We searched the electronic databases of the Cochrane Library, ClinicalTrials.gov, Embase, PubMed, Scopus, Web of Science, the World Health Organization’s Global Index Medicus (WHO-GIM), and Google Scholar for pre-clinical and clinical prophylactic EBV vaccine studies ( Table 2 ). The search was performed with a publication date limit of June 20, 2020, and duplicate articles were automatically removed after electronic comparison across databases (EndNote).

Articles were excluded if they were not in English or Spanish and if the full text was not available. Selection based on content was performed in two steps: first based on the title and abstract, then by reviewing the full text. In the first step, articles were excluded based on the nature of the article (review/commentary/antibody development study) and the type of targeted immune response (dendritic cell-targeted/T-cell-targeted vaccines). In the second step, pre-clinical studies were excluded based on the nature of the article (conference paper/case study/antibody generation or characterization study/diagnostic or detection study/in silico study/epitope mapping study/immunoglobulin prophylaxis study); the type of targeted immune response (T-cell-targeted vaccines); and lack of data (no in vitro data/no neutralization assessed). Clinical studies were excluded based on the nature of the article (study follow-up) or the type of targeted immune response (therapeutic vaccine). Both selection steps were performed by two independent reviewers (JGO and LZM), and any disagreements were resolved by a third reviewer (GME). A flowchart of this process is presented in Figure 2 .

Data extraction was performed by the JGO, GME, LZM, MM and ER. Review of extracted data was performed by JGO and GME, with any disagreements reconciled by LZM. For pre-clinical studies, the following information was extracted: bibliographic information (first author, publication year, title); animal model; characteristics of the population (animal strain, number of animals); characteristics of vaccine (type and source of antigen, dosage, adjuvant, administration route, vaccination schedule); characteristics of control treatments (type of placebo and non-placebo controls, dosage, administration route, vaccination schedule); characteristics of measurement techniques (sample collection schedule, type of antibody detection assay, type of neutralization assay and type of cells used for neutralization assay); and study outcomes. For clinical studies, the following information was extracted: bibliographic information (first author, publication year, title); trial characteristics (country of origin, phase, type, primary endpoint); study population characteristics (age, EBV status, sex, number of participants); characteristics of vaccine and control treatments (type and source of antigen, dosage, adjuvant, administration route, vaccination schedule); characteristics of measurement techniques (sample collection schedule, type of antibody detection assay, type of neutralization assay and type of cells used for neutralization assay, type of EBV diagnostic test); and trial outcomes. Extracted data is presented in Table 3 for pre-clinical studies, and Table 4 for clinical studies. For pre-clinical studies, the author list affiliations were individually checked to determine the country of origin for each study, and a global heat-map was prepared displaying the number of studies performed per country ( Figure 3 ). From the pre-clinical study extracted data, a list was prepared describing the immunogens tested, the type of vaccine delivery platforms used, and the animal models used to test vaccine immunogenicity and efficacy; the number of studies per variable in each category were graphed and are presented in Figures 4A–C .

The quality of all selected pre-clinical studies was assessed using a checklist of 18 items ( Table 5 ) that were modified and expanded from (84), which was developed based on the Collaborative Approach to Meta-Analysis of Animal Data from Experimental Stroke (CAMARADES).

These 18 items were divided into primary (#1-15) and secondary (#16-18) quality criteria. Each primary quality criterion carried a score of one point. In addition, a study was awarded two points for 3/3 of the secondary criteria met, one point for 2/3, and no points for 0-1/3, giving a possible total of 17 points. Because four of the primary criteria (#11-13, #15) did not apply to all studies, studies with total scores of 13 or above were considered very high quality. Studies with scores 10-12 were high quality, 7-9 were moderate quality, 5-8 were poor quality, and 0-4 were very poor quality. Tabulated results for each study analyzed are presented in Table S3 , and the overall quality distribution of the studies was graphed and presented in Figure 4D .

No institutional review board approval was required for this study.

Our electronic database search yielded a total of 5,614 unique articles. Of these, 36 pre-clinical studies and 4 clinical studies met the criteria for inclusion in the systematic review ( Figure 2 ). The pre-clinical studies were conducted between 1979 and 2020 ( Table 3 ) and the clinical studies were conducted between 1995 and 2009 ( Table 4 ). Most studies were conducted in the United States and the United Kingdom ( Figure 3 ).

In the 36 pre-clinical studies included in our systematic review, gp350 was the most commonly used individual immunogen, tested in a total of 33 studies ( Figure 4A ); gH/gL and gB were individually tested in 3 studies each. When multivalent combinations were tested, undefined viral membrane antigens were used in 3 studies, gp350-gH/gL and gp350-gB combinations were used in 2 studies each, and additional antigen combinations were used in 1 study each ( Figure 4A ). Vaccines were most commonly delivered as monomeric proteins (12 studies); other delivery platforms included nanoparticles (8 studies), virus-like particles (VLPs; 5 studies), multimeric proteins and vaccinia virus (4 studies each), and infected cell membranes (2 studies), as well as peptides, DNA, measles virus, and adenovirus (1 study each) ( Figure 4B ). Mice were the most commonly used animal for immunogenicity assessments (15 studies), followed by rabbits and cotton-top tamarins (10 studies each), common marmosets (3 studies), and cynomolgus macaques (2 studies), as well as rats and rhesus macaques (1 study each) ( Figure 4C ). Of these studies, 8 utilized cotton-top tamarins and 3 utilized common marmosets as EBV challenge models ( Figure 4C ).

Most pre-clinical studies (13) were rated as moderate quality, 11 were rated as poor quality, 5 were rated as high quality, 4 were rated as very high quality, and 3 were rated as very poor quality ( Table S3 and Figure 4D ).

All 4 clinical studies included in our systematic review used gp350 as the clinical immunogen ( Table 4 ). The first study (83), a Phase I study, used vaccinia virus (Tien Tan strain) to deliver the immunogen ( Table 4 ). The second (82), a Phase I/II study, and third (81), a Phase II study, used a monomeric gp350 protein formulation with Adjuvant System 04 (AS04) ( Table 4 ). The most recent study (80), a Phase I study, used a monomeric gp350 protein formulation of the same origin as the others but adjuvanted with alum alone ( Table 4 ). None of these clinical studies led to the prevention of EBV infection/sterile immunity.

The correlates of immune protection against primary EBV infection are not well-defined, partly due to a lack of an appropriate animal model to study EBV infection and test vaccine efficacy, as discussed below in Section 4.2. Thus, pre-clinical vaccine efficacy has been measured primarily using in vitro neutralization assays as a surrogate for in vivo protection. However, in our assessment, we found that there are no clear standards for performing neutralization assays, defining vaccine efficacy, or reporting outcomes.

Before the turn of the century, in vitro neutralization was assessed via transformation assays, in which primary human lymphoid B cells were exposed to EBV previously incubated with immune serum or plasma and then allowed to transform and immortalize over several weeks ( Table 3 ) (85). A reduction in or lack of transformation indicated the presence of nAbs in the serum/plasma sample. The advent of fluorescent recombinant EBV (86–88) allowed the emergence of a simpler and faster method that enables the neutralization assay to be performed on various virus-susceptible cells, including established epithelial and B cell lines—eliminating the required use of primary B cells—with a maximum turnaround of three days after infection. Thus, most pre-clinical studies, beginning with Ruiss et al. in 2011 (56) ( Table 3 ), have utilized a flow cytometry-based assay that was optimized by Sashihara et al. in 2009 to measure infection by quantifying the number of fluorescent cells in a given sample (89). Although these assays can provide invaluable insights, they are no substitute for in vivo studies and cannot provide true correlates of immune protection. Indeed, in the clinical trials analyzed in our study ( Table 4 ), the presence of nAbs could still be identified via in vitro assays in vaccinees who later became EBV-positive. Moreover, we found that there is extensive variability in the type of cell lines used in neutralization assays (e.g., Raji, HEK-293, SVKCR2, Akata, AGS) and in how outcomes are reported (e.g., % neutralization, % infected cells, IC₅₀ titer) ( Table 3 ). We also found that most studies did not provide any information regarding the infectivity of the EBV batch used for the assays (e.g., infectivity curve or virus titer) or the target level of infection used as the negative control for neutralization; based on our assessment, only 14/36 studies appropriately reported virus titer ( Table S3 ). Similarly, we found that only 21/36 studies appropriately reported the immune serum/plasma/antibody concentration or dilution used for neutralization assays ( Table S3 ).

Competitive ELISA against a known nAb has also been used as a surrogate in vitro neutralization assay ( Table 3 ). Although this assay provides important information regarding the types of known neutralizing epitopes a vaccine can target, it serves more as a complementary assay as it does not consider all potential targeted epitopes or any neutralizing mechanisms that can otherwise be probed using the traditional virus-based neutralization assay. Therefore, our review demonstrates a need to establish clear in vitro EBV neutralization protocols, reporting standards, and guidelines so that universal surrogate correlates of immune protection can be compared across in vitro efficacy studies. Moving forward, we propose that all studies perform virus-based neutralization assays using a full panel of established EBV-susceptible cell lines, when possible, especially if the vaccine tested targets multiple antigens that might be important for the infection of different cell types. In addition, viral titer information for the EBV batch used and the serum/plasma/antibody concentration in which the virus is incubated should be clearly reported; otherwise, an accurate assessment of vaccine efficacy and immune protection correlates is not possible.

Most studies in our analysis focused on assessing serum IgG responses, but it is important to note that mucosal IgA responses might also be a relevant source of protective nAbs, given that EBV is transmitted via the oral mucosa. In our assessment, we found that only 1/36 of the pre-clinical studies, and none of the clinical studies, assessed vaccine-induced IgA responses. In the one study that did assess IgA responses, these responses were measured in the serum [(45), Table 3 ]. A recent study assessing the presence of EBV glycoprotein-specific antibodies in patients with EBV-associated nasopharyngeal carcinoma, found that nAb activity in both epithelial and B cells was most strongly correlated with glycoprotein-specific IgGs rather than IgAs in the serum. However, IgA-related nAb activity might be most relevant in the saliva, as saliva is the first line of defense against EBV once the virus enters the oral mucosa. Since IgGs can also be found in saliva, both salivary IgAs and IgGs might be important to provide protection at the site of EBV infection (90). A recent study using the EBV-homologue rhesus lymphocryptovirus (rhLCV) model (described in Section 4.2 below) showed that oral transfer of glycoprotein-specific monoclonal IgGs can provide partial protection against rhLCV infection in rhesus macaques, demonstrating the importance of antibodies in the oral mucosa in infection prevention (91). While vaccine-induced salivary Ig responses might be difficult to study in small animal models, future pre-clinical non-human primate (NHP) studies and clinical studies should also assess both salivary and serum IgA/IgG responses. This will lead to better understanding of correlates of immune protection.

EBV is a human-tropic virus, making it difficult to study in vivo. Rodents, rabbits, and various NHPs have been historically used to test EBV vaccine immunogenicity ( Table 4 and Figure 4C ). However, only the cotton-top tamarin and common marmoset have been used to test in vivo vaccine efficacy. The cotton-top tamarin was the first animal to be established as an experimental model for EBV infection and the development of EBV-driven lymphoma (92). In 1985, the model was used for the first time to test the in vivo efficacy of a gp350 vaccine candidate (76). After that, 7 more studies used the model to test the efficacy of other gp350 vaccine candidates ( Figure 4C ), with the last study in 1994, as the model was abandoned in the early 2000s due to the endangered status of the species. Perhaps the greatest utility of this model was that it offered a direct readout of virus-associated disease (i.e., lymphoma) that was easy to measure in a relatively short timeline. However, the translatability of this model is limited because it did not truly recapitulate EBV infection in humans in terms of infection route, disease pathogenesis, or immune response to infection. The common marmoset, established as an experimental model of EBV infection in parallel with the cotton-top tamarin, was first used to test in vivo vaccine efficacy in 1989 (69, 93). Compared to the cotton-top tamarin model, this model is more translatable to human EBV infection (94, 95); nevertheless, only 2 more studies used it to test vaccine efficacy in the late 1990s, and it has since been abandoned for unclear reasons.

As a consequence of this dearth of in vivo models, in vitro neutralization assays using sera antibodies from immunized animals have been used as the primary vaccine “efficacy” readout in most pre-clinical EBV vaccine studies ( Table 3 ), as described above. Humanized mice with a reconstituted human immune system are also susceptible to EBV infection of B cells, and like cotton-top tamarins develop lymphoma upon infection, providing a relatively quick readout of disease. However, this model suffers from the same drawbacks as cotton-top tamarins in terms of dissimilarities to human EBV infection; their epithelial cells are not susceptible to infection and their use is more suited for testing therapeutics rather than prophylactics. Rabbits have also been reported as susceptible to EBV infection (96–99), but no EBV-associated pathologies have been shown, and as humanized mice, their epithelial cells are not susceptible to infection (100). Similarly, Chinese tree shrews were recently reported to be susceptible to EBV B cell infection (101), but a previous study by the same group suggests that their epithelial cells are not infected by EBV (102). Both rabbits and tree shrews have yet to be used as challenge models to test EBV vaccine efficacy.

A promising alternative to test the efficacy of EBV vaccine candidates is to use rhesus macaques infected with rhesus lymphocryptovirus (rhLCV) as a surrogate for EBV infection. rhLCV shares high amino-acid homology with EBV (103), and rhLCV genes can complement the viral activities of EBV orthologs (104). Importantly, rhLCV infection in rhesus macaques recapitulates most aspects of EBV infection in humans, including the routes of infection, use of homologous glycoproteins to infect the host, disease establishment, and host immune responses (105). Furthermore, rhLCV-infected rhesus macaques that become immunosuppressed develop virus-associated malignancies of both epithelial and lymphoid origin, similar to immunosuppressed EBV-infected humans (106). The similarities between EBV and rhLCV were underscored in a recent study from the McGuire group, in which AMMO1, an EBV-specific nAb, was able to reduce rhLCV infection in rhesus macaques that had been passively infused with the nAb and subsequently challenged with rhLCV (107). Thus, the rhLCV infection model in rhesus macaques offers an opportunity to test the efficacy of EBV vaccines in the form of surrogate rhLCV vaccines in both immunocompetent and immunosuppressed settings, and future EBV vaccines studies should consider using this model to validate their in vitro findings.

The field could also re-evaluate the use of the common marmoset as an EBV challenge model. Early EBV studies using this model demonstrated the establishment of EBV infection in challenged animals and uncovered pathogenic similarities between common marmoset and human EBV infection (93). For example, infected common marmosets sporadically shed the virus in the oral mucosa and can transmit it to EBV-naïve animals through close contact, similar to how EBV transmission is thought to occur in humans (60). Infected common marmosets also share similar EBV antibody kinetics to infected humans (94). With the support of more modern EBV diagnostic methods and the rapid growth of common marmosets as a research NHP, this model has the potential to provide a powerful tool for testing EBV vaccine efficacy.

Use of NHPs such as rhesus macaques and common marmosets in EBV research will also facilitate the study of salivary antibodies to arrive at a better understanding of correlates of immune protection. However, one commonly overlooked caveat to the use of NHPs in this field, whether to test general vaccine immunogenicity or vaccine efficacy in a challenge model, is that all NHPs should be screened for infection with EBV-homologue LCVs. Cynomolgus macaques, rhesus macaques, and common marmosets are all naturally infected by species-specific LCVs (108), and as is the case in humans, infection is usually pervasive due to the high transmissibility of these viruses. This presents a problem due to the potential antibody cross-reactivity between the viruses, which could skew the assessment of any anti-EBV response induced in immunized animals. In our evaluation, only 1/16 studies that used NHPs acknowledged this issue [(50), Table 3 ], and no NHP study reported the use of LCV-negative animals. Thus, it is critical that moving forward all NHPs should be screened for native LCVs prior to any EBV vaccine immunogenicity studies. Development of specific-pathogen-free NHP colonies worldwide that are LCV-negative would be ideal.

To date, there is still no consensus as to which glycoprotein(s) would serve as ideal vaccine antigens. Before the EBV glycoproteins were identified, early studies used whole virus or undefined membrane antigens from EBV-infected cells to begin exploring the general immunogenicity of EBV (109–111). By 1979, the analysis of purified cellular membranes from EBV-infected Raji cells allowed separation of various membrane-associated antigens with different molecular weights, which later became known as gp350/220, gB and gH (112). Despite this discovery and the identification of various nAbs specific to different glycoproteins in the 1980s and 2000s, most pre-clinical vaccine studies and all four clinical studies that followed focused on gp350 as the main vaccine target up until the 2010s, when other glycoproteins began to be seriously explored as additional potential vaccine antigens. The likely culprit for this focus seems to be the 1980 study by North, Morgan and Epstein, where neutralizing sera from rabbits immunized with whole virus was used to immunoprecipitate antigens from EBV producer cell lines, and gp350 was the only antigen precipitated at a detectable level (113). Furthermore, two gp350-specific nAbs were isolated and reported that same year ( Table 1 ), strengthening the support for a gp350-based vaccine. Most gp350 vaccine studies that followed did result in the generation of gp350-specific antibodies with neutralizing capabilities in vitro, but no pre-clinical in vivo vaccine efficacy study has definitively shown protection against primary EBV infection. Ragot et al. (64) and Morgan et al. (70) demonstrated that gp350-expressing adenovirus and ISCOM-based gp350 nanoparticles fully protected cotton-top tamarins from developing EBV-induced lymphoma, respectively, but only n=4 animals were tested in each study ( Table 3 ). Epstein et al. (76) similarly showed that cotton-top tamarins were fully protected from lymphoma when administered purified B95-8 membrane, but only n=2 animals were tested in this study. Thus, in these few cases showing positive results, too few experimental animals were tested to provide definitive conclusions, and Epstein et al. (76) also mentioned the presence of trace amounts of EBV in the administered vaccine product, which could have resulted in the emergence of nAbs against additional glycoproteins, skewing the results. Furthermore, there is no guarantee that protection from disease means primary infection did not take place; Morgan et al. (68) measured EBV infection in the blood and showed that although all cotton-top tamarins (n=4) were protected against lymphoma, only half were protected from infection ( Table 3 ). Epstein et al. (76) demonstrated that cotton-top tamarins administered gp350 liposomes were ultimately protected against disease (n=2), but they developed transient lesions during the observation period ( Table 3 ), suggesting that primary infection did take place.

This pattern continued in the clinical setting, and the first EBV vaccine clinical trial in 1995, which used a gp350 vaccinia-based vaccine, did not result in sterilizing immunity in vaccinated infants (83). Similar to the inconclusive results of the pre-clinical studies, this trial suggested that gp350 as a single vaccine antigen might not be sufficient to protect against EBV infection. Then, in 2000, the Delecluse group showed that recombinant EBV lacking gp350 could still infect both epithelial and B cells in vitro (114), suggesting that EBV could remain infectious even in the presence of gp350-specific nAbs. Nevertheless, three more clinical trials using gp350 vaccines were put forward in the 2000s (80–82). Although these vaccines reduced the incidence of IM in some cases, they did not provide sterilizing immunity against infection.

In the 2000s, several groups began to fully uncover the mechanisms of EBV entry in both epithelial and B cells, and the importance of the core fusion glycoproteins, gp42, gH/gL, and gB became apparent, as summarized in Figure 1 . Our group and others have shown that gp350 binds complement receptor type 1 (CD35) (115) or type 2 (CD21) (116) on B cells and triggers endocytosis of the virions (30). Although this interaction is not essential as discussed above, it does enhance infection (114). Once the virus is bound to its target cell, the fusion machinery is required to achieve virus–cell fusion. gB is highly conserved among herpesviruses and is considered the core fusogen, achieving fusion through attachment to the host receptor neuropilin 1 (117). gB exists in two conformations, the pre- and the post-fusion states (118), and it is unclear which conformation might be best suited as a vaccine antigen (i.e., is more immunogenic) in the case of vaccines that rely on conformational epitopes; this remains an area of active investigation. In epithelial cells, the fusogenic activity of gB is triggered by the interaction of the gH/gL complex with the host ephrin A2 receptor (117, 119, 120). In B cells, gp42, in complex with gH/gL, activates gB after interacting with major histocompatibility complex (MHC) class II; in this way, the level of gp42 expression on the virion confers host cell specificity, promoting the infection of B cells and inhibiting the infection of epithelial cells (121, 122). Importantly, although it can mature and egress upon lytic induction, recombinant EBV lacking gH cannot infect epithelial cells, and recombinant EBV lacking gp42 or gH cannot infect B cells, suggesting that these glycoproteins are essential for EBV infection (37, 123). nAbs against these glycoproteins were also discovered between 1982 and 2000 ( Table 1 ). Despite these discoveries, there have only been a few pre-clinical studies exploring non-gp350 glycoproteins as vaccine antigens ( Figure 4A ).

Given what we now know about the EBV infection process and how several glycoproteins collaborate to infect different types of target cells, we expect that the most effective vaccine strategy would be a multivalent approach targeting multiple important glycoproteins. Indeed, the fusion machinery constitutes the most obvious candidate to test in a multivalent EBV vaccine, but gp350 remains an important source of nAbs, and as such, it might still be an important vaccine target when used in combination with other antigens. Moreover, a recent study identified gp350 as the entry glycoprotein for T cells, which has implications for EBV-associated T and NK/T-cell lymphomas (124). Following this reasoning, the first glycoprotein combination vaccine was tested in 2008 by Lockey et al., which combined gp350 and gB (57). Although results were not conclusive on whether the combination was better than targeting gp350 alone, further combination studies that followed did provide support for a multivalent approach. For example, a recent pre-clinical study by the Cohen group reported better in vitro neutralization outcomes in mice after immunization with vaccines that combined the gH/gL and gp42/gH/gL complexes with gp350, compared to a vaccine that targeted gp350 alone; in the same study, a vaccine targeting the gp42/gH/gL complex resulted in better fusion inhibition outcomes than a vaccine targeting gH/gL complex alone (44). Similarly, in a recent study, our group reported that a multivalent vaccine incorporating gp350, gB, gp42, and gH/gL resulted in better neutralizing outcomes in immunized rabbits than a recombinant gp350 vaccine, with comparable results to immunization with UV-inactivated EBV (43). However, it is still not clear what combination of glycoproteins would best to prevent infection in vivo or whether incorporating all five glycoproteins is necessary. Thus, future studies should focus on identifying the required components of a multivalent EBV vaccine that can generate effective neutralizing activity both in vitro and in vivo. Indeed, ModernaTX Inc. has recently started recruiting participants for a Phase I clinical trial that will explore the safety and immunogenicity of their mRNA-based EBV vaccine, mRNA-1189, which targets multiple EBV glycoproteins (125–127).

While gp350, gB, gp42 and gH/gL have been the most studied entry glycoproteins, there are additional EBV envelope glycoproteins that could also prove to be important vaccine targets. BMRF2 has been reported to participate in viral entry by facilitating viral attachment to epithelial cells through its interactions with various host integrins (128, 129), and it has been implicated in cell-to-cell spread as well (130). There is data suggesting its potential to elicit nAbs (131), and it might be a relevant target if the virus escapes initial neutralization and breaches the oral mucosa, where it could spread from B cells to monocytes and to epithelial cells (132) in vivo. BDLF2 is co-expressed with BMRF2 (133) and thus might also be involved in cell-to-cell spread (134), but not much has been further uncovered about this glycoprotein. BDLF3/gp150 has been shown to bind to heparan sulfate proteoglyclans on the surface of epithelial cells, but this binding does not enhance infection (135), and its role in immune evasion may preclude any role in viral entry (136). BILF2 might be involved in glycoprotein transport (137), but otherwise is considered an orphan glycoprotein as no definite function has been uncovered for it (122). Future studies might reveal the potential utility of these additional glycoproteins as vaccine targets. Importantly, while the purpose of this review was to study previous prophylactic vaccine efforts focused on humoral immunity, we cannot ignore the possibility that a vaccine designed to target both humoral and cellular immunity could prove protective. Indeed, although we did not discuss the details, a few studies in our analysis [(46, 56, 57); Table 3 ] tested vaccines incorporating both glycoproteins and different targets of cellular immunity. The importance of these combinations remains to be tested in vivo. Furthermore, ideal vaccine regimen, dose, and route might differ between humoral and cellular immune targets, thus these parameters may be first tested separately for their immunogenicity before combining immune targets.

In our analysis, most pre-clinical trials ( Figure 4B , Table 3 ) and clinical trials ( Table 4 ) delivered EBV antigens as monomeric proteins, the most widely used platform in the earlier half of EBV vaccine research. Monomeric proteins are perhaps the least immunogenic of the vaccine platforms available, requiring the use of adjuvants to induce a robust immune response. The pre-clinical studies in our analysis used a wide variety of adjuvants, including aluminum hydroxide/alum, glucopyranosyl lipid A/GLA, stable emulsion/SE, SAF-1, and Freund’s adjuvant ( Table 3 ) (138). In addition to low immunogenicity, another limitation to the use of monomeric proteins is that the higher-level protein structure of the antigens might be impaired in the monomeric state, inhibiting the induction of conformation-dependent nAbs (139). It is perhaps for these reasons that, after a lag in pre-clinical prophylactic EBV vaccine development between 1999 and 2008, several additional antigen delivery approaches were adopted to incorporate protein structural support and to provide the ability to present multiple antigens per unit. For example, two pre-clinical studies by the Snapper group explored the use of multimeric proteins [(48, 54), Table 3 ]. In both studies, the multimeric protein vaccines were more immunogenic than their corresponding monomeric protein vaccines. The use of self-assembling nanoparticles incorporating EBV glycoproteins has also been explored. The Cohen group (44, 50) showed that ferritin-based nanoparticles incorporating gp350, gH/gL, or gH/gL/gp42 were more immunogenic than the monomeric soluble forms of these glycoproteins, and the National Institute of Allergy and Infectious Diseases is currently recruiting participants to test the safety and immunogenicity of the gp350-ferritin-based nanoparticle in a Phase I clinical trial (140). Nanoparticles in the form of liposomes were also tested in the early stages of EBV vaccine research [(74, 76–78), Table 3 ], which focused on gp350. The platform was not pursued further, perhaps due to its ultimate inability to prevent lymphoma in cotton-top tamarins after viral challenge; however, as discussed in Section 4.3, this result may have occurred due to the use of gp350 as a single immunogen rather than insufficiency of the platform. Other recent studies have explored the use of VLPs, a special type of nanoparticle that resembles viruses in structure and, like other nanoparticles, can present multiple antigens or epitopes in a repetitive array, increasing antigen immunogenicity (141). Importantly, VLPs lack viral DNA, which is critical in the development of vaccines for oncogenic viruses such as EBV. The success of VLPs as a vaccine platform is underscored by the many licensed VLP-based vaccines on the market, including vaccines against human papillomavirus and hepatitis B and E. Five pre-clinical studies have investigated the use of VLPs for EBV, including VLPs produced from EBV-packaging cell lines (56), VLPs based on Newcastle disease virus (49), and VLPs produced from the hepatitis B virus core antigen (42). Although vaccines that present proteins in a multimeric form, whether as multimeric proteins or nanoparticles, are a better alternative to monomeric proteins, they might be expensive to produce in large quantities and still require adjuvants to achieve an adequate level of immunogenicity.

Viral vectors have the advantage of possessing intrinsic immune-stimulating capabilities and thus do not require administration with adjuvants; in addition, they can be combined with VLP platforms to produce VLPs in vivo (142–144). The 6 pre-clinical EBV vaccine studies using viral vectors tested measles virus (55), adenovirus (64), and vaccinia virus vectors (57, 61, 70, 75) ( Table 3 ); the latter was also tested in one of the 4 EBV vaccine clinical trials [(83), Table 4 ]. In these studies, the vectors were engineered to express single antigens, but viral vectors with the necessary genetic insert capacity can be engineered to express multiple antigens. Measles virus and adenovirus vectors have good clinical track records, but they do not possess the genetic capacity to express multiple EBV antigens (145, 146). Vaccinia virus vectors, on the other hand, have the largest genetic insert capacity of all available viral vaccine vectors (>30 kb), and although the original vaccinia strains can pose serious health complications, the highly attenuated modified vaccinia Ankara (MVA) strain is safe even in immunocompromised populations and is highly immunogenic (147–149). Hence, future multivalent vaccine studies looking to use a viral vector-based platform should consider an MVA vaccine platform.

Although we did not encounter any studies within our search limits that utilized mRNA-based vaccines, the new Phase I clinical trial by ModernaTX Inc. that is currently recruiting participants to test a multivalent mRNA-based EBV vaccine will open the way for this technology to the field of EBV vaccine research (127). Having only recently been clinically tested at a large scale against SARS-CoV-2, questions about the response durability of mRNA-based vaccines remain (150), and whether this will be a successful approach for targeting multiple antigens in the context of EBV remains to be studied.

While not a concern for viral or nucleic-acid-based vaccines, an important consideration when producing protein antigens for vaccination is the choice of expression system. Post-translational modifications, such as glycosylation, are an important element that can affect antigen immunogenicity. Indeed, the Hayes group in 2015 studied the conformational requirements of gp350 to generate a nAb response [(51), Table 3 ], and found that glycosylation is essential to gp350 nAb generation, which can only properly be achieved in mammalian expression systems. This was previously demonstrated by two studies in 1991 and 1988 [(67, 72), Table 3 ] which explored expression of gp350 in bacteria and yeast, respectively, and found the resulting products incapable of generating nAbs, presumably due to the inability of these systems to perform complex glycosylation (151). Protein production in simpler systems such as bacteria and yeast do offer various advantages in terms of large-scale production feasibility, but unless glycosylation engineering approaches are applied in this context (152), mammalian cells remain the ideal production system for EBV glycoproteins.

A final consideration when it comes to vaccine platforms pertains to the feasibility of large-scale manufacturing, including processing, purification, and quality testing. An early EBV vaccine study by Epstein and his group [(74), Table 3 ] demonstrates the potential relevance of establishing appropriate processing methods. In this study, the authors used a protocol established by Randle et al., 1985 (153) to perform large-scale purification of gp350 from EBV producer cells using immunoaffinity chromatography; however, they found that the resulting product had a specific activity of 20% of that of a product prepared by molecular weight-based separation methods in small-scale, and failed to protect animals against EBV disease when delivered in liposomes. The authors attribute this to either the type of purification used resulting in antigen denaturation, or the isolated protein fraction not being immunogenic enough to confer protection as the previous formulation did. Today, biological manufacturing techniques have greatly improved, but the fact that different vaccine platforms possess different biochemical and structural qualities that require unique processing needs and quality testing should still be considered when designing a new vaccine that can be produced in enough quantity and with high enough quality for large-scale immunization (154). Thus, an ideal vaccine platform should not only be sufficiently immunogenic and protective, but also allow for the necessary infrastructure to facilitate future clinical translation.

Most of the pre-clinical studies included in our systematic review were of moderate, high, or very high quality ( Figure 4D and Table S3 ); all pre-clinical studies of poor or very poor quality were performed before the year 2000. Although there was a clear shift toward higher-quality studies after 1999, our analysis demonstrated that there is still room for improvement. For example, none of the studies analyzed performed a blinded analysis, which would reduce potential bias and enhance scientific rigor. In addition, only 3 studies explicitly evaluated the toxicity of the vaccine tested ( Table S3 ), which is essential for selecting vaccine candidates that are suitable for clinical use. As discussed in Section 4.1, it is particularly important to report neutralization assay details to accurately assess vaccine efficacy, including titer/infectivity information for the viruses used, but this was not the case for many of the analyzed studies. The use of both positive and negative control treatment groups and appropriately powered group sizes are also critical but were frequently not reported. Future studies should focus on these and other qualities ( Table S3 ) to generate high-caliber studies with accurate, generalizable conclusions.

While we aimed for our report to be as robust as possible, we do acknowledge a few limitations in our study. The small number of clinical trials included in our analysis, together with the variability in the quality ( Table S3 ) and content of the pre-clinical trials, did not allow us to perform a meta-analysis or any additional meaningful statistical analysis or comparison between the studies, so our results lack statistical relevance. In addition, although our search was thorough to the best of our efforts, it is possible that we might have missed articles that fit our selection criteria.

Finally, while we structured our discussion towards developing a humoral-based vaccine to prevent primary EBV infection, this goal might be unrealistic. No sterilizing immunity has yet been achieved in pre-clinical or clinical studies against any herpesvirus. In the case of EBV, recent passive immunization studies in humanized mice (155) and in rhesus macaques using the rhLCV model (91, 107), did not achieve sterilizing immunity, although these studies were not focused on multivalent approaches. Cellular immunity might also be an important component to improve protective vaccine responses, which we did not consider in our review. Nevertheless, even if sterilizing immunity cannot be achieved against EBV via vaccination, an EBV vaccine might still be effective in reducing the rates of EBV diseases and their associated morbidities and mortality, such as is the case for the approved varicella zoster virus vaccines (156), and the EBV gp350-based vaccine tested in a Phase II clinical trial that was 78% successful in preventing infectious mononucleosis [(81), Table 4 ].