The Epstein-Barr virus (EBV) is an oncogenic γ-herpesvirus that is endemic in human populations worldwide (1). The oncogenic potential of EBV was first suggested through its association with Burkitt lymphoma (2) and by its ability to transform B cells in vitro (3, 4). It is now understood that EBV is able to transform cells through the expression of Epstein-Barr nuclear antigens (EBNA) that are endowed with transactivating properties and the latent membrane proteins (LMP) that provide proliferative and survival signals (5). These proteins are expressed during non-productive (latent) infection along with several viral microRNAs (6). EBV-infected cells are also capable of supporting productive (lytic) infection, which also contributes to the development of malignancies (7–9) and is characterized by the expression of more than 80 viral genes (10) and enables the production of infectious progeny.

EBV predominantly spreads via saliva and EBV virions target epithelial cells and B cells of the oropharynx upon entering new hosts (11). Primary EBV infection usually occurs during early childhood and is not accompanied by any overt signs or symptoms. However, when EBV is acquired during adolescence or adulthood, it commonly results in infectious mononucleosis (IM) (12), a self-limiting disease whose clinical features include pharyngitis, cervical lymphadenopathy, fatigue, and fever (13). Most individuals recover from IM within a couple of weeks, but a notable portion of individuals experience fatigue that lasts for 2–6 months (14). Thus, IM is directly connected to a significant reduction in quality of life and imposes a sizable financial burden on wider society. Additionally, the occurrence of IM has been linked to an increased risk for the development of Hodgkin lymphoma (HL) (15) and multiple sclerosis (MS) (16). This suggests that a prophylactic vaccine against EBV able to prevent IM could potentially decrease the disease burden associated with HL and MS. The development of an EBV vaccine is further encouraged by the association of EBV with several other malignancies of hematopoietic or epithelial origin (17). Moreover, since EBV causes post-transplant lymphoproliferative disease (PTLD) in immunosuppressed hosts, it suggests that vaccination against EBV might be useful in hematopoietic stem cell or solid organ transplant candidates, if possible prior to transplantation (18).

EBV infection and the ensuing lifelong persistence is a complex, multistep process that starts with the infection of permissive cells within the oropharynx, and culminates in the maintenance of EBV in circulating memory B cells (19). The germinal center model (GCM) of EBV persistence suggests that EBV utilizes the normal pathway of B-cell differentiation to achieve this feat. Since EBV is capable of implementing various latency and lytic transcription programs, it suggests that EBV assumes distinct antigenic states within infected individuals (Table 1). Moreover, since the nature of these antigens varies, they offer unique challenges to the adaptive immune system (Figure 1). Yet, despite the wide variety of antigens that predominate throughout the EBV life cycle, EBV vaccines candidates have traditionally only focused on a limited number of EBV antigens (See the review by Cohen (24) for a summary on these vaccine candidates. We now consider the various antigenic states of EBV during a single infection cycle and how vaccination may aid their recognition and elimination (Figure 2).

The struggle toward a licensed prophylactic vaccine is not limited to EBV and has also been the case for the majority of human herpesviruses. The only exception to this trend is the alpha herpesvirus Varicella-zoster virus (VZV), which is responsible for the development of varicella and zoster (85, 86). VZV establishes latency after primary infection, is carried lifelong and is controlled in healthy individuals through the coordinated activities of humoral and cellular immunity (87). Vaccination against VZV is carried out with a live-attenuated virus that induces immune responses comparable to wild-type VZV infections (88). Vaccinated individuals develop humoral and cellular immune responses that target numerous VZV proteins (88–90). Vaccination against herpesviruses has also been successfully carried out in animals, with Marek's disease virus (MDV) representing an exemplary case. MDV is a highly oncogenic poultry pathogen that causes the development of lymphomas (91, 92). Vaccination against MD is also carried out with a live-attenuated virus and successfully prevents the development of tumors (93).

Whilst vaccination against VZV and MDV successfully prevents the manifestation of disease, they do not prevent infection with wild-type strains (94, 95). Similarly, vaccination against EBV might be able to reduce EBV-associated diseases and malignancies without achieving sterile immunity (96). Considering what is known about successful vaccination against VZV, it is reasonable to assume that the ideal EBV vaccine should generate immune responses that mimic those observed during wild-type infections. However, since live-attenuated herpesvirus vaccines persist in infected individuals (97), it is unlikely that EBV, an oncogenic virus, would be suitable as a live-attenuated vaccine. Considering the complexity of EBV infection, in terms of the different cell types that are infected, the ability of EBV to spread directly from cell-to-cell (30, 98) and the diverse number of antigens that predominate during the EBV life cycle, EBV vaccines have to deal with the challenging task of being efficacious and highly safe.

The potential of vaccination to mitigate EBV-associated diseases was eluded to within a decade of EBVs discovery (99), with so-called membrane antigen being suggested as a potential target (100). This was followed by three decades of research in which gp350 was championed as the vaccine candidate of choice (24). Since gp350-specific antibodies block B-cell infection, gp350 was marked as a promising vaccine target. However, when a subunit gp350 vaccine was finally tested in a phase 2 clinical trial, it failed to prevent EBV infection despite inducing neutralizing antibodies in vaccinees (20). This was succeeded by the development of numerous vaccine platforms that vary in antigenic content and structural complexity (Figure 3, Table 2).

The simplest vaccine platform that has been considered for EBV infection is an EBNA3A epitope peptide mixed with tetanus toxin as a water-in-oil emulsion (Figure 3A) (23). In contrast to recombinant gp350, which was developed to elicit neutralizing antibodies, the EBNA3A epitope vaccine was designed to induce EBV-specific T-cell responses that recognize infected B cells displaying latency III. Vaccination with the EBNA3A epitope peptide successfully induced the relevant CD8+ T-cell responses in ~90% of vaccinees, but failed to prevent EBV infection. Although vaccination appeared to reduce incidence of IM, the number of participants was too few for statistical analysis.

The next group of EBV vaccine candidates is based on antigen-antibody conjugates, or antigen-armed antibodies (AgAbs), which specifically deliver antigenic peptides to antigen-presenting cells (APCs) (Figure 3B) (111). αDEC-205 equipped with EBNA1 is processed by dendritic cells and enables the expansion of EBNA1-specific CD4+ and CD8+ T cells from the blood of EBV-positive individuals (101). Also, humanized mice vaccinated with αDEC-205-EBNA1 successfully developed EBNA1-specific IgM and T-cell responses. EBV latent antigens have also been fused to αCD19, αCD20, αCD21, and αCD22 to enable B-cell-mediated stimulation of EBV-specific T cells (102). Antigenic epitopes fused to B-cell-specific antibodies are up to ~4,000-fold more antigenic than peptide alone and enable the stimulation of cytolytic CD4+ T cells (102, 103). The ability of antigen-antibody conjugates to potently stimulate latent protein-specific T cells suggests that they might induce protective T-cell responses upon vaccination.

Owing to the weak immunogenicity of monomeric subunit vaccines (112–115), several groups have moved toward the development of multimeric vaccines that arrange EBV antigens in a repetitive manner. This enables multivalent, long-lasting, stimulatory interactions with the immune system, and generates considerably stronger B- and T-cell responses than monomeric proteins (116). A direct approach for increasing the immunogenicity of individual EBV antigens is through the use of multimerization domains (Figure 3C). One group has successfully generated multimeric gp350 (104) and gH/gL (105) by fusing the individual antigens to mutimerization domains (e.g., the T4 bacteriophage foldon). They demonstrated that rabbits vaccinated with the multimeric proteins developed superior neutralizing antibody responses compared to their monomeric counterparts. However, both monomeric and multimeric gH/gL were shown to outperform all other antigens, with multimeric gH/gL inducing serum neutralization titers >100-fold higher than monomeric gp350.

EBV antigens have also been introduced into virus-like particles (VLPs) (107, 108) and self-assembling nanoparticles (106). These platforms enable monomeric antigens to be displayed in a manner highly comparable to EBV virions, but since they lack viral DNA they are incapable of causing disease. One group has developed chimeric VLPs by fusing EBV antigens to Newcastle disease virus (NDV) structural proteins (Figure 3D) (107, 108). Chimeric NDV VLPs containing gp350 have been shown to induce superior neutralizing antibody responses upon vaccination in mice compared to monomeric gp350 (107). Subsequently, the NDV VLP platform was utilized to generate immunogenic particles that contained multiple EBV antigens, including envelope proteins (gB, gH/gL, gp350) and latent proteins (EBNA1 and LMP2) (108). The potential of these particles to stimulate structural protein- and latent protein-specific immune responses suggests they might enable the targeting of EBV virions and EBV-infected cells in vaccinated individuals. The gp350 antigen has also been fused to Helicobacter pylori-bullfrog hybrid ferritin to generate highly immunogenic self-assembling nanoparticles (Figure 3E) (106). Nanoparticles containing gp350 induced significantly higher neutralizing antibody titers in mice and monkeys in comparison to monomeric gp350. The incorporation of gp350 into ferritin nanoparticles was elegantly shown to enhance the presentation and recognition of the CD21-binding site on gp350. Moreover, mice vaccinated with the gp350 nanoparticles were successfully protected against a recombinant vaccinia virus that expressed gp350. However, since gp350 did not play a functional role in the infection process and vaccination success or failure was solely based on the weight of the animals, it is difficult to relate these findings to vaccination against EBV in humans.

Another approach to EBV vaccination is through the use of EBV VLPs. Since EBV VLPs are structurally similar to EBV and comprise numerous viral antigens (viz., envelope, capsid, and tegument), they have the potential to elicit immune responses against multiple EBV antigens. By deleting genetic elements (109, 117, 118) or proteins (63) involved in DNA packaging, the generated EBV VLPs contain little to no viral DNA (Figure 3F). One group has developed a VLP-producing packaging cell line by deleting the terminal repeats (TR), involved in DNA packaging, and six viral genes that contribute to transformation or production of virions (EBNA2, EBNA3-A, -B, -C, LMP1, and BZLF1) (109). Vaccination of mice with these VLPs induced polyvalent EBV-specific humoral and cellular immune response, highlighting the potential of EBV VLPs to elicit broad immune responses. Impressively, the potency of vaccine-induced neutralizing antibodies was comparable to the anti-gp350 MAb 72A1. This finding is supported by studies that have shown UV-inactivated EBV to induce potent neutralizing antibodies (107, 108). We have developed an alternate EBV VLP packaging cell line by deleting BFLF1/BFRF1A and gB from the EBV genome (9, 63). By deleting BFLF1/BFRF1A we improved the purity of the VLPs relative to those obtained with a ΔTR genome in that they contained no detectible DNA (63), while deletion of gB prevents VLPs from fusing with host cells and greatly reduces their pathogenic potential (9, 119, 120). We subsequently increased the antigenic spectrum of EBV VLPs by fusing latent antigens to the major tegument protein BNRF1 (Figure 3G) (110). VLPs containing latent antigen are capable of stimulating both structural protein- and latent protein-specific T cells and afford increased protection against EBV infection in humanized mice. Since EBV encodes more than a dozen tegument proteins, EBV VLPs have the potential to accommodate multiple immunodominant latent antigens. By fine-tuning the antigenic cargo of EBV VLPs, they might enable the induction of immune responses that recognize viral antigens that predominate during the early stages of the infection and in latently infected cells.

Therapeutic EBV vaccines aim at boosting and sustaining antiviral adaptive immune responses in patients with virus-associated disorders. For several reasons, most therapeutic vaccination approaches have focused on NPC. First, almost all cases of the non-keratinizing subtype of nasopharyngeal carcinoma (NPC), which represents >95% of NPC in endemic regions, are EBV-positive and consistently express EBNA1, LMP2, and to variable degrees LMP1 (121). Second, EBNA1 and LMP2 are major targets of the virus-specific T-cell response in healthy virus carriers, and CD4⁺ and CD8⁺ T cells against these antigens have been detected in tumor patients (122, 123). Moreover, virus-specific T cells can be cultured from NPC tissues suggesting that T-cell function is either maintained or not irreversibly impaired in these patients (123). Third, HLA loss is uncommon in NPC and clinical responses have been observed after infusion of EBV-specific T-cell preparations, suggesting that tumor growth can be controlled by the immune system (124, 125). Based on these findings, different therapeutic vaccines have been designed and tested in NPC patients. All of these vaccines have been well-tolerated with minimal side effects and no evidence of dose-limiting toxicity.

In the first therapeutic vaccination trial for NPC, 16 patients with residual disease received four cycles of autologous monocyte-derived dendritic cells (DC) loaded with LMP2-specific CD8⁺ T-cell epitope peptides (126). In more than half of these patients, increases in LMP2-specific CD8⁺ T cells were noted and these increases were sustained for 3 months before declining. Partial clinical responses were observed in two patients that had shown heightened virus-specific T-cell frequencies. This study used a small number of defined CD8⁺ T-cell epitope peptides selected on the basis of the patients' HLA type. In order to boost a wider range of T-cell specificities including those still undefined or presented on HLA class II molecules, autologous DC expressing truncated LMP1, and full length LMP2 protein were used (127). Partial clinical response or stable disease was achieved in three of the 16 vaccinated patients. Based on the widely used modified vaccinia Ankara (MVA) vector, a therapeutic vaccine was designed that expresses a fusion protein of full length LMP2 and the C-terminal half of EBNA1. Two dose escalation phase IA clinical trials were conducted with this MVA-EBNA1/LMP2 vaccine on NPC patients in Hong Kong and the United Kingdom (128, 129). Patients who had received the highest dose responded to EBNA1, LMP2, or both. Moreover, increased CD8⁺ and CD4⁺ T-cell responses against LMP2 and EBNA1 were observed in both patient cohorts in Hong Kong and UK, demonstrating that the vaccine was immunogenic in different ethnicities with different HLA types and EBV strain variants. In ongoing phase IB and II trials, immunogenicity, and clinical efficacy of the vaccine are further examined.

Despite these encouraging results, therapeutic vaccination alone is unlikely to reduce disease recurrence in the majority of patients. Additional studies are needed to test whether treatment outcomes can be improved by combining therapeutic vaccination with other forms of immunotherapy, e.g., adoptive T-cell therapy or checkpoint inhibition. Besides, issues of optimal combination need to be addressed for incorporating immunotherapy into standard treatment protocols.

The number of EBV vaccine candidates has greatly increased over the last decade. The latest advances in vaccine technology, coupled with our growing understanding of EBV biology and immunology, have enabled emerging EBV vaccine candidates to directly address the shortcomings of a monomeric gp350 vaccine. However, since the correlates of protection against EBV have not been clearly defined, it is hard to reliably predict the ideal EBV vaccine targets and whether humoral immunity or cellular immunity or both should be engaged. Vaccines comprising a limited set of EBV antigens would certainly be easier to manufacture and safer compared to vaccines that contain a large combination of EBV antigens. However, if vaccination with a limited set of antigens is unable to preclude the establishment of latency, vaccines might have to increase their antigenic spectrum to include multiple structural antigens and perhaps even latent antigens. However, since latent proteins are highly polymorphic compared to structural proteins (130, 131), it suggests that they are considerably more challenging to target across multiple geographic regions. Nevertheless, since all EBV-associated malignancies express latent antigens, it suggests that they are worth exploring as vaccine targets. Also, different vaccine types might end up having different aims and target populations. While vaccines based on a one or several antigens might be sufficient to prevent the occurrence of IM and its complications such as multiple sclerosis without necessarily being able to confer sterile immunity, a cocktail of EBV antigens from different EBV strains, carrying latent and lytic proteins might be able to achieve a higher level of protection against infection and its malignant and non-malignant consequences. The first type of vaccine would probably be indicated in immunocompetent western populations at low risk of EBV-associated tumors and the second in populations vulnerable to endemic EBV-associated malignancies or persons awaiting organ transplantation. However, only vaccination campaigns will be able to determine which types of vaccines confer protection and to what degree.

All authors have made a substantial contribution to the work. DvZ wrote the original draft and generated the graphics. JM wrote the section on therapeutic vaccines. H-JD and JM reviewed and edited the final manuscript.