The simplest method of supplying graft-supporting factors is the repeated injection of recombinant human cytokines. This approach was used in an early-generation SCID model and the injection of cytokines, including human SCF, IL-3, and GM-CSF, supported enhanced engraftment levels of human BM cells as well as their myeloid differentiation (26). Because human NK cell differentiation is limited in mice, humanized BRG mice were treated with recombinant human IL-15 coupled to IL-15Rα, which is an essential growth factor for NK cell development and homeostasis. This treatment resulted in a significant increase in the differentiation and homeostasis of human NK cells in mice transplanted with human CD34⁺ cells (94).

A second cytokine delivery approach relies on hydrodynamic injection of a human cytokine-encoding DNA plasmid, resulting in the in vivo “transfection” of hepatocytes. In turn, hepatocytes produce high levels of the encoded cytokine and release it into the circulation for up to 5 days (95, 96). This method was used to demonstrate that human cytokines, such as GM-CSF and M-CSF, support myeloid differentiation of human CD34⁺ cells in NSG mice, while IL-15 and Flt-3 ligand supported NK cell differentiation (96, 97).

The administration of human cytokines can be combined with genetic opening of the niche. In BRGF mice, which lack the Flt3 gene, injection of human recombinant Flt-3 ligand boosts the development of human dendritic cells, in the absence of mouse dendritic cells (66).

Recombinant cytokine injection and hydrodynamic plasmid delivery are easy to implement, irrespective of the recipient mouse and protocol of human cell transplantation used. However, they both result in transient, systemic and generally supra-physiological cytokine expression.

To circumvent the requirement for repeated cytokine administrations, the genome of the recipient mouse can be engineered to express human cytokines (Table 2). The initial method relied on transgenic overexpression of a human cytokine-encoding cDNA, under the control of a strong promoter. Such transgenic mice are still in use, based on NSG or similar genetic backgrounds. However, results obtained with these mice need to be interpreted cautiously as the systemic overexpression of human cytokines frequently results in non-physiological hematopoiesis. For example, the transgenic pCMV promoter-driven overexpression of human SCF, GM-CSF, and IL-3 results in the mobilization of CD34⁺ HSPCs in the NSG-SGM3 mouse and the loss of their functional properties (68). As a consequence, although high-level human hematopoietic engraftment is achieved in NSG-SGM3 mice, hematopoietic progenitors lose their stemness and long-term hematopoiesis is deficient, as demonstrated by their inability to serially engraft (18, 69). Nevertheless, transgenic overexpression of cytokines can support the development of specific cell lineages and provide useful models to study those cells. In NSG-SGM3, the overexpression of human SCF results in the development of abundant and functional human mast cells, which are effective in models of passive cutaneous and systemic anaphylaxis (98). Similarly, human IL-15 overexpression in NOG mice supports the maintenance and function of human NK cells isolated from human peripheral blood (99, 100). Such models can provide useful experimental systems to evaluate the effect of candidate drugs on specific human immune cell populations.

More physiological expression of human cytokines can be achieved by bacterial artificial chromosome (referred to as “BAC”) transgenesis, inserting an entire human gene, including its regulatory elements, into the mouse genome. This method has been used to express human IL-6 and human IL-7 in NSG mice (101, 102). In another strategy, the mouse gene can be eliminated and corresponding human gene inserted in its place, including introns and exons from the start to the stop codon (Figure 3A) (103). This knock-in strategy has been used for a number of cytokine-encoding genes, including Csf1, Csf2, Il3, Il6, Il15, and Thpo (46, 65, 67, 72, 104). A slightly different approach was used for replacement of the Il7, Il15, and Tnfsf13b genes, through insertion of a human cDNA in frame with the start codon of the corresponding mouse gene, followed by a poly-adenylation signal (Figure 3B) (101, 105).

Because of its role in supporting NK cell development, human IL-15 is highly featured in the humanized mouse literature, and it has been delivered to recipient mice by each of the methods described above (46, 94, 96, 99, 101). Since these approaches have been reported by different groups, their direct comparison and the evaluation of respective merits and limitations are difficult. However, the SRG-15 recipient mouse stands out as a definitive solution to the limited development of human NK cells in humanized mice, owing to its simplicity of use (no cytokine administration is needed), the physiological expression of the cytokine by knock-in gene replacement, the comprehensive phenotypic comparison to human peripheral blood cells and rigorous functional in vivo characterization (46).

In addition to expressing a human cytokine, knock-in gene replacement abrogates expression of the corresponding mouse cytokine (Figure 3). If the human cytokine is not fully cross-reactive, this can result in the absence of the cytokine-dependent mouse cell population and niche opening (103). This is best illustrated in the case of GM-CSF (encoded by Csf2/CSF2), where only 56% of amino acids are shared between human and mouse (1, 67). In mice with homozygous humanization of the Csf2 locus, the absence of mouse GM-CSF induces alveolar macrophage deficiency in the lung and a resulting pathology described as pulmonary alveolar proteinosis (67), recapitulating the phenotype of Csf2 knockout mice (88–90). Upon human CD34⁺ cell transplantation, human GM-CSF supports the development of human alveolar macrophages and partial rescue of the proteinosis pathology (67).

Importantly, because hematopoiesis is a stepwise process in which stem cells gradually differentiate toward more committed progenitors and multiple lineages of mature cells (1, 2), humanization of several cytokines is required to support the maintenance and multilineage differentiation of human HSPCs. This is exemplified in the MISTRG mouse in which M-CSF, IL-3, GM-CSF and THPO are humanized by knock-in replacement, on the SRG genetic background (17, 71). Niche opening and four human cytokines synergize to support the engraftment of human CD34⁺ cells without radiation preconditioning (17, 18). The long-term maintenance of functional HSPCs is demonstrated by their serial transplantation for up to four generations of MISTRG recipients (18). These mice also support multilineage differentiation of human B, T and dendritic cells (similarly to NSG mice) as well as different subsets of functional myelo-monocytic cells in lymphoid and non-lymphoid tissues (17, 18). The myelo-monocytic cells themselves express human IL-15/IL-15Rα, which in turn supports the development and function of human NK cells (17). Overall, MISTRG mice emphasize the benefits of combining multiple strategies for the provision of human cytokines in the development of improved recipient mice by: niche opening by elimination of the corresponding mouse cytokines; physiological expression of each cytokine; synergy between multiple cytokines; and, indirectly, cross-support between different human immune cell lineages.

Although providing specific cytokines can be effective, it would be a long and arduous journey to humanize the entire spectrum of hematopoiesis-supporting cytokines (and possibly other required factors). Therefore, co-transplantation of supporting human cells or tissues, naturally found in the BM niche or in other primary lymphoid organs, can be used to provide all required factors and more fully recapitulate human hematopoietic development and homeostasis.

A well-established protocol of such co-transplantation is the BM, liver, thymus (BLT) humanized mouse model (106, 107). BLT mice are generated by co-transplanting human fetal liver and thymus under the murine renal capsule in preconditioned adult immunodeficient mice, before injecting syngeneic CD34⁺ HSPCs intravenously. The liver and thymus implants form a liver-thymus organoid that contains stromal microenvironments and provides cytokines at local physiological concentrations, requisite for the differentiation of functional B, T, and NK cells, dendritic cells and monocytes/macrophages, as well as long-term maintenance of human hematopoiesis and lymphopoiesis (106–108). Because human T cells develop in the context of thymic epithelial cells, the human thymus organoid contains significantly higher absolute numbers of thymocytes, compared to mouse thymus (109) and a repertoire of immunocompetent human leukocyte antigen (HLA) class I- and class II-restricted T lymphocytes is selected (106, 107, 110). The immune cell distribution in BLT humanized mice is well-described for both primary and secondary lymphoid organs, including BM, human and mouse thymus, intestines (lamina propria), mesenteric lymph nodes, vaginal tissues, liver and lungs (106, 107, 109, 111–113). Although the BLT protocol was first described using the NOD-SCID background (106, 107), this technique can be applied to any recipient mouse, and has been used successfully in NSG (109, 113), NSG-SGM3 (98), and B6RG-CD47 (48) mice. Thus, investigators can use the BLT protocol to enhance human T cell development in a mouse strain with immune characteristics that are appropriate for their study.

Intravenous or intrafemoral co-transplantation of human BM-derived MSCs along with CD34⁺ cord blood cells is another strategy to improve engraftment of CD34⁺ and CD45⁺ cells in immunodeficient mice (114–117). Overall, co-transplantation of the BM-derived MSCs into humanized mice can improve human HSPC maintenance and expansion, and improve reconstitution of the human hemato-lymphoid system in humanized mice. Furthermore, recent studies have used genetically engineered MSCs to deliver additional human factors to support human hematopoiesis, or to facilitate the expansion and maintenance of HSPCs to allow serial transplantation and generation of larger quantities of humanized mice (118, 119).

Recent studies have used human bone organoid (ossicles) as a method to add human BM microenvironment for engrafted HSPCs to occupy (120, 121). Ossicles are often generated by seeding 3D polymer scaffolds with human BM-derived MSCs. These are then implanted subcutaneously into NSG mice and become colonized with subsequently injected human HSPCs, enabling their expansion and differentiation (122–128). Humanized ossicles contain increased human immature and mature hematopoietic cells as compared to the bones of host mice implanted with only human CD34⁺ cells, indicating homing of HSPCs (122, 128). Additionally, self-renewing HSPCs from humanized ossicles can reconstitute hematopoiesis in secondary recipient mice, demonstrating maintenance of their functional properties (129). Most importantly, high engraftment levels of hCD45⁺ cells were measured in the blood, spleen and mouse bone (130, 131). High numbers of human erythroid lineage cells and robust differentiation of mature myeloid cells were also detected (132).

With the development of MISTRG and MISTRG6 mouse strains that express essential human cytokines, and protocols for co-transplantation of human fetal bone chips or ossicles, major progress has been made in transplanting patient-derived hematologic malignancies into humanized BM niches. Samples from patients with myelodysplastic syndromes, myeloproliferative neoplasms, low risk acute myeloid leukemia, diffuse large B cell lymphoma or multiple myeloma have been successfully engrafted, using diverse recipient mice and protocols (72, 123, 125, 133–136).

Humanized mice generated by transplantation of CD34⁺ cells, or following the BLT protocol, can mount antigen-specific adaptive immune responses (21, 107, 141, 142). Evidence of protective immunity, capable of controlling pathogen replication and resulting disease, was provided in the context of Epstein-Barr virus (EBV), dengue virus (DENV), and human immunodeficiency virus-1 (HIV-1) infection. Indeed, in CD34⁺-humanized NSG mice, antibody-mediated depletion of T cells led to the development of EBV-associated tumors, suggesting that T cell-mediated immunity can control EBV infection in this setting (143). Robust CD4⁺ and CD8⁺ T cell responses were also detected after DENV and HIV-1 infection. Notably, in HIV-1-infected BLT mice, CD8⁺ T cell responses were HLA-restricted and directed against epitopes previously described as immunogenic in humans. Furthermore, CD8⁺ T cell-mediated viral recognition led to viral epitope evolution, closely resembling clinical observations (142).

Despite having detectable anti-viral T cells, CD34⁺-humanized and BLT mice generally exhibit weak humoral responses after viral infection (144, 145). While neutralizing antibodies were detected in a subset of DENV-infected mice, in both CD34⁺-only and BLT humanized mice, the anti-EBV or anti-HIV-1 antibody responses were either weak, equivocal or delayed (111, 141). This was also the case in humanized MISTRG mice, where enhanced innate immune cell engraftment led to improved T cell function after Listeria monocytogenes infection, but humoral immunity remained weak (17). Across these models, immunoglobulin class-switching and somatic hypermutation are rarely achieved, likely because most B cells fail to reach a fully mature phenotype in the periphery, T cell repertoire is selected on mouse and human MHC and is suboptimal, and recipient mice have disorganized lymph nodes and little capacity for germinal center formation (146).

While human B cells are detectable in high frequencies in humanized mice, most exhibit an immature phenotype, and remain blocked at the transitional stage of B cell development (1, 146). Since immature B cells have a reduced capacity to respond to antigen (147), several groups sought to improve B cell maturation and by extension, their function by knocking-in human cytokines that are known to support B cell development. Surprisingly, human B cell activation factor (BAFF, encoded by TNFSF13B) knock-in resulted in reduced numbers of mature naïve B cells, and reduced antibody production after immunization (105). Human IL-7 knock-in had no effect on B cell numbers and any effect on B cell function remains to be determined (101). In contrast, human IL-6 positively impacted B cell differentiation into plasmablasts and memory cells after immunization with model antigen, and promoted somatic hypermutation and class switching, albeit to a lower level than that observed in humans (73). IL-6 knock-in mice also had improved T cell development; therefore, it is possible that enhanced B cell responses were in part due to increased T cell help.

Thymic lymphopoiesis is suboptimal in mice humanized by transplantation of CD34⁺ HSPCs, in part due to a lack of thymopoiesis-supporting human cytokines. As a result, thymic cellularity is extremely low (only a few million cells) and CD4⁺ CD8⁺ double positive cells, which represent the vast majority of thymocytes in a healthy thymus, are frequently underrepresented. Human IL-6 knock-in produces increased thymic cellularity (73), but additional cytokines are likely required to restore the normal size of a mouse thymus, populated by human thymocytes. While a human IL-7 knock-in recipient mouse has been developed, a thorough characterization of thymic cellularity upon HSPC engraftment has not been reported to date (101).

Human T cell repertoire selection in the mouse thymus involves interactions with both mouse MHC and human HLA-expressing cells-expressing cells (1, 10, 21). Specifically, developing human T cells are positively selected by mouse epithelial cells and negatively selected by both mouse epithelial and mouse and human hematopoietic cells (148). The resulting T cell receptor (TCR) repertoire is weakly reactive to autologous human leukocyte antigen (HLA) class I and II, and tolerant to mouse MHC. Indeed, in in vitro re-stimulation assays, T cells from humanized mice were shown to proliferate better in response to allogeneic human DCs compared to autologous human DCs or mouse DCs (21). To improve the repertoire and function of human T cells, several recipient strains have been engineered to express HLA class I and class II (143, 144, 149–154). HLA-restricted T cell responses have been reported in these mice (143, 149). Additional characterization and comparison of these mice are required to rigorously evaluate how transgenic HLA expression in recipient HSPC-humanized mice impacts human T cell selection and function.

In BLT mice, T cells develop in the human thymus organoid and are positively and negatively selected on human autologous HLA molecules (106, 107, 155). Consequently, BLT mice are generally accepted as having a more diverse T cell repertoire, capable of mounting more robust adaptive immune responses, as discussed above. However, because the induction of tolerance for mouse MHC and mouse tissue-restricted antigens is incomplete, BLT mice are prone to the development of xenogeneic graft-vs.-host disease (xGvHD) (156, 157), as discussed below.

Few direct comparisons between BLT and HSPC-engrafted mice have been reported to date (144, 145). In both models, it is apparent that thymic lymphopoiesis and the selection of a diverse and tolerant T cell repertoire is a complex issue, and additional work is needed.

Secondary lymphoid organs including lymph nodes (LNs), spleen, Peyer's patches, and mucosa-associated lymphoid tissues normally provide niche microenvironments where T and B cells interact with hematopoietic antigen-presenting cells and stromal follicular dendritic cells (FDCs) to initiate the adaptive immune response (158, 159). LN formation requires IL-7- and IL-2Rγ-dependent lymphoid tissue inducer (LTi) cells (160). However, because most immunodeficient recipient mice lack Il2rg to promote tolerance to the human graft through NK cell depletion, they also lack LTi cells and defined LN structures. This represents a major obstacle in recapitulating the human adaptive immune response in mice. While engraftment with human CD34⁺ cells can partially rescue the LN anlagen, T and B cell zones remain poorly organized compared to those in normal human or normal mouse LNs. Furthermore, the germinal center formation is impaired, in part because it involves human B cells interacting with stromal FDCs of mouse origin (1, 159).

Given that the germinal center is the primary site of B cell-T cell collaboration, antibody affinity maturation and class switching, multiple groups sought to improve LN formation and germinal center development (158). Adoptive transfer of autologous human DCs lentivirally transduced to express GM-CSF, IFNα, and human cytomegalovirus (HCMV) viral antigen enhances LN development in humanized NRG mice, and by extension promotes T cell-B cell collaboration and the production of neutralizing anti-HCMV antibodies (161, 162). Human fetal organ co-transplantation also partially rescues secondary lymphoid organ development. BLT mice have improved lymphoid structure development in the spleen and lymph nodes compared to CD34⁺-only engrafted mice, and as a result, they also have more robust antigen-specific adaptive immunity (106, 107). However, a comparison of IL-2Rγ-sufficient NOD-SCID BLT humanized mice and IL-2Rγ-deficient NSG BLT humanized mice showed that only NOD-SCID BLT mice contain mucosal tissue-associated T cells (109). This result emphasizes the critical nature of IL-2Rγ in the development of LN and other secondary lymphoid organs that are essential for adaptive immune responses and mucosal immunity. The importance of structured secondary lymphoid organs for adaptive immunity is further demonstrated by the addition of fetal spleen to BLT mice (163). Human fetal spleen implants grow into spleen organoids with prominent follicular lymphoid structure; implanted mice have improved B cell and T cell engraftment, compared to control mice without human spleen implants, and can mount antigen-specific responses to immunization (163).

Another strategy to overcome the near-absence of LNs in Il2rg^−/− humanized mice involves transgenic overexpression of murine thymic stromal lymphopoietin (TSLP). TSLP is another cytokine of the IL-2 family, its receptor is independent of the IL-2Rγ chain and, when overexpressed under a keratin 14 promoter (specific for epithelial/mesenchymal cells), it rescues LN development in Il7 deficient mice (164, 165). Crossing the transgene to BRGS resulted in the BRGST model, in which mouse LTi cells and LN structures are restored (74). Upon transplantation of CD34⁺ HSPCs, BRGST mice support the development of a human immune system, including sizeable LNs with compartmentalized human T and B cell zones. BRGST mice also have more mature B cells and IL-21 producing follicular helper T cells, essential to promoting adaptive immunity. Consequently, BRGST mice mounted enhanced antigen-specific humoral immune responses upon immunization with an experimental antigen (74). Overall, this novel model successfully addresses a major limitation that had hampered immune function in most humanized mouse models to date.

Adoptive transfer of human T cells from donor PBMCs into immunodeficient recipient mice results in xGvHD, thereby limiting the potential duration of experiments with these mice. To prevent xGvHD, two strains of immunodeficient mice lacking mouse MHC-I and MHC-II have been developed, which prevented the onset of xGvHD, while retaining the functional properties of human T cells (166, 167). But, as discussed above, PBMC transfer results in an incomplete human hematopoietic system in the mice.

When mice are humanized by transplantation of CD34⁺ HSPCs, developing human T cells undergo positive and negative selection in the mouse thymus. Consequently, they are tolerized for mouse MHC-I and MHC-II, and xGvHD is not a limitation in this model (1, 10, 21).

Finally, in BLT mice, T cells are educated in the human thymus organoid and once they reach the periphery, they are allogeneic to the mouse MHC molecules and can induce xGvHD (156, 157). Different levels of xGvHD are reported by different groups, suggesting that the disease could be affected by subtle differences in protocols, the microbiota of the mice, or the recipient strain used (48). The development of xGvHD in the BLT model is attributed to residual mature T cells present in the fetal human thymus grafts, and these passenger T cells can be removed by treating the thymic implants with 2′-deoxyguanosine, or by treating the mice with anti-human CD2 antibody post-surgery (168).

The efficient development of human phagocytic cells of the myelo-monocytic lineage in MISTRG creates a new challenge; i.e., the absence of phagocytic tolerance from the human graft toward the mouse host. Mouse red blood cells are particularly susceptible to destruction by human phagocytic cells, and highly engrafted MISTRG mice develop lethal anemia (17). Consequently, the transplantation protocol in MISTRG needs to be optimized so that engraftment allows mouse erythropoiesis in the BM (or extramedullary erythropoiesis) to sufficiently compensate for the loss of mouse red blood cells by phagocytosis, as long as a specific experiment requires (17, 18, 58, 133). Long-term solutions will need to be implemented, either to establish human-to-mouse phagocytic tolerance or to enable human erythropoiesis to reach healthy human RBC counts and maintain mouse homeostasis, as discussed above.

Xenogeneic GvHD and hemophagocytosis remind us that activation and tolerance are two equally important features of the immune system and are inseparable. As human immune function improves in mouse recipients, it is likely that parallel strategies will need to be developed to maintain tolerance of the human immune graft for the mouse host.