Owing to the adverse effects of immunosuppression and an inability to prevent chronic rejection, there is a pressing need for alternative strategies to control alloimmunity. In three decades, regulatory T cells (Tregs) have evolved from a hypothetical mediator of adoptively transferred tolerance to a well-defined population that can be expanded ex vivo and returned safely to patients in clinical trials. Herein, we review the historical developments that have permitted these advances and the current status of clinical trials examining Tregs as a cellular therapy in transplantation. We conclude by discussing the critical unanswered questions that face this field in the coming years.
cytotoxic T lymphocyte antigen 4
fluorescence-activated cell sorting
glucocorticoid-induced tumor necrosis factor receptor
good manufacturing practice
hematopoietic stem cell
hematopoietic stem cell transplantation
induced regulatory T cell
naturally occurring regulatory T cell
peripheral blood mononuclear cells
solid organ transplantation
transforming growth factor β
regulatory T cell
Treg-specific demethylated region
umbilical cord blood
Solid organ transplantation (SOT) and hematopoietic stem cell transplantation (HSCT) are two of the most successful medical innovations of the twentieth century. Nevertheless, in the twenty-first century, we are still faced with hindrances to their success. These issues include the complications of long-term nonspecific immunosuppression and, despite a wide array of immune modulating drugs, an inability to prevent chronic allograft dysfunction.
Consequently, attention has turned to efforts aimed at achieving allograft tolerance while avoiding the need for ongoing immunosuppression. The use of CD4+CD25+FoxP3+ regulatory T cells (Tregs) as a cellular therapy given at or near the time of transplantation is one approach that may address these concerns. Data from animal models suggest that Tregs can exert dominant tolerance to alloantigens in vivo by inducing regulatory properties in alloreactive T cells, and may establish a tolerant state that could obviate or minimize the need for immunosuppressive drugs . Further, Tregs may prevent chronic allograft pathology such as transplant vasculopathy [2-4].
In this review, we discuss animal data that have guided the development of clinical studies. We then consider methods of expanding human Tregs, followed by a discussion of anticipated problems, unanswered questions and outstanding concerns that will need to be addressed in future work.
Historical Perspective: Origins of the Treg Concept
Pioneering work by Billingham, Brent and Medawar in the 1950s hinted at the involvement of mechanisms in addition to deletion of donor-reactive immune cells in allograft acceptance . The existence of suppressor cells was demonstrated in 1971 by Gershon and Kondo , and later, Kilshaw et al  identified a role for suppressor cells in transplant tolerance. Yet, there remained much skepticism surrounding the existence of Tregs until Hall et al showed that CD4+ , and specifically CD4+CD25+  T cells, could mediate transplant tolerance. Contemporaneously, it was observed that allogeneic blood transfusion in the rat generated CD4+ T cells capable of inhibiting anti-donor but not third-party responses . Waldmann's group then demonstrated that adoptively transferred tolerance was “infectious,” in that CD4+ T cells from tolerant animals could recruit naïve nonregulatory cells into the regulatory pool .
Studies showing that tolerance could be induced by treatment of animals with anti-CD4 antibodies with or without donor cells [11-13] led to a subsequent finding that induction and/or activation of regulatory CD4+ T cells were likely responsible , with the demonstration that naïve CD4+CD25+ T cells, but not CD4+CD25− T cells, could inhibit allogeneic skin graft rejection and autoimmunity. Sakaguchi et al  showed that a specific subset of CD4+ T cells could transfer tolerance. Later, the discovery that expression of FoxP3 defines thymus-derived, naturally occurring (nTregs) in vivo [16-18] as well as transforming growth factor β (TGFβ)-induced Tregs (iTregs) in vitro , it became possible to confirm that CD4+CD25+FoxP3+ cells could control autoimmunity , inhibit graft-versus-host disease (GVHD) [20-22], and prevent or delay allograft rejection [15, 23] in animal models.
In parallel, human Tregs have been characterized [24, 25], and in the last several years, Phase I and II clinical trials of Treg therapy in GVHD [26-28] and Type I diabetes  have been reported. More recently, the ONE Study, a multi-center trial assessing Tregs and other regulatory cells in SOT, has begun . There are many other regulatory cell populations (reviewed in []); however, this review will focus on CD4+FoxP3+ Tregs. A timeline of the key events in the history of Tregs in transplantation is shown in Figure 1.
Insights From Mouse Models of Transplantation
Mechanisms employed by Tregs in controlling alloimmune responses
FoxP3+ Tregs can utilize multiple mechanisms to inhibit effector T cells. Broadly these include (1) modulation of antigen-presenting cell (APC) function; (2) metabolic disruption (IL-2 deprivation, adenosine secretion); (3) direct cytotoxicity toward effector T cells; and (4) secretion of inhibitory cytokines such as IL-10, IL-35 and TGFβ (reviewed in []). Suppression of T cell responses by Tregs is primarily cell contact dependent . Which mechanisms are most important under homeostatic and inflammatory conditions in vivo is still a matter of debate, but some are indispensable for allograft tolerance in the mouse (Figure 2).
Tregs suppress colitis and maintain immune homeostasis in mice partly via IL-10 [34, 35] and cytotoxic T lymphocyte antigen 4 (CTLA-4) [36-38]. Therefore, it is not surprising that CD4+CD25+ T cells from mice rendered tolerant to cardiac allografts can transfer tolerance to naïve mice in an IL-10-dependent fashion [23, 39]. Similarly, CTLA-4 blockade abrogates graft acceptance mediated by adoptively transferred Tregs . In keeping with observations in other settings [40-42], adenosine generation by CD39 and CD73 can contribute to Treg-mediated allograft acceptance . The glucocorticoid-induced tumor necrosis factor receptor (GITR) has well-established roles in Treg function [44, 45]. GITR ligation with antibody abrogates Treg-mediated allograft tolerance , while GITR blockade has the opposite effect . Induction of indoleamine 2,3-dioxygenase and other enzymes in APCs hinders T cell proliferation and is a mechanism for Treg control of graft rejection in vivo [48, 49]. Thus, several of the major Treg pathways involved in the control of autoimmunity are also relevant to Treg-mediated suppression of the allograft response.
Observations that interferon gamma (IFNγ) plays an important role in the control of skin graft rejection and GVHD by Tregs in mouse models [50, 51] were more unexpected. Rapid, transient IFNγ production by alloantigen-reactive Tregs allows them to inhibit skin allograft rejection in vivo . Furthermore, Tregs use IFNγ in an autocrine manner, promoting their function . These findings may partly explain the observation that exogenous IFNγ can inhibit GVHD , but IFNγ has other immunoregulatory effects . For example, Treg-secreted IFNγ might also promote effector T cell apoptosis [55, 56], but this has not yet been described as a Treg mechanism.
Another intriguing finding was that IL-9 secretion by Tregs activates mast cells in mice rendered tolerant with donor-specific transfusion and anti-CD154 antibody ; further, mast cell–deficient mice could not be rendered tolerant. These data should remind us that the effectiveness of Treg therapy may be critically dependent upon additional, previously unsuspected in vivo determinants.
Data from in vitro assays suggest that human Tregs can also employ IL-10, CTLA-4, TGFβ, IL-35 and adenosine to control effector T cells [58-63]. The extent to which these mechanisms operate in vivo is obviously more difficult to determine in man than in the mouse but given the similarity of the in vitro data, it seems likely that human and mouse Tregs use common functional pathways in vivo.
Nevertheless, since memory T cells are a component of the human alloimmune response  and are less susceptible to control by Tregs [65, 66], it is conceivable that the relevant Treg mechanisms in transplant patients will be quite different than those discussed here. Nonhuman primate studies may shed some light in this area (reviewed in []).
Where do Tregs control allograft rejection?
Tregs can interfere with T cell priming in organ-draining lymph nodes. They achieve this by limiting the ability of effector T cells to form stable contacts with APCs [68, 69] and by constraining T cell activation . Further, they undergo antigen-stimulated proliferation within lymphoid tissue, favoring local immune regulation . In keeping with these findings, Tregs inhibited the alloantigen-driven proliferation of T cells in graft-draining lymph nodes in models of skin allograft rejection [72, 73], and in secondary lymphoid organs in a model of GVHD .
Tolerated grafts are infiltrated by recipient lymphocytes, suggesting that the allograft might also be a site of immune regulation . Indeed, Waldmann's group  showed that tolerated grafts contained T cells that, when transplanted onto secondary T cell–depleted hosts, could exit the graft and establish donor-specific tolerance. Our group found that intragraft regulation of donor-reactive CD8+ T cells resulted in a failure to generate memory cells capable of rejecting the graft . Using mice deficient in various chemokine receptors and adhesion molecules, Bromberg's group has demonstrated in an islet allograft model that Tregs migrate first to the allograft, where they prevent DC migration, and then to the draining lymph node, where alloreactive T cell priming and migration are prevented . Thus, Treg-mediated control of the allograft response occurs in both secondary lymphoid organs and the graft itself.
Expanding Tregs for Clinical Use in Humans
The low frequency of nTregs necessitates their expansion ex vivo prior to clinical use, and the past decade has seen rapid progress in this area. It has been estimated that adult humans have ∼13 × 109 Tregs, or <10% of the total CD4+ pool . Lymphopenic mouse models suggest that a Treg to effector T cell ratio of 1:1 to 1:2 is required to regulate graft rejection [39, 79]. Extrapolating from these observations, Tang and Lee  estimate that 5–8 × 1010 Tregs (7–11 × 108 Tregs/kg in a 70 kg patient) would be needed to regulate the alloimmune response in nonlymphodepleted humans. Expansion of nTregs on this scale is now possible , but it should be noted that limited data demonstrate control of allograft rejection with adoptively transferred Tregs in lymphoreplete animals . Our unpublished data suggest that a suboptimal dose of abatacept in combination with iTregs (but neither treatment alone) can significantly prolong cardiac allograft survival, at a dose of ∼2–8 × 106 Tregs/kg. This dose is similar to the maximum 10 × 106 Tregs/kg dose that has been approved for UK sites in the ONE Study, although this study is not primarily designed to assess efficacy. Other factors, such as the presence of difficult-to-control alloreactive memory T cells in humans, will also need to be considered [65, 66].
Multiple studies have reported the expansion of nTregs from peripheral blood (PB) and umbilical cord blood (UCB) [3, 27, 28, 80-82]. Administration of expanded nTregs at a dose of 3 × 106/kg is safe , and although 10- to 100-fold higher doses are now possible  these have not been tested in clinical studies. Human iTregs have also been generated and expanded from PB CD4+CD25− cells . Major human Treg expansion studies are summarized in Table 1.
|Levings et al ||PB CD4+CD25+||Anti-CD3, autologous PBMCs, EBV-transformed feeder cells||First report of human Treg expansion|
|Hoffmann et al ||PB CD4+CD25+||Anti-CD3/-CD28-loaded artificial APCs||First large-scale expansion of human Tregs|
|Trzonkowski et al ||PB CD4+CD25+CD127lo||Anti-CD3/-CD28-coated beads||First-in-human study of Treg cellular therapy in two patients|
|Di Ianni et al ||PB CD4+CD25+||No expansion||Showed safety of Tregs in high-risk HSCT patients|
|Nadig et al ||PB CD4+CD25+CD127lo vs. CD4+CD25hi||Anti-CD3/-CD28-coated beads||Greater potency of expanded CD4+CD25+CD127lo Tregs in a humanized mouse model of transplant arteriosclerosis|
|Hippen et al ||PB CD4+CD25hi||Anti-CD3/-CD28-coated beads or artificial APCs rapamycin||Massive expansion of PB Tregs, capable of suppressing xeno-GVHD|
|Sagoo et al ||PB CD4+CD25+||Allogeneic DCs||Protection of human skin grafts from alloimmune injury in humanized mice|
|Putnam et al ||PB CD4+CD25+CD127lo||CD154-stimulated allogeneic B cells||Protection of human skin grafts from alloimmune injury in humanized mice|
|Brunstein et al ||UCB CD4+CD25+||Anti-CD3/-CD28-coated beads||Reduction in Grades I–II GVHD and demonstration of safety in 23 UCB HSCT recipients compared with historical controls|
|Hippen et al ||PB CD4+CD25− or CD4+CD25−CD45RA+||Anti-CD3/-CD28-loaded artificial APCs|
|TGFβ rapamycin||Large-scale expansion of human induced Tregs|
Recently, the conditions required to expand nTregs have become increasingly well defined, and have been translated into good manufacturing practice (GMP) conditions. Most protocols use anti-CD3 antibody attached to beads [3, 27, 28] or artificial APCs expressing high-affinity Fc receptors [80, 82] and recombinant human IL-2 ranging from 300 to 1000 U/mL [3, 27, 28, 80]. Allogeneic APCs can also be used to expand nTregs. By enriching for alloantigen-reactive cells, rather than expanding the entire nTreg pool, this approach may result in a product with greater potency [84-86].
For successful expansion, CD28 costimulation is required . Anti-CD28 antibody may be attached to beads or artificial APCs. Given the multitude of costimulatory molecules, it might make sense to exploit other pathways to expand Tregs. Unfortunately, OX40, CD40L, ICOS and 4-1BB cannot substitute for CD28, and furthermore the addition of OX40 costimulation to anti-CD28 antibody improves the yield of expansion but reduces FoxP3 expression and suppressive capacity . This finding has led investigators to focus exclusively on the CD28 pathway in expansion protocols.
All-trans-retinoic acid (ATRA) and rapamycin are often added to Treg expansion cultures [80, 83, 88-90]. Both prevent the outgrowth of effector T cells and promote Treg expansion. It has recently become apparent that ATRA enhances the effect of rapamycin, but ATRA's usefulness on its own is limited .
Levings et al  first described the expansion of magnetically purified human CD4+CD25+ T cells from PB, using soluble anti-CD3 antibody and autologous peripheral blood mononuclear cells (PBMCs) admixed with irradiated Epstein-Barr virus–transformed feeder cells. In this early proof-of-concept study, a 40-fold expansion of CD4+CD25+ nTregs was achieved. Then, Edinger's group achieved 4 × 104-fold expansion using anti-CD3 and anti-CD28 antibodies bound to artificial APCs, combined with high-dose IL-2 . The expanded nTregs inhibited alloantigen-stimulated proliferation of CD4+CD25− T cells and retained expression of CD62L and CCR7, suggesting that they would traffic to lymph nodes in vivo. Although a subsequent mouse study suggested that CD62L expression by Tregs is dispensable for GVHD protection , the presence of lymph node homing molecules on the Treg surface is a relevant consideration because lymphoid organs are the important site for Treg-mediated regulation of alloreactive T cells in the mouse.
Since human CD4+ T cells transiently express FoxP3 upon activation , and nTregs are mainly CD4+CD25+CD127lo , Trzonkowski et al  expanded this population in the first clinical study of nTreg administration to two HSCT patients, one with acute and one with chronic GVHD. CD127lo cells were magnetically enriched from family members' PB followed by fluorescence-activated cell sorting (FACS), and expanded with anti-CD3/anti-CD28-coated beads and IL-2 for up to 3 weeks. The chronic GVHD patient had steroid-responsive bronchiolitis obliterans but was intolerant of prednisone withdrawal. After receiving a single dose of 1 × 105/kg expanded nTregs, the patient was able to discontinue immunosuppressive drugs and bronchodilators, concomitant with improved lung function. The acute GVHD patient had severe hepatic and gastrointestinal GVHD, and the pace of clinical deterioration slowed in response to three doses of ∼6 × 107 nTregs. Unfortunately, despite ongoing tacrolimus, solumedrol, mycophenolate mofetil (MMF) and five doses of anti-thymocyte globulin (ATG), the patient died . It is unlikely that Tregs would have had a marked benefit in this refractory case. Thus, this study was the first to show that ex vivo expanded human nTregs could be given safely to patients with GVHD, and that they may have a therapeutic effect.
Di Ianni et al  tested whether purified unexpanded nTregs from donor leukapheresis products could be used to prevent GVHD and improve immune reconstitution in 28 high-risk acute leukemia patients. Such patients receive aggressive myeloablation followed by T cell–depleted CD34+ PB hematopoietic stem cells (HSCs) from multiply HLA-mismatched donors. This approach delays recovery of pathogen-specific immunity, which depends on donor T cells early post-HSCT . The authors examined whether adding effector T cells to T cell–depleted CD34+ HSCs along with nTregs would permit reconstitution and avoid opportunistic infections without inducing GVHD, compared with historical controls . After conditioning from day −10 to day −6, nTregs were given on day −4, and effector T cells along with PB HSCs were given on day 0. In the presence of up to 4 × 106 nTregs/kg, the T cell dose was titrated up to 2 × 106 cells/kg, far higher than the dose known to induce GVHD [96, 97]. Despite this high dose, only 2 of the 28 patients experienced Grade II or higher GVHD, while none developed chronic GVHD, even though GVHD prophylaxis was not given. Further, reconstitution of pathogen-specific T cells was superior to 150 historical controls, and a lower rate of cytomegalovirus reactivation was observed in treated patients . This study therefore provided indirect evidence of immune regulation by human nTregs in vivo and revealed their safety in high-risk patients. In these heavily conditioned patients receiving titrated T cell doses, it appears that unexpanded nTregs had a clinically relevant effect. However, in SOT, unexpanded Tregs are unlikely to be effective given the large pool of alloreactive T cells in the recipient .
With regard to the cell population that should be used to expand human nTregs, our group compared the expansion and function of human CD4+CD25hi T cells with that of CD4+CD25+CD127lo cells in a humanized mouse model of transplant arteriosclerosis . nTregs were expanded using anti-CD3/anti-CD28-coated beads and IL-2, restimulated at day 7 for another 7 days before culture in IL-2 alone for 2 days. More than 600-fold expansion was achieved for both populations with >75% FoxP3+, and most expressing CTLA-4 and GITR. Expanded CD127lo cells expressed higher levels of CD62L and CCR7, however, and more potently inhibited autologous T cells in vitro and in vivo . In BALB/c RAG−/− γc−/− mice in which a segment of human internal mammary artery had been interposed into the abdominal aorta, CD127lo expanded nTregs more potently inhibited transplant arteriosclerosis caused by allogeneic (but autologous to the nTregs) PBMCs, than did expanded CD25hi cells, by a factor of 5 . The expanded CD4+CD25hi product likely contained some nonregulatory cells absent from the CD127lo compartment, accounting for the potency difference. Subsequent studies have demonstrated the efficacy of expanded CD4+CD25+CD127lo nTregs in humanized models of skin  and islet  allograft rejection, but given their low frequency in human PB, further work will be needed to determine whether expansion of this specific subset is a viable strategy when their increased potency is taken into account. Moreover, the flow cytometry-based sorting required to obtain CD127lo nTregs does not meet GMP standards in many jurisdictions at this time, unlike magnetic bead-based sorting.
Massive expansion of PB nTregs has been reported by Blazar's group . In their GMP protocol, CD4+CD25hi nTregs from leukapheresis preparations were stimulated with artificial APCs or anti-CD3/anti-CD28-coated beads plus rapamycin . Five stimulations led to ∼50 × 106-fold expansion, such that more than 6 × 1011 cells could be obtained from one leukapheresis. Inclusion of rapamycin prevented the emergence of large numbers of FoxP3− cells. The expanded cells retained suppressive properties in a xeno-GVHD model even after cryopreservation. Decreased methylation of the Treg-specific demethylated region (TSDR) upstream of the FoxP3 gene is a feature of bona fide Treg cells and correlates with stable FoxP3 expression . Importantly, the nTregs expanded by these investigators had extensive TSDR demethylation and did not express proinflammatory cytokines. These findings are provocative and demonstrate that large-scale human PB CD4+CD25hi nTreg expansion is feasible, although it remains to be seen whether the expanded cell product will retain its beneficial properties in patients.
Expansion using alloantigen
While polyclonal expansion generates large numbers of nTregs, it may be more desirable to use allogeneic APCs. Alloantigen-reactive nTregs may be more potent than polyclonally expanded nTregs (reviewed in []). Its major disadvantage is that it is not easily applicable in deceased donor SOT, since there would not be a sufficient period of time to expand recipient nTregs using donor APCs in advance of transplant surgery.
Sagoo et al  screened human nTregs for activation markers following stimulation with allogeneic PB or dermal CD1c+ DCs. Expression of CD69 and CD71 peaked 3–5 days after activation and defined alloantigen-activated nTregs. Alloantigen-activated CD69+CD71+ nTregs more potently suppressed donor-specific but not third-party responses by effector T cells than did polyclonally activated CD69+CD71+ nTregs. Human skin grafts on NOD.SCID.γc−/− mice showed less allograft damage and effector T cell infiltration in mice given alloantigen-activated nTregs compared with polyclonally activated nTregs. In another protocol, flow sorted CD4+CD25+CD127lo PB nTregs were expanded using activated allogeneic B cells . These B cells expressed high levels of costimulatory molecules and efficiently expanded nTregs that were more potent than polyclonally expanded nTregs in vitro and in the same model of human skin allograft injury reported by Sagoo et al . While these findings are promising, graft outcome data were not provided in either report, making a definitive conclusion regarding potency elusive.
Veerapathran et al  have expanded both directly and indirectly alloreactive nTregs. CD4+CD25+CD127lo nTregs sorted from PB were cultured with allogeneic DCs, rapamycin, IL-2 and IL-15, resulting in an 800-fold expansion of directly alloreactive nTregs over 12 days. Allogeneic fibroblast lysates were used to pulse autologous DCs, which expanded indirectly alloreactive nTregs in the same conditions by a factor of 2000 over 27 days . As observed by Sagoo et al , the alloreactive nTregs were far more potent in vitro than polyclonally expanded nTregs from the same donor. The same group has recently reported the expansion of nTregs specific for minor histocompatibility antigens, which are relevant in HLA-matched HSCT .
Thus, expansion of alloantigen-reactive nTregs is feasible and preliminary data suggest these cells may be more potent than polyclonally expanded nTregs. As discussed below, a comparison of these types of Treg will be a component of the ONE Study. However, data comparing the abilities of these two products to extend graft survival in a humanized mouse model would also help to inform future work.
For patients with hematologic malignancies, unrelated donor UCB represents a useful HSCT graft because it is more readily available than matched unrelated donor bone marrow . One UCB unit is insufficient for adult recipients, mandating double UCB transplantation . Since two donors contribute to the graft, GVHD risk is elevated . Thus, employing UCB-derived nTregs as a “companion product” to the UCB graft has received considerable attention.
UCB nTregs differ in important ways from PB nTregs. Since UCB is of fetal origin, it contains few activated T cells, such that CD25+ UCB lymphocytes are almost exclusively nTregs [105, 106]. Consequently, highly pure nTregs can be obtained from UCB with a single magnetic selection step [28, 105-107]. The presence of CD4+CD127+CD25+ activated effector T cells in adult PB means that nTregs are primarily CD127lo . Indeed, PB contains CD45RA−FoxP3lo nonregulatory memory T cells that secrete inflammatory cytokines, making it desirable to exclude them from clinical products [108, 109]. Thus, UCB is an attractive potential source of nTregs.
Brunstein et al  assessed whether expanded UCB nTregs could be used to prevent GVHD. In a Phase I dose escalation study , 23 recipients of double UCB HSCT received up to 3 × 106/kg expanded nTregs on day +1 following HSCT; some received an additional dose on day +15. All patients received GVHD prophylaxis with either cyclosporine and MMF or rapamycin. CD25+ nTregs from a third unrelated UCB unit were expanded for 18 days using anti-CD3/anti-CD28-coated beads and IL-2 . The expanded cells were suppressive in vitro and transiently prolonged dual chimerism in vivo, reflecting inhibition of alloimmunity. Three-quarters of the patients received the target dose; no dose-limiting toxicities were observed. In this small study, neither increased risk of opportunistic infection nor of leukemia relapse was observed in nTreg recipients compared with 108 matched historical controls . Treated patients and controls had similar rates of nonrelapse mortality; the incidence of Grades II–IV acute GVHD was reduced, but Grades III and IV GVHD were not. Intriguingly, treated patients had a much lower rate of chronic GVHD than controls. The main conclusion to be drawn from this study, however, is that it is possible that expanded UCB nTregs can be infused safely into adult HSCT recipients. Conclusions about the efficacy of expanded UCB nTregs in GVHD will need to await randomized controlled trials.
Pooling multiple units of UCB might help to overcome the small number of nTregs obtainable from single units . Further, since activation and expansion of UCB nTregs may alter their phenotype, our group compared freshly isolated nTregs from multiple pooled UCB units (using a one-step CD25 positive selection protocol) with PB nTregs (using a two-step protocol for isolating CD4+CD25+CD127lo cells) . The two products had similar suppressive capacity in vitro, but UCB nTregs had superior persistence in a humanized mouse model. Pooled UCB nTregs, but not pooled PB nTregs, prolonged the survival of allogeneic human skin grafts on BALB/c RAG2−/− γc−/− mice co-transferred with PBMCs . Since the pooled UCB nTregs exhibited diminished MHC Class I expression compared with PB-derived nTregs, they may have avoided elimination by co-transferred PBMCs, perhaps explaining their persistence. Thus, UCB-derived nTregs pooled from multiple donors may have more robust in vivo regulatory function than PB-derived nTregs. It is unclear, however, how such a product might perform compared with ex vivo expanded UCB nTregs . Furthermore, it is not known whether the salutary properties of pooled UCB nTregs would be retained after expansion, which is likely to be required in most settings.
Thus, the use of UCB-derived nTregs is clinically feasible in HSCT, and these cells may confer benefits. Further work in the field of UCB-derived nTregs will need to focus on defining whether they compare favorably to PB-derived nTregs.
Induction and expansion of iTregs from naïve PB precursors
Generation of iTregs from naïve precursors is another strategy under consideration. Blazar's group converted PB CD4+CD25−CD45RA+ T cells to iTregs and expanded them over 14 days using artificial APCs loaded with anti-CD3 antibodies plus IL-2, TGFβ and rapamycin . As in their PB nTreg expansion protocol , rapamycin prevented effector cell outgrowth and enhanced iTreg FoxP3 expression . That the cells were iTregs rather than nTregs was confirmed by their lack of expression of the latency-associated peptide, which distinguishes these two cell types in humans . The cell yield was 70–560 × 109 per leukapheresis, with 56 ± 15% of the cells being CD25+FoxP3+; correcting for this composition, the authors estimated that two- to fivefold more iTregs could be obtained than nTregs expanded using the same method . Induction of iTregs using ATRA is inefficient in the presence of memory cells ; however, the presence of CD45RA− memory cells in the starting population did not impair iTreg generation with rapamycin in this study. The iTregs were suppressive in vitro and performed as well as expanded PB nTregs at suppressing xeno-GVHD in NOD.SCID.γc−/− mice . Hence, iTreg induction from CD4+CD25− precursors and their expansion using rapamycin and TGFβ is feasible, and may be capable of generating more (or at least as many) iTregs as the various nTreg expansion protocols.
Treg therapy in SOT
The ONE Study (www.onestudy.org []), currently under way, is a Phase I/IIa clinical trial designed to test the safety and practicality of seven different regulatory cell populations in living donor kidney transplantation. Funded by the European Commission's Seventh Framework, the ONE Study involves eight clinical centers in Europe and the United States and compares polyclonally expanded PB CD4+CD25+ nTregs (FACS or magnetically sorted), type 1 regulatory cells, tolerogenic dendritic cells, regulatory macrophages, alloantigen-driven nTregs and alloantigen-driven T cells anergized by costimulation blockade. Clinical trial authorization has thus far been given to two groups for 1, 3, 6 and 10 × 106 Tregs/kg. This is in the same dose range shown to be effective in lymphoreplete mice ([] and our own unpublished data).
The immunosuppression protocol includes prednisone, tacrolimus and MMF based on the Symphony Study . Importantly, the same protocol will be used at all trial sites, allowing a direct comparison of transplant outcomes. In the initial phase of recruitment, Reference Group patients will receive standard-of-care immunosuppression without cellular therapy to establish baseline transplant and immunological data and these will be followed by the Cell Therapy groups. Although the primary goal of the ONE Study is to assess the safety and production feasibility of each cell product, the trial also includes immune monitoring of each patient in a central laboratory. A key objective is that all cell therapy patients will have been transplanted by the end of 2014, so that by the end of 2015 or early 2016, it should be possible to suggest which cell populations might be considered for large-scale trials powered for efficacy.
Very recently, Yamashita et al  have reported the administration of iTregs generated in vitro to 10 patients undergoing living donor liver transplantation. The iTregs were generated by co-culturing recipient PBMCs with irradiated donor PBMCs in the presence of costimulatory blockade. Administration of 0.6–2.6 × 109 iTregs was safe and appeared to facilitate early weaning and discontinuation of immunosuppression in 5 of the 10 patients. These findings are exciting but preliminary and will need to be compared to those of control patients not receiving Tregs. Longer-term follow-up in these patients will be essential.
The data from this and the ONE Study will therefore be of critical importance in defining the key clinical, practical and financial considerations involved in introducing regulatory cell therapy to the management of transplant patients.
Unanswered Questions and Future Directions
As we have seen, tremendous progress has been made in human Treg expansion. Clinical studies support the concept that expanded nTregs are safe in HSCT recipients, but several important questions must now be addressed.
Source and timing of administration
The most appropriate source of Tregs for cell therapy is unknown. As we have discussed, PB and UCB are both viable sources of Tregs, and Tregs may be expanded with alloantigens or anti-CD3 and anti-CD28 antibodies. Will expanded PB CD4+CD25+ cells be a safe and effective product, given that this population may contain effector T cells? Or, will it be necessary to expand only CD127lo cells or UCB nTregs? Assessing the comparative effectiveness of alloantigen-reactive and polyclonally activated PB nTregs is a goal of the ONE Study, but further comparisons will have to await future larger-scale trials.
Another area of uncertainty is the timing of Treg administration, which will, to some extent, be determined by the type of donor (deceased vs. living donor) and the Treg product. Expanded Tregs can be cryopreserved, effectively allowing for infusion of cells into the patient any time [3, 80]. As previously discussed, in deceased donor SOT, alloantigen-reactive Tregs would not be available until several days posttransplant, meaning they would not be present at the point of first exposure to alloantigens. This might reduce their effectiveness, but they could still have a role to play in allowing for drug minimization, for example. Additional clinical and preclinical studies will need to address these questions.
Stability and in vivo fate
The stability of Tregs in vivo is an important concern. Duarte et al  found that nTregs can lose FoxP3 expression in lymphopenic mice. This phenomenon might be important in patients treated with lymphodepleting agents such as ATG or alemtuzumab. Tregs that lose FoxP3 can develop an effector memory phenotype and become pathogenic in animal models . However, only a small proportion of nTregs lose FoxP3 in lymphoreplete mice, and such cells may represent transiently FoxP3+ activated effector cells . Further, although a minor population of true nTregs may lose FoxP3 expression, they can re-express it and remain suppressive when challenged with antigen .
Instability of FoxP3 expression is a more serious concern with TGFβ and IL-2-generated iTregs . After 6 days in culture, TGFβ-induced iTregs lost 90% of their FoxP3 expression, compared with 7% in nTreg . Loss of FoxP3 expression correlated with partial TSDR demethylation in TGFβ-induced iTregs compared with near complete demethylation in nTregs. Although rapamycin may mitigate it [83, 117], unstable FoxP3 expression can result in the conversion of iTregs into effector cells in vivo in mice undergoing GVHD [117, 118], which underscores the need for careful evaluation of Treg cell therapy products before administration.
Tracking transferred Tregs in vivo in transplant patients would be very useful for assessing their stability and survival, and could guide future treatment protocols. Durable labeling of cells in a manner that permits their long-term tracking in a minimally invasive manner has been challenging, although a recent study has shown that mouse Tregs transfected with the human sodium/iodide symporter could be detected in syngeneic mice using technetium-based SPECT/CT . Human Tregs can be labeled with gadolinium-based contrast agents, transferred into mice and detected ex vivo using mass spectroscopy . Hence, it seems likely that a clinically suitable method to detect adoptively transferred Tregs may be available in the near future.
Effects of immunosuppression on Tregs
Owing to the large pool of potentially alloreactive T cells in individual patients, Treg therapy alone is unlikely to prevent the rejection of a major MHC-mismatched graft . Therefore, immunosuppression will also be needed, but its effects on Treg efficacy and survival are poorly understood.
Rapamycin inhibits mTOR, blocking IL-2 signaling in T cells . Despite this, rapamycin has beneficial effects on Treg expansion and function in vitro and in vivo [83, 117, 122-124]. Rapamycin increases the frequency of CD62Lhi Treg in the PB of lung transplant recipients  and importantly, adoptive transfer of alloantigen-driven Tregs in combination with rapamycin induced long-term survival of cardiac allografts in mice  and prevented arteriosclerosis , suggesting that rapamycin and Tregs may have additive or synergistic effects.
Calcineurin inhibitors (CNIs) such as tacrolimus and cyclosporine block the activation of T cells by preventing IL-2 transcription , but some data suggest that they may adversely affect Treg therapy . CNIs have been associated with reduced numbers and decreased suppressor function of Tregs in renal transplant patients, in comparison with patients receiving rapamycin [128-130]. However, there is in vivo animal evidence to suggest that at low concentrations, tacrolimus can be used in conjunction with Tregs to provide long-term graft survival . Furthermore, low-dose cyclosporine promoted allograft tolerance while leaving Tregs unaffected in another study . MMF prevents purine biosynthesis, inhibiting T cell proliferation. MMF preserved Treg function in vivo  and reverted CNI attenuation of Treg function in vitro . Thus, although the data are conflicting, in vivo studies suggest that MMF and low-dose CNIs may not be harmful.
Alemtuzumab is an mAb that binds CD52, depleting mature lymphocytes . It may promote Treg survival, since it preferentially depletes effector T cells over Tregs . However, this conclusion was confounded by co-administration of rapamycin. In a later study, Treg numbers remained low in alemtuzumab-treated patients until the introduction of rapamycin . Lymphopenia resulting from alemtuzumab therapy may result in the loss of FoxP3 expression in Tregs . Therefore, whether alemtuzumab will be permissive to Treg therapy is unclear. Many of these uncertainties will take many years of clinical trials to resolve.
Taken together, existing data suggest that rapamycin is the least deleterious immunosuppressive drug for Tregs. Low-dose CNIs are unlikely to be harmful, although more data are needed. It is hoped that improved animal and human studies will address these issues in the near future.
Determining efficacy in clinical trials
One of the goals of Treg cellular therapy in SOT and HSCT is to minimize the need for immunosuppression. How future efficacy trials for Treg therapy might be designed is therefore potentially problematic. At present, it would be unethical to randomize patients to receive either Tregs or immunosuppressive drugs, since the latter are standard-of-care. Thus, the first randomized trials will have to test Tregs as add-on therapy to a standard immunosuppression protocol, as in the ONE Study. Since contemporary acute rejection rates are relatively low in adherent patients , the statistical power to detect a beneficial effect of Tregs as an adjunctive therapy could be prohibitive. Measures of donor reactivity or tolerance might be considered as surrogate endpoints, but these also require further validation .
Another approach might be to compare Treg therapy with immunosuppression in patients at high risk for or who have already experienced significant adverse effects from the drugs. Since children undergoing transplantation face decades of immunosuppression and its attendant complications, there may be greater clinical equipoise in withholding or withdrawing their drugs in favor of Tregs. In these contexts, Tregs could be tested as an adjunct to early weaning of patients from immunosuppression. Finally, patients in whom there is early evidence that immunosuppression is failing to prevent allograft loss (e.g. detection of cardiac allograft vasculopathy using intravascular ultrasound []) might be another group in which to test Tregs compared with augmented or altered immunosuppression.
Assessment of the cost-effectiveness of using Tregs will ultimately be required. Kidney transplant patients in the UK typically receive basiliximab induction followed by tacrolimus, MMF and corticosteroids, costing about £6000 for 1 year (A. Devaney, personal communication). We estimate that the cost of administering autologous PB Tregs in the ONE Study is about £30 000 (US$45 000), which will be over and above the cost of maintenance immunosuppression. However, Tregs may permit lower doses of immunosuppression to be used [122, 131]. The lowered immunosuppression doses might result in fewer incident cases of diabetes mellitus, hypertension, malignancy and opportunistic infections, which could drastically reduce indirect costs. Therefore, although the initial costs of Treg therapy are likely to be very high, if it permits weaning from immunosuppression and improves graft outcomes, then long-term Treg therapy could be economically viable.
Since Hall et al  first demonstrated that CD4+CD25+ T cells from tolerant rats could regulate allograft rejection, a great deal has been learned about the nature of Tregs, including their genetic and epigenetic properties and mechanisms of action. We now have at our disposal the tools to isolate them from patients and expand them in large numbers, and such cell therapy products seem safe . By 2016, the results of the ONE Study will be available to help guide the design of future studies. Yet, many questions regarding the efficacy, safety, mechanisms of action, tolerability and cost-effectiveness of Tregs remain to be answered. Moreover, how best to assess Treg therapy in the context of a randomized trial is a crucial open question that will require careful consideration. Indeed, we have not reached the beginning of the end, but rather the end of the beginning.
SCJ is the recipient of a Research Fellowship from the International Society for Heart and Lung Transplantation, and of a Detweiler Traveling Fellowship from the Royal College of Physicians and Surgeons of Canada. The work from the authors' own laboratory referred to herein was supported by the Wellcome Trust, the British Heart Foundation, and the European Commission's Seventh Framework Programme funding for the ONE Study and BioDRIM network.
The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.