Enhanced human cell engraftment in mice deficient in RAG2 and the common cytokine receptor γ chain


Dr Jacki PGoldman Molecular Immunology Unit, Institute of Child Health, 30 Guilford Street, London WC1N 1EH.


Xenotransplantation of human cells into immunodeficient mice has been used to develop models of human haemopoiesis and lymphoid cell function. However, the utility of existing mouse strains can be limited by shortened life-spans, spontaneous production of functional lymphocytes with ageing, and residual innate immunity leading to variable levels of engraftment. Mice with a deletion of the common cytokine receptor γ chain (γc) gene have reduced numbers of peripheral T and B lymphocytes, and absent natural killer cell (NK) activity. A genetic cross with a recombinase activating gene 2 (RAG2)-deficient strain produced mice doubly homozygous for the γc and RAG2 null alleles (γc/RAG2). These mice have a stable phenotype characterized by the absence of all T lymphocyte, B lymphocyte and NK cell function. Injection of human B-lymphoblastoid cells resulted in earlier fatal metastatic lymphoproliferative disease than in NOD/LtSz-scid controls. This was particularly evident in animals injected intravenously, possibly because of residual NK activity in NOD/LtSz-scid mice. Levels of engraftment with peripheral-blood-derived human lymphocytes were also increased and associated with higher CD4/CD8 ratios. These findings demonstrate that this new strain of immunodeficient mice has significant advantages over existing strains for engraftment of human cells, and may be useful for study of adoptive immunotherapy and novel therapies for GvHD and HIV infection.

Mice with a severely restricted ability to reject xenogeneic cells have become invaluable tools for the study of human haemopoietic and immune cell function. By virtue of a spontaneous mutation in the gene encoding DNA protein kinase catalytic subunit (DNA-PKcs), homozygous severe combined immunodeficient (SCID) mice (scid/scid) have a generalized double-strand DNA breakpoint repair defect resulting in increased cellular sensitivity to ionizing radiation and an inability to resolve immunoglobulin (Ig) and T-lymphocyte receptor (TCR) gene rearrangement ( Bosma & Carroll, 1991). Engraftment of normal and malignant human stem cells and human T lymphocytes cells in scid/scid mice has been successful ( Kamel-Reid & Dick, 1988; Dick et al, 1991 ; Mosier et al, 1988 ). Initially, however, levels of engraftment of both progenitor and lymphoid cells, restricted by residual innate mechanisms, were generally low.

NK function in particular is detrimental to engraftment by semi-syngeneic (hybrid resistance), and MHC-disparate allogeneic bone marrow in irradiated immunocompetent mice and SCID mice ( Murphy et al, 1987 ; Kumar et al, 1997 ). Depletion of NK function with antibodies directed against NK cell-surface markers has been shown to remove this barrier to murine cell engraftment ( Lotzova et al, 1983 ; Tiberghien et al, 1990 ) and more recently to facilitate engraftment of human lymphocytes ( Murphy et al, 1992 ; Shpitz et al, 1994 ; Sandhu et al, 1994 ). This strategy has been used to develop models for adoptive immunotherapy ( Lacerda et al, 1996 ), GvHD ( Sandhu et al, 1995 ), HIV infection ( Koyanagi et al, 1997 ) and vaccine-specific responses ( Albert et al, 1997 ).

Attempts have been made recently to develop novel immunodeficient mouse strains which, in addition to T- and B-lymphocyte deficiency, incorporate selective defects of innate immunity. The inbred non-obese diabetic (NOD) mouse strain is characterized by deficiency of complement (2 base-pair deletion in a 5′ exon of the murine C5 gene), molecularly undefined defects of macrophage function (defective regulation of colony-stimulating factor-1 and interferon-γ receptors, and reduced secretion of IL-1) and NK function. Introduction of the scid mutation onto the NOD background has produced strains of mice (NOD/LtSz-scid ( Shultz et al, 1995 ) and NOD/Shi-scid ( Koyanagi et al, 1997 )) which incorporate these defects of innate immunity. In contrast to SCID mice, NOD/LtSz-scid mice engraft both normal and malignant haemopoietic progenitor cells to high levels which proliferate and differentiate into multiple myeloid, erythroid and B-cell lineages without the need for administration of exogenous cytokines ( Pflumio et al, 1996 ; Cashman et al, 1997 ; Bonnet & Dick, 1997). This has provided an accessible assay system for testing repopulating ability of very primitive human haemopoietic precursors ( Bhatia et al, 1997 ) and for testing the efficiency of somatic gene transfer to these cells ( Larochelle et al, 1995 ). Engraftment of human PBLs is also enhanced in NOD/LtSz-scid and NOD/Shi-scid strains ( Hesselton et al, 1995 ; Koyanagi et al, 1997 ). However, residual NK activity, and a marked tendency to develop thymomas, limit the utility of these strains.

An immunodeficient mouse which, in addition to the absence of B and T cells, lacks NK cell function, may have significant advantages over these existing strains. The common cytokine receptor γ chain (γc) is a functional subunit of the receptors for interleukin-2 (IL-2), IL-4, IL-7, IL-9 and IL-15, and plays an important role in lymphoid development ( Sugamura et al, 1996 ). Inactivation of this gene is responsible for human X-linked SCID ( Noguchi et al, 1993 ; Leonard et al, 1994 ), and in homozygous γc mutant mice results in abnormal T-cell lymphopoiesis and dysfunction of residual peripheral T lymphocyte cell populations ( DiSanto et al, 1995 , 1996; Cao et al, 1995 ; Nakajima et al, 1997 ). More significantly, there is an absolute dependence on γc for development of NK cells (probably as part of the IL-15 receptor) and for export and/or survival of NK T lymphocytes from the thymus ( Lantz et al, 1997 ; DiSanto, 1997). Inactivation of lymphoid-specific recombinase activating genes (RAG) 1 and 2 in mice and in humans (some autosomal recessive forms of TBSCID), completely abrogates the production of thymus-derived T lymphocytes and peripheral B lymphocytes, although numbers of splenic macrophages, and splenic NK activity are increased ( Shinkai et al, 1992 ; Mombaerts et al, 1992 ; Schwarz et al, 1996 ).

In this study we have utilized the defined molecular lesions of RAG2-deficient and γc-deficient mice to produce a novel strain homozygous for γc and RAG2 null alleles (γc/RAG2), resulting in a complete lack of all T, B and NK lymphocyte function. These mice retain stability of immunophenotype over time and have significant advantages over conventional immunodeficient mouse strains in human cell engraftment experiments.



Mice were bred and maintained in the Western Laboratories (Institute of Child Health). They were housed in sterile microisolator cages in a laminar flow caging system (Thoren, Hazleton, Pa.) and supplied with sterile food, water and bedding. All manipulations were conducted in a laminar flow hood. Original stocks of NOD/LtSz-scid/scid mice (NOD/LtSz-scid) were kindly provided by John E. Dick, Hospital for Sick Children, Toronto. Mice carrying a null mutation of cytokine receptor common gamma chain (γc) on the X chromosome have been described previously ( DiSanto et al, 1995 ). γc−/− females were crossed with males homozygous for a mutation disrupting the RAG2 gene ( Shinkai et al, 1992 ). F1 males heterozygous for RAG2 deletion and hemizygous for γc deletion were backcrossed onto γc−/−RAG2+/+ females. The RAG2 genotype of the resulting offspring was determined by tail DNA PCR ( Horton et al, 1995 ). RAG2 heterozygotes were bred to produce the γc−/−RAG2−/− female and γc−/YRAG2−/−male mice (γc/RAG2). These mice have a mixed background of 129 Ola, Balb/c and C57BL/6.

Cytotoxicity assay

NK activity was determined by standard 51Cr-release assay. Balb/c, NOD/LtSz-scid and γc/RAG2 mice were injected i.p. with 100 μg poly I:C (Sigma Chemical Co.) and the spleens removed 36 h later. Spleen cells were used as effectors at various E:T ratios with 51Cr-labelled YAC-1 tumour target cells in triplicate in V-bottom plates. Supernatants harvested after 4 h incubation at 37°C were counted in a gamma counter to measure experimental 51Cr release (E). Spontaneous release from target cells (S) was measured after incubation in the absence of effector cells and total release (T) was measured by treating target cells with 2% SDS. Specific 51Cr-release was calculated as: % specific release = E−S/T−S × 100.

Tumour cell engraftment

The EBV-immortalized B lymphoblastoid cell line (B-LCLs) and the erythroleukaemia cell line K562 were grown in RPMI supplemented with 10% fetal calf serum and antibiotics. Cells were washed once and resuspended in RPMI prior to injection. B-LCLs were injected either intraperitoneally (i.p.) or intravenously (i.v.) via the tail vein into previously unmanipulated and unirradiated γc/RAG2 or NOD/LtSz-scid mice. K562 cells were injected i.v. Mice were killed by CO2 inhalation when they became sick, as shown by hunched posture, loss of activity, weight loss and/or ruffled fur. Organs were harvested for macroscopic and histological examination and/or analysis by flow cytometry.

PBL engraftment

Heparinized whole blood was collected from healthy volunteers, and PBMCs were isolated by centrifugation on Ficoll-Hypaque (Pharmacia, Uppsala, Sweden). Published data has shown that prior irradiation enhances engraftment of human PBLs in CB17-scid/scid mice ( Shpitz et al, 1994 ). On the basis of this, 8–10-week-old mice were irradiated with 325 cGy (NOD/LtSz-scid) or 600 cGy (γc/RAG2) from a 137Cs source 0–1 d prior to i.p. injection of 20–40 × 106 human PBMCs. The amount of preconditioning radiation had been determined as the maximum sublethal dose for each strain. To determine the effect of radiation on engraftment, eight γc/RAG2 mice were given human PBMCs without prior irradiation. After 4 weeks, mice were killed by CO2 inhalation. Single-cell suspensions from spleen were subjected to NH4Cl red cell lysis and analysed by flow cytometry for the presence of human lymphocytes.

Flow cytometry

Directly conjugated antibodies against human cell surface antigens were purchased from Becton Dickinson, Oxford (CD45-PerCP, CD4-FITC/CD8-PE, HLA-DR-PE) or DAKO Ltd, High Wycombe (CD3-FITC). Anti-mouse IgM was purchased from Pharmingen (SanDiego, Calif.). Anti-mouse CD3-FITC was a gift from Rose Zymoyska (NIMR, Mill Hill, London). 106 cells were incubated for 30 min on ice with saturating amounts of antibodies in staining buffer (PBS, 5% FCS, 0.01% sodium azide). A sample from each tissue was also stained with directly conjugated isotype-matched control antibodies (Becton Dickinson). Cells were washed three times and fixed in 1% paraformaldehyde. Labelled cells were analysed within 1 week on a FACSCalibur using CellQuest software (Becton Dickinson). A total of 1–2 × 104 events were acquired per sample. The lymphocyte population was defined on the basis of forward and side scatter, and percentage of cells positive for a given marker was determined for cells falling within the lymphocyte gate.


Characterization of γc/RAG2 homozygous mutant mice

γc/RAG2 mice survive and breed well in the barrier colony in which they are maintained. Of 18 mice randomly selected at birth, all remained alive with no apparent health problems > 12 months later. Interestingly, colitis and rectal prolapse suffered by γc single mutants and γc/RAG2+/− heterozygotes has not been observed in γc/RAG2 homozygotes, suggesting that this is an immunoreactive process ( Cao et al, 1995 ; DiSanto, 1997).

To confirm the predicted immunophenotype of γc/RAG2 mice, single-cell suspensions from spleen and bone marrow were labelled with fluorescent antibodies against mouse IgM and CD3. In contrast to γc/RAG2+/−and control mice, no positive cells were detected (data not shown). Serum immunoglobulin, measured by ELISA, was at or below background level (<10 ng/ml, data not shown) in all mice tested. As expected from the stable phenotype of RAG2 single mutants, spontaneous production of antibody in the γc/RAG2 mice was not observed. To determine the presence of NK activity, in vivo poly I:C-stimulated splenocytes were tested against YAC-1 target cells. Activity in γc/RAG2 mice was totally absent. In contrast, significant activity could be induced in NOD/LtSz-scid animals compared with Balb/c controls (Fig 1).

Figure 1.

target cells at various E:T ratios. Each point represents the mean ± SD of triplicate wells.

Tumour cell engraftment

The ability of γc/RAG2 and NOD/LtSz-scid mice to engraft a normal human EBV-immortalized B lymphoblastoid cell line (B-LCL) was examined. Clinical disease was apparent in all γc/RAG2 mice 26–41 d after intraperitoneal injection of 1–5 × 106 B-LCLs, with survival time inversely proportional to the cell dose (Fig 2A). Engrafted animals developed ascites with large intra-abdominal solid tumours, splenomegaly, and enlargement of lymph nodes in the porta hepatis and mesentery. Metastatic tumours were universally found in the liver and kidneys and occasionally in the lung. Mice given a lower dose of cells (105) had extensive solid tumour in the peritoneal cavity, but no evidence of dissemination at the time of necropsy. More than 80% of NOD/LtSz-scid mice engrafted with equivalent numbers of B-LCLs developed a similar disease pattern to γc/RAG2 mice. However, two NOD/LtSz-scid mice failed to engraft and died of spontaneous thymomas 120–127 d post-injection (Fig 2B). There was no incidence of thymoma in the γc/RAG2 mice.

Figure 2.

) cells.

Intravenous administration of B-LCLs resulted in significant differences between the two mouse strains. All γc/RAG2 mice injected with between 106 and 107 cells died within 35 d (Fig 2C). Necropsy of these animals revealed splenomegaly, lymphadenopathy and solid tumours in kidney, liver and lungs. 50% of NOD/LtSz-scid mice injected with equivalent doses of B-LCLs cells survived for > 100 d, but died of spontaneous thymomas. The remaining NOD/LtSz-scid mice died 23–67 d after injection from indeterminate causes (Fig 2C). However, this did not appear to be caused by the B-LCLs as the spleen, lymph nodes and other organs were of normal size and macroscopically free of tumour. To determine whether residual NK function in NOD/LtSz-scid animals was responsible for the observed differences, B-LCLs were used as targets for poly I:C-stimulated murine splenocytes in vitro. However, detection of activity above background in both immunodeficient strains, even at high effector to target cell ratios, was not possible, probably because of limitations to the sensitivity of the assay. Similar patterns of engraftment were seen in unirradiated γc/RAG2 and NOD/LtSz-scid mice following intravenous injection of the human erythroleukaemia line K562. NOD/LtSz-scid mice given 5 × 105 cells remained healthy for > 6 months after injection. In contrast, γc/RAG2 mice became ill 35–42 d after injection, and were found to have multiple tumours in the kidney (data not shown).

Engraftment with human PBL

To determine the efficiency of engraftment of human primary lymphocyte populations, 20–40 × 106 PBMCs from normal volunteers were injected intraperitoneally into γc/RAG2 and NOD/LtSz-scid mice. Levels of human cell engraftment were determined 4 weeks later. Red-cell-depleted single-cell suspensions of spleen were labelled with antibodies recognizing human T lymphocyte markers (CD45, CD3, CD4, CD8, HLA-DR) (Fig 3). Mice in which > 0.1% of spleen cells were positive for human CD45 and in which human CD3+cells as well as CD4+ and/or CD8+ cells could be detected were considered to have been successfully engrafted. A similar proportion of γc/RAG2(27/35, 77.1%) and NOD/LtSz-scid mice (13/17, 76.5%) were positive for human cell engraftment by these criteria (Fig 4A). The average % human CD45 positive spleen cells of engrafted γc/RAG2 mice (9.5%) was higher than the average of the engrafted NOD/LtSz-scid mice (5.5%), although this difference was not statistically significant (Fig 4A). There was, however, a marked difference in the proportion of highly engrafted animals. Of the engrafted NOD/LtSz-scid mice only one (7.7%) had > 10% human CD45-positive spleen cells. In contrast, 10 γc/RAG2 mice (37.0%) were highly engrafted (Fig 4A). Levels of human CD45+ cells in the spleens of irradiated and unirradiated γc/RAG2 mice were comparable, indicating that engraftment with human cells does not depend on prior irradiation (Fig 4A). Human immunoglobulin could be detected by ELISA in the serum of engrafted mice in concentrations ranging from 0.3–7.5 mg/ml (data not shown).

Figure 3.

-FITC v HLA-DR-PE (A, C, E) or CD4-FITC v CD8-PE (B, D, F) in a non-engrafted γc/RAG2 mouse (A, B), an engrafted NOD/LtSz-scid mouse (C, D) and an engrafted γc/RAG2- mouse (E, F). Numbers indicate % of spleen lymphocytes which fall within the quadrant or region. Representative mice from different experiments are shown.

Figure 4.

:CD8 ratios calculated for the individual mice used for (B). Average ratio indicated by horizontal lines (1.58 for γc/RAG2 and 0.50 for NOD/LtSz-scid). The difference between γc/RAG2 and NOD/LtSz-scid mice was significant (P = 0.0013 by Mann-Whitney U test).

Consistent with earlier studies ( Hesselton et al, 1995 ; Christianson et al, 1997 ), the vast majority of human PBLs in the spleens of engrafted mice were CD3 positive in both γc/RAG2 (95.8 ± 3.1%) and NOD/LtSz-scid mice (93.9 ± 7.5%) (Fig 4B). There was, however, a significant difference with regard to engraftment with CD4-positive cells. An average of 51.0 ± 11.4% of human cells in γc/RAG2 mice were CD4+ as opposed to only 24.7 ± 9.3% in NOD/LtSz-scid mice (Fig 4B). This concurs with earlier reports which showed that CD4-positive T cells did not engraft well in NOD/LtSz-scid mice ( Hesselton et al, 1995 ; Christianson et al, 1997 ). CD4/CD8 ratios were determined for mice engrafted with > 1% human cells in the spleen at 4 weeks. Regardless of the level of human lymphocyte engraftment, the CD4/CD8 ratio in NOD/LtSz-scid was < 1 in every case (range 0.16–0.70, average 0.50) ( Figs 3D and 4C). In contrast, CD4/CD8 ratios in engrafted γc/RAG2mice ranged from 0.47 to 4.10 with an average value of 1.58, which more closely approximated a normal CD4:CD8 ratio in the periphery ( Figs 3F and 4C).

In all engrafted γc/RAG2- mice and half of engrafted NOD/LtSz-scid mice, the majority of CD45+CD3+ spleen cells were HLA-DR positive indicative of a memory/activated phenotype ( Figs 3C and 3E). This is reflected in our preliminary observations that PBMC-engrafted NOD/LtSz-scid and γc/RAG2 mice developed mild xenogeneic graft-versus-host disease (GvHD), characterized by lymphocyte infiltration in the liver and fibrosis of the spleen, 4–10 weeks after injection of human PBMCs (unpublished observations).


Since it was first demonstrated that CB17-scid mice could engraft human cells, significant progress has been made in the development of models of human haemopoietic progenitor and lymphoid cell function. These improvements for the most part have been aimed at reducing the innate immunity in recipient strains. Several studies have shown that depletion of NK activity using antibodies directed against cell surface molecules such as asialo-GM1 or IL2Rβ, facilitates more consistent and higher level engraftment of human lymphoid cells ( Murphy et al, 1992 ; Shpitz et al, 1994 ; Sandhu et al, 1994 ; Koyanagi et al, 1997 ). This increase in human cell engraftment effected by removal of NK cells in recipient scid mice indicates that NK cells present a significant barrier to engraftment. We have demonstrated that the γc/RAG2 mice developed for this study are completely deficient in NK activity. Although the NOD/LtSz-scid mice have reduced NK activity relative to normal or CB17-scid mice, we and others have shown that these mice retain a significant level of poly I:C-induced NK activity ( Shultz et al, 1995 ). γc/RAG2 mice have been shown to engraft human B-LCLs and K562 cells more efficiently than NOD/LtSz-scid mice. The difference between the two strains was particularly noticeable following intravenous injection of cells, suggesting that viability was compromised in the circulation of NOD/LtSz-scid mice, possibly because of residual NK activity, although it has not been possible to demonstrate this in vitro.

Engraftment of human PBMCs in γc/RAG2 mice was also improved compared to NOD/LtSz-scid mice. 4 weeks after injection the average level of human CD45 positive spleen cells was higher. Although this difference was not statistically significant, it was apparent that a larger proportion of γc/RAG2 mice were highly engrafted (i.e. had > 10% human CD45-positive spleen cells) and engrafted a significantly higher proportion of CD4-positive cells. This would make the γc/RAG2 mice more useful for in vivo studies involving HIV infection. The recently described β2-microglobulin-deficient (β2mnull) NOD/LtSz-scid strain, in which poly-I:C-induced NK activity is absent, also showed enhanced engraftment of human PBLs relative to NOD/LtSz-scid with regard to both absolute human cell number and increased CD4+ T-cell. engraftment ( Christianson et al, 1997 ). However, B2mnull mice have been shown to mount IL2-driven NK responses in some circumstances ( Su et al, 1994 ), which may limit engraftment in β2mnull-NOD/LtSz-scid mice.

There are other advantages to the γc/RAG2 strain due to limitations, inherent in the genetic background of NOD/LtSz-scid mice, which restrict their utility as recipients for human cells. Spontaneous development of thymomas associated with the expression of an ecotropic murine leukaemia virus Emv30, limits the longevity of strains derived from the NOD background ( Serreze et al, 1995 ). This was particularly true for the β2-microglobulin-deficient (β2mnull) NOD/LtSz-scid strain ( Christianson et al, 1997 ). The genotype responsible for the defects in innate immunity in NOD mice have not been defined. Any genetic manipulations in these mice are therefore complicated by the requirement for extensive backcrossing. Additionally, in mice with the scid mutation, some aberrant joining of the V(D)J gene segments can occur in the absence of DNA-PKcs activity, leading to the production of functional TCR or immunoglobulin genes. For reasons that are unclear, this occurs less frequently in NOD/LtSz-scid mice ( Shultz et al, 1995 ). However, because the block in V(D)J recombination caused by the RAG2 mutation cannot be circumvented ( McBlane et al, 1995 ), immunocompetence in the form of circulating T lymphocytes and/or serum immunoglobulin, which develops with ageing in a significant number of scid mice ( Bosma et al, 1988 ; Carroll et al, 1989 ), did not occur in γc/RAG2 mice.

Consistent with previous studies, the majority of human cells in the spleens of engrafted γc/RAG2 mice were HLA-DR+CD3+, indicative of a memory/activated phenotype ( Murphy et al, 1992 ; Tary-Lehmann & Saxon, 1992). We have seen evidence for a mild GvHD in PBL-engrafted mice after several weeks. These mice may therefore provide a useful in vivo model to investigate novel therapies for the treatment of GvHD. The utility of PBL-engrafted γc/RAG2 mice for measurement of immune responses may be limited by xenoreactive responses and may fail to engraft a representative T-lymphocyte repertoire ( Tary-Lehmann et al, 1995 ; Garcia et al, 1997 ). However, a recent study has shown that early engraftment facilitated by depletion of NK cells can result in the persistence of vaccine-responsive human T lymphocytes with a naïve phenotype (CD45RA+RO and HLA-DR), and which reflect a repertoire similar to that of the donor ( Albert et al, 1997 ).

γc/RAG2 mice are a new strain of immunodeficient mice derived from animals with defined molecular lesions and have distinct advantages over existing strains. They engrafted human B-LCLs and tumour cells more efficiently than NOD/LtSz-scid animals under the same conditions, and facilitated engraftment of T lymphocytes with a greater proportion of CD4+ cells. γc/RAG2 mice may therefore be useful for study of therapies for GvHD, adoptive immunotherapy for haemopoietic malignancies and HIV infection. Selective introduction of additional defects of innate immunity is simplified by the defined γc/RAG2 genotype and may further facilitate engraftment of human cells. For this reason C5-deficient γc/RAG2 mice have been created, and are currently being tested for engraftment of human haemopoietic progenitor and lymphoid cells.


The authors thank Ms Marguerita Evans for her assistance with the breeding and maintenance of the immunodeficient mouse colonies. This work was supported in part by the Primary Immunodeficiency Association (J.P.G.), the Insititut National de la Santé et de la Recherche Medicale (J.P.D.), the Association pour le Recherche sur le Cancer (J.P.D.), the Ligue Nationale Contre le Cancer (J.P.D.) and the Wellcome Trust (A.J.T.).