Research ArticleType 1 Diabetes

A Model for Personalized in Vivo Analysis of Human Immune Responsiveness

See allHide authors and affiliations

Science Translational Medicine  14 Mar 2012:
Vol. 4, Issue 125, pp. 125ra30
DOI: 10.1126/scitranslmed.3003481


Studies of human immune diseases are generally limited to the analysis of peripheral blood lymphocytes of heterogeneous patient populations. Improved models are needed to allow analysis of fundamental immunologic abnormalities predisposing to disease and in which to assess immunotherapies. Immunodeficient mice receiving human fetal thymus grafts and fetal CD34+ cells intravenously produce robust human immune systems, allowing analysis of human T cell development and function. However, to use humanized mice to study human immune-mediated disorders, immune systems must be generated from adult hematopoietic cells. Here, we demonstrated robust immune reconstitution in mice with hematopoietic stem cells (HSCs) aspirated from bone marrow of adults with type 1 diabetes (T1D) and healthy control volunteers. In these humanized mice, cryopreservation of human leukocyte antigen allele–matched fetal thymic tissue prevented allogeneic adult HSC rejection. Newly generated T cells, which included regulatory T cells (Tregs), were functional and self-tolerant and had a diverse repertoire. The immune recognition of these mice mimicked that of the adult CD34+ cell donor, but the T cell phenotypes were more predominantly “naïve” than those of the adult donors. HSCs from T1D and control donors generated similar numbers of natural Tregs intrathymically; however, peripheral T cells from T1D subjects showed increased proportions of activated or memory cells compared to controls, suggesting possible HSC-intrinsic differences in T cell homeostasis that might underlie immune pathology in T1D. This “personalized immune” mouse provides a new model for individualized analysis of human immune responses that may provide new insights into not only T1D but also other forms of immune function and dysfunction as well.


Although large-scale studies of human populations have provided important clues to the genetic basis of immune diseases and responses, little is known about the mechanisms by which these genes exert their effects. The ability to dissect these mechanisms in patient populations is currently limited largely to the analysis of peripheral blood lymphocytes from individuals with diverse disease characteristics, duration, treatments, and environments and in whom immunological causes and effects of inflammatory cascades cannot be readily distinguished. Thus, there is a need for models that eliminate all of these interindividual variables while allowing analysis of individuals with demonstrated disease. Although human peripheral blood mononuclear cells (PBMCs) can populate immunodeficient mice (1), the function of T cells is limited in this setting and complicated by xenogeneic graft-versus-host reactivity (2). Human T cells develop in human fetal thymus grafts implanted with fetal liver under the kidney capsule (3). The combination of intravenous human hematopoietic stem cell (HSC) infusion with human fetal thymus and liver grafts under the kidney capsule allows human immune reconstitution with high levels of peripheral human T cells, B cells, and both myeloid and plasmacytoid dendritic cells (4), with antigen-specific immune responses in vivo (46). Normal thymic development of regulatory T cells (Tregs) with suppressive function (7) and homeostatic peripheral expansion of human T cells occurs (8). However, this model is limited to the analysis of fetal HSC-derived immune systems.

Humanized mice provide the opportunity to analyze the effects of autoimmunity-associated genetic polymorphisms on immune regulation. Recently defined non–human leukocyte antigen (HLA)–linked genes collectively confer substantial autoimmune disease risk (912). In humans with autoimmune diseases, however, underlying immunoregulatory defects arising from non–HLA-associated genes are largely undefined. Because many of these loci contain immunoregulatory genes, such as cytokines and costimulatory and inhibitory molecules (912), intrinsic abnormalities in the cells of the immune system, which originate from HSCs, likely contribute to the development of autoimmunity. Consistently, diabetes disease susceptibility is transferred via hematopoietic cells in nonobese diabetic (NOD) mice (13) and possibly in humans (14). However, studies of patients with disease cannot distinguish underlying causes from effects of disease evolution, disease treatment, or precipitating environmental factors.

Fulfillment of the above basic research potentialities of humanized mice would require achievement of human immune reconstitution and function with adult HSCs obtained from patients. However, these cells are not available in large quantities from study volunteers, and adult HSCs engraft less efficiently than fetal CD34+ cells in immunodeficient mice (15). Furthermore, even if obtained in large quantities for therapeutic applications, adult HSCs may be rejected by allogeneic thymocytes preexisting in fetal thymus grafts. Here, we report the development of a new humanized mouse model that supports robust peripheral reconstitution of T cells and antigen-presenting cells (APCs) from small numbers of adult, allogeneic bone marrow CD34+ cells. This “personalized immune” (PI) mouse model could potentially be used to identify HSC-intrinsic immune abnormalities predisposing to autoimmunity as well as model individual human immune responses in both health and disease.


Overcoming the immune barrier imposed by mature T cells in fetal thymus grafts

To assess human immune reconstitution from adult HSCs in immunodeficient mice, we isolated CD34+ cells from discarded human bone marrow infusion filters and gave them intravenously to sublethally irradiated NOD/SCID (severe combined immunodeficient) mice receiving fetal thymus transplantation. Recipients of untreated fetal human thymus grafts showed low peripheral T cell reconstitution during the first weeks after transplantation, which declined markedly over time, suggesting that these cells emigrated from the graft (average CD3+ cell reconstitution of PBMCs at 6 weeks, 7.0 ± 8.22%, n = 5, versus 2.34 ± 2.26% at 16 weeks after transplantation). Non-T cells did not reconstitute from injected allogeneic CD34+ cells, suggesting that these were rejected. Moreover, some animals reconstituted with fetal human thymus grafts and CD34+ cells have developed a late-onset graft-versus-host disease (GVHD)–like wasting syndrome that resulted in mortality (fig. S1). We hypothesized that thymocytes preexisting in the thymus grafts might reject allogeneic CD34+ cells and expand to attack recipient tissues, preventing immune reconstitution and causing xenogeneic GVHD, respectively. We therefore tested methods for depleting graft thymocytes in an effort to prevent these phenomena.

Fetal thymus organ culture (FTOC) with 2′-deoxyguanosine (dGuo) depletes thymocytes while preserving stromal elements (16) that can support thymopoiesis (17). NOD/SCID mice received allogeneic adult CD34+ cells plus fetal thymus tissue cultured for 7 or 21 days in the presence of dGuo. Control animals received fetal liver CD34+ cells from the thymic tissue donor. Mice that received 5 × 105 adult CD34+ cells without a thymus graft reconstituted an average of 20% human PBMCs by week 10 (Fig. 1A), with robust B cell reconstitution (Fig. 1B), but CD3+ cells were undetectable (Fig. 1C). In mice that received 7-day dGuo-cultured thymus tissue plus allogeneic CD34+ cells, CD3+ levels averaging ~7% of PBMCs were detectable by 6 weeks and subsequently declined (Fig. 1C), but CD19+ cells did not appear (Fig. 1B). These data suggest that mature T cells escaping dGuo depletion may have rejected the infused allogeneic CD34+ cells. Cells within the thymus grafts did not achieve human non-T cell reconstitution or a high level of T cell reconstitution (Fig. 1C).

Fig. 1

Peripheral human cell reconstitution in NOD/SCID mice after transplantation of dGuo-treated thymic tissue. Fetal thymic tissue was treated with dGuo for 7 (black diamonds, n = 4) or 21 days (black triangles, n = 5) before transplantation into sublethally irradiated NOD/SCID that received 5 × 105 allogeneic, adult CD34+ cells (open squares and black diamonds) or 4 × 105 autologous fetal liver CD34+ cells (open circles, n = 5) intravenously. Age-matched control animals received 5 × 105 adult CD34+ cells alone (open squares, n = 5). The mice were bled to measure human (Hu) cell reconstitution in (total mouse plus human) PBMCs at the indicated time points. (A) Total human chimerism. (B) Percentage of human B cells. (C) Percentage of human T cells among PBMCs at indicated time points.

In contrast, successful thymic engraftment with human thymopoiesis as well as peripheral CD19+ cell reconstitution occurred after intravenous infusion of 5 × 105 allogeneic adult CD34+ cells with a 21-day dGuo-cultured thymus graft (average human CD3+ cells among PBMCs at 20 weeks, ~25%; Fig. 1, B and C). Because recipients of CD34+ cells without a thymus graft did not reconstitute T cells and thymus grafts treated with dGuo for even 7 days did not contain enough progenitors to permit long-term (>14 weeks) T cell reconstitution (Fig. 1C), we conclude that progenitors derived from peripherally infused allogeneic adult CD34+ cells populated 21-day dGuo-treated thymi, underwent thymopoiesis, and emigrated to the periphery. Control recipients of dGuo-treated fetal thymus tissue with 4 × 105 autologous fetal liver CD34+ cells instead of allogeneic adult marrow CD34+ cells showed more rapid T cell reconstitution (Fig. 1C) than was achieved with allogeneic adult CD34+ cells.

Although the above recipients of 21-day FTOC grafts plus 5 × 105 adult CD34+ cells exhibited high levels of long-term (>20 weeks) human chimerism (Fig. 1), a lower dose of adult CD34+ cells did not achieve robust immune reconstitution (average human cells in PBMCs at 20 weeks in recipients of 2 × 105 CD34+ cells, 3.89 ± 9.89%; n = 9). Only limited HSCs are available through volunteer bone marrow aspiration. Therefore, we used NOD/SCID/IL-2 (interleukin-2) receptor γ chainnull (NSG) mice, which lack natural killer cells and are more permissive for engraftment of human HSCs (18), for the ensuing experiments.

We evaluated irradiation of thymus grafts to deplete preexisting thymocytes. NSG mice receiving 7-Gy–irradiated thymus plus 3 × 105 adult CD34+ cells showed excellent B cell and monocyte reconstitution, but low numbers of peripheral T cells by 20 weeks (fig. S2). Thymus grafts were so small as to be barely visible under the kidney capsule upon laparotomy (fig. S3A). T cells eventually reconstituted the periphery by 34 weeks after implantation (fig. S2).

Cryopreservation and thawing of fetal thymus grafts to allow peripheral reconstitution of T cells and multiple hematopoietic lineages from allogeneic, adult human HSCs

Transplantation of cryopreserved and thawed mouse thymus tissue can restore immune function (19, 20). Because the studies above suggested that human fetal thymus tissue contains viable, alloreactive, and xenoreactive thymocytes, we evaluated the ability of cryopreservation of intact fetal thymic tissue fragments to deplete these thymocytes. As shown in fig. S4, cryopreservation of thymic tissue indeed led to marked depletion (1567-fold decrease in total cellularity) of all thymocyte subsets from fetal thymic tissue.

To test the use of cryopreserved fetal thymic tissue, sublethally irradiated NSG mice received cryopreserved and thawed fetal human thymus grafts plus 3 × 105 or 5 × 105 allogeneic, adult CD34+ cells intravenously. To further ensure depletion of T cells derived from preexisting graft thymocytes, the mice received a depleting anti-human CD2 monoclonal antibody (mAb). As shown in Fig. 2, all mice achieved human B cell and monocyte chimerism by 6 weeks. Unlike recipients of CD34+ cells alone, which showed minimal T cell reconstitution, recipients of cryopreserved and thawed thymus grafts generated peripheral T cells, which appeared by 6 weeks and peaked at ~10% and ~30% of PBMCs at 16 weeks after infusion with 3 × 105 or 5 × 105 CD34+ cells, respectively (Fig. 2A, lower right panel). Although it took 20 weeks to achieve T cell reconstitution with 21-day dGuo-treated grafts, similar T cell levels were reconstituted by 10 weeks with cryopreserved and thawed thymus grafts. At the time of animal killing, these cryopreserved grafts were markedly enlarged, showing evidence of robust human thymopoiesis with predominant CD4/CD8 double-positive thymocytes (Fig. 2, B and C).

Fig. 2

Multilineage human cell reconstitution in NSG mice receiving cryopreserved and thawed thymic grafts and allogeneic, adult CD34+ cells. (A) Sublethally irradiated NSG mice that received a cryopreserved and thawed fetal thymus graft in combination with 3 × 105 (black squares, n = 6) or 5 × 105 (black triangles, n = 6) adult CD34+ cells and two doses of anti-CD2 mAb intravenously (i.v.) were bled to measure human cell reconstitution in (total mouse plus human) PBMCs at the indicated time points. Age-matched control animals received 3 × 105 adult HSCs alone (white squares, n = 6). Single-cell suspensions of PBMCs were stained for markers of human hematopoietic cells (CD45), T cells (CD3), B cells (CD19), and monocytes (CD14). Dead cells and mouse red blood cells were excluded from the analysis. (B) Representative thymus graft appearance 20 weeks after transplantation. NSG mice that were transplanted with cryopreserved and thawed fetal thymus tissue had abundant, viable thymic tissue underneath the kidney capsule. (C) Flow cytometry analysis of thymocytes (representative of 12 grafts in a single experiment).

Control animals receiving cryopreserved thymus grafts without intravenous CD34+ cells [with (n = 5) or without (n = 4) anti-CD2 mAb] did not repopulate significant human T cells or non-T cells in the periphery (0.46 ± 0.34% human cell reconstitution 20 weeks after transplantation). Thus, most of the preexisting graft thymocytes were depleted by cryopreservation, and administration of CD34+ cells was necessary for human T cell and non-T cell reconstitution. Additional control NSG animals receiving fresh thymic tissue with allogeneic CD34+ cells showed human T cells (mean of PBMCs, 2.4%) but no human chimerism in any other lineage by 5 weeks (<0.006%), whereas recipients of autologous CD34+ cells showed significant human T cell (mean, 3.4%), B cell (mean, 2.1%), and monocyte (mean, 0.2%) reconstitution by this time (fig. S5). These data are consistent with the interpretation that T cells from fresh thymic tissue implanted into NSG mice rejected allogeneic CD34+ cells.

None of the long-term animals (17 of 17 followed for 20 weeks) that received cryopreserved and thawed thymus plus anti-CD2 mAb developed wasting syndrome or other evidence of GVHD. Mouse class II+ cells were present in the long-term human thymus grafts (fig. S6). Anti-CD2 mAb was not required to prevent rejection of allogeneic donor stem cells, because no difference was seen in the level of human reconstitution by 15 weeks when groups of mice receiving cryopreserved thymic grafts and allogeneic CD34+ cells with or without anti-CD2 mAb treatment were compared (fig. S7).

Human immune reconstitution from a bedside bone marrow aspirate from control and type 1 diabetes volunteers

We next evaluated reconstitution capabilities of adult CD34+ cells isolated from bedside bone marrow aspiration. An aspiration of 15-ml bone marrow yielded 3.6 × 105 and 2.7 × 106 CD34+ cells from an initial healthy control and type 1 diabetes (T1D) volunteer subject, respectively. Sublethally irradiated NSG mice received 1.8 × 105 adult CD34+ cells each plus a cryopreserved and thawed human fetal thymus graft and anti-human CD2 mAb. Control irradiated mice received CD34+ cells without thymus tissue. Human chimerism was detectable by week 6 and peaked at ~25 to 80%. Recipients of thymus grafts plus intravenous CD34+ cells from the control and T1D volunteers developed substantial CD3+ cell levels by 8 weeks, whereas control mice (no thymus graft) had markedly delayed T cell reconstitution (Fig. 3). CD19+ and CD14+ cells also developed from the HSCs of the T1D and control volunteers. Similar results were obtained in three additional experiments, in each of which 4 to 14 NSG mice were each reconstituted with 2 × 105 to 3 × 105 CD34+ cells from a single volunteer aspirate. Composite data from these experiments are presented in Fig. 3A. Splenic T cell reconstitution was also rapid and robust, with a mean of 2.47 × 106 (SEM, 0.6 × 106) human CD3 cells per spleen in a group of three T1D cell–reconstituted mice and 106 CD3 cells in a healthy control–reconstituted mouse spleen that were killed at 9 weeks. Figure 3B shows the robust thymopoiesis from adult CD34+ cells, including normal proportions of CD4/CD8 double- and single-positive cells and similar proportions of CD45RO+ and CD45RA+ cells among single-positive thymocytes, as was seen with fetal thymus and fetal CD34+ cells (7). Although thymocyte numbers tended to be lower in T1D compared to healthy control CD34 cell–reconstituted animals, no statistically significant differences were seen between the two groups.

Fig. 3

Multilineage human cell reconstitution in NSG mice receiving cryopreserved/thawed thymic grafts and allogeneic, adult CD34+ cells isolated from bedside bone marrow aspirates. (A) Sublethally irradiated NSG mice received cryopreserved/thawed fetal thymus (Thy) tissue in combination with 1.8 × 105 to 3.0 × 105 adult CD34+ cells isolated from bone marrow aspirates from healthy volunteers (black squares, n = 3 donors, 6 recipients) and T1D subjects (black triangles, n = 4 donors, 29 recipients). Mean levels of human cell reconstitution in (total mouse plus human) PBMCs are shown over time. Control animals received adult HSCs alone from the T1D subjects (open circles, n = 2 donors, 7 recipients). Thymus grafts and bone marrow donors were HLA-typed for T1D-associated DRB and DQB alleles and HLA A*201 using single-nucleotide polymorphism genotyping assays. The thymic tissue and bone marrow donors shared at least HLA*A201 and DRB*0302 and/or DQB*0301. ANOVA revealed an effect of thymus transplant on CD3 reconstitution comparing T1D CD34+ cells alone with T1D CD34+ plus thymus transplantation at early (6 to 14 weeks) time points. Individual time-point comparisons with Mann-Whitney U test revealed significant differences at 8 to 14 weeks (*P < 0.05). (B) Graft thymocytes were analyzed 22 to 25 weeks after transplantation in T1D (n = 5) and control (n = 3) HSC-reconstituted animals. Means + SEM are shown. No significant differences between T1D and control animals were noted.

T cell function and self-tolerance in mice reconstituted with volunteer donor bone marrow CD34+ cells

We assessed T cell function by transplanting allogeneic human and xenogeneic (pig) skin to thymus-grafted mice that received adult CD34+ cells. These mice rapidly rejected allogeneic human and xenogeneic pig skin grafts (Fig. 4A), whereas naïve, untreated NSG mice accepted allogeneic human and xenogeneic skin grafts for the duration of follow-up (106 and 50 days, respectively) (Fig. 4A), with no infiltrates or evidence for rejection on histology (fig. S8).

Fig. 4

Functional and self-tolerant immune systems in NSG mice receiving fetal thymus graft and adult CD34+ cells. (A) NSG mice (n = 3) that received a 7 Gy irradiated thymic graft plus 3 × 105 adult CD34+ cells reconstituted peripheral T cells >30 weeks after transplantation. Thirty-nine weeks after transplantation, they were grafted with allogeneic human and xenogeneic pig skin. Survival of pig and human skin grafts (n = 3 and 4, respectively) on untreated control NSG mice (“naïve NSG”) is also shown. (B) Human T cells (>90% pure) were enriched from the spleen and peripheral lymph nodes of NSG mice 20 weeks after transplantation of cryopreserved and thawed thymus grafts and allogeneic bone marrow CD34+ cells from a healthy control (black bar) or T1D subject (dotted bars). Control T cells were isolated from PBMCs of the same healthy control volunteer (open bar). Table S1 shows the naïve/memory cell distribution of CD4 cells in the three mice reconstituted from T1D CD34+ cells.

To assess self-tolerance of T cells generated from adult CD34+ cells of T1D and healthy volunteers, we performed mixed lymphocyte reactions (MLRs) using purified T cells isolated from the spleens and lymph nodes. T cells from mice reconstituted from T1D and control subjects showed self-tolerance along with strong responses to allogeneic human stimulators in MLRs (Fig. 4B and table S1). Notably, fresh adult donor T cells and T cells from a mouse reconstituted from the same healthy control bone marrow donor showed similar, robust responses to the allogeneic stimulator and similar self-tolerance. Thus, immune responsiveness and self-tolerance to the adult volunteer were recapitulated in these mice. Because they generate immune function from an individual adult bone marrow donor, we refer to these animals henceforth as PI mice.

Similar Treg development from T1D and control bone marrow CD34+ cells

We assessed the presence of Tregs in thymus grafts and the periphery of reconstituted mice. As shown in Fig. 5A, CD25highFoxP3+ Tregs were present among CD4+CD8 thymocytes of PI mice. Analysis of CD4+CD8CD25+CD127lo thymocytes demonstrated that most FoxP3+ cells were also Helios+, indicative of thymically derived “natural” Tregs. Similar numbers and proportions of Tregs were detected in thymus grafts reconstituted from control and T1D volunteers. Furthermore, although some studies have suggested that Treg numbers are reduced in the blood of T1D subjects compared to healthy controls (21), similar proportions of Tregs were detected in the peripheral blood of both groups of reconstituted mice (Fig. 5B).

Fig. 5

Tregs in thymus grafts and periphery of PI mice. (A) Twenty to 22 weeks after transplantation, single-cell suspensions were prepared from thymus grafts of NSG mice that received a cryopreserved and thawed thymus graft and allogeneic CD34+ cells from one of two healthy volunteers (circles) or one T1D subject (squares) and analyzed for CD4+CD8CD25+FoxP3+ cells by flow cytometry (top row). As a marker for natural Tregs, Helios expression in CD4+CD8CD25+CD127loFoxP3+ thymocytes is shown in NSG mice derived from a second human donor pair in the bottom row. (B) Similar proportions of Tregs in PBMCs 20 weeks after transplantation of cryopreserved and thawed fetal thymus grafts with CD34+ cells from one of two healthy controls (black squares) or one T1D subject (black triangles) compared to two healthy humans (black circles). Left plots show CD25 and FoxP3 staining on CD4+ T cells from PI mice generated from control and T1D donors.

Diverse T cell receptor repertoire in single-positive thymocytes derived from adult donor CD34+ cells

At 20 weeks after transplantation, we performed spectratyping analysis on CD4 and CD8 single-positive thymocytes of mice reconstituted from T1D CD34+ cells and a normal volunteer (Fig. 6). These human T cells showed a diverse repertoire, with similar use of the BV families and a polyclonal CDR3 length distribution for each BV. The reconstituted repertoires resembled those of the average CD4 T cell repertoires of 12 healthy adults, with average Hamming distances for all analyzed BV families in each sample ranging from 14.2 to 26.2 (mean, 20.6) (table S2). This is indicative of typical T cell polyclonality as seen in healthy control peripheral blood lymphocytes.

Fig. 6

Diverse repertoire of T cells in PI mice. Spectratyping (β chain CDR3 length distribution) of human CD4 and CD8 single-positive T cells in thymus grafts reconstituted with CD34+ cells from one T1D donor or one of two healthy controls 20 weeks after transplantation. Spectratype from one representative animal (#5700) of six (mice) is shown. The vertical axis is relative fluorescence units (full scale = 6000 units). The horizontal axis is nucleotide size. Reference size markers are low fainter peaks. Representative BVs are shown from a total of 12 analyzed per sample. The Hamming distances of all six samples, a measure of the relative distances of the observed TCR β chain length distribution from a reference distribution of healthy adult CD4 T cells, are shown in table S1, and each indicates the reconstitution of a polyclonal repertoire.

Naïve versus memory T cell phenotype in PI mice

The T cells populating the peripheral tissues of mice reconstituted from CD34+ bone marrow cells of healthy control and T1D volunteers included both “naïve”-type CD45RA+CD45RO and “memory”-type CD45RACD45RO+ cells (Fig. 7A). Comparison of T cells in the blood of a control CD34+ cell donor revealed a marked increase in the proportion of naïve-type CD45RA+CD45RO CD4, CD8, and Treg subsets in the PI mouse reconstituted from the same donor (Fig. 7, A and B). Thus, a rejuvenated version of the adult donor’s immune system is generated in PI mice. As shown in Fig. 7C, the human thymus was necessary for this rejuvenation, because the proportion of naïve-type T cells in PBMCs of recipients of CD34+ cells alone was markedly lower than that in mice that also received thymus grafts.

Fig. 7

Naïve versus memory phenotype of T cells in PI mice. (A and B) Proportions of CD45RA+ CD4 and CD8 T cells and Tregs (B) in PBMCs of healthy volunteers and of PI mice 20 weeks after thymus implantation plus intravenous infusion of CD34+ cells from one T1D subject (black squares) or one of two healthy controls (black triangles), including (open circle) the donor of CD34+ cells for the control mouse indicated with an open triangle (*P < 0.05, excluding the outlier in the CD8 population of controls from statistical analysis). (C) T cell populations were assayed by flow cytometry in NSG mice injected with CD34+ fetal liver cells with or without allogeneic thymus at 7 weeks after transplant. Means ± SEM are shown (n = 4 for each group). *P < 0.05; **P < 0.005.

When we compared the proportions of naïve- and memory-type CD4 and CD8 T cells in the blood of PI mice generated simultaneously from T1D or healthy control donors, the T cells derived from T1D CD34+ cells showed significantly reduced proportions of naïve-type cells compared to those derived from healthy controls (Fig. 7A). Tregs derived from CD34+ cells of T1D donors tended more toward the memory phenotype than those from healthy controls, but this trend did not achieve statistical significance (Fig. 7B; P = 0.07).


Here, we sought to develop a means of generating functional immune systems from HSCs of adult humans in immunodeficient mice to overcome the limitations of current models for the analysis of human immunoregulation in health and disease. We demonstrate that adult, bone marrow–derived CD34+ cells can reconstitute NSG mice grafted with cryopreserved and thawed allogeneic thymus tissue, generating multiple hematopoietic lineages, including T cells, B cells, and myeloid cells. Cryopreserving and thawing the fetal thymus plus administering anti-CD2 mAb depletes mature T cells from the graft, preventing rejection of allogeneic CD34+ cells and GVHD while preserving thymic function. Thymopoiesis, growth of the thymus graft, and reconstitution of a functional, diverse, and rejuvenated immune system are achieved. Self-tolerance of the adult donors is recapitulated. Although fetal liver fragments were included in the humanized mouse model upon which our studies are based (4, 6), these fragments are not required, because thymocyte progenitors from infused CD34+ cells populated the human thymic grafts in the current study.

Although in vivo thymopoiesis and peripheral reconstitution were also achieved from dGuo-treated human thymi, T cell reconstitution from infused adult CD34+ cells was slow when thymi were dGuo-treated sufficiently long (21 days) to prevent rejection of allogeneic CD34+ cells. Slow T cell recovery has also been observed in patients with complete DiGeorge syndrome receiving thymic tissue cultured for several weeks in dGuo (22). Our results suggest that cryopreservation of thymic tissue might support more rapid T cell recovery while preventing GVHD.

Cryopreservation of fetal thymus tissue permits HLA typing of tissue for use with adult CD34+ cells sharing HLA alleles, which is important for optimal immune function. The use of NSG mice allows the engraftment of relatively small numbers of allogeneic adult HSCs, allowing reconstitution of multiple mice from a bedside bone marrow aspirate.

The specific tolerance to CD34+ cell donor “self” antigens and the absence of GVHD in our studies most likely reflect intrathymic deletion due to the presence of APCs from the human HSC donor and the murine recipient, respectively, in the human thymus graft, as previously suggested in another thymic xenograft model (23). Although not tested directly, we hypothesize that the inclusion of anti-CD2 mAb was important for the prevention of a wasting syndrome induced by residual xenogeneic GVH-reactive mature T cells emigrating from fetal human thymus grafts. This possibility was suggested by the development of a late-onset (at 22 weeks) GVHD-like syndrome (severe alopecia, skin inflammation, hunched posture, and weight loss) in the only mouse that did not receive anti-CD2 mAb within a group of NSG mice receiving cryopreserved thymus grafts (plus allogeneic CD34+ cells intravenously).

Immune reconstitution from adult bone marrow CD34+ cells of patients in NSG mice provides an immune system unaltered by disease, allowing comparison of individuals in a controlled and prospective manner. Human immune analyses are typically limited to peripheral blood samples, and underlying immune dysregulation cannot be distinguished from the ensuing cascade of inflammatory events that culminate in disease. Defects in Treg numbers and function have been reported for T1D (21, 2426), systemic lupus erythematosus (27), and rheumatoid arthritis (28), but this has been controversial in T1D (2931). We observed no gross abnormalities in the T cell populations generated from T1D subjects’ CD34+ cells, which generated Tregs intrathymically in proportions similar to that of healthy control CD34+ cells. However, we observed significantly reduced proportions of naïve-type T cells in the blood of PI mice generated from T1D compared to healthy control donors, suggesting that abnormalities of T cell homeostasis, as described in NOD mice (32), might be a feature of T1D-derived HSCs. Our model will allow assessment of genetically programmed, HSC-intrinsic immunoregulatory abnormalities in T1D in relation to predisposing gene alleles.

HLA-transgenic immunocompetent mice have provided insight into the pathogenesis of autoimmune diseases such as rheumatoid arthritis (33), multiple sclerosis (34), celiac disease (35), and T1D (3639). However, none of these models permit analyses of human HSC-intrinsic, genetically determined immune abnormalities that may contribute to autoimmune pathogenesis. In contrast, the combined administration of intravenous CD34+ cells and fetal thymus tissue in immunodeficient mice generated functional human T cells, T-B interactions, and class-switched antibody responses, with secondary lymphoid organs containing both plasmacytoid and myeloid dendritic cells (46). Because Tregs develop normally (7) and T cell homeostasis can be studied in this model (8), the model here should allow assessment of HSC-intrinsic immunoregulatory abnormalities associated with autoimmune diseases in HSC donors. The ability to HLA-type the thymus before transplantation allows selection for thymi with disease-associated HLA alleles.

Our model has not yet been developed to allow the study of diabetes pathogenesis, because animals reconstituted with HSCs from T1D patients did not develop evidence of insulitis (fig. S9). Although we would not expect transplantation of T1D HSCs to cause autoimmune disease in unmodified NSG mice, further development of the model using HLA-transgenic NSG mice might permit studies of autoimmune disease pathogenesis.

Our PI mouse model will also allow the analysis of individual responsiveness of an adult marrow donor to immunotherapeutic agents. In addition, the reconstitution of multiple mice with naïve T cells with a diverse repertoire derived from adult HSCs could potentially provide patients with thymic insufficiency due to immunosuppressants, chemotherapy, irradiation, or HIV, with functional, self-tolerant T cells for adoptive transfer. Mice receiving human fetal thymus and CD34+ cell grafts generate anti-HIV and other antigen-specific immune responses (5, 40), suggesting the immunotherapeutic potential of this approach. However, robust methods of overcoming the infectious risks of murine pathogens, including endogenous retroviruses (41), would be needed before this approach could be considered.

In summary, we have established a model that permits the development of multilineage peripheral human hematopoietic cells from adult HSCs. The PI mouse provides an immune system unaltered by disease or its treatment that should allow the analysis of intrinsic defects in immunoregulation associated with autoimmune disorders and of genetically controlled responses to immunotherapies.

Materials and Methods

Animals and human tissues and cells

NOD/SCID and NSG mice were obtained from the Jackson Laboratory and housed in a specific pathogen-free microisolator environment. Human fetal thymus and liver tissues (gestational age, 17 to 20 weeks) were obtained from Advanced Biosciences Resource. Fetal thymus fragments were cryopreserved in 10% dimethyl sulfoxide and 90% human AB serum (Atlanta Biologicals), irradiated, or cultured, depending on the experimental design. CD34+ cells were isolated from a 15-ml bone marrow aspirate, from discarded human bone marrow filters obtained from the Massachusetts General Hospital (MGH) Bone Marrow Processing Laboratory, or from fetal human liver tissue by magnetic-activated cell sorter (MACS) separation with anti-human CD34+ microbeads (Miltenyi Biotec). Human skin was obtained from the National Disease Research Interchange, and pig skin was provided by D. H. Sachs (MGH). The use of human tissues/cells and animals was approved by the MGH and Columbia University Medical Center (CUMC) Human and Animal research review committees, respectively, and the experiments were performed in accordance with the approved protocols.

Blood samples and bone marrow aspirates from T1D and control volunteers were recruited through the Human Studies Core of the Harvard Juvenile Diabetes Research Foundation (JDRF) Autoimmunity Center from the Joslin Diabetes Institute or from the Naomi Berrie Diabetes Center at CUMC and were collected with written informed consent under protocols approved by the Institutional Review Boards of the MGH and the CUMC, respectively.

Fetal thymus organ culture

Human fetal thymus culture was performed as previously published (16). Briefly, thymus fragments were placed on 0.8-μm isopore membrane filters (Millipore) on 1-cm2 Gelfoam sponges (Pharmacia & Upjohn Co.). To eliminate endogenous thymocytes, we grew organ cultures in the presence of 1.35 mM dGuo (Sigma-Aldrich) in Dulbecco’s modified Eagle’s medium (Sigma-Aldrich) at 37°C for 7 or 21 days.

Human tissue transplantation

Mice were conditioned with sublethal (2.5 Gy) total-body irradiation. Human fetal thymus fragments measuring about 1 mm3 were implanted underneath the recipient kidney capsule. Within 24 hours, 1 × 105 to 5 × 105 human CD34+ cells were injected intravenously. Some recipients were treated intravenously with anti-human CD2 mAb [BTI322 (42); 100 μg] on days 0 and 7.

Skin grafting

Split-thickness (2.3 mm) skin samples from an MGH (Massachusetts General Hospital) miniature pig and an allogeneic human donor were grafted on the lateral thoracic wall 39 weeks after human tissue transplantation. Skin grafts were evaluated daily from day 7 onward to 4 weeks followed by at least one inspection every third day thereafter. Grafts were defined as rejected when less than 10% of the graft remained viable.

Flow cytometry

Levels of human hematopoietic cells in transplanted mice were assessed by multicolor flow cytometry. Mice were tail-bled at regular intervals after transplantation to obtain PBMCs, which were prepared with Histopaque-1077 (Sigma-Aldrich). Fluorochrome-labeled mAbs, purchased from BD Pharmingen, were used in different combinations: anti-mouse CD45, anti-mouse Ter119, anti-human CD4, anti-human CD8, anti-human CD14, anti-human CD19, anti-human CD45, anti-human CD3, anti-human CD45RA, anti-human CD45RO, anti-human CD127, anti-human FoxP3, anti-human CD25, and isotype control mAbs. Flow cytometry analysis was performed with a FACSCalibur, FACSCanto, or LSRII (BD), and analysis was performed by FlowJo software (TreeStar). Dead cells were excluded from the analysis by gating out low forward scatter and high propidium iodide–retaining cells. Murine erythroid cells were excluded by gating out mouse Ter119+ cells.

Mixed lymphocyte reactions

Splenocytes and lymph nodes were harvested from humanized mice, and mononuclear cell suspensions were isolated by Ficoll separation. Human T cells were enriched by depletion of mouse cells with anti-mouse CD45 and anti-Ter119 microbeads (Miltenyi Biotec) followed by T cell purification with the Pan T Cell Isolation Kit II (Miltenyi Biotec) according to the manufacturer’s instructions. Purity was >90%. Responder T cells (105 per well) were cultured with irradiated human allogeneic PBMCs (30 Gy, 105 cells per well) as stimulators for 5 days, and proliferation was measured via [3H]thymidine incorporation as we have described (43). In self-stimulated control cultures, responder cells were incubated with autologous PBMCs from the same humanized mouse depleted of mouse CD45+ and Ter119+ cells. Data are shown as mean [3H]thymidine incorporation in triplicate cultures.


Total RNA was extracted directly from 1 × 104 to 2 × 104 CD4 or CD8 single-positive thymocytes (purity, >80%) and reverse-transcribed, and single-strand complementary DNA synthesis was performed as described (44). Amplification reactions were performed with a T cell receptor (TCR) β chain constant region primer and individual variable region primers as described (44). Products were then used in run-off reactions with a Cβ-specific FAM (6-carboxyfluorescein)–labeled primer (Integrated DNA Technologies) as described (44). The labeled products were then used to determine the length distribution of the TCR β chain length. The size and area of the peaks corresponding to the DNA products were determined with an ABI 3100 Genetic Analyzer (Applied Biosystems) and analyzed with Applied Biosystems Genotyper 3.7 NT. Hamming distances to assess the quantitative difference between the experimental and the reference β chain length distributions of peripheral blood CD4 T cells in normal humans were calculated as described (44).

Statistical analysis

Statistical analysis and comparisons were performed with GraphPad Prism version 4.0 (GraphPad Software). Data in bar graphs are expressed as means ± SEM. Student’s t test for parametric data sets or Mann-Whitney test for nonparametric data sets was used to compare groups. Analysis of variance (ANOVA) was used to resolve overall effects between transplant groups over time, and Mann-Whitney test was used for individual time-point comparisons. A P value of less than 0.05 was considered to be statistically significant.

Supplementary Materials

Methods (Immunohistochemistry)

Fig. S1. Death due to late-onset GVHD-like wasting syndrome in humanized mice.

Fig. S2. Late T cell reconstitution in NSG mice receiving irradiated thymus transplants and adult CD34+ cells.

Fig. S3. Irradiation of thymus graft inhibits growth.

Fig. S4. Cryopreservation depletes thymocytes in human fetal graft.

Fig. S5. Human cell reconstitution with autologous versus allogeneic CD34+ fetal liver cells and transplantation of fresh fetal thymus.

Fig. S6. Antigen-presenting cells from the recipient mouse in human thymic graft of PI mouse.

Fig. S7. Effect of anti-CD2 mAb BTI322 on chimerism in humanized mice.

Fig. S8. Human allografts are accepted by unmanipulated NSG mice.

Fig. S9. Normal islet histology in PI mice reconstituted with T1D CD34+ cells.

Table S1. T cell percentages in PI mice reconstituted with T1D marrow used for MLRs in Fig. 4B.

Table S2. TCR Vβ distribution and Hamming distances for Fig. 6.

References and Notes

  1. Acknowledgments: We thank J. Sachs and H. Wang for critical review of this manuscript, O. Moreno for outstanding animal husbandry, S. Washington for expert assistance with the manuscript, and G. Eisenbarth and T. Armstrong at the Barbara Davis Center for Childhood Diabetes for HLA typing. Funding: Supported by the JDRF Autoimmunity Center at Harvard University and by NIH grant RO1 AI084903. H.K. was supported by the Deutsche Forschungsgemeinschaft (German Research Foundation) and N.D. was supported by the American Diabetes Association. Author contributions: H.K. designed and performed the experiments, analyzed the results, and wrote the paper; N.D. designed and performed the experiments, analyzed the results, and wrote the paper; T.O. assisted with the design and performance of the experiments; T.F. participated in the performance of experiments; R.W. designed, conducted, and analyzed spectratyping experiments; R.G. assisted with regulatory approval, volunteer selection criteria, and recruitment; E.G. assisted with and coordinated volunteer recruitment and procurement of tissue samples; T.R.S. assisted with regulatory approval and procured human bone marrow samples; D.G.S. procured human bone marrow samples; H.T. designed, performed, and analyzed skin grafting experiments; Y.-G.Y. provided significant intellectual input; M.S. provided funding, oversaw the design, conduct, and interpretation of all experiments, and, with H.K. and N.D., wrote the paper. Competing interests: M.S. and H.K. are authors on patent IR2871/19240-021US1, “Generation of autologous T cells in mice,” submitted by Columbia University. The other authors declare that they have no competing interests.

Stay Connected to Science Translational Medicine

Navigate This Article