Despite many theoretical incompatibilities between mouse and human cells, mice with reconstituted human immune system components contain nearly all human leukocyte populations. Accordingly, several human-tropic pathogens have been investigated in these in vivo models of the human immune system, including viruses such as human immunodeficiency virus (HIV) and Epstein-Barr virus (EBV), as well as bacteria such as Mycobacterium tuberculosis and Salmonella enterica Typhi. While these studies initially aimed to establish similarities in the pathogenesis of infections between these models and the pathobiology in patients, recent investigations have provided new and interesting functional insights into the protective value of certain immune compartments and altered pathology upon mutant pathogen infections. As more tools and methodologies are developed to make these models more versatile to study human immune responses in vivo, such improvements build toward small animal models with human immune components, which could predict immune responses to therapies and vaccination in human patients.
The complexity of infections and the corresponding elicited immune responses are best investigated in animal models that allow the manipulation of the timing and dose of infection, as well as of the responding immune compartments. Small animal models, such as the mouse, are preferred for these types of investigations due to low costs and ease of handling. However, divergent evolution between these small mammals and humans in the past 65 million years has rendered the immune system the third most different organ system between the two species, after olfaction and reproduction . Many of these differences are found in the innate immune system, which ensures the initial survival of the infected host and also recognizes pathogens by their molecular patterns . This divergence probably results from the different infectious disease challenges associated with the respective ecological niches that these two species inhabit. Unfortunately, these differences between the mouse and human immune systems also result in dissimilar inflammatory responses to burns, trauma, and endotoxemia at the gene expression level, such as integrin, ICOS-ICOSL, CD28, and PKCΘ signaling . Therefore, alternatives to classical mouse models, which more closely model human immune system behavior during infection in vivo, would be of significant benefit for the development of immunomodulatory treatments.
The category of new models, which comes closest to achieving this goal, is mice with reconstituted human immune system components. These mice are mainly generated by neonatal injection of human hematopoietic progenitor cells in mice that lack murine innate and adaptive lymphocytes, namely NOD-scid γc−/− (NSG), NOD-scid γctm1sug, NOD Rag1−/− γc−/−, or BALB/c Rag2−/− γc−/− (BRG) mice  (Fig. 1). For some studies, a fetal organoid of liver and thymic tissue is implanted under the kidney capsule, which together with the i.v. injection of human hematopoietic progenitor cells generates BM liver thymic mice . In all of these models, cellular components of the human immune system develop over several months, including human T cells, B cells, natural killer (NK) cells, monocytes, macrophages, and dendritic cells (DCs) [6-8]. However, the degree of human immune system component reconstitution differs significantly between these mouse strains, with 60% of mononuclear cells being of human origin in the spleen and blood of NSG, NOD-scid γctm1sug, and NOD Rag1−/− γc−/− mice 3 months after human hematopoietic progenitor cell transfer, while in BRG mice only 20% are of human origin at this time point [9, 10]. This difference in the proportion of mononuclear cells of human origin among the various mouse models results at least in part from the polymorphism among mouse strains in signal regulatory protein-α (SIRP-α), an inhibitory receptor on mouse myeloid cells. This receptor recognizes human CD47 in the NOD mouse background and thereby prevents phagocytosis of human cells by the mouse myeloid compartments, which are still intact in all these mouse backgrounds . Indeed, when human or NOD-mouse signal regulatory protein-α is transgenically introduced into BRG mice, or when BRG mice are reconstitute with human hematopoietic progenitor cells that are transduced to express mouse CD47, human immune system reconstitution is similar to that in NSG mice [12, 13]. In particular, human T-cell and NK-cell reconstitution is very sensitive to optimal reconstitution of the other human immune compartments, such as dendritic cells, but comprise up to 60 and 5% of human CD45-positive cells, respectively [9, 14, 15]. All subsets of these lymphocyte populations are present after reconstitution, including αβ and γδ T cells, thymic T-cell developmental stages, and CD56bright as well as CD56dim NK cells [8, 15]. Although all these human immune system compartments can be reconstituted in NSG and BRG mice, it is important to point out that reconstitution can greatly vary between laboratories and even within the same laboratory, due to variations in the CD34+ hematopoietic progenitor cell donors and, especially, when limiting numbers of these cells are used for reconstitution. Nevertheless, reconstitution can reach 1–2 × 107 human leukocytes per mouse spleen  and, therefore, match cellularities that are observed in WT C57BL/6 and BALB/c animals . Thus far, human DC, NK-cell, and T-cell responses against human pathogens can be modeled effectively in mice with human immune system components, and their in vivo responses to human pathogens will be discussed in this review.
Viral infections of humanized mice
Among viruses that infect humans, human immunodeficiency virus (HIV) and Epstein-Barr virus (EBV) infection have been most extensively investigated in mice with human immune system components. However human cytomegalovirus (HCMV), hepatitis C virus (HCV), human T-cell leukemia virus (HTLV), John Cunningham virus (JC virus), herpes simplex virus (HSV), and dengue virus have also been investigated in these reconstituted mice  (Table 1).
Table 1. Pathology and immune responses after infection with human pathogens in mice with reconstituted human immune system components
Prolonged HIV infection (up to 300 days) and HIV-mediated CD4+ T-cell depletion have both been reported in mice with reconstituted human immune system components [18-22]. Both C-C chemokine receptor 5 (CCR5)- and C-X-C chemokine receptor 4-tropic HIV-1 virus strains have been examined in these mice, with C-X-C chemokine receptor 4-tropic HIV targeting CD4+ T cells broadly and CCR5-tropic HIV preferentially infecting memory CD4+ T cells and macrophages . Most of these infections were performed i.v. or i.p., but a few studies have also suggested that the more physiological mucosal HIV transmission through rectal or vaginal routes also leads to infection in mice with human immune system components [24-26]. Furthermore, these in vivo models allow the characterization of HIV dissemination after mucosal transmission. In a recent study, HIV-driven syncytia and virological synapse formation between HIV-infected T cells was observed in secondary lymphoid tissues of infected mice . These infected T cells also served as vehicles for systemic distribution of the infection, because inhibition of T-cell egress from secondary lymphoid tissues by blocking the sphingosine 1-phosphate receptor compromised systemic viral load . This systemic HIV infection in mice with human immune system components can even reach the brain via human mononuclear phagocytes, resulting in meningitis and less frequently encephalitis, especially under immunosuppressive conditions . Finally, HIV latency can be observed in infected mice [29-31]. Because the characteristics of HIV infection in mice with human immune system components are reminiscent of HIV infection in human patients, both new and currently approved treatment options against this virus can be evaluated in more detail. For example, antiretroviral drugs as either preexposure prophylaxis or treatment of established infection have been examined in mice with reconstituted human immune system components, and preexposure prophylaxis with these reagents has been shown to block rectal transmission [26, 32-34]. In addition, experimental therapies against HIV infection using either antiviral siRNA delivery to T cells, siRNA-mediated silencing of the CCR5 coreceptor and of viral proteins, or cyclin-dependent kinase blockade to inhibit viral replication have been successfully employed in these mouse models [35-37]. Thus mice with reconstituted human immune system components recapitulate HIV infection and can be used as a preclinical model for therapies against this viral infection.
Besides HIV, infection with the human tumor virus EBV has been studied in this in vivo model of the human immune system [6, 38-40]. For these studies the viral strain B95–8 was used almost exclusively, which was originally isolated from a patient with symptomatic primary EBV infection, called infectious mononucleosis . i.p. infection with increasing infectious doses of EBV leads to asymptomatic persistent infection, lymphoproliferative disease, or even hemophagocytic lymphohistiocytosis [40, 42]. During persistent infection, B cells primarily harbor the virus and strong evidence exists for both latent EBV infection as well as a low level of lytic EBV replication . These persistently infected B cells can be purified from EBV-carrying animals and cultured in vitro as immortalized lymphoblastoid cell lines. They express all eight latent EBV antigens in so-called latency type III. However, it is much less clear if other EBV latencies also develop in mice with reconstituted human immune system components, such as latency 0, which is found without any EBV protein expression in memory B cells of healthy virus carriers; latency I, which is found in Burkitt's lymphoma and homeostatic proliferating memory B cells in humans; and latency II, which is present in Hodgkin's lymphoma and germinal center B cells in healthy EBV carriers . Immunohistochemical studies provide some evidence to support the development of latencies 0, I, and II in reconstituted mice [44, 45]. However, false-negative immunohistochemistry for EBV gene products might erroneously suggest the presence of latency types other than latency III. Interestingly, EBV-encoded miRNAs are required to establish systemic persistent infection . Furthermore, a latent nuclear antigen of the virus, called Epstein-Barr nuclear antigen 3B (EBNA3B), suppresses tumor formation in vivo . EBNA3B-deficient EBV causes diffuse large B-cell lymphomas in mice with reconstituted human immune system components and the gene expression profile of the in vivo transformed tumor cells is similar to that of immortalized lymphoma lines from a subset of diffuse large B-cell lymphoma patients, which harbor EBNA3B-mutated EBV in their tumor cells. Transformation of human B cells by EBV infection in vivo might, however, require not only these EBV latent antigens, but also the low level of lytic EBV replication that has been observed in B cells. EBV, which can no longer switch into lytic infection by virtue of a deficiency in BZLF1, the main transactivator that induces EBV replication, was reported in one study to cause less EBV-associated lymphomas after infection . Therefore, hallmarks of EBV infection, such as persistence and tumorigenesis, can be recapitulated in mice with reconstituted human immune system components, but it remains unclear if all latency stages, which are finely attuned to human B-cell differentiation , can be modeled in this system.
In addition to HIV and EBV, several other viral infections have been tested in mice with reconstituted human immune system components. Among these, dengue virus was also found to establish infection in this in vivo model and a third of the infected animals developed weight loss and skin rash [49-51]. However, the identity of the infected human cells could not be clearly determined, but might be DC precursors . Nevertheless, around half of the infected animals developed viral loads, which reached 103–105 viral copies/μg RNA in the spleen, 104–107 viral copies/μg RNA in the blood, and 104–109 viral copies/μg RNA in the liver [49-51]. Similarly, i.p. injection of JC virus resulted in an infection of reconstituted mice, which could be followed by JC virus DNA in blood and urine up to 100 days after infection, but the identity of the infected cells in this study remained unclear as well . Furthermore, HSV-2 infection was observed in reconstituted BRG mice by intravaginal inoculation . In contrast, ex vivo infection of hematopoietic progenitor cells with HTLV-1 and in vivo reconstitution from these cells produced CD4+ T-cell lymphomas . From this study, the authors concluded that human hematopoietic progenitor cells could constitute a HTLV-1 reservoir in the BM, from which HTLV-associated T-cell lymphomas can develop. Similarly to HTLV-1, infection with HCMV cannot simply be achieved by injecting the virus into reconstituted mice . Instead, HCMV-infected fibroblasts had to be transferred into the peritoneal cavity of reconstituted mice. G-CSF treatment to mobilize monocytes was then able to increase HCMV viremia and systemic dissemination, and viral antigen expression was found exclusively in human monocytes and macrophages of these mice . Finally, i.v. HCV infection has been attempted in mice with reconstituted human immune system components; these mice were then additionally injected with human hepatocyte progenitors . HCV infection caused liver inflammation, hepatitis, and fibrosis in the infected mice. Therefore, mice with reconstituted human immune system components can be challenged with a variety of human viral pathogens, and constitute promising models to test pharmacological treatments and immuno-therapies against viral diseases. However, except for HIV and EBV, the other human viral pathogens have often only been tested in one or two studies in mice with human immune system components. The obtained information is often too sparse to judge whether these infections faithfully recapitulate pathogenesis in patients. Moreover, the low number of animals analyzed in the respective experiments begs for further characterization, in greater detail.
Immune responses to human viruses in humanized mice
Although the reconstitution of human immune system components has been catalogued in detail and the T cells as well as B cells arising in these systems have been shown to possess a highly diversified antigen receptor repertoire [8, 57-59], the characterization of the immune competence of the reconstituted immune system lags behind. While primary lymphoid tissues such as thymus and BM are populated with human cells and support the development of B-cell and T-cell compartments [60, 61], the development of secondary lymphoid tissues is compromised, with only few lymph nodes developing and a disorganized white pulp structure in the spleen . Given that isotype switching and affinity maturation of B-cell responses depend much more on these secondary lymphoid structures than T-cell responses , isotype-switched B-cell responses are difficult to achieve in these models. This is in contrast to T-cell responses, which develop readily in response to pathogen challenge in mice with reconstituted human immune system components.
Accordingly, specific antibody responses to HIV, HSV-2, JC virus, dengue virus, and EBV were mostly IgM after infection and only a minor subset of reconstituted mice developed IgG responses against viral antigens [22, 40, 49, 50, 52, 53, 64]. Improved interactions of human B cells with CD4+ follicular helper T cells might overcome this shortcoming , but currently no protocol that would consistently ensure these interactions has been established. Moreover, no studies have so far addressed the protective value of the observed B-cell responses by B-cell depletion, for example. Thus, it remains unclear to which extent protective anti-viral humoral immune responses can be modeled in mice with reconstituted human immune system compartments. Partly due to this limitation, the protective value of antibodies is starting to be assessed by passive immunization in these in vivo models. It was recently documented that HIV was able to escape neutralizing antibody monotherapy during infection of reconstituted mice, but that a pool of five HIV neutralizing antibodies controlled HIV viral load . This protection lasted 60 days after cessation of therapy. Taking it one step further, a HIV-neutralizing IgA antibody was expressed in human hematopoietic progenitors by lentiviral transduction and following reconstitution, a protective effect was observed against mucosal transmission of HIV . These studies suggest that, while mice with reconstituted human immune system components are not yet capable of consistently mounting human humoral immune responses after vaccination, they can indeed be used to assess the protective value of passive immunization against human viral infections.
Many more studies have been done on the human T-cell responses to viral infections in mice with reconstituted human immune system components, particularly on CD8+ T-cell responses. In both HIV and EBV infection of reconstituted mice viral antigen-specific T-cell responses were detected, but their frequency as assessed by IFN-γ production usually did not exceed 0.1%, despite the fact that a substantial proportion of the expanded CD8+ T-cell population could be detected by MHC class I/viral peptide tetramer staining [5, 38, 40, 64, 68]. This inability of most expanded antiviral CD8+ T cells to secrete cytokines might result from infection-induced differentiation of these cells and concomitant upregulation of the inhibitory receptor PD-1. Indeed, PD-1 blockade was able to rescue proinflammatory cytokine secretion in HIV-infected reconstituted mice . However, this terminal differentiation of the expanded CD8+ T cells might not negatively affect their cytotoxicity, and indeed significant perforin and granzyme B upregulation as well as cytolytic activity was found in expanded CD8+ T cells after HIV and EBV infection [38, 64, 68, 70]. Nevertheless, the viral peptide epitopes that were recognized by these responding T cells seemed to strongly depend on the MHC class I context, in which the CD8+ T-cell repertoire is educated in the thymus. In mice with human thymic transplants, the reconstituted CD8+ T-cell compartments can readily recognize immunodominant dengue virus and HIV derived epitopes [49, 64]. In reconstituted mice, in which the T-cell repertoire gets selected through the mouse thymus, these immunodominant epitopes are only recognized if the murine host transgenically expresses the respective HLA class I molecules [38, 50, 70]. In the absence of these HLA class I molecules from the murine thymic stroma, presumably unusual and in humans subdominant epitopes are recognized by the expanding CD8+ T cells. However, this has only been documented for one clonal EBV specificity so far . Although the epitope specificities of the expanding CD8+ T-cell response are still being unraveled in reconstituted mice, this adaptive immune response clearly exerts protective immune control. HIV, for example, accumulates escape mutations in response to primed CD8+ T cells . Moreover, the presence of the protective HLA-B57 molecule on the reconstituted human immune system components and on the thymic transplant allowed better HIV-specific immune control and restricted CD8+ T-cell responses similar to those found in human patients . Without a thymic transplant, the HLA-B57 transgene also allowed the reconstituted human CD8+ T cells to differentiate further in response to HIV infection . In addition to mutational immune escape from CD8+ T-cell responses, the protective value of the expanding CD8+ T-cell responses has also been shown by CD8+ T-cell depletion. Higher viral titers were observed in the absence of CD8+ T cells during HIV and EBV infection [38, 73, 74], which led to decreased CD4+ T-cell counts in HIV infection and increased tumorigenesis as well as elevated mortality of EBV-infected animals after high-dose infections. Thus, protective CD8+ T-cell responses are successfully primed during viral infections in mice with reconstituted human immune system components. While less data have been generated for CD4+ T-cell responses in reconstituted mice, viral antigen-derived peptide pool-specific CD4+ T-cell responses have been detected by intracellular cytokine staining in HCV, HIV, and JC virus infection [52, 56, 64]. Clonal CD4+ T cells that had been primed during EBV infection were able to target autologous EBV transformed B cells by cytotoxicity . Moreover, vaccination by targeting the EBNA1 via an antibody fusion construct to a receptor on DCs, together with a TLR3 agonist as adjuvant, was able to prime EBNA1-specific HLA class II-restricted CD4+ T cells, which secreted cytokines and degranulated in response to an autologous EBV-transformed B-cell line . Finally, a protective role for these CD4+ T cells has been established by CD4+ T-cell depletion during EBV infection, which resulted in elevated viral titers . Moreover, only reconstituted, but not mice without human immune system components, could restrict intravaginal HSV-2 infection, and this immune control was associated with HSV-2-specific proliferating and IFN-γ-secreting T cells at the site of infection and in draining lymph nodes . Thus, both protective CD4+ and CD8+ T-cell responses seem to be primed during viral infections of mice with reconstituted human immune system components. However, the respective CD4+ T-cell responses have been more difficult to monitor due to their limited expansion during infection.
In contrast to these adaptive immune compartments, innate immune responses have not been studied as extensively in reconstituted mice. Innate restriction of HIV by apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3 was deduced from characteristic mutations that accumulated after infection [75, 76]. Furthermore, the viral protein that targets apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3 for degradation, called Vif, reverted to WT after infection with HIV that encoded a catalytically inactive mutant of Vif . Apart from these cell-intrinsic innate immune responses, DC responses to viral infections have been analyzed in mice with reconstituted human immune system components. HIV was found to compromise plasmacytoid DC responses by diminishing their function, although the numbers of plasmacytoid DCs were not affected . Moreover, following influenza A virus infection, the major conventional CD1c+ DC subset was found to prime T cells and induce their lung homing capacity . Since innate immune responses in particular differ between mice and humans, these responses should be investigated more intensively after viral infection of mice with reconstituted human immune system components.
Bacterial infections of humanized mice
Two bacterial pathogens in particular have been explored in mice with reconstituted human immune system components, namely Mycobacterium tuberculosis (Mtb) and Salmonella enterica Typhi (S. Typhi), the etiological agents of tuberculosis and typhoid fever, respectively (Table 1). Intranasal Mtb infection led to lung granuloma formation in mice with reconstituted human immune system components [79, 80]. These granulomas were quite similar to granulomas of tuberculosis patients in that they were comprised of human giant cells and macrophages in a necrotic core, surrounded by human T cells and encapsulated by a fibrotic response. Mouse leukocytes of the NSG hosts were sparse in these granulomas and restricted to the periphery. Moreover, no granulomas were observed in nonreconstituted mice. Apart from Mtb, i.p. or i.v. injection of S. Typhi established this infection in reconstituted, but not BRG or NSG mice without reconstitution [81-83]. Infection was documented by colony-forming units (cfu) in the spleen, liver, BM, gall bladder, and blood. Mutant S. Typhi strains were also explored in this setting, and a strain that was avirulent in human volunteers replicated to lower cfu levels, while a typhoid toxin mutant showed increased infection. Therefore, both Mtb and S. Typhi infections can be explored in ice with reconstituted human immune system components.
Immune responses to bacterial infections in humanized mice
Interestingly, while the reported S. Typhi immune response was only analyzed for bacteria-specific antibody responses of an undefined isotype in a subset of mice (25%) , the CD4+ T-cell responses to Mtb infection seemed to serve an unexpected purpose . CD4+ T-cell depletion compromised granuloma formation and this diminished bacterial load . In contrast, TNF neutralization preserved granuloma formation and diminished Mtb load. These data suggest that granulomas promote Mtb replication and TNF mediates protective functions, which are independent of granuloma formation. These studies mark the beginning of investigations of antibacterial immune responses in mice with human immune system components. The limited information that has been generated thus far already leads to a better understanding of bacterial pathogenesis in humans and allows exploring mutants as vaccine candidates to elicit immune responses in this preclinical model of human immune responses.
Born out of the need for new in vivo models for infection with human pathogens and the immune responses raised against them, which might be better translatable to human patients than the classical animal models, mice with reconstituted human immune system components are increasingly being explored. Many important human pathogens such as HIV, EBV, dengue virus, JC virus, HCMV, HSV-2, HCV, Mtb, and S. Typhi, can infect these mice and cause aspects of the pathology that is observed in human patients. However, with respect to the elicited human immune responses, more needs to be done to evaluate the immune competence of these models. While it has become clear thus far that isotype-switched humoral immune responses are difficult to achieve, cell-mediated T-cell immunity can be detected in most of the investigated infections. In contrast to adaptive immune responses, innate immunity is still largely unexplored in most of these infectious settings and remains an interesting and promising topic for examination.
Therefore, further studies are required to characterize in detail the immune competence of human reconstituted innate leukocyte populations. Moreover, apart from the evaluation of genetically modified pathogens, which the field is starting to explore, genetic modifications by viral transduction of transferred hematopoietic progenitor cells have to be established. In addition, more information on the donor variability of reconstitution in relation to genetic polymorphisms needs to be gathered. Furthermore, a set of antibodies that not only deplete reconstituted human leukocyte populations, but instead block distinct receptors, needs to be established. Finally, treatments that robustly induce secondary lymphoid tissues in mice with reconstituted human immune system components would be of great value. While several additional methodological developments are needed to improve the versatility of in vivo models of human immune responses, combining these efforts with recent and ongoing studies of infection and immunity in vivo promises to result in new preclinical models that are more predictive than current models for immune reactivity and therapy in patients.
Work in our laboratory is supported by the National Cancer Institute (R01CA108609), Sassella Foundation (10/02, 11/02, and 12/02), Cancer Research Switzerland (KFS-02652–08–2010), Association for International Cancer Research (11–0516), KFSPMS and KFSPHLD of the University of Zurich, Vontobel Foundation, Baugarten Foundation, EMDO Foundation, Sobek Foundation, Fondation Acteria, Novartis, and Swiss National Science Foundation (310030_143979 and CRSII3_136241).
Conflict of interest
The authors declare no financial or commercial conflict of interest.