastric cancer reflect EBV′s potential to infect epithelial cells. Burkitt lymphoma is a malignant lymphoma arising from EBV genome-carrying memory B cells.

In vivo latency (latency 0) in memory B cells and viral reactivation

Although EBV-infected proliferating B cells are eliminated by cytotoxic T cells in vivo, the virus is not totally eliminated from the body. Instead, EBV is able to establish in vivo latency (also called latency 0) in memory B cells (Hochberg et al, 2004). The switch from proliferation to in vivo latency in memory B cells is still poorly understood. The presence of EBV does not appear to interfere with the physiological germinal centre reaction, including somatic hypermutation, class switch recombination and development of memory B cells (Souza et al, 2005). No viral gene product appears to be required to maintain the virus in its latent state in latently infected resting memory B cells in vivo (except for EBNA1 during rare cell divisions as part of the homeostatic regulation of the B cell pool) and likewise, no viral gene product appears to be expressed (except EBNA1 in these rare mitoses). In normal individuals, about 1 in 105–106 peripheral blood cells are latently infected by EBV and the viral genome load is remarkably constant in one individual over time (Khan et al, 1996). B cells are the only site of viral latency in vivo, as elegantly demonstrated by Gratama et al (1989) in EBV-seropositive bone marrow transplant recipients, who became seronegative upon transplantation of EBV-seronegative bone marrow. It is the B cell memory compartment from which the virus can be reactivated to enter the lytic cycle, leading to the production of infectious virus particles that can then be transmitted to other cells or individuals (Laichalk & Thorley-Lawson, 2005). It is at this stage of reactivation that the CD4+ T cells contribute to the elimination of reactivated or newly infected B cells. In immunocompromised individuals the number of in vivo latently infected cells increases by one to three orders of magnitude. Remarkably, this is not accompanied by an induction of lytic viral gene expression (Babcock et al, 1999). Viral reactivation is most likely mediated by binding of a cognate antigen to its B cell receptor (Tovey et al, 1978). This triggers a physiological B cell receptor response, which results in virus reactivation.

Epithelial cells as a reservoir for virus amplification in vivo

The first evidence that EBV may have access to epithelial cells was provided by Wolf et al (1973), who used in-situ hybridization to show that EBV DNA is harboured in epithelial rather than lymphoid cells of NPC biopsies (Wolf et al, 1973). Unequivocal evidence that EBV replicates in epithelial cells of the tongue in lesions called oral hairy leukoplakia in acquired immunodeficiency syndrome patients was provided by Greenspan et al (1985). Such lesions have also been observed in human immunodeficiency virus (HIV)-negative immunocompromised transplant recipients but not in healthy individuals (Itin et al, 1988). Nevertheless, EBV reactivation and release of infectious virus into saliva is common in healthy individuals as revealed by the substantial increase in EBV seroprevalence as well as infectious mononucleosis during early adolescence in countries with high socioeconomic status (infectious mononucleosis is also designated ‘kissing disease’). Most likely, the infection of epithelial cells and replication of the virus in these cells is part of the normal life cycle of EBV in vivo, yet below the detection limit in healthy normal individuals. In rare cases EBV can also be found in T- and NK/T-cells. How EBV accesses T and NK/T cells is still unknown.

Several EBV-associated tumours mimic physiological virus-host interactions in vivo

When considering the association of EBV with various tumours in light of the physiological virus-host interaction in vivo, the role of EBV in PTLD appears to be self-evident. In the absence of a functional T cell response, EBV-infected B cells proliferate in an unlimited fashion, leading to the development of a poly- or monoclonal, rapidly progressive, life-threatening lymphoproliferative disease (Fig 2). In both PTLD and EBV-immortalized cells the full spectrum of EBV latent transcripts and proteins is expressed (designated latency III). Unfortunately the term ‘latency’ is used for the description of two different phenomena in the EBV field: on one hand it describes the mode of virus persistence in vivo (in vivo latency), on the other hand, it describes viral gene expression programmes in vitro or in vivo that are not associated with lytic virus production. The causative role of EBV in PTLD has been illustrated by the pioneering work of Cliona Rooney and collaborators who showed that EBV-specific T cells, expanded in vitro using EBV-immortalized B cells as stimulator cells, are able to cure PTLD (Heslop et al, 1994; Rooney et al, 1995). It is, however, important to note that therapy has to be initiated quickly, as the success of EBV-specific T cell therapy correlates inversely with progression of the disease. Therefore, various strategies have been designed to improve early diagnosis and enable the administration of EBV-specific T cells at the onset of the disease, thereby improving therapeutic outcome (Wilkie et al, 2004; Moosmann et al, 2010).

The association of EBV with epithelial tumours like NPC, gastric cancer and parotid tumours reflects the physiological interaction of EBV with epithelial cells (Fig 2), but the analogy may end here as the viral genome is present in episomal form in NPC cells (Kaschka-Dierich et al, 1976), whereas episomal EBV DNA has not been found in normal epithelial cells in vivo. In EBV-associated epithelial tumours, EBNA1, LMP1 and LMP2 are usually expressed together with the small noncoding RNAs EBER1 and EBER2 and the viral microRNAs (a pattern of viral gene expression designated latency II). The same pattern of latency II gene expression is also observed in classical Hodgkin lymphoma.

The pattern of viral gene expression in BL cells mimics that in latently infected resting memory B cells in vivo with the only exception that these cells rapidly divide and all express EBNA1 for the maintenance of the viral genome. In memory cells EBNA1 expression is observed only in rare mitoses that maintain B cell memory homeostasis. This pattern of viral gene expression is designated latency I (Fig 1).

The putative role of EBV in the development of BL

The BL paradox

Although the discovery of EBV was a result of investigations carried out because of its suspected causative role in the development of BL and although EBV has proven to be a tumour virus that causes malignant B cell proliferations in humans, we are left with the paradox that the role of EBV in the development of BL has still remained enigmatic. First of all, EBV is not required for the development of BL. BL is observed with particularly high frequency in tropical areas of the world including Equatorial Africa and New Guinea, but a histologically indistinguishable disease is found with 10- to 20-fold lower incidence all over the world. Virtually all of the tropical cases harbour the viral genome in the tumour cells, yet, only about 10–20% of the sporadic cases are associated with EBV. With the advent of HIV in the 1980′s, a third form of BL has come into focus, of which about 40–50% are associated with EBV. Most importantly, all the genes that drive B cells into continuous proliferation, namely EBNA2, EBNA-LP, EBNA3A, -3B, -3C, LMP1, and LMP2 are not expressed in BL cells in vivo and ex vivo (Rowe et al, 1986, 1987) (Fig 1). Likewise, the in vivo and ex vivo phenotype of BL cells reflects that of resting memory B cells. They are negative for B cell activation markers and adhesion molecules (e.g. CD23, CD48, and CD54), positive for CD10 and CD77 and high for IgM, and express BCL-6. BL cells grow in single cell suspension, whereas EBV-immortalized B cells grow in large clumps, are negative for BCL-6 and express a variety of adhesion and costimulatory molecules and activation markers, are negative for CD10 and CD77 and low for IgM. When BL cells grow out as cell lines from biopsies, only a minority maintains the in vivo phenotype. Most cells in culture start expressing a latency III programme and tend to adopt an activated phenotype like EBV-immortalized cells. The notion that BL cells represent memory B cells (or at least post-germinal centre B cells) is corroborated by the fact that EBV-positive as well as EBV-negative BL cells harbour somatic hypermutations in their immunoglobulin genes (Klein et al, 1995). BL cells and EBV-immortalized cells are thus biologically fundamentally different, as already described in 1975 by Nilsson and Ponten, based on morphological criteria (Nilsson & Ponten, 1975).

MYC activation is a hallmark of BL

Remarkably, in all forms of BL, regardless of its geographic origin and its association with EBV, chromosomal translocations have been found that juxtapose the human MYC oncogene on the long arm of chromosome 8 into the vicinity of the regulatory elements of one the immunoglobulin heavy or light chain genes on chromosomes 14, 2 or 22 (Dalla-Favera et al, 1982; Klein, 1983). The chromosomal translocations most likely arise as aberrant rearrangements through the action of activation induced deaminase (AICDA, also known as AID), the enzyme that drives somatic hypermutation (SH) as well as class switch recombination (CSR) during the germinal centre reaction (Goossens et al, 1998; Nussenzweig & Nussenzweig, 2011). Activation of the Myc gene is also the hallmark of B cell malignancies in avians and rodents, but its role as an oncogene is not restricted to haematological malignancies, it also plays a pivotal role in the development of a variety of human epithelial tumours. The product of the MYC gene, MYC, is a global regulator of transcription that affects thousands of genes involved in cell cycle control, metabolism, regulation of RNA processing, microRNA expression, signal transduction, cell-cell interaction, immune function, and apoptosis. MYC appears to be bound to about 10–15% of all gene loci as revealed by chromatin immunoprecipitation (Fernandez et al, 2003; Li et al, 2003).

Secondary events most likely affect the regulation of apoptosis in BL

Most importantly, the MYC protein is not only a potent inducer of proliferation, it also induces as a fail-safe mechanism a large number of pro-apoptotic and inhibits expression of anti-apoptotic genes, thereby inducing apoptosis or predisposing cells to apoptosis (Evan et al, 1992) (Fig 3). As a consequence, MYC-driven tumours usually have acquired additional genetic mutations or epigenetic modifications that promote cell survival and shift the balance between proliferation and apoptosis towards proliferation. In the Eμ-myc transgenic mouse model, the vast majority of tumours have acquired secondary changes in the Tp53-ARF (Eischen et al, 1999), yet, in human BL only about one-third of the cases was found to harbour TP53 mutations in vivo. The pro-apoptotic gene BCL2L11 (also known as BIM) is a particularly interesting candidate for secondary changes for a number of reasons: (i) BCL2L11 is a MYC target gene and is not a tumour suppressor gene on its own, yet, (ii) loss of even one Bcl2l11 allele predisposes Eμ-myc mice to lymphoma development (Bcl2l11 haplo-insufficiency) (Egle et al, 2004), (iii) BCL2L11 is required for the elimination of B cells with low-affinity B cell receptors in the germinal centre (Fischer et al, 2007), and (iv) most importantly, mutations in the transactivation domain of MYC (P57S, T58A), as found in human tumours, have lost the ability to induce BCL2L11 without affecting proliferation (Chang et al, 2000; Hemann et al, 2005). These mutations thus uncouple proliferation from induction of BCL2L11 expression. It is not known whether BCL2L11 is mutated or epigenetically silenced in BL cells in vivo and whether differences in BCL2L11 mutations or expression exist in EBV-negative versus EBV-positive tumours.

Figure 3.

 Model for the pathogenesis of Burkitt lymphoma. MYC induces proliferation as well as apoptosis. Presumably by secondary events, the balance is shifted towards proliferation in BL cells. Activation-induced deaminase (AICDA) is presumably a driver of primary events (chromosomal translocations leading to MYC activation) as well as secondary events (deletions, mutations) taking place during the germinal centre reaction. Several observations have converged on BCL2L11 (also known as BIM) as a potentially important player in B cell lymphomagenesis. Several mechanisms may be envisaged for EBV as a cofactor that are not mutually exclusive: contribution of EBNA1, EBERs, and viral microRNAs in the regulation of apoptosis, epigenetic silencing of cellular promoters, and modulation of the activity of AICDA. BBC3 is also known as PUMA and AICDA as AID.

Holoendemic malaria and HIV as cofactors for BL

The epidemiological data collected by Denis Burkitt during the famous safari together with Ted Williams and Cliff Nelson in subsaharian Africa indicated that the incidence of BL coincides with temperature and humidity (Burkitt, 1962), which correlates with the holoendemicity of malaria (high load of malaria transmission in over 75% of the population all over the year) (Dalldorf et al, 1964; Kafuko & Burkitt, 1970). This suggested initially that an arthropod-borne virus might be involved in the pathogenesis of BL. With the discovery of EBV and its transmission through saliva (in the case of small infants from the mother) it became evident that the hypothesis of an arthropod-borne virus did not hold true. Rather, these new findings suggested that malaria per se is a cofactor of BL in conjunction with EBV, and that malaria eventually would modulate the interaction of EBV with its host in vivo. Indeed, this appears to be the case. Malaria parasites are strong polyclonal stimulators of the B cell system, thereby increasing the likelihood of chromosomal translocations. Moreover, certain plasmodium falciparum antigens and exposure to a large number of antigens during multiple infections will reactivate the virus from memory B cells, thereby increasing viral load and consequently the number of EBV-infected B cells in vivo (Donati et al, 2004, 2006; Chene et al, 2007). It is likely that similar mechanisms may operate early after HIV infection. Remarkably, BL incidence is increased early but not late after HIV infection, at a time when the T cell compartment does not yet seem to be affected (Kalter et al, 1985).

The prospective study in the West Nile District in Uganda provides epidemiological evidence for a role of EBV in African BL

In the 1970s, a large prospective study was initiated in the West Nile District of Uganda by the International Agency for Research on Cancer (IARC) under the conductance of Guy de Thé. Between 1971 and 1974, 42·000 blood specimens were collected from children under the age of 8 years and the children were followed-up and periodically surveyed until 1978. Fourteen new BL cases were recorded amongst the pre-bled children. The serological study provided evidence that (i) infection with EBV precedes the onset of BL by several years, (ii) that BL is not linked to primary infection with the virus and most importantly, (iii) those who later developed BL had significantly higher antibody titres against EBV antigens than control patients long before the onset of BL. In other words, patients with high antibody titres against EBV antigens (who most likely harbour higher viral loads) were at increased risk for BL. This indicated that EBV is indeed a risk factor for the development of BL (de Thé et al, 1978).

EBV, AICDA and somatic hypermutations in endemic BL

Another attractive hypothesis is that the presence of EBV during the germinal centre reaction increases the rate of somatic hypermutation by modulating AICDA activity. Although there is no direct evidence for a crosstalk between EBV and AICDA, an increased frequency of somatic hypermutations has been observed not only in EBV-positive as compared to EBV-negative memory cells in normal individuals (Souza et al, 2007) but also in endemic as compared to sporadic BL cases (Klein et al, 1995).

EBV gene products expressed in BL cells: viral gene expression pattern of latency I

The hypothesis that EBV is contributing to the development of BL in EBV-positive BL has been corroborated by the finding that in many, but not all, EBV-positive BL cell lines the presence of the viral genome appears to be selected for. This is indicated by the fact that partitioning of the viral genomes onto daughter cells occurs at only <90% fidelity (Nanbo et al, 2007). These observations have been extended by forcing the loss of the viral genome in BL cells through inducible expression of a dominant negative EBNA1, which inhibits episomal replication. These studies revealed a variable dependence of different BL cell lines on the presence of the viral genome, whereas cell lines from PTLD patients were absolutely dependent on EBV. Thus, there appears to be a graded dependence that correlates with the expression pattern of viral genes (Vereide & Sugden, 2011).

Most BL cases exhibit a viral latency I expression profile, which includes EBNA1, EBER1 and EBER2, and the viral microRNAs of the BART0 locus. In these cases, EBNA1 is expressed from the Q-promoter (Fig 1, second panel).

EBNA1

EBNA1 has been reported to be involved in the regulation of TP53 (Komano et al, 1999; Ruf et al, 1999; Kennedy et al, 2003; Saridakis et al, 2005), but there is no evidence that the frequency of TP53 mutations is lower in EBV-positive than in EBV-negative cases as one would predict if EBV inactivates TP53. Also the potential ability of EBNA1 to induce malignant B cell lymphomas in transgenic mice (Wilson et al, 1996) has not yet been corroborated by other investigators (Kang et al, 2008). A new role for EBNA1 has been proposed in the generation of reactive oxygen species and induction of genomic instability through transcriptional induction of the catalytic subunit of the NADPH oxidase NOX2 (Gruhne et al, 2009). A function of EBNA1 in chromatin remodelling that may lead to dramatic changes in long-term, but much less so in short-term transcription has recently been reported by the same group (Sompallae et al, 2011). If EBNA1 does indeed exhibit such a long-term chromatin remodelling function in vivo, it should be possible to identify distinct differences in the expression profile of EBV-negative and EBV-positive BL latency I cell lines (as well as in EBV-positive and EBV-negative BL biopsies). It will be interesting to see whether this prediction will be verified experimentally.

EBER1 and EBER2

The small noncoding RNAs EBER 1 and EBER2 are found in complexes with La protein, the autoantigen of Lupus Erythematosus, and with the ribosomal protein L22 (Lerner et al, 1981; Toczyski et al, 1994), but their mechanism of action is not or only poorly understood. They have been reported to regulate apoptosis, expression of interleukin-10 through activation of RIG-I and the interferon response (Nanbo et al, 2002; Samanta et al, 2008; Gregorovic et al, 2011), but their oncogenic potential has not been demonstrated unequivocally.

Viral microRNAs of the BART0 locus

Polyadenylated transcripts spanning the BamHI-BART0 fragment into the right direction were first described in NPC cells by Hitt et al (1989). Similar transcripts were also found at lower abundance in EBV-immortalized cells, BL cell lines of latency I, as well as in BL biopsies (Brooks et al, 1993; Xue et al, 2002). Despite numerous attempts, the in vivo coding capacity of these transcripts could not be demonstrated unequivocally (Smith et al, 2000), leaving the significance of these transcripts enigmatic. The transcripts came into the spotlight with the discovery of two clusters of stem loops in the first intron and an additional single stem loop in the last intron of the BART0 locus that encode for a total of 44 microRNAs (Pfeffer et al, 2004; Barth et al, 2011) (Fig 1). All except six are encoded in the region deleted in the B95-8 virus strain. BL latency I cell lines that are not dependent on the viral genome and tend to loose it (e.g. Akata, MutuI, and Daudi) were found to have the lowest microRNA expression level, suggesting that the viral microRNAs contribute to proliferation and/or survival of the other latency I BL cell lines (Pratt et al, 2009). By now, a variety of viral and cellular targets of the viral microRNAs have been described (Barth et al, 2011), the most prominent including the anti-apoptotic proteins BBC3 (also known as PUMA) and BCL2L11, both of which are targeted by the BART0 miRNA clusters 1 and 2 (Choy et al, 2008; Marquitz et al, 2011). It will be particularly interesting to see whether unbiased target identification using the Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation (PAR-CLIP) technology will confirm preferential targeting of apoptosis-related genes, or may identify new players that have not been previously considered.

Contribution of additional viral genes to the pathogenesis of BL in cases with Wp-restricted latency

A second pattern of viral gene expression called Wp-restricted latency was described in about 5–10% of BL cases in vivo and in freshly established cell lines (Kelly et al, 2002). These cases exhibit a deletion in the EBNA2 gene in vivo and are characterized by usage of the W-promoter for expression of BHRF1, EBNA3A, -3B and -3C, as well as EBNA1 from the Wp-initiated transcript by differential splicing (Fig 1, lower panel). BL cells with Wp-restricted latency additionally express the EBERs, the viral microRNAs from the BART0 locus as well as the microRNAs of the BHRF1 locus. The cases with Wp-restricted latency provide a straight-forward explanation for a causal role of EBV. Two different mechanisms appear to converge on inhibition of apoptosis by the viral gene expression programme of Wp restricted latency: BHRF1 is a potent Bcl-2 homologue (Kelly et al, 2009), and EBNA3A and EBNA3C have been shown to be involved in epigenetic silencing of the BCL2L11 locus (Clybouw et al, 2005; Anderton et al, 2008). This is most likely not the only mechanism by which the EBNA3 proteins contribute to the pathogenesis of BL in cases with Wp-restricted latency. As one might predict, BL cells with Wp-restricted latency are highly dependent on the presence of EBV when the cells are forced to express dominant negative EBNA1. In contrast, BL cell lines with latency I did not exhibit a uniform pattern; EBV was dispensable in some cell lines and indispensable in others (Vereide & Sugden, 2011).

A scenario for the involvement of EBV in a hit-and-run mechanism in the pathogenesis of BL

The finding that EBNA3A and EBNA3C are involved in epigenetic silencing of cellular and viral genes has opened the highly interesting possibility that EBV might contribute to the pathogenesis of BL by a hit-and-run mechanism. Given that the EBNA3 proteins contribute to the termination of EBNA2 expression (and thus viral latency III) during the switch to in vivo latency (latency 0), through epigenetic silencing of the C-promoter, it seems plausible that the EBNA3 proteins may also target cellular genes, e.g. those involved in the regulation of apoptosis, like BCL2L11 (Anderton et al, 2008). If such a cell, in which BCL2L11 has been epigenetically silenced through the cooperative action of EBNA3 proteins, is hit by a chromosomal translocation leading to MYC activation, the BCL2L11 promoter may have been rendered unresponsive to induction by MYC. As epigenetic silencing will be fixed by promoter methylation, a cell harbouring EBV in latency 0 may be characterized by epigenetic footprints rendering the cell more susceptible to lymphomagenesis. Models of hit-and-run carcinogenesis are not very popular, because it is difficult (if not impossible) to test them experimentally. Here, this is different. A number of predictions can be made which can indeed be tested experimentally: (i) comparative transcriptional profiling should reveal genes that are expressed in EBV-negative, but not in EBV-positive cases, (ii) the promoter region of these genes should be methylated in EBV-positive, but not, or not regularly in EBV-negative cases, (iii) such genes are candidates for being mutated in EBV-negative, but not in EBV-positive BL cases, (iv) silencing and promoter methylation of such genes may also be selected for in EBV-negative cases, but more than one mechanism would be expected to be operative in these cases, i.e. epigenetic silencing or inactivating mutations. Genes following this pattern of expression, methylation and mutation should be highly indicative of being actively silenced by EBV.

Outlook

BL represents an amazing case for the heuristic process in science. When Denis Burkitt described BL as a clinical entity (Burkitt, 1958, 1962), he addressed very specific questions for a very specific disease that was seemingly restricted to particular areas of Tropical Africa. The search for an arthropod-borne virus was unsuccessful, yet a ubiquitous herpes virus was discovered in BL cells whose role in the pathogenesis of the disease is still not fully understood. The concept of an arthropod-borne transmissible agent has nevertheless been most fruitful as it paved the way for the identification of holoendemic malaria as the most important cofactor for the development of BL in endemic areas in Central Africa and New Guinea.

It was completely unforeseeable, though, that BL would become a paradigm for other most important aspects of cancer research in general. The Philadelphia chromosome had already been described as a recurrent chromosomal marker in chronic myeloid leukaemia (Nowell & Hungerford, 1960), yet, the concept of chromosomal aberrations as causes, rather than by-products, of cancer gained wide acceptance only with the discovery of a second example, i.e. the discovery of the t(8;14) and the variant translocations in BL (Manolov & Manolova, 1972; Dalla-Favera et al, 1982; Klein, 1983). This was a milestone in cancer research because several decades’ avian retrovirus research converged with human clinical research for the first time.

The third area of general interest (besides infections and cancer as well as chromosomal translocations and cancer) is the early phase of chemotherapy. The early reports of Denis Burkitt (Burkitt et al, 1965) had a great impact on chemotherapy in general and have fostered systematic studies rendering most forms of childhood leukaemias and lymphomas curable.

Taken together, BL has evolved as one of the most important and fruitful paradigms in cancer research, and it may be predicted that this will remain so in forthcoming years. It will be exciting to unravel the role of secondary events including the pattern of gene expression, methylation/silencing and mutations in EBV-positive versus -negative cases by next generation sequencing. This will be accompanied by a comprehensive description of the viral and cellular microRNA expression pattern in BL biopsies (and in other EBV-associated malignancies as well) combined with the identification of targets of the viral and cellular microRNAs by an unbiased approach. It will be most interesting to see whether distinct patterns will emerge that may enable clear function(s) to be assigned to EBV in the pathogenesis of BL.

Acknowledgement

We are grateful to Lars Dölken, Michael Hummel, and Dido Lenze for critically reading the manuscript.

Ancillary