In various fields of biomedical research, disease models using immunocompetent mice are beneficial and widely used tools. However, there are limitations in recapitulating human biological systems using mouse models because of the genetic and immunological differences between mice and humans, especially in studies involving the human immune response (1). To address these limitations, humanized mouse models, which are engrafted with human hematopoietic cells or lymphoid tissues in immunodeficient mice, have been developed (1, 2).

Humanized mouse models can be classified into the following three categories; human peripheral blood lymphocytes (hPBL), human CD34+ hematopoietic stem cells (hCD34+), and human bone marrow-liver-thymus (BLT) engrafted models (2, 3). The hPBL model, which is generated by intravenous or intraperitoneal injection of human peripheral blood mononuclear cells (PBMC) into immunodeficient mice, is considered the fastest, simplest, and most economic model. However, there is a narrow experimental window (4~6 weeks after PBMC injection) because of the occurrence of lethal graft-versus-host disease (GvHD), and this is the main limitation of the model (4, 5). The BLT model generated by implantation of human fetal liver and thymus tissues into immunodeficient mice provides the microenvironment of the human thymus and represents a complete human immune system, including the development of human T cells. However, the availability of human tissues is seriously restricted, and GvHD is more severe in the BLT model than the hCD34+ model due to T cell affinity for the mouse major histocompatibility complex (3, 4, 6). The hCD34+ model is created by the injection of human CD34+ hematopoietic stem cells (HSCs) after total body irradiation (TBI) or the administration of myelosuppressive agents such as busulfan (BSF). This model has advantages, including the immune reconstitution of all human hematopoietic lineages, a low incidence of GvHD as compared to other models, and a relatively long-term experimental window. On the other hand, the relatively long period for humanization of more than 10 weeks, low accessibility of TBI, and a variable success rate in humanization are disadvantages of this model (3, 4). To overcome the limitations of each humanized mouse model, various research trials such as additional human cytokine treatments, modifications of human cells, and the application of new genetically engineered mice strains have been performed (1, 4).

Numerous immunodeficient mouse strains are applied for generating humanized mouse models, and NOD-scid IL2rγ^null, NOD-Rag1/2^nullIL2rγ^null, Balb/c-Rag2^nullIL2rγ^null, and related strains are representative models (7). The signal-regulatory protein alpha (SIRPA) gene of the NOD mouse strain has a strong affinity for human CD47, which could facilitate engraftment of human leukocytes (8). The Rag1/2 mutation has the advantage of irradiation tolerance and no occurrence of T/B cell leakiness (9). Therefore, the NOD-Rag1/2^nullIL2rγ^null strain is a more suitable platform for generating the hCD34+ model than the NOD-scid IL2rγ^null strain.

SIRPA, also known as CD172a, is an immunoglobulin superfamily protein that is abundantly expressed mainly on macrophages and myeloid cells. CD47, also called integrin-associated protein, is expressed ubiquitously, including on leukocytes, and it interacts with SIRPA. The interaction of CD47 and SIRPA plays a role in inhibiting host cell phagocytosis. Therefore, CD47 is functionally known as a “don’t-eat-me” signal (10). Furthermore, CD47 of the SIRPA signaling system plays an important role in the engraftment of human tissues and HSCs (11). Recent studies of humanized mouse models using CD47 knockout immunodeficient mice demonstrated enhanced engraftment of human cells and reduced GvHD occurrence (12–15). However, there are few studies about CD47 gene modification using NOD background mice (8).

In this study, we generated the hCD34+ model using NOD-Rag2^nullIL2rγ^null (RID) and NOD-CD47^nullRag2^nullIL2rγ^null (RTKO) mice, and their immunological and pathological features were examined in detail. RID mice are suitable for generating hCD34+ models (9), and we expected that human cell engraftment and GvHD occurrence would be improved in RTKO mice. At the same time, two pre-conditioning methods, TBI and BSF injection, were compared to improve on the low accessibility of TBI due to the cost of the irradiator and strict regulations (16), and to evaluate whether BSF injection is also effective in the RTKO mouse, a new genetically engineered strain.

In this study, we generated hCD34+ humanized mice using RID and RTKO mice, and two pre-conditioning methods, TBI and BSF injection, were compared. The production of humanized mice is generally determined as successful if the hCD45 percentage in mouse PBMC is more than 25% (3). Considering this standard and the health of our mice, including the body weights and survival rates ( Figures 1C, D ) and histopathological analysis ( Figure 3 ), humanized mice were successfully generated in all groups. At 8 wpt, the hCD45 percentage of the RTKO BSF was more than 30%, and in the other groups, it was subsequently over 25%. The engraftment of human leukocytes was maintained for a long period of time. Greater than 25% of the hCD45 percentage was maintained in most groups until 40 wpt, and those of the RID TBI group with the lowest results were more than 20% during the same time frame ( Figure 2B ). After 40 wpt, the ratio of hCD45 decreased ( Figure 2B ) and the fatalities increased ( Figure 1D ). Therefore it was decided to finish the experiment at 48 wpt. However, it is predicted that the experimental window of the RTKO BSF group could be extended beyond 48 wpt, as the ratio of hCD45 was maintained at 23.0% at the last measurement in that group ( Figure 2B ).

Considering the body weight ratio and the histological evaluation, the RID BSF group had the best health condition, and the RTKO BSF group, which had the highest histopathological scores, was also healthy, with a body weight increase of 156.5% by the end of the experiment ( Figures 1C , 3B ). The final survival rate at 48 wpt was the lowest in the RID BSF group (50.0%), and the highest in the RTKO TBI group (71.4%), with no significant difference among the groups ( Figure 1D ). Although mortality was observed in all groups until 48 wpt, only two animals died before 40 wpt in all groups (one mouse in each of the RID TBI and the RTKO BSF groups) ( Figure 1D ). And there were no major health problems other than skin lesions except for mortalities ( Figure 3 ). Considering the animal health and human leukocytes engraftment in this study, as well as the experimental period in other studies (21–24), the humanized mouse models of all applied conditions in this study is expected to be suitable models for long-term experiments.

Human leukocytes reconstructed in vivo were mostly hCD19 + B cells at the beginning, but they gradually decreased to less than 2% after 40 wpt. The percentage of hCD19 + B cells in the RTKO BSF group was the highest until 20 wpt, and after that, it was the highest in the RID BSF group until 40 wpt ( Figure 2D ). These results were attributed to the earlier differentiation of human leukocytes in RTKO mice than RID mice, resulting in an increase in hCD3+ T cells. The percentage of hCD3+ T cells increased from 12 wpt. Human leukocytes were differentiated more rapidly in the RTKO groups, and the percentage of hCD3 cells exceeded 10% at 16 wpt, which was eight weeks earlier than those of the RID groups ( Figure 2F ). After 16 wpt, hCD3+ T cells were higher in the RTKO BSF group, and the percentage of hCD4 to hCD8 was also the highest in the RTKO BSF group at almost all measurement points ( Figures 2H, I ). In addition, the human immune cells in the spleen were significantly engrafted in the RTKO groups ( Figure 4B ). Interpreting the results of leukocytes engraftment and differentiation, the RTKO mouse is a more suitable platform than the RID mouse, and BSF injection is a more appropriate pre-conditioning method than TBI for generating the hCD34+ humanized mice.

In histopathological analysis, GvHD lesions were prominent in the skin, not severe in the lungs and liver ( Figure 3 ), and there were no life-threatening lesions except in the dead mice ( Figure 1D ). Most of the immune cells aggregated in the tissues were hCD3+ T cells ( Figure 4A ). The expression of hIL-6, hIL-8, hIL-22, hTNFα, and hIFNγ was increased in the RTKO mice ( Figure 5 ). Associated with T lymphocytes, IFNγ and TNFα are core cytokines of chronic GvHD pathogenesis (25), and IL-6 and IL-22 aggravate skin lesions (26). IL-8 has a function of lymphocyte recruitment (27). These results suggest that cytokines secreted by the human T cells infiltrating tissues played a crucial role in the development of GvHD lesions and this mainly affecting skin pathogenesis. On the other hand, hIL-5 was decreased in RTKO mice, as hIL-5 is decreased in a CD47 blockade (28). hIL-5, classified as a Th2-type cytokine (29), is known to be involved in both acute and chronic GvHD pathogenesis, but the exact mechanism is not yet clearly understood (30, 31). hIL-5 is produced by CD4 T cells (32) and the enhancement of hIL-8 could reduce the activation of CD4 T cells (33). So, it is interpreted that the increased levels of hIL-8 ( Figure 5A ) in the RTKO groups may reduce the activation of CD4 T cells, thereby reducing the secretion of hIL-5 from CD4 T cells. We also hypothesize that the intensity of GvHD might be alleviated due to the reduction of hIL5 in RTKO, resulting in no significant differences in the survival rate among groups.

Recently, mice modified with the CD47-SIRPA axis have been studied to improve humanized mouse models (13, 14, 34). This is because the CD47 deficient condition could enhance tolerance for transplanted human leukocytes (35). However, due to the strong affinity between human SIRPA and mouse CD47 (8), such trials have been rarely applied in NOD mice. In this study, hCD34+ humanized mice were generated using CD47 KO NOD mice for the first time, and the CD47 deficiency enhanced human immune cell engraftment ( Figures 2 , 4B ), like in studies using the C57BL/6 strain (13, 14). Compared to humanized mice studies applying NSG mice, it was also confirmed that the engraftment and the differentiation of human immune cells were enhanced in RTKO mice (3, 23, 36, 37). Considering the improvement in reproducing human immune system and the additional benefits of rag2 gene mutation, including irradiation tolerance and the deficiency of T/B cell leakiness (8, 9), RTKO mice could be an alternative platform for generating humanized mice.

There are two errors in this study. The first is an error in the FACS analysis. The human leukocyte antigen analysis in the RTKO groups was underestimated at 44 wpt. When checking the results at that time, we judged that it was a simple decrease, but considering the high values at 40, 46, and 48 wpt, we expected an underestimation of the results due to experimental error. Therefore, it is possible that the RTKO BSF group maintained a more stable human leukocytes engraftment until the end of the experiment. The second error is the death of mice due to breeding management faults. On the weekend of 44 wpt, two mice in the RID BSF group died in the same cage due to water supply problems ( Table 2 ). Considering the weight change ( Figure 1C ) and the clinical signs of these mice, they were expected to have survived, and if so, the RID BSF group might have had the highest survival rate (83.3%, n=6), and might have been the most stable model to maintain a good health condition.

In conclusion, CD34+ humanized mouse models were successfully established in all four groups using RID and RTKO mice with two pre-conditioning methods, TBI and BSF. Among them, the RTKO BSF group was identified as the most suitable model considering the improvement of human leukocytes reconstruction and the extended experimental window beyond 48 wpt. This model is expected to be a novel and useful platform for various immune research, including cancer immunotherapy, virology, hematology, and autoimmunity. However, further studies are needed reduce the mortality and enhance the expansion of immune cells other than B and T cells.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

Ethical approval was not required for the studies on humans in accordance with the local legislation and institutional requirements because only commercially available established cell lines were used. The animal study was approved by The Institutional Animal Care and Use Committee of Asan Institute for Life Sciences (Seoul, Korea, IACUC No. 2020-12-093). The study was conducted in accordance with the local legislation and institutional requirements.

KK: Formal analysis, Writing – original draft, Validation, Investigation. SL: Conceptualization, Supervision, Writing – review & editing. IB: Supervision, Writing – review & editing, Resources. HS: Formal analysis, Validation, Writing – original draft, Investigation. SJ: Formal analysis, Investigation, Writing – original draft. JR: Formal analysis, Investigation, Methodology, Writing – original draft. SR: Formal analysis, Investigation, Methodology, Writing – original draft. JS: Conceptualization, Writing – review & editing, Methodology, Validation. JK: Conceptualization, Supervision, Validation, Writing – review & editing. SH: Conceptualization, Supervision, Validation, Writing – original draft, Writing – review & editing.