Due to the inherent tumorigenicity of EBV and the difficulty to achieve high virus production in cell culture, inactivated or attenuated vaccines are not available. From the 1970s onwards, multiple EBV vaccine studies encompassed subunit vaccines, epitope vaccines, DNA vaccines, nanoparticle-based vaccines, viral vector vaccines, virus-like-particles (VLPs), or dendritic cells (DC) vaccines (Table 1). The animal models that can be infected by EBV include humanized mice, rabbits, rhesus macaques (Macaca mulatta), common marmosets (Callithrix jacchus), cottontop tamarins (Saguinus oedipus), and owl monkeys (Aotus trivirgatus). In addition, five human clinical trials were completed, but none of these vaccines successfully prevented EBV infection in humans.
Table 1 Different types of EBV vaccine studies. Full size table
Vaccine candidates
Considering the complexity of the EBV life cycle, EBV glycoproteins, lytic proteins, and latent proteins are all potential immunogens in EBV vaccine design. It is worth noting that the oncogenic potential of latent proteins should be avoided through proper modification of their immunogenic forms. It is likely that combinations of antigens will induce a more protective immune response, but much needs to be done to define the optimal selections of antigens or their combinations.
Vaccines using lytic glycoproteins as immunogens
gp350
gp350 is the most abundant glycoprotein on the EBV envelope and most previous vaccine studies focused on this antigen55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79. The selection of an adjuvant is one of the pivotal parts to develop an effective subunit and several combinations have been tested in various models.
Vaccines comprising monomeric gp350 (mono-gp350-based vaccines) have been combined with various adjuvants, including alum55,57,60,63, glucopyranosyl lipid A incorporated into the stable emulsion (GLA/SE)62, Syntex adjuvant formulation (SAF-1)60,61,70, immune-stimulating complexes (ISCOMs)72, Freund’s adjuvant 57,63, and incomplete Freund’s adjuvant (IFA)57,67 (Table 1).
The immune response to monomeric gp350 has been influenced by these different adjuvants. Mono-gp350 adjuvanted with alum protected three out of five cottontop tamarins from lymphoma and reduced secretion of EBV DNA in common marmosets55,57,60. Additionally, mono-gp350 adjuvanted with alum induced more robust protective responses than Freund’s adjuvant and IFA in common marmosets57. In a different study, mono-gp350 adjuvanted with alum elicited the same antibody levels in a rabbit model compared with SAF-160. Cottontop tamarins inoculated with mono-gp350 and SAF-1 were protected from lymphoma (two out of three were free of lymphoma61; four out of four were free of lymphoma70). Furthermore, vaccines incorporating mono-gp350 into glycoside Quil A-based ISCOMs required a lower antigen dose to protect four out of four cottontop tamarins from tumorigenesis after EBV challenge72. Besides, after inoculation of mono-gp350 with the Toll-like receptor 4 (TLR4) agonist GLA/SE, a gp350-specific T cell response was elicited, and anti-gp350 antibodies were detected for more than a year, indicating a durable immune response in vaccinated mice62. In addition, mono-gp350 adjuvanted with Freund’s adjuvant required less antigen dose and induced higher neutralizing titers than that adjuvanted with alum in rabbit63. Finally, sera from rabbits and owl monkeys inoculated with gp350 alone could mediate antibody-dependent cellular cytotoxicity (ADCC)76.
To sum up, although the various combinations with mono-gp350 were not systematically compared, they illustrate the critical role of adjuvants to address the requirements for lower gp350 antigen doses, less frequent inoculations, and durable immune responses. One study showed that levels of neutralizing antibodies do not reflect the protective effect of a vaccine in common marmosets57 while other studies emphasized the essential role of neutralizing antibodies at prevention tumor prevention in cottontop tamarins60,61. Among the adjuvants formulated with mono-gp350, SAF-1 seems to be better than alum to attain protection61,70. GLA/SE is beneficial to induce cellular immune responses62. Importantly, even with the applications of various adjuvants, these vaccines still need to be inoculated several times in order to elicit immune responses that protect animals from lymphoma. With the development of novel adjuvants, mono-gp350-based EBV vaccines may be successful in preventing EBV infection and associated diseases.
Sera of mice immunized with non-adjuvanted liposomes incorporating monomeric soluble gp350 (lipo-gp350) neutralize EBV infection in vitro74. However, multiple inoculations of lipo-gp350 adjuvanted with lipid A (fraction from E.coli lipopolysaccharide) induced high titers of neutralizing antibodies in mice and cottontop tamarins71. After 17 immunizations, cottontop tamarins vaccinated with lipo-gp350 were protected from lymphoma, while those immunized six times with lipo-gp350 still developed lymphoma58,59. Therefore, liposome delivery of gp350 combined with an efficient adjuvant may be another potential strategy to develop an effective EBV vaccine.
The use of multimeric gp350 has also been explored because of its higher immunogenicity compared to monomers. Mice immunized with tetrameric gp3501–470 using alum and CpG oligonucleotides (CpG ODN) as adjuvants elicited much higher anti-gp350 antibody and specific CD4+ T cell responses than mice immunized with monomeric gp35056. The enhanced immunogenicity may be due to enhanced B cell receptor (BCR) binding and signaling, vaccine uptake, or presentation and trapping by follicular dendritic cells. The first step of B cell activation is BCR recognition and cross-linking, thus, multimeric antigens are more effective because they better mimic the natural arrangement of multiple copies of the antigen on the virion surface. In addition, a heterodimeric antigen consisting of a mouse IgG2a crystallizable fragment (Fc) fragment and gp350 induced higher neutralizing antibody titers in mice compared to monomeric gp35079.
As different approaches to gp350-based EBV vaccination, vaccinia virus and adenovirus were used as viral vectors to express gp350. The WR strain of vaccinia virus expressing gp350 (VV-gp350) induced humoral immune responses in rabbits, cottontop tamarins, and common marmosets68,69,73. It is remarkable that although no anti-gp350 antibodies and low levels of neutralizing antibodies were detected in cottontop tamarins inoculated with VV-gp350, three out of four animals were still free of lymphoma after an EBV challenge with a dose of 105.3 lymphocytes-transforming doses that cause tumors in 100% of unvaccinated tamarins73. Similarly, all cottontop tamarins, which were vaccinated with a serotype 5 adenovirus expressing gp350 (Ad-gp350), were protected from lymphoma in vivo, even though their sera did not neutralize EBV in vitro77.
A DNA vaccine targeting antigen-presenting cells (APC) showed a good ability to elicit T-cell responses to gp350. Mice immunized with a recombinant pcDNA3.1 vector encoding gp350 induced not only gp350-specific antibodies but also cellular immune responses64. In a different study, sera from mice immunized with a plasmid expressing a gp3501–470 tetramer delivered with the PowderJect-XR-1 system showed higher antibody titers than those immunized with monomeric gp3501–47056. Those studies showed that nucleotide vaccines are attractive to improve immunogenicity and induce a stronger T-cell response that is crucial for killing EBV-infected cells. Note that only DNA vaccines have been studied and no data on RNA-based vaccines are currently available. mRNA vaccines for SARS-CoV-2 showed potent protective effects80,81. Recently, Moderna Inc. announced the initiation of a phase I study for its EBV mRNA vaccine mRNA-1189 (NCT05164094). The efficacy of such mRNA-based vaccines will likely influence the design of future EBV vaccines.
Other gp350-based vaccines include nanoparticle vaccines, epitope vaccines, and VLPs. Ferritin nanoparticles self-assemble to display 24 copies of gp350 (ferritin-gp350)65. These nanoparticles adjuvanted with Sigma Adjuvant System (SAS) elicited neutralizing antibodies in both mice and cynomolgus macaques65. Additionally, the immunized mice were protected from challenges with a recombinant vaccinia virus expressing gp35065. Nanoparticles of lumazine synthase (LS) or I3-01 displaying gp350 domain I/II/III induced higher titers of neutralizing antibodies than the monomeric form of gp35066.
Similarly, mice immunized with gp350 Cytotoxic T lymphocytes (CTL) epitopes combined with IFA and tetanus toxoid were also protected against the challenge of recombinant vaccinia virus expressing gp35067. The data highlight the importance of gp350 CTL epitopes and suggest that such epitopes are beneficial in the design of EBV vaccines.
Virus-like particles (VLPs) provide another attractive delivery system for EBV gp350 antigens. VLPs are multimeric self-assembled particles consisting of one or more structural proteins without a viral genome, which have no pathogenicity. Because their morphology and organization patterns are similar to natural viruses, VLPs can induce both cellular and humoral immune responses82. Chimeric VLPs based on the self-assembling hepatitis B capsid fragment hepatitis B core antigen (HBc149) were constructed to display three immunodominant epitopes of gp35078. In this system, these three peptides from the receptor binding domain of gp350 induced neutralizing antibodies in mice78. Interestingly, the humoral immune response was highly dependent on the sequential order in which these peptides were inserted in the HBc149 backbone78. VLPs based on Newcastle disease virus (NDV) capsid were constructed to display the ectodomain of gp350 (NDV-VLPs-gp350). These VLPs elicited a robust and durable neutralizing antibody response in mice75.
Other glycoproteins and combinations of glycoproteins
EBV entry into target cells is a well-organized and complex process. In addition to gp350, other glycoproteins are involved in virus entry and targeted by neutralizing antibodies. Central to the process of membrane fusion is the herpesvirus core fusion apparatus comprising gB trimers and gH/gL heterodimers16. Additionally, EBV B cell tropism is determined by the expression of gp4216. All these glycoproteins are potential antigens for vaccines aimed at neutralizing infection. Sera from rabbits inoculated with a mixture of glycoproteins prepared from the plasma membrane of EBV-positive P3HR-1 cells neutralized EBV in vitro83. Mice immunized with an epitope-based vaccine comprising gp85 (gH) and gp350 epitopes were protected from challenges with a recombinant vaccinia virus expressing gp85 or gp35067. Neutralizing titers of sera from rabbits immunized with trimeric or monomeric gH/gL, trimeric gB, and tetrameric gp3501–470 were much higher than those elicited by monomeric gp3501–470 (100-fold, 20-fold, 18-fold, and 4-fold higher, respectively)84. Ferritin nanoparticles containing gH/gL/gp42 elicited 2.5-fold higher neutralizing antibody levels against B cells infection and 250-fold higher neutralizing antibody titers against epithelial cells infection compared to ferritin-gp35085. In addition, nanoparticles displaying 60 copies of gH/gL instead of monomeric gH/gL induced neutralizing antibodies protected humanized mice from lethal EBV challenge86. Besides, a pentavalent vaccine based on NDV-VLPs containing EBV gp350, gB, gp42, gH, and gL was used with alum and monophosphoryl lipid A (MPLA) as adjuvants87. This cocktail induced neutralizing antibodies against infection of both B and epithelial cells in vitro87. Moreover, gH/gL/gp42 and gH/gL ferritin nanoparticles induced neutralizing antibodies in mice, ferrets, and nonhuman primates88. No immune competition was observed when combined with gp350D 123 ferritin nanoparticles88. Besides, the passive transfer of antibodies purified from mice immunized with gH/gL/gp42 + gp350D 123 or gH/gL + gp350D 123 ferritin nanoparticles protected humanized mice from EBV-associated lymphoma88. These results clearly support the fact that gH/gL and gB are promising immunogen candidates. Recently, a clinical trial has been launched to evaluate an mRNA vaccine (mRNA-1189), which includes four mRNAs encoding gH, gL, gp42, and gp220 (NCT05164094). Overall, combining glycoprotein antigens is a promising approach for successful EBV vaccine development.
Vaccines using latent proteins and other lytic proteins as immunogens
Proteins that are not involved in virus entry should also be taken into consideration to develop effective vaccines for their expression in infected cells. These targets include other proteins of the lytic cycle as well as proteins expressed in various stages of latency.
In particular, EBNA-1 proved to be a robust immunogen. This antigen is expressed in almost all EBV-linked diseases and its role in maintaining the EBV genome in infected cells is a key factor in viral persistent infections89. EBNA-1 can be recognized by CD4+ T cells from almost all healthy carriers and EBNA-1-specific CD4+ and CD8+ T cells react with EBV-transformed B cells90,91,92,93,94. To use EBNA-1 as a vaccine, its C-terminus was fused with DEC-205 (a human endocytic receptor) and adjuvanted with poly (I:C)95. This vaccine candidate induced robust anti-EBNA-1 CD4+ and CD8+ T cell responses as well as anti-EBNA-1 IgM antibodies in humanized mice95. A heterologous prime-boost vaccination that combined a primary immunization with a recombinant adenovirus expressing EBNA-1 and a boost with a modified vaccinia virus Ankara (MVA) expressing EBNA-1 protected mice from EBNA-1 positive lymphoma after challenge96. Another nuclear antigen, EBNA-2 is one of the first viral proteins expressed during the initial stage of B cell immortalization97. EBV-infected B cells are recognized by EBNA-2-specific CD8+ T cells within 1-day post-infection and their proliferation can be prevented97.
BZLF-1 (Zta) has also been investigated as an immunogen. In a model of EBV-associated lymphoproliferative disease (LPD), survival rates of humanized mice significantly increased due to the specific CD8+ T cell response induced after inoculation of dendritic cells (DCs) transfected to express BZLF-198. This result suggests that the BZLF-1-based vaccine could potentially prevent or delay EBV-associated diseases98.
Combinations of membrane glycoproteins, latent, and lytic proteins
The above studies indicate that proteins involved in virus entry, lytic infection, as well as latency, can contribute to an effective vaccine against EBV. It is, therefore, worth considering different combinations of latent and lytic proteins to develop a comprehensive cocktail vaccine. Toward that goal, a multivalent vaccine was devised by combining recombinant vaccinia viruses, each expressing gp350, gp110, EBNA-2, or EBNA-3C99. This cocktail induced CD4+ T cell responses and antibody responses in mice, indicating that the combination of different EBV proteins into a single dose produces the desired immune response99.
Heterologous VLP is another platform of choice to combine various antigens. NDV-VLPs-gH/gL-EBNA-1 and NDV-VLPs-gB-LMP-2 induced potent neutralizing antibodies as well as EBV-specific cellular responses in mice100. A different approach is to produce EBV-VLPs in non-transforming, virus-free packaging cell lines, using EBV genomes with deletions of some genes101. For instance, EBV-VLPs lacking major oncoproteins EBNA-2, LMP-1, EBNA-3A, -B and -C, and BZLF-1 can be produced in engineered 293-VII+ cells102. Such EBV-VLPs elicited potent humoral and cellular responses in mice102. Alternatively, EBV-VLPs with deletions of BFLF-1/BFRF-1A or of BBRF-1 induced a CD4+ T cell response103. The above EBV-VLPs usually contain many lytic proteins instead of latent proteins. Van Zyl et al.104 constructed more immunogenic particles by overexpressing EBNA-1 in producer cells. Humanized mice immunized with these EBNA-1-VLPs were successfully protected against EBV challenge104. However, except for latent proteins, BNRF1 and viral particles can also induce genetic instability and chromosome defects in infected cells105,106. Hence, safety evaluation of genetic instability and chromosome defects is needed for VLPs generated from non-transforming, virus-free packaging cell lines.
Animal models
The lack of suitable animal models greatly hinders the research and development of EBV vaccines. Animal models which can be used to assess the protective effect of EBV vaccine candidates after EBV challenge, include humanized mice, rabbits, as well as nonhuman primates such as rhesus macaques, owl monkeys, cottontop tamarins, and common marmosets.
Humanized mice
Humanized mice are a novel model to investigate EBV infection and pathogenesis, study EBV-associated diseases as well as evaluate EBV vaccine candidates107. Humanized mice are based on immunodeficient mice, such as non-obese diabetic mice with scid, RAG, and/or IL-2 receptor γ chain mutations. These mice are transplanted with human CD34+ hematopoietic progenitor cells (HPCs) or peripheral blood mononuclear cells (PBMCs) from healthy donors. Infected cells in humanized mice express both latent and lytic EBV antigens after viral challenge108. Importantly, they can develop asymptomatic EBV infections, IM-like syndromes, or tumors depending on the EBV challenge dose, thus, they are useful models to study protection against EBV pathologies108,109. EBV-specific cellular immune responses are observed in humanized mice following EBV infection and the immune responses elicited by vaccines are similar to those of humans110,111. Furthermore, the innate immunity generated by reconstituted human NK cells also plays a significant role in the control of EBV lytic infections in this model112. However, humanized mice lack human epithelial cells, which are instrumental in the whole EBV infection cycle. Additionally, the development of “human” germinal centers and secondary lymphoid tissues is poor in this model109,113. Hence, the humoral immune responses cannot be reliably evaluated in the current humanized mice models. IgM antibody production against the viral capsid antigen BFRF-3 is detected in humanized mice114. Therefore, humanized mouse model is more suitable to evaluate the passive protective effect of antibodies purified from immunized mice, rabbits, or nonhuman primates. Studies using humanized mice to evaluate EBV vaccines are compiled in Table 2.
Table 2 Humanized mice models for EBV vaccines. Full size table
Rabbits
Evidence showed that Japanese White rabbits can be persistently infected by EBV through intravenous inoculation since viral DNA and anti-EBV-VCA antibodies were both detected for 15 months115. Notably, persistent infections were also observed following infection of New Zealand White rabbits and Japanese White rabbits via the oral route, which is also the natural infection route in humans116,117. Furthermore, cells from New Zealand White rabbits infected intravenously proliferated in vivo following immunosuppression by cyclosporine A, which is reminiscent of observations in human post-transplantation lymphoproliferative disorder (PTLD) patients118. Together, these studies indicated that rabbit models are potential platforms for EBV vaccine evaluation.
Nonhuman primates
Rhesus lymphocryptovirus (rhLCV) is a homolog of EBV that only infects rhesus macaques and shares the same infectious features with EBV119. Experimental rhLCV infection in rhesus macaque causes either asymptomatic persistent latent infection or IM-like syndrome in immunocompetent macaques. However, in immunosuppressed macaques previously infected by simian immunodeficiency virus, rhLCV infection can lead to tumor formation119,120,121. Differences between rhLCV and EBV cannot be ignored, however, rhLCV vaccines and challenges performed in rhesus macaques can be considered as an indirect surrogate model to assess EBV vaccines122. Rhesus monkeys immunized with soluble rhLCV gp350 combined with alum as the adjuvant were protected against rhLCV oral challenge123. Interestingly, 72A1, a strong neutralizing monoclonal antibody targeting EBV gp350, protected rhesus macaques from oral challenge with a recombinant rhLCV carrying EBV gp350124. Such a chimeric virus may provide an interesting model to assess the in vivo protective effect of antibodies elicited by vaccine candidates.
Cottontop tamarins, common marmosets, and owl monkeys can be experimentally infected by EBV and recapitulate different aspects of human disease (Table 3). Cottontop tamarins are susceptible to experimental EBV infection and can develop malignant lymphomas after challenge with high doses of EBV125,126. Cleary and colleagues127 determined the 100% tumorigenesis dose of EBV strain B95-8 in cottontop tamarins and confirmed that tumors consisted of large-cell lymphomas with multiple copies of the EBV genome, which resembles the condition of PTLD patients. In addition, when cottontop tamarins recovered from tumors after the first challenge, cellular immune responses were observed after a second challenge, and these subjects remained healthy without any EBV-associated diseases128. Common marmosets can be infected by either the M81 strain (derived from an NPC patient) or the B95-8 strain (derived from an IM patient)129,130,131,132. The symptoms of infected common marmosets include lymphocytosis, the production of heterophile antibodies and the long-term production of EBV-specific antibodies are similar to those in humans132. After the EBV challenge, a persistent antibody response against EBV-VCA and early lytic proteins was observed132. However, antibodies against EBNA-1 were not detected and there were no viral antigens in the lymphocytes of infected animals, which differs from human cases133. In terms of pathologies, chronic infectious mononucleosis instead of LPD or lymphoma was observed in common marmosets. Owl monkeys also developed LPD after the experimental EBV challenge, and, interestingly, the EBV genome was found in a cell line established from an infected owl monkey134,135.
Table 3 Outcomes of Nonhuman primate after EBV infection reflect different aspects of human diseases. Full size table
From 1980 to 2000, various EBV vaccines were assessed in these nonhuman primate models for efficacy (Table 4). Notably, sterilizing immunity was not achieved in any of these studies. Another limitation lies in the fact that experimental infection in nonhuman primate models is quite different from natural routes in humans. In some studies, data showed that there was no direct correlation between neutralizing antibody levels and vaccine protective effects, for some of the immunized animals with high neutralizing antibody levels still developed lymphoma after a 100% tumorigenesis virus challenge while those without high neutralizing titers free of lymphoma59,73,77. However, another study demonstrated that neutralizing antibodies is one of the key attributes of tumorigenesis prevention58,60,61,70. Therefore, additional factors may be involved to confer complete protection, such as cellular immune responses and ADCC.
Table 4 Nonhuman primate models for EBV vaccines. Full size table
Finally, one should note that nonhuman primates are expensive and not necessarily amenable to large and preliminary studies of vaccine candidates. In addition, specific models for EBV, such as marmosets and owl monkeys, are rare and not readily accessible. Cottontop tamarins are not available since they are an endangered species.
Clinical trials
From 1990 onwards, seven human clinical trials have been launched utilizing EBV gp350 or EBNA-3. For instance, Gu et al.136 utilized a live recombinant vaccinia virus (Tien Tan strain) expressing gp350 to immunize three groups of volunteers, including 11 adults (EBV seropositive and vaccinia seropositive), six juveniles (EBV seropositive and vaccinia seronegative), and 19 infants (EBV seronegative and vaccinia seronegative). In the adult group, antibody titers against EBV did not change after inoculation, while neutralizing antibody titers increased in young children and infants. Three out of nine infants still became naturally infected by EBV later. Meanwhile, ten out of ten control infants also became naturally infected. Moutschen and colleagues137 compared three vaccine formulations (recombinant gp350 alone, recombinant gp350 with alum, or recombinant gp350 with AS04) in seronegative and seropositive youths. All formulations were safe and well-tolerated. The formulation containing gp350 alone showed the weakest immunogenicity. Despite the detection of neutralizing antibodies and cellular immune responses, some subjects still became naturally infected. These observations are partly consistent with results obtained in common marmosets and cottontop tamarins, as discussed above. A phase II trial enrolled 181 seronegative young volunteers to test an EBV vaccine formulated with recombinant EBV gp350 and AS04 as an adjuvant138. Although anti-gp350 antibodies were detected over 18 months, this vaccine only prevented IM but not asymptomatic EBV infection. Another phase I trial recruited children with chronic kidney disease waiting for organ transplantation139. After inoculating two different doses (12.5 and 25 μg) of recombinant gp350 with alum, specific IgGs were found in all subjects. However, neutralizing antibodies were only detected in 1/4 of subjects who received the low dose and in 3/9 of subjects who received the high dose. Nevertheless, titers dropped quickly and vaccination did not affect the post-transplant immune condition of these children. Recently, a phase I clinical trial for a gp350-ferritin nanoparticle vaccine started to recruit subjects to evaluate vaccine safety and immunogenicity (NCT04645147). Another phase I clinical trial for an mRNA-based vaccine (mRNA-1189) containing four mRNA encoding gH, gL, gp42, and gp220 has been launched to evaluate to safety and tolerability of EBV mRNA vaccine in healthy adults ages 18 to 30 (NCT05164094).
Aside from glycoprotein-based vaccines, an EBNA-3 epitope-based vaccine was tested in 14 HLA B*0801-positive EBV-seronegative adults in a phase I trial140. This vaccine consisted of an EBNA-3 CD8+ epitope (FLRGRAYGL) combined with tetanus toxoid as CD4+ T cell helper and Montanide ISA 720 as the adjuvant. The vaccine proved to be safe and epitope-specific responses were observed, however, some immunized subjects seroconverted asymptomatically.