Mice have played a central role in biomedical researches including those on infectious diseases, as an easily accessible and versatile animal model. Abundant knowledge of their genetics promoted their use as a model system. Many human viruses, however, do not infect mice. Although murine viruses such as the mouse herpesvirus 68 (MHV-68) that is related to Epstein-Barr virus (EBV) and Kaposi’s sarcoma-associated herpesvirus (KSHV), and murine cytomegalovirus (CMV) homologous to human CMV have been considered as models of human viruses, there are considerable diversity in biological activity and pathogenesis between murine and human viruses [1,2,3]. Introduction of transgenes encoding virus receptors to mice enabled viruses such as adenoviruses [4], measles virus [5], coronavirus [6,7], and hepatitis C virus [8], to infect mice, but this approach did not remove the species barrier in many other viruses including human immunodeficiency virus 1 (HIV-1). A breakthrough in this issue was brought by the development of the scid-hu thy/liv mouse and the scid-hu PBL mouse in 1988 both of which are based on the C.B-17 scid mouse. Due to a mutation in the gene coding for a subunit of the DNA-dependent protein kinase that is essential for recombination of both the T-cell and B-cell receptor genes, C.B-17-scid mice lack both T and B cells and do not have the ability to reject human tissues and cells [9]. Scid-hu thy/liv mice were prepared by transplanting tissues of human fetal thymus and liver into the renal capsule of C.B-17 scid mice and allowed de novo generation of human T cells including the CD4⁺ compartment [10]. Scid-hu thy/liv mice were primarily used as an In vivo model of HIV-1 infection that targets mainly CD4⁺ T cells and monocytes/macrophages. In addition, scid-hu thy/liv mice were used for modeling of human herpesvirus 6 [11], human cytomegalovirus [12], Kaposi’s sarcoma-associated herpesvirus (KSHV) infections [13], measles virus [14], and varicella-zoster virus [15]. Various key features of human HIV-1 infection, including selective depletion of CD4⁺ T cells, were reproduced in scid-hu thy/liv mice [16,17,18]. Scid-hu PBL mice, on the other hand, were prepared by transplanting human peripheral blood mononuclear cells (PBMC) to the peritoneal cavity of C.B-17 scid mice and efficient engraftment of most components of the human immune system including T and B cells was observed [19]. Scid-hu PBL mice were used for modeling of HIV-1, human T-lymphotropic virus 1 (HTLV-1), and EBV infections and recapitulated many features of their pathogenesis [20,21,22,23]. There were, however, limitations in these early-generation humanized mouse models: human lymphocytes generated in scid-hu thy/liv mice did not distribute systemically and therefore intrathymic or intranodal HIV-1 inoculation was necessary and virus infection was mostly confined in the inoculated organs; scid-hu PBL mice were prone to graft versus host disease; both scid-hu thy/liv and scid-hu PBL mice did not mount primary immune responses to viruses. These disadvantages were overcome in new-generation humanized mice developed in this decade [24,25,26,27]. New-generation humanized mice were prepared by transplanting human hematopoietic stem cells (HSC) to mice of various severely immunodeficient strains (e.g., NOD/Shi-scid Il2rg ^null (NOG), Balb/c Rag2^−/−Il2rg^−/− (BRG), and NOD/LtSz-scid Il2rg^−/− (NSG)) and major components of the human immune system, including T cells, B cells, NK cells, dendritic cells, and monocytes/macrophages, were reconstituted. In one protocol (BLT (bone marrow, liver, thymus) mouse), tissues of human fetal liver and thymus, as well as HSC derived from the same liver were transplanted and enabled proper intrathymic selection of human T cells [26]. These new-generation humanized mice can be infected with EBV [24,26,28,29,30,31], HIV-1 [32,33,34,35,36], HTLV-1 [37,38,39], dengue virus [40,41], herpes simplex virus type 2 (HSV-2) [42], and Kaposi’s sarcoma-associated herpesvirus (KSHV) [43]. In addition, humanized mice provided an opportunity to analyze human anti-influenza virus immune responses in mice, although infection with influenza virus per se had been feasible with regular mice [44,45]. Besides these immune system-humanized mice, immunodeficient mice harboring functioning human hepatocytes were developed and shown to be an excellent model for both hepatitis B and C virus infections and pathogenesis [46,47,48]. Furthermore, humanized mice harboring both human hepatocytes and human immune system components were prepared that could reproduce HCV-specific immune responses and pathogenesis [49]. Various immunodeficient mouse strains used in In vivo modeling of human virus infections are summarized in Table 1. In this review, we focus on humanized mouse models of EBV infection, especially new-generation humanized mouse models, and describe how they have been utilized for the reproduction of key features of human EBV infection, including pathogenesis, immune responses, and latent infection. Applications of these models, including preclinical studies on therapeutic strategies, are also discussed.

EBV is a ubiquitous lymphotropic γ herpesvirus infecting more than 90% of the adult population in the world [64]. Although primary EBV infection in childhood is mostly asymptomatic, that in adolescents and young adults results in infectious mononucleosis (IM) in at least 25% of cases [64]. Whether or not accompanied by IM, EBV infection results in life-long persistent infection. In a restricted fraction of hosts, EBV is etiologically involved in a variety of diseases, including lymphoproliferative diseases (LPDs) in immunocompromised patients such as posttransplant lymphoproliferative disease (PTLD) and AIDS-associated lymphoma, malignant lymphomas such as endemic Burkitt lymphoma and Hodgkin lymphoma, epithelial malignancies such as nasopharyngeal carcinoma and gastric carcinoma, and T/NK-cell LPDs such as chronic active EBV infection (CAEBV) and EBV-associated hemophagocytic lymphohistiocytosis (EBV-HLH). In addition, substantial evidence, although largely indirect at present, has accumulated implicating EBV in the pathogenesis of autoimmune diseases such as rheumatoid arthritis (RA) and multiple sclerosis. The list of most EBV-associated diseases is presented in Table 2. Some of these EBV-associated diseases show uneven geographical distribution: EBV-positive Burkitt lymphoma is endemic in equatorial Africa and New guinea; nasopharyngeal carcinoma is highly prevalent in southern China; EBV-associated T/NK LPDs are particularly prevalent in east Asia and natives in Central and South America; and parotic carcinoma is endemic in Alaskan Inuits. Obviously, etiological cofactors in addition to EBV, both environmental and host factors, are involved in the pathogenesis of these EBV-associated diseases, although majority of them are still unknown. EBV has a unique biological activity to transform human B lymphocytes into lymphoblastoid cell lines (LCLs) with the capacity of indefinite proliferation in culture. These EBV-transformed lymphoblasts are readily removed by the virus-specific cytotoxic T lymphocytes (CTL) in imunocompetent hosts [65,66]. In immunocompromised hosts like transplant recipients and AIDS patients, however, these transformed cells may proliferate unlimitedly to cause B-cell LPDs such as PTLD and AIDS-associated lymphoma.

EBV genome encodes more than 80 genes and nine proteins (EBV nuclear antigens (EBNAs) 1, 2, 3A, 3B, 3C, and LP, and latent membrane proteins (LMPs) 1, 2A, and 2B) and two sets of untranslated RNAs (EBV-encoded small RNAs (EBERs) 1 and 2, and BamHI-A rightward transcripts (BARTs)) are expressed in B lymphoblastoid cells transformed by EBV. All or a part of these latent-cycle EBV proteins and RNAs are expressed in EBV-associated malignancies and LPDs. The pattern of EBV gene expression can be classified into three types: latency I, characterized by the expression of EBNA1, EBERs, and BARTs, is seen in Burkitt lymphoma and gastric carcinoma (some gastric carcinomas express LMP2 in addition); latency II, characterized by the expression of EBNA1, LMP1, LMP2A, EBERs, and BARTs, is found in CAEBV, Hodgkin lymphoma, and nasopharyngeal carcinoma; latency III, characterized by the expression of all nine EBV proteins and two sets of EBV RNAs expressed in LCLs, is seen in PTLD and AIDS-associated lymphoma [67].

EBV naturally infects only humans and has been reported to infect cotton-top tamarins, common marmosets, owl monkeys, and rabbits in experimental setting [68,69,70,71]. Cotton-top tamarins develop B-cell lymphomas following inoculation with EBV and the efficacy of candidate vaccines against the virus was evaluated in this model [72]. In spite of their potential value as an In vivo model of human EBV infection, they are endangered species and cannot be used in large numbers. A rhesus monkey lymphocryptovirus highly homologous to EBV was shown to reproduce key features of human EBV infection upon primary infection in naive monkeys, including oral transmission, atypical lymphocytosis, lymphadenopathy, expression of the B-cell activation marker in the peripheral blood, and sustained antibody responses to viral proteins [73]. This animal model has thus made significant contribution in the studies on EBV pathogenesis and treatment, although it has disadvantages common to primate models such as limited accessibility and high costs [74]. Intrinsic differences between the rhesus lymphocryptovirus and EBV needs also to be considered. Mouse models for EBV infection have not been available until the development of scid-hu PBL mice. Since the major target of EBV is lymphocytes, humanized mice are readily infected with the virus and EBV was actually the first human virus shown to infect new-generation humanized mice.

Analysis on immune responses to EBV infection in humanized mice was made possible by the development of new-generation humanized mice [24,26]. Reconstitution of human B cells, T cells, and dendritic cells was achieved in BRG mice transplanted with human HSC as described above. Upon infection with EBV the number of T cells significantly increased in the spleen and CD8⁺ T cells isolated from infected mice were shown to proliferate vigorously when stimulated with autologous EBV-transformed LCL, although EBV specificity of this response was not confirmed [24]. In BLT mice developed by Melkus and others [26], the presence of human thymus tissue enabled restriction of T cells by human MHC and efficient EBV-specific T-cell responses was demonstrated.

Subsequent to the pioneering works described above, EBV-specific T-cell responses have been studied in humanized mice of various strains [30,31,52,53]. EBV-specific T-cell clones of both CD4⁺ and CD8⁺ phenotypes were established from hu-NSG mice infected with the virus and specific lysis of HLA-matched LCL was demonstrated [31]. A few studies confirmed that antibodies specific to human class-I MHC inhibits production of IFN-γ by CD8⁺ T cells following stimulation with autologous LCL, indicating that the T cells isolated from infected mice recognize EBV epitopes presented by the human MHC class I [26,30]. Furthermore, human MHC class I tetramers presenting EBV epitopes identified EBV-specific CD8⁺ T cells in hu-NOG mice [53]. Since depletion of either CD4⁺ or CD8⁺ T cells by administration of anti-CD4 or anti-CD8 antibodies, respectively, caused more aggressive proliferation of EBV-infected cells and reduced the life span of infected mice, T-cell responses induced in humanized mice are thought to play a protective role [31,52]. One study further showed that CD8⁺ cells isolated from EBV-infected mice have the ability to suppress transformation of autologous B cells by EBV [52].

Induction of EBV-specific T-cell responses in BLT mice is readily acceptable because the presence of human thymus tissue is expected to allow proper selection of T cells by human MHC. However, T-cell responses specific to EBV have been demonstrated not only in BLT mice but also in other humanized mice without transplanted human thymus tissues [30,31,52,53]. Furthermore, T-cell responses specific to HIV-1 [104], an adenovirus vector expressing HCG glycoproteins [59], HSV-2 [42], and influenza virus [44] have been demonstrated in humanized mice without human thymic tissue. Since the positive selection of T cells in the thymus is thought to occur on the surface of epithelial cells and epithelial cells in humanized mice are of the murine origin, it is puzzling how T cells have developed with the capacity to recognize epitopes in the context of human MHC. Watanabe and others recognized a significant delay in the development of human T cells in humanized NOG I-A^−/− mice and this suggests that murine MHC is involved in the positive selection of human T cells [105]. Nevertheless, the fact that the development of human T cells was retained in humanized NOG I-A^−/− mice, although delayed and in a reduced number, suggests that human HSC-derived cells, possibly B cells or dendritic cells, are also involved in the positive selection of human T cells, explaining the T-cell responses restricted by human MHC in humanized mice [105]. One way to clear the problem of T-cell education is to introduce a human MHC transgene to humanized mice. Two NSG-derived mouse strains with human HLA-A2 transgene were developed and transplanted with HSC having the same HLA allele. Humanized mice thus obtained mounted efficient T-cell responses following EBV infection and restriction by HLA-A2 was demonstrated [31,93]. Although not tested in EBV infection so far, the Rag1^−/−γc^−/− mouse with HLA-DR4 transgene were shown to have more human CD4⁺ T cells than control Rag1^−/−γc^−/− mouse following reconstitution with HLA-DR4-positive HSC [106].

Antibody responses to EBV were analyzed in hu-NOG mice infected with EBV [30]. Compared with T-cell responses, antibody responses were much less effective; by examining 40 infected mice only three mice were found producing IgM antibody to P18^BFRF3, a major component of the viral capsid antigen (VCA). No IgG antibody to EBV-encoded proteins were detected. Antibody responses to HIV-1 [32,33,104], dengue virus [41], and HSV-2 [42] have also been described in humanized mice. Although IgG response was reported for dengue virus and HSV-2 [41,42], most of these antibody responses have been restricted to IgM. This inefficiency in antibody response, especially that in IgG response, is probably not due to intrinsic defects in B cells in humanized mice because transfer of functional human T cells expressing TCR specific to hemagglutinin (HA) improved HA-specific IgG response in hu-NOG mice [105]. It was suggested that suboptimal interaction between T and B lymphocytes underlay the inefficient production of antigen-specific IgG antibodies in hu-NOG mice [105]. The lack of human follicular dendritic cells in humanized mice may be also involved in the inefficiency of class switching [24].

EBV infection in immunocompetent hosts is mostly asymptomatic and EBV-infected lymphoblastoid cells with potential to unlimited proliferation are efficiently removed by the virus-specific CTL because they express highly immunogenic proteins such as EBNAs 3A, 3B, and 3C. In EBV latency in humans maintained by T-cell immunosurveillance, EBV reside in memory B cells where all viral protein expression is shut down, rendering them invisible to CTL. Persistent infection reminiscent of EBV latency in humans was reproduced in hu-NOG mice inoculated with low doses of the virus. Majority of hu-NOG mice inoculated with EBV of less than 10¹ TD₅₀ remained normal and survived for more than six months without apparent signs of diseases [30]. EBV DNA was detected in the peripheral blood only transiently for the several weeks following infection. When these mice were sacrificed more than six months post-infection, no macroscopic pathological changes were observed, yet a low number of EBER-positive cells were found in the tissues of spleen and lymph nodes. These cells were shown CD20-positive, but their morphology, with rather large cytoplasm, was not consistent with that of resting memory B cells [30]. RT-PCR analysis of RNA obtained from the spleen or liver of these persistently infected mice showed the expression of EBNA1, EBNA2, LMP1, and LMP2A, being consistent with the presence of latency III cells (Yajima et al. unpublished results). Thus, persistent EBV infection in hu-NOG mice did not appear to recapitulate all aspects of EBV latency in humans. Since a number of deficiencies have been observed in the functions of B cells in humanized mice [105], reproduction of bone fide EBV latency in memory B cells may require more sophisticated humanized mice. Nevertheless, it is an interesting question how immune responses are involved in the induction and maintenance of this persistent EBV infection in hu-NOG mice. Interestingly, EBV DNA level in the peripheral blood fluctuated in a few persistently infected mice and there the rise in EBV DNA level was immediately followed by the increase in CD8⁺ T cells and subsequent decline of EBV DNA level, suggesting an effective T-cell control of EBV-infected cells [52]. Trials to disrupt this persistent infection by immunosuppressive measures and induce EBV-associated LPD are underway. Cocco and others characterized EBV gene expression, surface marker expression, and hypermutation of immunoglobulin variable region in a single cell level in lymphoid tissues of humanized BRG mice infected with EBV [29]. They could identify EBV-infected cells of all three types of EBV latency (I, II, and III) in specific correlations with the location in the tissue and presence of hypermutation. Thus, the exact route by which EBV establishes latent infection in memory B cells, which is at present unclear, may be clarified in experiments using humanized mice.

Among the nine EBV proteins and the two sets of untranslated RNAs expressed in immortalized lymphoblastoid cells, EBNA2, EBNA3A, EBNA3C and LMP1 have been shown to play essential roles in the process of transformation, whereas knocking-out of the EBNA3B gene by homologous recombination did not affect the in vitro transforming ability of the virus [67]. Virus replication was not affected either. Nevertheless, as EBNA3B is well conserved in fresh clinical EBV isolates, a critical role for EBNA3B in the life cycle of EBV had been supposed. Recent work by White and others demonstrated that an EBV mutant with its EBNA3B gene knocked out induces more aggressive LPD in hu-NSG mice, suggesting that EBNA3B functions as a tumor suppressor gene [82]. B cells infected with this mutant virus secreted less T-cell chemoattractant CXCL10 and thereby escaped T-cell mediated killing. EBNA3B may thus help the host to control EBV-induced lymphoproliferation so that the virus should not give a life-threatening harm to the host. These findings were possible only in In vivo experiments with humanized mice and points to an important area for their application, namely In vivo characterization of virus mutants.

BZLF1 is an immediate-early gene of EBV and acts as a switch from the latent to lytic cycle of EBV infection. Knocking-out of BZLF gene did not affect the virus’ ability to transform B cells in vitro and the involvement of BZLF1 in lymphomagenesis was not expected until experiments with humanized mice became feasible. Ma and others prepared an EBV recombinant with the BZLF1 gene knocked-out and that with enhanced BZLF1 expression and compared the efficiency of lymphoma genesis in BLT NSG mice [62,63]. The results clearly indicated that BZLF1 enhances lymphoma genesis by inducing abortive lytic infection.

There are a number of EBV genes such as BHRF1 (encoding an Bcl-2-like anti-apoptotic protein) [107], BXLF1 (encoding EBV thymidine kinase) [108], BCRF1 (encoding viral IL-10) [109], loss-of-function mutants of which exhibited no or only minor phenotypic alteration in in vitro studies. Examination of these EBV mutants in humanized mice may reveal critical roles for these genes in EBV life cycle and pathogenesis.

Although B cells are the major target of EBV, in a group of diseases termed EBV-associated T/NK-cell LPDs, including CAEBV and EBV-HLH, the virus is mainly found in T or NK cells proliferating oligoclonally or monoclonally. CAEBV is characterized by prolonged IM-like symptoms, unusual patterns of antibody responses to EBV, and elevated EBV DNA load in the peripheral blood [110,111,112]. In the WHO classification of lymphomas [113], CAEBV corresponds largely with the systemic EBV⁺ T-cell lymphoproliferative diseases of childhood. Although monoclonal proliferation of EBV-infected cells implies malignant nature of the disease, chronic clinical time course and absence of morphological atypia in proliferating cells contradicts this notion, and the pathogenesis of this disease is largely unresolved. Overproduction of cytokines by EBV-infected T or NK cells and reacting T cells and macrophages is thought to be responsible for systemic inflammatory symptoms in CAEBV. Although it is still not possible to transform human T and NK cells in vitro with EBV to establish immortalized cell lines, the nature of these diseases strongly suggests that in a specific condition EBV can infect T and NK cells and induce their proliferation. EBV infection of T and NK cells has not been reproduced so far in humanized mice and recapitulation of EBV-associated T/NK LPD in mice required xenotransplantation of PBMC derived from patients. Imadome and others transplanted PBMC isolated from patients with CAEBV and EBV-HLH to NOG mice and succeeded in reproducing major features of these diseases including systemic monoclonal proliferation of EBV-infected T or NK cells and hypercytokinemia [99]. Many features were common to CAEBV and EBV-HLH model mice, but the findings of hemorrhagic lesions and extreme hypercytokinemia were unique to the latter model. Importantly, these models revealed an essential role of CD4⁺ T cells (whether or not infected with EBV) in the In vivo proliferation of EBV-infected T and NK cells and depletion of CD4⁺ T cells by administrating OKT-4 antibody just following transplantation of PBMC effectively prevented the engraftment of EBV-infected cells [99]. Furthermore, administration of OKT-4 antibody after engraftment of EBV-infected cells reduced peripheral blood EBV DNA load to undetectable level (Imadome and others, unpublished results). These results suggest therapeutic approaches targeting CD4⁺ T cells may be possible.