Research ArticleGraft-Versus-Host Disease

Human CD8+ Regulatory T Cells Inhibit GVHD and Preserve General Immunity in Humanized Mice

See allHide authors and affiliations

Science Translational Medicine  16 Jan 2013:
Vol. 5, Issue 168, pp. 168ra9
DOI: 10.1126/scitranslmed.3004943

Abstract

Graft-versus-host disease (GVHD) is a lethal complication of allogeneic bone marrow transplantation (BMT). Immunosuppressive agents are currently used to control GVHD but may cause general immune suppression and limit the effectiveness of BMT. Adoptive transfer of regulatory T cells (Tregs) can prevent GVHD in rodents, suggesting a therapeutic potential of Tregs for GVHD in humans. However, the clinical application of Treg-based therapy is hampered by the low frequency of human Tregs and the lack of a reliable model to test their therapeutic effects in vivo. Recently, we successfully generated human alloantigen-specific CD8hi Tregs in a large scale from antigenically naïve precursors ex vivo using allogeneic CD40-activated B cells as stimulators. We report a human allogeneic GVHD model established in humanized mice to mimic GVHD after BMT in humans. We demonstrate that ex vivo–induced CD8hi Tregs controlled GVHD in an allospecific manner by reducing alloreactive T cell proliferation as well as decreasing inflammatory cytokine and chemokine secretion within target organs through a CTLA-4–dependent mechanism in humanized mice. These CD8hi Tregs induced long-term tolerance effectively without compromising general immunity and graft-versus-tumor activity. Our results support testing of human CD8hi Tregs in GVHD in clinical trials.

Introduction

Bone marrow transplantation (BMT) is now widely accepted as an effective treatment for malignant and nonmalignant hematologic diseases. However, graft-versus-host disease (GVHD) is a lethal complication of allogeneic BMT. The current strategy of controlling GVHD by depletion (1) or general inhibition (2) of donor T cells using immunosuppressive agents may cause general immune suppression, resulting in tumor relapse or opportunistic infection, and limiting the effectiveness of BMT (2). The ideal for BMT would be to induce a sustained state of specific tolerance to alloantigen with minimal or no conventional immunosuppression.

Alloantigen-specific regulatory T cells (Tregs), which are the negative regulators of immune responses to alloantigen, are critical for maintaining alloantigen-specific tolerance (35). In addition to the well-described role of CD4+ Tregs in suppressing excessive immune responses (6, 7), CD8+ Tregs have also been reported to contribute in maintaining immune tolerance (812). Adoptive transfer of murine alloantigen-specific Tregs can prevent GVHD and allograft rejection in mouse models (5, 13), indicating that Treg-based therapy has a great therapeutic potential for these diseases in humans. However, the clinical application of Treg-based therapy is limited by the low frequency of human Tregs and the lack of a reliable model to evaluate their therapeutic effects in vivo.

Although several protocols have been developed for generation of human CD8+ Tregs (1419), none has been robust in terms of being of a practical scale for clinical use. Recently, using allogeneic CD40-activated B cells as the tolerogenic antigen-presenting cells (2022) to stimulate naïve CD8+CD25 T cells, we developed a simple, cost-effective protocol to rapidly induce and expand large numbers of functional human alloantigen-specific CD8+ Tregs. These induced CD8+ Tregs expressed higher levels of CD8 on their surface than their precursors and are thus identified as CD8hi Tregs (23).

In vivo studies of human CD4+ Tregs in immunodeficient mice showed that human CD4+ Tregs can prevent the rejection of human skin allograft and the development of transplant atherosclerosis (24, 25). However, because the immunodeficient mice used in these studies did not contain a stable human immune system before the adoptive transfer of human Tregs, the relevance of these results to human disease is unclear. In addition, the in vivo function of ex vivo–induced human CD8+ Tregs still remains unknown. Therefore, developing more reliable models to mimic human diseases and evaluate the function of the ex vivo–induced human Tregs is urgently required.

Previously, we established a complete human immune system in C57BL/10SgAiRag2−/−γc−/− (Rag2−/−γc−/−) mice reconstituted with human peripheral blood mononuclear cells (hPBMCs) (26). Here, we further established a human allogeneic acute GVHD model on these “humanized mice” and investigated the therapeutic potential of CD8hi Tregs in preventing GVHD in vivo. We demonstrate here that human CD8hi Tregs induced ex vivo by allogeneic CD40-activated B cells can ameliorate acute GVHD in an allospecific manner via reduction of alloreactive T cell proliferation and inflammatory cytokine secretion within target organs through a CTLA-4–dependent mechanism. These CD8hi Tregs can induce long-term tolerance effectively without compromising general immunity and graft-versus-tumor (GVT) activity in humanized mice. Our results support testing of ex vivo–induced human CD8hi Tregs in preventing and treating GVHD in clinical trials.

Results

A human allogeneic acute GVHD model was established in humanized mice

To mimic GVHD in humans after BMT, we established a human allogeneic acute GVHD model by injection of 1.0 × 107 allogeneic donor hPBMCs into humanized mice with stable reconstitution of recipient hPBMCs (Fig. 1A). Acute lethal GVHD was observed in humanized mice receiving allogeneic donor hPBMCs, as evidenced by weight loss, disease score (hunching, activity, ruffling, and diarrhea) (27), and death during 1 to 2 weeks after transplantation (Fig. 1B). Similar to humans (28), humanized mice with GVHD showed severe inflammation, leukocyte infiltration, fibrosis, necrosis, and tissue damage in target organs such as lung, liver, kidney, gut, and spleen (Fig. 1C). In contrast, injection of the same amount of either autologous hPBMCs or CD3+ T cell–depleted allogeneic hPBMCs into humanized mice failed to induce acute GVHD (Fig. 1, B and C). These results indicated that the acute GVHD induced by allogeneic donor hPBMCs in humanized mice was mediated mainly by donor T cells.

Fig. 1

Establishment of a human allogeneic GVHD model in humanized mice. (A) Protocol for establishment of GVHD model. Rag2−/−γc−/− mice were injected with hPBMCs. After 4 weeks, humanized mice with stable reconstitution of hPBMCs were established. Humanized mice were then irradiated sublethally and transplanted with autologous hPBMCs, allogeneic hPBMCs, CD3-depleted allogeneic hPBMCs, or PBS (phosphate-buffered saline). i.p., intraperitoneally; i.v., intravenously. (B) Survival, weight change, and disease score in humanized mice (allogeneic hPBMCs versus PBS, autologous hPBMCs, or CD3-depleted allogeneic hPBMCs, P < 0.001; PBS versus CD3-depleted allogeneic hPBMCs or autologous hPBMCs, P > 0.05; n = 8 per group). Data represent three independent experiments. (C) Representative histology of the target organs harvested on day 6 after transplantation. (D) Survival and weight change in humanized or Rag2−/−γc−/− mice after transplantation of allogeneic hPBMCs (n = 10 per group). Data represent three independent experiments. (E) Proliferation of donor hPBMCs against irradiated spleen cells of Rag2−/−γc−/− mice (mouse Ag) or recipient hPBMCs (human Ag). Data are means ± SEM and represent three independent experiments (*P < 0.05).

Different from that in humanized mice, injection of the same amount of donor hPBMCs into Rag2−/−γc−/− mice caused only minor weight loss (about 2%) without death (Fig. 1D), suggesting that no significant xenogeneic response was involved in this acute GVHD model. No proliferative response of donor hPBMCs to xenogeneic murine antigen was also observed in vitro (Fig. 1E). To further exclude potential influences of exogeneic responses, we generated A2+ and A2 humanized mice with hPBMCs from human leukocyte antigen (HLA)–A2+ and HLA-A2 donors, respectively. We then isolated A2+ human T cells from the spleen and blood of A2+ humanized mice and used them as the stimulant to induce GVHD in A2 humanized recipient mice (Fig. 2A). These A2+ allogeneic human T cells remained stably engrafted in Rag2−/−γc−/− mice for at least 4 weeks; we subsequently refer to this as an “education” process and to the engrafted T cells as educated human CD3+ (A2+ eduCD3+) T cells. Like hPBMCs, purified A2+ eduCD3+ T cells showed no proliferative responses to xenogeneic murine antigen while maintaining their responses to human alloantigens (Fig. 2B). Allogeneic A2+ eduCD3+ T cells induced a lethal acute GVHD in A2 humanized recipient mice but not in Rag2−/−γc−/− mice, just as conventional human CD3+ T cells did from the same A2+ donors (Fig. 2C). These results demonstrated that the acute GVHD model established here is mediated mainly by human CD3+ T cell–mediated allogeneic responses but not by xenogeneic responses.

Fig. 2

Human allogeneic GVHD induced by educated CD3+ T cells in humanized mice. (A) Humanized mice were reconstituted with hPBMCs from human donor HLA-A2+ (A2+) or HLA-A2 (A2). Four weeks later, human educated A2+ CD3+ T cells (hCD3+) were isolated from the peripheral blood and spleen of A2+ humanized mice. GVHD was induced in A2 humanized mice with allogeneic educated A2+ hCD3+ or conventional A2+ hCD3+ T cells from A2+ donor. (B) Proliferation of conventional human CD3+ (conCD3+) and educated CD3+ (eduCD3+) T cells from A2+ donor against irradiated spleen cells of Rag2−/−γc−/− mice (murine Ag) or A2 hPBMCs (human Ag). Data are means ± SEM and represent three independent experiments. ns, no significant difference. (C) Survival and weight change of A2 humanized mice receiving conventional A2+ hCD3+ (group 1) or educated A2+ hCD3+ (group 4) T cells, and Rag2−/−γc−/− mice receiving conventional A2+ hCD3+ (group 2) or educated A2+ hCD3+ (group 3) T cells (n = 6 per group). For survival and weight change, group 1 versus group 2 or group 3, P < 0.001; group 4 versus group 2 or group 3, P < 0.001; group 1 versus group 4, P > 0.05. Data represent three independent experiments.

CD8hi Tregs suppress the activation and proliferation of alloreactive T cells in vitro

We then generated human CD8hi Tregs according to the protocol we described previously (23). Human CD40-activated B (hCD40-B) cells were generated from A2 B cells (A2 donor) in large scale by activation and expansion through engagement of CD40 using CD40 ligand–transfected murine fibroblast cell line (NIH3T3-CD40L). Highly purified naïve CD8+CD25 T cells from A2+ donors were then cocultured with these allogeneic A2 hCD40-B cells (23). On day 9 of coculture, CD8hi Tregs with high levels of CD25, CTLA-4, and Foxp3 expression were induced and then purified by fluorescence-activated cell sorting (FACS) (Fig. 3A). As shown in Fig. 3B, the purified CD8hi Tregs potently suppressed the proliferation of autologous hPBMCs stimulated by irradiated allogeneic hPBMCs in vitro. Moreover, CD8hi Tregs significantly inhibited the secretion of inflammatory chemokines (MCP-1, IP-10, and RANTES) and cytokines [interleukin-1β (IL-1β), IL-2, IL-6, IL-17a, interferon-γ (IFN-γ), and tumor necrosis factor–α (TNF-α)] by autologous hPBMCs (Fig. 3, C and D). With intracellular cytokine staining, CD8hi Tregs were further found to suppress the expression of IFN-γ, TNF-α, and IL-2 in autologous CD4+ T cells significantly but only showed a similar suppressive effect on IFN-γ expression in autologous CD8+ T cells during initial 24 hours of coculture (fig. S1). These results indicated that CD8hi Tregs induced by hCD40-B inhibit the activation and proliferation of alloreactive T cells in vitro.

Fig. 3

CD8hi Tregs suppress the activation and proliferation of alloreactive T cells in vitro. (A) Protocol of induction and purification of CD8hi Tregs. CD8hi Tregs were induced from naïve CD8+CD25 T cells (A2+ donor) by coculture with allogeneic hCD40-B cells (A2 donor) for 9 days. The surface expression of CD8, CD25, and CTLA-4 and the intracellular expression of Foxp3 in naïve CD8+ T cells, CD8hi Tregs, and CD8mid subset were detected by flow cytometry. (B) Effect of donor CD8hi Tregs (regulator) on the proliferation of donor hPBMCs (responder) to irradiated recipient hPBMCs (stimulator). Data are means ± SEM of four replicates (*P < 0.05). (C and D) Donor CD8hi Tregs (regulator) were cocultured with donor hPBMCs (responder) and irradiated recipient hPBMCs (stimulator) for 24 hours. The concentrations of inflammatory chemokines (C) and cytokines (D) in the supernatant were measured. Data are means ± SEM and represent four independent experiments (*P < 0.05).

CD8hi Tregs inhibit acute GVHD in an allospecific manner in vivo

For evaluation of the in vivo effects and antigen specificity of these ex vivo–induced CD8hi Tregs, recipient-specific A2+ CD8hi Tregs were induced by hCD40-B cells from an A2 donor whose hPBMCs were used for establishing the humanized recipient mice. Nonrelated A2+ CD8hi Tregs were induced by A2 hCD40-B cells from a third party. Highly purified (>97% purity) 1.0 × 106 nonrelated A2+ CD8hi Tregs, recipient-specific A2+ CD8hi Tregs, or conventional A2+ CD8+ T cells were transplanted with 1.0 × 107 autologous A2+ hPBMCs into A2 humanized recipient mice (Fig. 4A). As shown in Fig. 4B, adoptive transfer of recipient-specific A2+ CD8hi Tregs not only significantly ameliorated the severity of acute GVHD in terms of weight loss and disease score but also protected mice from death during 100 days of observation. Moreover, recipient-specific A2+ CD8hi Tregs prevented leukocyte infiltration and reduced pathology in the lung, liver, gut, kidney, and spleen on day 6 after transplantation (Fig. 4C). In contrast, neither nonrelated A2+ CD8hi Tregs nor conventional A2+ CD8+ T cells had such protective effects (Fig. 4B). The antigen-specific protection of CD8hi Tregs was also confirmed in an acute GVHD model induced by allogeneic eduCD3+ T cells (Fig. 4, D and E). Collectively, these results demonstrated that CD8hi Tregs inhibit human allogeneic acute GVHD in an allo-specific manner in vivo.

Fig. 4

CD8hi Tregs inhibit human allogeneic acute GVHD in an allospecific manner in vivo. (A) Nonrelated or recipient-specific donor CD8hi Tregs (A2+) were induced by hCD40-B cells from third-party (nonrelated) donors or those whose hPBMCs (A2) were used for establishing humanized recipient mice, respectively. Nonrelated CD8hi Tregs, recipient-specific CD8hi Tregs, or conventional CD8+ T cells (convCD8) were transplanted with hPBMCs from same A2+ human donors into A2 humanized recipient mice. (B) Survival, weight change, and disease score. (C) Representative histology and histology score of target organs (lung, liver, kidney, gut, and spleen) on day 6 after transplantation in A2 humanized recipient mice. For survival, weight change, and disease score: hPBMCs + CD8hi Tregs (recipient-specific) versus hPBMCs + PBS, hPBMCs + CD8hi Tregs (nonrelated), or hPBMCs + convCD8, P < 0.001. hPBMCs + PBS (n = 7); hPBMCs + CD8hi Tregs (recipient-specific) (n = 7); hPBMCs + CD8hi Tregs (nonrelated) (n = 5); hPBMCs + convCD8 (n = 4); untreated (n = 4). *P < 0.05. (D) Protocol for inhibiting educated CD3+ T cell–mediated human allogeneic GVHD by CD8hi Tregs. (E) Survival, weight change, and disease score in A2 humanized recipient mice receiving educated A2+ hCD3+ T cells (eduCD3) alone or educated A2+ hCD3+ T cells and recipient-specific or nonrelated CD8hi Tregs (n = 6 per group). For survival, weight change, and disease score, eduCD3 + Tregs (recipient-specific) versus eduCD3 or eduCD3 + Tregs (nonrelated), P < 0.0001. Data represent three independent experiments.

CD8hi Tregs inhibit alloreactive T cell proliferation and inflammatory chemokine/cytokine secretion

To investigate the mechanisms underlying the prevention of GVHD by CD8hi Tregs, we injected allogeneic donor hPBMCs (A2+) labeled with a lipophilic dye (Dil) into A2 humanized recipient mice with A2+ recipient-specific donor CD8hi Tregs that were distinguished with another lipophilic dye (Dir). Organ imaging ex vivo detected hPBMCs and CD8hi Tregs in the lung and liver only, where their distribution overlapped (Fig. 5A). After transplantation, the accumulation of donor hPBMCs in these two organs gradually increased to peak levels on day 6 and then decreased on day 9 (Fig. 5A and fig. S2A). The treatment of CD8hi Tregs significantly reduced the accumulation of allogeneic hPBMCs in target organs from day 1 to day 9 after transplantation (Fig. 5B and fig. S2A). By injecting carboxyfluorescein succinimidyl ester (CFSE)–labeled allogeneic donor hPBMCs (A2+) into A2 humanized recipient mice, we further found that CD8hi Treg treatment significantly inhibited the proliferation of donor hCD3+ T cells (A2+) in these target organs in vivo from day 3 to day 9 after transplantation (Fig. 5C and fig. S2B). Furthermore, CD8hi Treg treatment significantly inhibited the secretion of human inflammatory chemokines and cytokines, such as RANTES (CCL5), IP-10 (CXCL10), MCP-1 (CCL1), IL-1β, IL-2, IL-6, IL-17A, IFN-γ, and/or TNF-α, from the lung, liver, or gut on day 6 after transplantation (Fig. 5, D and E), whereas the amounts of IL-10 and TGF-β (transforming growth factor–β) in target organs were actually lower from day 3 to day 9 day after CD8hi Treg treatment in humanized mice compared with mice receiving hPBMCs only (fig. S3). Specifically, the percentages of IFN-γ–, TNF-α–, or IL-2–secreting CD4 and CD8 T cells in the lung and liver were significantly decreased in the humanized recipient mice after treatment of CD8hi Tregs (fig. S4). These results demonstrated that CD8hi Tregs ameliorate human allogeneic acute GVHD by reduction of alloreactive T cell proliferation and inflammatory chemokine and cytokine secretion in target organs.

Fig. 5

Recipient-specific donor CD8hi Tregs suppress alloreactive T cell proliferation and inflammatory chemokine and cytokine secretion in target organs after donor hPBMC transplantation. (A and B) Distribution of hPBMCs (Dil-labeled, green) and CD8hi Tregs (Dir-labeled, red) (A) and accumulation of donor hPBMCs (shown as intensity of Dil signal) (B) in target organs on day 6 after transplantation in humanized GVHD mice with or without recipient-specific donor CD8hi Treg treatment. Data are representative of four independent experiments. (C) Proliferation of donor CD3+ T cells determined by CFSE staining (original histogram represents the expression level of CFSE in donor CD3+ T cells before injection into recipient) in target organs on day 6 after transplantation in humanized GVHD mice with or without recipient-specific donor CD8hi Treg treatment. Data represent four independent experiments. (D and E) Concentration of inflammatory chemokines (D) and cytokines (E) in lungs, livers, and guts on day 6 after transplantation in humanized GVHD mice with or without recipient-specific donor CD8hi Treg treatment. Data are means ± SEM and represent four independent experiments (*P < 0.05).

The prevention of acute GVHD by CD8hi Tregs is dependent on CTLA-4

Because CD8hi Tregs express high levels of CTLA-4 (Fig. 3A), we further determined whether the control of acute GVHD is mediated by CTLA-4. With pretreatment of CTLA-4–neutralizing antibody, the expression of CTLA-4 on CD8hi Tregs was completely blocked (Fig. 6A). The suppression of the allogeneic proliferative response by CD8hi Tregs was also significantly reversed after pretreatment with CTLA-4–neutralizing antibody in vitro (Fig. 6B). Blockade of CTLA-4 expression on CD8hi Tregs abolished their protection from acute GVHD in humanized mice in terms of mice survival, weight loss, and disease score (Fig. 6C). Moreover, blockade of CTLA-4 on CD8hi Tregs significantly reversed their suppression on the production of human IL-2 and TNF-α and the accumulation of hCD3+ T cells in the lung, liver, or gut, but did not affect the distribution of CD8hi Tregs in these target organs on day 6 after transplantation (Fig. 6, D to F). These results indicated that the prevention of acute GVHD by CD8hi Tregs in humanized mice is mediated mainly by CTLA-4.

Fig. 6

Prevention of acute GVHD by recipient-specific donor CD8hi Tregs depends on CTLA-4. (A) Expression level of CTLA-4 on CD8hi Tregs pretreated with CTLA-4–neutralizing monoclonal antibody (mAb) or isotype control (mIgG1). (B) Effect of CTLA-4 blockade on the suppression by CD8hi Tregs on allogeneic proliferative responses. Untreated CD8hi Tregs (UT), CD8hi Tregs pretreated with CTLA-4–neutralizing mAb (αCTLA-4), and its isotype control (IC, mIgG1) were added to coculture of donor hPBMCs [responder (R)] and irradiated recipient hPBMCs [stimulator (S)]. (C) CD8hi Tregs pretreated with CTLA-4–neutralizing mAb (αCTLA-4) or mIgG1 were transplanted with allogeneic hPBMCs into humanized mice. Survival, weight change, and disease score in humanized mice are shown (n = 6 per group). For survival, weight change, and disease score: hPBMCs + Tregs or hPBMCs + mIgG1-treated Tregs versus hPBMCs or hPBMCs + αCTLA-4–treated Tregs, P < 0.05; hPBMCs versus hPBMCs + αCTLA-4–treated Tregs, P > 0.05. Data represent two independent experiments. (D) Concentration of IL-2 and TNF-α in target organs of humanized mice accepting different treatments on day 6 after transplantation. Data are means ± SEM and represent four independent experiments (*P < 0.05). (E and F) Accumulation of donor CD3+ T cells and CD8hi Tregs (labeled with Dir) in target organs of humanized mice accepting different treatments on day 6 after transplantation. Data are means ± SEM and represent four independent experiments (*P < 0.05; ns, no significant difference).

CD8hi Tregs induce long-term tolerance and preserve general immunity

To monitor chimerism of donor- and recipient-original cells, we injected hPBMCs and CD8hi Tregs from A2+ donor into humanized recipient mice reconstituted with A2 hPBMCs. As shown in Fig. 7A, a mix chimerism was established in A2 humanized recipient mice within 9 days after transplantation of hPBMCs and CD8hi Tregs. On day 100, most reconstituted lymphoid cells in A2 humanized recipient mice were originated from A2+ donor in the peripheral blood, spleen, lung, liver, and gut after treatment with CD8hi Tregs (Fig. 7A and fig. S5). We then further examined the general immune function in A2 humanized recipient mice after 14 days of transplantation with A2+ hPBMCs and CD8hi Tregs by immunization with tetanus toxoid (TT) vaccine (Fig. 7B). As shown in Fig. 7C, the vaccination induced TT-specific IFN-γ secretion by human CD4+ and CD8+ T cells and serum TT-specific human antibody in CD8hi Treg-treated humanized recipient mice (group II) (Fig. 7C). The T cell responses and antibody production in these mice were comparable with those in humanized mice reconstituted with hPBMCs from the same donor (group I) (Fig. 7C). Moreover, donor-origin CD3+ T cells isolated from CD8hi Treg-treated humanized recipient mice on day 100 did not respond to recipient antigen but had a robust proliferative response to nonrelated antigens from a third party (Fig. 7D), indicating that the alloantigen-specific tolerance can be maintained up to 100 days after a single dose of CD8hi Treg treatment. These results demonstrated that CD8hi Tregs can induce long-term tolerance and retain general immune function to foreign antigens in humanized recipient mice.

Fig. 7

Recipient-specific donor CD8hi Tregs induce long-term tolerance and preserve general immunity in humanized mice. (A) Chimerism in humanized recipient mice (A2) before and after transplantation of allogeneic hPBMCs (A2+) and CD8hi Tregs (A2+). Percentages of donor (A2+) and recipient (A2) original cells in human CD45+ cells in peripheral blood and spleen from humanized GVHD mice at the indicated time after CD8hi Treg treatment, shown as means ± SEM (n = 9). (B) Protocol for evaluation of general immunity in humanized GVHD mice after CD8hi Treg treatment. (C) Percentages of IFN-γ–producing cells in donor CD4+ and CD8+ T cells (A2+) from peripheral blood and levels of serum TT-specific antibodies (Ab) after a booster vaccination of TT (n = 4 per group). (D) Long-term alloantigen-specific tolerance of donor-origin CD3+ T cells. Donor-origin CD3+ T cells (A2+) were isolated from humanized mice transplanted with hPBMCs and CD8hi Tregs on day 100 after transplantation and stimulated with irradiated recipient hPBMCs (specific) or hPBMCs from a third party (nonrelated). The proliferative responses of donor-origin T cells to specific or nonrelated hPBMCs are shown. Data are means ± SEM and represent four independent experiments (*P < 0.05; ns, no significant difference).

To further determine whether the treatment of CD8hi Tregs affects GVT activity, we injected humanized mice intravenously with 1.0 × 105 green fluorescent protein (GFP)–expressing Epstein-Barr virus (EBV)–transformed autologous lymphoblastoid cell line (LCL) cells 4 days before transplantation of 1.0 × 107 donor hPBMCs with or without 1.0 × 106 recipient-specific donor CD8hi Tregs or recipient-specific CD8+ T cells (Fig. 8A). All mice that received donor hPBMCs alone, or donor hPBMCs and recipient-specific CD8+ T cells died, whereas steady LCL cell growth was seen in the mice treated with PBS. In contrast, about 90% of humanized mice that received donor hPBMCs and recipient-specific donor CD8hi Tregs survived, with neither detectable tumor cells in peripheral blood (Fig. 8B) nor visible solid tumor during the observation of 100 days after transplantation. These results demonstrated that GVT response is preserved after the long-term tolerance induced by CD8hi Tregs. CD8hi Tregs exhibited direct cytotoxic activity against LCL cells both in vitro and in vivo, whereas CD4+CD25+ Tregs expanded by anti-CD3/CD28 antibodies had no such cytotoxicity against LCL cells (Fig. 8, C and D). The cytotoxicities of CD8hi Tregs were mediated by Fas-FasL (Fas ligand) and perforin–granzyme B pathways because either blockade of FasL, inhibition of perforin, or inactivation of granzyme B could significantly abrogate their cytotoxicities (fig. S6).

Fig. 8

GVT activity in humanized GVHD mice after CD8hi Treg treatment. (A) Protocol for evaluating GVT activity in humanized GVHD mice. (B) Survival, weight change, and tumor recurrence of humanized mice receiving PBS (n = 15), donor hPBMCs (n = 10), donor hPBMCs with donor CD8hi Tregs (n = 15), or recipient-specific CD8+ T cells (n = 10). hPBMCs + Tregs versus hPBMC, hPBMCs + recipient-specific CD8, or PBS, P < 0.001. Data represent two independent experiments. (C) Cytotoxicity of CD8hi Tregs and CD4+CD25+ Tregs against allogeneic LCL in vitro. CD8hi Tregs and anti–CD3/CD28-expanded CD4+CD25+ Tregs [effector cells (E)] from same HLA-A2+ donors were cocultured with LCL [target cells (T)] generated from HLA-A2 donors, whose hCD40-B cells were used for generating CD8hi Tregs, at the indicated E:T ratios, and death of target cells was determined by propidium iodide (PI) staining. Data are means ± SEM of four independent experiments. *P < 0.05, **P < 0.01. (D) Cytotoxicity of CD8hi Tregs and CD4+CD25+ Tregs against allogeneic LCL in vivo. GFP-labeled A2 LCL cells (1.0 × 105) were cotransplanted with 1.0 × 106 A2+ CD8hi Tregs or anti–CD3/CD28-expanded CD4+CD25+ Tregs into Rag2−/−γc−/− mice. Tumor recurrence of humanized mice is shown as mean ± SEM and represents four independent experiments. LCL + CD8hi Tregs versus LCL + CD4+CD25+ Tregs or LCL alone, P < 0.05; LCL alone versus LCL + CD4+CD25+ Tregs, P > 0.05.

Discussion

One of the major obstacles for Treg-based therapy is the lack of reliable models to test the therapeutic effects of human Tregs in vivo. More recently, the in vivo function of the ex vivo–induced human CD4+ Tregs has been evaluated in immunodeficient mice (24, 25). However, the relevance of the results to human disease remains unclear because the immunodeficient mice do not contain a stable human immune system before the adoptive transfer of human Tregs. Here, by adoptive transfer of allogeneic hPBMCs into humanized mice with a stable reconstitution of human immune system (26), we establish a human allogeneic acute GVHD model in humanized mice. Similar to humans (28, 29), the acute GVHD is mediated mainly by donor CD3+ T cells and characterized by disease appearance (hunching, activity, ruffling, and diarrhea), recruitment of alloreactive cells in target organs, and dysregulation of proinflammatory chemokines and cytokines. Using mouse-educated human CD3+ T cells, we further demonstrated that the acute GVHD is mediated mainly by human allogeneic responses but not by xenogeneic response. The human allogeneic GVHD model established here may provide a more relevant approach for studies of human immunopathogenesis and therapeutics for GVHD after BMT.

Different from other antigen-specific CD8+ Tregs, which are difficult to be expanded (1419), human CD8hi Tregs induced ex vivo by allogeneic hCD40-B cells have highly secondary proliferative capacity and, therefore, they are easy to be expanded in large scale (23). It is unnecessary to add any exogenous cytokines for inducing and expanding CD8hi Tregs because hCD40-B cells can secrete substantial amounts of IL-2 (20). This lack of requirement for exogenous cytokines could significantly reduce the cost for the generation of human CD8hi Tregs. In addition, these CD8hi Tregs not only express high levels of Foxp3 and CTLA-4 but also have higher levels of CD8 and CD25 expression on their surface compared to their precursors, thus making it easy to purify them from coculture.

The importance of CD8+ Tregs in the induction of tolerance during transplantation has been confirmed recently in rodents. It has been shown that murine CD8+Foxp3+ Tregs were induced during GVHD after allogeneic BMT, and the induction of these Tregs was correlated positively with the protection of GVHD in mice (30, 31). In a heart transplant model, the accumulation of rat CD8+ Tregs in allograft was found to be associated with tolerance induction in allograft recipients (32). By adoptive transfer, the ex vivo–induced CD8+Foxp3+ Tregs prevented the skin allograft rejection in mice (13). Here, using human CD8hi Tregs induced by allogeneic hCD40-B cells ex vivo, we found that they can suppress the proliferation and inflammatory cytokine and chemokine secretion in alloreactive T cells in vitro. In the human allogeneic acute GVHD model, we further demonstrated that these ex vivo–induced human CD8hi Tregs can effectively control acute GVHD in an allospecific manner by reduction of alloreactive T cell proliferation and inflammatory cytokine and chemokine secretion in target organs.

The major challenge of allogeneic BMT is to maintain long-term tolerance to allograft without compromising both general immunity and GVT activity. Our results showed that a rapid immune reconstitution and a high donor chimerism were achieved in humanized mice after treatment with CD8hi Tregs. On day 100 after transplantation, more than 80% of reconstituted human cells in the peripheral blood, spleen, lung, liver, and gut were originated from donor cells. In addition, the alloantigen-specific tolerance can be maintained up to 100 days after CD8hi Treg treatment. Together, these results demonstrated that CD8hi Tregs induced a stable tolerance rather than simply eliminate responder cells. Furthermore, our data showed that CD8hi Treg treatment did not suppress the general immune function of cotransplanted conventional hPBMCs against foreign antigen as evidenced by normal antigen-specific CD4+ and CD8+ T cell responses and antibody production. Using a tumor-loaded humanized mouse model, we also demonstrated that the GVT activity is preserved after the long-term tolerance induced by CD8hi Tregs. Therefore, our study provided proof of concept of using ex vivo–induced human CD8hi Tregs to control GVHD while preserving both general immunity and GVT activity after BMT. We found that CD8hi Tregs have direct cytotoxic activity against tumor cells, whereas conventional CD4+CD25+ Tregs do not have such activity, suggesting that CD8hi Tregs may provide more advantage than CD4+CD25+ Tregs to control GVHD and avoid tumor relapse (33).

Although some in vitro studies suggest that IL-10, TGF-β, or CTLA-4 may be involved in the suppression of CD8+ Tregs (16, 17), it remains unknown whether these molecules participate in the suppression mediated by human CD8+ Tregs in vivo. Here, we found that the amounts of IL-10 and TGF-β in target organs in humanized GVHD mice decreased after CD8hi Treg treatment, suggesting that IL-10 and TGF-β are not involved in the suppression in vivo. The indispensable role of CTLA-4 in the suppression of murine CD4+Foxp3+ Tregs has been demonstrated in vitro and in vivo (7, 34, 35). By blocking CTLA-4 expression on CD8hi Tregs, here we demonstrated that the suppression of allogeneic proliferative response by human CD8hi Tregs in vitro and the prevention of acute GVHD by human CD8hi Tregs in humanized mice are mediated mainly by CTLA-4. Consistent with that in murine CD4+ Tregs (35), here we found that blockade of CTLA-4 on human CD8hi Tregs significantly increased the secretion of human IL-2 and TNF-α and the accumulation of human CD3+ T cells in the lung, liver, or gut, but did not affect the distribution of CD8hi Tregs in these target organs during the progress of acute GVHD in humanized mice. In support of these findings, other studies also showed that high level of IL-2 might favor the exacerbation of T cell–mediated inflammation rather than the survival of Tregs under proinflammatory conditions (36).

This study had some limitations. Similar to CD8+ Tregs reported by other groups (37, 38), human CD8hi Tregs also have alloantigen-specific cytotoxicity at a high ratio of Tregs to target cells in vitro (23). Here, we also found that the blockade of CTLA-4 could not completely abolish the CD8hi Treg-mediated protection from acute GVHD. Therefore, we cannot exclude the possibility that the cytotoxicity of CD8hi Tregs may partially contribute to preventing acute GVHD in humanized mice. Because human nonhematopoietic cells also express major histocompatibility complex molecules, the profile of target cells in human GVHD should be broader than that in our model. To determine whether CD8hi Tregs could also induce tolerance on nonhematopoietic cells, the solid organ transplantation models established in humanized mice could help evaluate the efficacy of CD8hi Treg-based therapy. In addition, although we demonstrated that the acute GVHD model established in this study is mediated mainly by human CD3+ T cell–mediated allogeneic responses, we cannot completely exclude the involvement of xenogeneic responses in this GVHD model.

In summary, using humanized mice with a complete human immune system, we successfully established a human allogeneic acute GVHD model. Using this model, we demonstrated that human CD8hi Tregs induced ex vivo by allogeneic hCD40-B cells can control acute GVHD in an allospecific manner via reduction of alloreactive T cell proliferation and inflammatory cytokine secretion within target organs through a CTLA-4–dependent mechanism. These CD8hi Tregs not only can induce long-term tolerance effectively without compromising general immunity and GVT activity but also have potent antitumor activity. Therefore, our study provided proof of concept of using ex vivo–induced human CD8hi Tregs to control GVHD after BMT. This strategy could readily be extended to human clinical trials using human CD8hi Tregs alone or in combination with minimal conventional immunosuppression to control GVHD. The GVHD model established here may also provide a more relevant platform for further studies of human immunopathogenesis and therapeutics for GVHD after BMT.

Materials and Methods

Animals

C57BL/10SgAiRag2−/−γc−/− (Rag2−/−γc−/−) mice were purchased from Taconic and maintained in the Laboratory Animal Unit of the University of Hong Kong. All manipulations were performed in compliance with the guidelines for the use of experimental animals by the Committee on the Use of Live Animals in Teaching and Research, Hong Kong.

Cell isolation and preparation

hPBMCs were isolated from the buffy coats of healthy donors from Hong Kong Red Cross by Ficoll-Hypaque (Pharmacia) gradient centrifugation as described before (22). The research protocol was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster. Human CD8hi Tregs were generated as described previously (23). In brief, hCD40-B cells were induced from hPBMCs by NIH3T3-CD40L cells, whereas naïve CD8+CD25CD45RA+CD45RO T cells were isolated from hPBMCs with a naïve CD8+ T cell isolation kit (Miltenyi Biotec). Naïve CD8+ T cells were cocultured with allogeneic CD40-activated B cells at a T cell/B cell ratio of 10:1. After 9 days of incubation, CD8+ T cells expressing high levels of CD8 and CD25 were isolated by FACS (BD). The CD3+ T cell–depleted hPBMCs were prepared with anti-CD3 microbeads (Miltenyi Biotec).

GVHD model

Human allogeneic GVHD models were established in humanized mice prepared with a protocol similar to that described in our previous study (26). In brief, 4- to 5-week-old Rag2−/−γc−/− mice pretreated with liposome-clodronate (VU Medisch Centrum) were sublethally irradiated (1 Gy/6 g) and transplanted intraperitoneally with 3.0 × 107 hPBMCs (26). After 4 weeks, these humanized mice were treated as recipients and injected intravenously with 1.0 × 107 autologous hPBMCs, 1.0 × 107 allogeneic hPBMCs with or without 1.0 × 106 CD8hi Tregs, or 1.0 × 107 CD3+ T cell–depleted allogeneic hPBMCs 1 day after sublethal irradiation (1 Gy/6g). In some experiments, hPBMCs and CD8hi Tregs from HLA-A2+ donor were injected into humanized recipient mice reconstituted with HLA-A2 hPBMCs. For induction of GVHD by eduCD3+ T cells, humanized mice were reconstituted with hPBMCs from human donor A2+ or A2. Four weeks later, humanized mice reconstituted with hPBMCs from donor A2+ were sacrificed, and human eduCD3+ cells were isolated from their peripheral blood and spleen. GVHD was induced in humanized mice reconstituted with donor A2 hPBMCs by 1.0 × 106 purified allogeneic eduCD3+ or conventional human CD3+ T cells from donor A2+. GVHD disease was scored with weight change, posture, activity, fur texture, skin integrity, and diarrhea as described by others (27).

Tumor model

For tumor model in humanized mice, LCL cell lines were established by infecting HLA-A2 hPBMCs with GFP-expressing EBV and purified by FACS. Humanized mice reconstituted with HLA-A2 hPBMCs were injected intravenously with 1.0 × 105 GFP-LCL cell lines established from the same donors 4 days before lethal irradiation. These LCL-injected mice were then transplanted with 1.0 × 107 allogeneic HLA-A2+ hPBMCs. For tumor model in Rag2−/−γc−/− mice, 1.0 × 105 GFP-labeled A2 LCL cells were intravenously injected alone or cotransplanted with 1.0 × 106 A2+ CD8hi Tregs or CD4+CD25+ Tregs into Rag2−/−γc−/− mice. Tumor reoccurrence was assessed as the percentage of GFP-LCL in their peripheral blood.

Vaccination protocol

Humanized mice reconstituted with HLA-A2 hPBMCs were transplanted with 1.0 × 107 hPBMCs and 1.0 × 106 CD8hi Tregs from HLA-A2+ donors. At 14 days after transplantation, mice were primed with 1.5 limits of flocculation (lf) of TT vaccine (Adventis-Pasteur) subcutaneously in the inguinal pouch region. A booster of 0.25 lf of TT was injected in the right hind footpad 10 days later. On day 30 after transplantation, IFN-γ–producing CD4+ and CD8+ T cells in the peripheral blood of these vaccinated mice were counted with FACSAria II (BD), and the concentration of total TT-specific immunoglobulin G (IgG) in the serum of humanized mice was determined with a commercial ELISA kit (Bethyl Laboratories) as we described before (26).

In vivo imaging

hPBMCs and CD8hi Tregs were stained with Dil and Dir (Invitrogen), respectively. Dil-labeled hPBMCs were injected into recipient humanized mice with Dir-labeled or unlabeled CD8hi Tregs. The migration and accumulation of hPBMCs and/or CD8hi Tregs were visualized and analyzed with a TM 2 in vivo imaging system (CRI Maestro).

Histology

Lungs, livers, spleens, kidneys, and guts from humanized mice were harvested at the indicated times. Sections were prepared according to standard protocols and stained with hematoxylin and eosin. The histopathology score was calculated on the basis of inflammation and cell infiltration in lung, liver, kidney, and gut (each organ ranked 0 to 5) and analyzed by two independent experienced pathologists who were blinded to the treatment.

Mixed lymphocyte reaction assay

The mixed lymphocyte reaction system was established as follows: Mouse spleen cells or hPBMCs (recipient) were irradiated and used as stimulator cells, whereas allogeneic hPBMCs (donor) were used as responder cells. Responder cells were cocultured with stimulator cells at a 1:1 cell ratio with or without regulator (CD8hi Tregs) for 5 days, and [3H]thymidine (Perkin Elmer) was added to the culture at a concentration of 5.0 μCi/ml for the last 16 hours of incubation. Proliferation of the responder cells was analyzed by [3H]thymidine incorporation as we described before (20).

Blocking assay

To block the effects of cytokines and granules, we used the following reagents and antibodies: anti-human IFN-γ (2 μg/ml, goat IgG), anti-human TNF-α (2 μg/ml, 28401, mouse IgG1), anti-human FasL (10 μg/ml, 100419, mouse IgG2b), Bcl-2 (granzyme inhibitor, 2 μg/ml) (R&D Systems), anti-human CTLA-4 (10 μg/ml, ANC152.2, mouse IgG1κ) (Ancell), and concanamycin A (CMA) (perforin inhibitor, 10 μg/ml) (Sigma-Aldrich). For blocking, antibodies against cytokines were directly added into the culture at the indicated final concentrations, whereas effector cells were incubated with Bcl-2 and CMA 1 hour before coculturing with target cells to exclude the effects of preserved granule as we described before (39).

In vivo proliferation assays

HLA-A2+ allogeneic hPBMCs (5.0 × 106 cells/ml) were stained with 22 μl of 0.05 mM CFSE (Sigma) at 37°C for 5 min, washed with PBS three times, and injected into humanized recipient mice reconstituted with HLA-A2 hPBMCs. On days 3, 6, and 9 after transplantation, the lung and liver of recipient mice were harvested and prepared as a single-cell suspension. The levels of CFSE in HLA-A2+ CD3+ T cells were analyzed with FACSAria II and FlowJo software (Tree Star).

Cytotoxicity assays

For determining the cytotoxicity of CD8hi Tregs and CD4+CD25+ Tregs [effector (E)] against GFP-LCL [target (T)], effector cells and target cells were cocultured at different E:T ratios in 37°C for 4 hours with the addition of PI at the final 15 min. The apoptosis of target cells was analyzed with FACSAria II by back gating on GFP- and PI-positive cells.

Flow cytometric analysis

Cells were stained for surface markers with the following mAbs: anti-human CD3-FITC (fluorescein isothiocyanate) (HIT3a), anti-human CD19-APC (allophycocyanin) (HIB19), anti-human CD25-APC (2A3), and anti–HLA-A2–FITC (BB7.2) (BD Biosciences); anti-human CD4–Alexa-405 (S3.5), anti-human CD8–phycoerythrin-Cy7 (3B5), and anti-human CD45-APC (HI30) (Invitrogen); and anti-human IFN-γ–FITC (4S.B3) (R&D Systems). All samples were acquired on FACSAria II with FlowJo software as described previously (20).

FlowCytomix assay

For the detection of cytokines and chemokines, the lungs, livers, and guts from recipient humanized mice were harvested at the indicated times and homogenized in PBS. The concentrations of human proinflammatory cytokines and chemokines in these samples were detected and analyzed with human cytokine and chemokine assay kits (Bender MedSystems) as we described before (26).

Statistical analysis

Data are means ± SEM. Multiple regression analysis was used to test the differences in the body weight changes between groups adjusted for time after transplantation. The differences in cell percentage and concentrations of proinflammatory cytokines/chemokines among groups were analyzed by unpaired, two-tailed Student’s t test. The significance of differences in survival was determined by the Kaplan-Meier log-rank test. P < 0.05 was considered to be significant.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/5/168/168ra9/DC1

Fig. S1. Effects of CD8hi Tregs on inflammatory cytokine production by CD4+ and CD8+ T cells from donor PBMCs in vitro.

Fig. S2. Recipient-specific donor CD8hi Tregs suppress the acute proliferation of donor T cells in target organs after donor hPBMC transplantation.

Fig. S3. Concentration of human TGF-β and IL-10 does not increase in the target organs of CD8hi Treg-treated humanized GVHD mice.

Fig. S4. Effects of CD8hi Tregs on inflammatory cytokine production by CD4+ and CD8+ T cells from donor PBMCs in vivo.

Fig. S5. Chimerism in humanized recipient mice (A2) before and after transplantation of allogeneic hPBMCs (A2+) and CD8hi Tregs (A2+).

Fig. S6. Molecular mechanisms of CD8hi Treg-mediated cytotoxicity against allogeneic LCL.

References and Notes

  1. Funding: This work was supported in part by National Natural Science Foundation of China/Research Grants Council Joint Research Scheme (N_HKU 747/11), General Research Fund, Research Grants Council of Hong Kong (HKU 781211M), Area of Excellence program (AoE/M-12/06), and University Grants Committee of Hong Kong Special Administrative Region. Author contributions: W.T. initiated the project and wrote the manuscript; J.Z. and Yinping Liu designed the experiments and wrote the manuscript; J.Z., Yinping Liu, and Yuan Liu performed the experiments; M.L. and K.-T.L. accomplished the histology experiments; Z.X. helped accomplish some of the animal experiments; D.B.L. and Y.-L.L. monitored the experiments and revised the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article