CD4+CD25+ T regulatory cells (Tregs) can actively suppress immune responses and thus have substantial therapeutical potential. Clinical application is, however, frustrated by their scarcity, anergic status, and lack of defined specificity. We found that a single injection of a small number of expanded but not fresh HY-specific Tregs protected syngeneic male skin grafts from rejection by immune-competent recipients. The expanded Tregs were predominantly located in the grafts and graft-draining lymph nodes. In vitro expanded Tregs displayed a phenotype of CD25highCD4lowFoxp3+CTLA4+, and also up-regulated IL10 and TGFβ while down-regulating IFN-γ, GM-CSF, IL5, and TNF-α production. Furthermore, expanded Tregs appeared to express a reduced level of Foxp3, which could be prevented by adding TGFβ to the culture, and they also tended to lose Foxp3 following the repeated stimulation. Finally, a proportion of expanded HY-specific Tregs secreted IL2 in response to their cognate peptide, and this finding could be confirmed using Tregs from Foxp3GFP reporter mice. We not only demonstrated that expanded Tregs are superior to fresh Tregs in suppressing T cell responses against alloantigens, but also revealed some novel immunobiological properties of expended Tregs which are very instructive for modifying current Treg expansion procedures.
Regulation of T cells actively contributes to the control of potentially self Ag-reactive T cells that have escaped thymic deletion and exist in the periphery of healthy individuals. In mice and humans, CD4+25+ T cells represent one of the best-characterized CD4+ T cell subpopulations with immunoregulatory properties in the context of autoimmunity (1, 2). However, CD25 is not a specific marker for CD4 regulatory T cells (Tregs)3due to following limitations. First of all, CD25, as a traditional marker of T cell activation, is up-regulated in naive, nonregulatory CD4 T cells following TCR engagement. Secondly, although the CD4+CD25+ T cell population is enriched for suppressive activities, CD4+CD25− T cells are also potentially regulatory after in vitro clonal expansion. Thirdly, CD25 expression by Tregs is not stable: CD25+ Tregs can lose CD25 whereas CD25− Tregs can gain CD25 following antigenic and homeostatic stimulation in vivo (3). Although a panel of molecules (GITR, CTLA4, CD45RB, CD62L, Nrp-1, CD103, and CD127) has been described to be associated with Tregs (1, 2), none of them qualify as a specific marker for Tregs. More recently, the folate receptor 4 has been found to constitutively express at high amounts on Tregs, and the use of CD25 and folate receptor 4 appears to be able to distinguish naive, activated, and Tregs (4).
The discovery of Foxp3 as the key transcription factor controlling Treg development and their function represents one of the most significant advances in Treg immunobiology. Because both murine and human CD4+CD25− T cells can transiently up-regulate Foxp3 expression during in vitro activation, the value of Foxp3 as a Treg-specific marker has been questioned. A recent study conducted by Gavin and coworkers (5) demonstrated that activation-induced Foxp3 expression in human T cells is transient and insufficient to induce Tregs. Thus, sustained high level of Foxp3 expression appears to be essential for induction of Treg phenotype and function. Indeed, this conclusion is supported by a mouse study from von Boehmer’s group (6). They showed that mouse naive T cells can be converted into Tregs, which stably express high level of Foxp3, indicating that T cells which stably express Foxp3 are Tregs, independent of their origin.
CD4+CD25+ Tregs are professional suppressive cells capable of actively inhibiting T cell responses, and thus have substantial potential for treating graft-vs host disease (GVHD), autoimmune diseases, and transplant rejection (1, 2). However, therapeutic application of Tregs has been limited by several features, including their limited number, anergic status, and ill-defined Ag specificity. Therefore, optimizing large-scale in vitro expansion of Tregs is an essential requirement for their effective clinical intervention (7). Several clinical trials of the use of expanded human Tregs or IL10-secreting type 1 regulatory T cells to prevent GVHD in patients receiving hemopoietic stem cell transplants have been recently launched (8, 9).
Consistent with their hyporesponsiveness to TCR-mediated stimulation, Tregs have been found to express high levels of growth inhibitory genes including suppressor of cytokine signaling 1 and 2, cytotoxic T lymphocyte associated Ag 4 (CTLA4), and programmed death molecule 1 (1, 2). Therefore, strong signals through TCR/CD28 and IL2R are thought to be required to overcome the cell cycle arrest of Tregs (7). Indeed, a high concentration of IL2 would appear to be one of the essential components in all published protocols aimed at in vitro expansion to generate very large numbers of human and murine Tregs (7, 10, 11, 12). In addition, protocols using bone marrow-derived dendritic cells (BMDC) show that only fully matured BMDC are able to induce Tregs to proliferate in vitro (12). Although a combination of IL2 and beads coated with anti-CD3/CD28 gives rise to up to a 100–500-fold increase, the yield of expanded Tregs can be further enhanced when beads are replaced by artificial APCs over-expressing CD86 and FcR (8, 13). Although protocols for production of large numbers of Tregs within a limited time period are established, the challenge is to prepare high quality functional Tregs with a minimum contamination of effector T cells. One solution is the use of IL7 receptor α-chain (CD127) negative Tregs as the input cell population for expansion (14, 15), and to introduce rapamycin to the culture, which appears to selectively suppress the expansion of contaminating effector T cells (13). However, several studies (16) as well as our own unpublished data (Chai et al., manuscript in preparation), have revealed that Foxp3 expression in Tregs is not as stable as previously predicted: down-regulation of Foxp3 by Tregs was observed following homeostatic proliferation. The level of Foxp3 in Tregs is directly related to the efficiency of suppression (17), and Tregs can convert to anergic or even effector cells once they lose Foxp3 (16). Taking these findings into account, we wished to modify the existing in vitro Treg expansion protocols by introducing TGFβ. TGFβ is a growth factor for Tregs and is required for maintaining Foxp3 expression (18). In addition, TGFβ can protect T cells from apoptosis (19), and convert naive CD4 T cells into Foxp3-expressing Tregs (20).
Expanded islet Ag-specific Tregs are highly effective in suppressing autoimmune diabetes in NOD mice (11, 12), however, whether alloantigen-specific Tregs expanded in vitro have therapeutic potential in the context of transplantation has not been directly investigated, especially regarding the mechanisms by which expanded Tregs protect allografts from rejection by host T cells (21, 22). In addition, it should be noted that in most adoptive transfer experiments, immune-deficient mice have been used as recipients, and the use of T cell-free hosts could potentially complicate the interpretation of data. To avoid this, we have transferred expanded Tregs into wild-type (WT) mice and followed their expansion, homing, and localization.
The male specific minor histocompatibility Ag, HY, is comprised of multiple MHC class I and II restricted peptide epitopes (23). Syngeneic male skin and bone marrow grafts are rejected by female mice of high responder H2b strains such as C57BL/6 (23). In this study, we used the Rag2+/− TCR-transgenic strain Marilyn, specific for a HY/Ab/Dby peptide presented by H2Ab (24), as a source of HY-specific Tregs. Fresh Tregs were isolated by cell sorting and expanded in vitro allowing a large number of HY-specific Tregs to be obtained. The in vitro phenotype and function of expanded Tregs have been examined and their potential to control graft rejection is evaluated.
Materials and Methods
Mice and peptides
Thy1.2-B6 and Thy1.1-B6 mice were purchased from Harlan Breeders and The Jackson Laboratory, respectively. Rag2−/− Marilyn mice (TCR transgenic for HY/Dby peptide) (24) were provided by Dr. O. Lantz (Paris, France). Rag2+/− Marilyn mice were (B6 × Marilyn) F1 mice and used as a source of HY-specific Tregs. The spleen and lymph node (LN) cells of Foxp3GFP mice (25) were the gifts from Prof. R. Maizels (Edinburgh University, Edinburgh, U.K.) with the permission of Prof. A Rudensky (Washington University, St. Louis, MO). HY/Dby peptide (26) was synthesized by the Central Research Resources unit at the MRC Clinical Sciences Centre. The spleen and LN cells of Rag2+/− OT-II mice (TCR transgenic for cOVA323–339 peptide) (27) were provided by Dr. B. Stockinger (National Institute of Medical Research, London).
Purification of naive and Tregs
Rag2+/− Marilyn female spleen and LN cells were depleted of CD8 and B cells by CD8− and B220-Dynabeads, and stained with anti-Vβ6FITC, anti-CD45RBPE, anti-CD4PerCP, and anti-CD25APC before sorting on a BD FACSAria. The purity of resulting naive (defined by CD4+Vβ6+CD25−CD45RBhigh) and Tregs (defined by CD4+Vβ6+CD25+CD45RBlow) was routinely 98–99%. For experiments where only Tregs were needed, the cells were stained with anti-CD25PE followed by anti-PE-microbeads. After MACS separation, the enriched CD4+CD25+ cells were stained with anti-Vβ6FITC, anti-CD25PE, and anti-CD4PerCP before FACS sorting.
Treg expansion protocol
For the bead-based expansion, naive or Tregs (5–10 × 103/well) were incubated with anti-CD3/CD28-coated-beads (1:1 ratio) plus recombinant human IL2 (rHuIL2, 1000 IU/ml) in 96 round-bottom plates. Similar conditions were used for a dendritic cell (DC)-based protocol, in which beads were replaced by male BMDC, generated by culturing male B6 bone marrow cells with GM-CSF (5% supernatant (vol/vol) from X63-Ag8 cells transfected with murine GM-CSF) and IL4 (5% supernatant (vol/vol) from X63-Ag8 cells transfected with murine IL4) and followed by maturation with LPS (50 ng/ml). In Fig. 6, B and C, TGFβ was used at 1 ng/ml.
Phenotypic analysis of expanded T cells
For analysis of intracellular Foxp3 and CTLA4 expression, cells were stained by anti-CD4PerCP or anti-CD25APC, then fixed and permeabilized before staining with anti-Foxp3FITC or anti-CTLA4PE. All mAbs were purchased from BD Biosciences except anti-Foxp3, provided by eBioscience.
Intracellular cytokine analysis of expanded T cells
This was performed as described previously (28, 29). In brief, the expanded cells were harvested at day 7, and viable T cells were restimulated with HY/Dby peptide and B6 female BMDC in the presence of GolgiStop for 6 h. In some experiments, the cells were stimulated with PMA and ionomycin for 4 h. After staining with anti-CD4PerCP, the cells were fixed and permeabilized. The cells were then stained with APC-conjugated Abs (anti-IFN-γ or anti-IL10) together with PE-conjugated Abs (anti-IL-2, anti-GM-CSF, anti-IL5, anti-TNF-α, or anti-TGFβ). All anti-cytokine mAbs were purchased from BD Biosciences except for anti-TGFβ, which was provided by IQ Products. In some experiments, after fixation and permeabilization, the cells were stained with anti-IL2PE and anti-Foxp3FITC to determine whether IL2-producing cells coexpressed Foxp3.
Adoptive cell transfer, skin grafting, and in vivo CTL assays
Both skin grafting and in vivo CTL assays were conducted as described previously (28, 29). In brief, Tregs (1–10 × 105/per mouse, fresh or expanded) were adoptively transferred to B6 females by i.v. injection. The following day, the mice were grafted with syngeneic male skin, or i.v. injected with a mixture of equal numbers of male and female spleen cells (107/per mouse) labeled with a low (2 μM) or high (20 μM) concentration of CFSE. The ratio of male-to-female cells was determined by measuring CFSElow and CFSEhigh by FACS analysis of PBL samples taken at various time points. The selective loss of cells expressing the low level of CFSE indicated the rejection of male cells.
Isolation of T cells from skin grafts and tail skin
The accepted skin grafts and control host tail skin were chopped into small pieces and incubated in a mixture containing trypsin, collagenase XI, and DNase. The cell suspensions were prepared from digested tissues. After washing and removing the dead cells, the cells were then subjected to the same staining protocol as cells isolated from the spleen and lymph node.
This was performed using the Mann-Whitney U test and GraphPad Prism version 3.02 software.
Tissues were embedded in OCT, sectioned, and fixed. Frozen sections were incubated overnight anti-Thy1.1FITC and anti-Foxp3PE. Sections were examined on an immunofluorescence microscope. All sections were also counterstained with 4′,6′-diamidino-2-phenylindole.
Tregs in Rag2+/− Marilyn females are largely HY specific
Tregs are virtually absent in Rag2−/− Marilyn females (0.5 ± 0.1%, n = 3), but present in Rag2+/− Marilyn females (5 ± 0.2%, n = 3). Similar results were observed in A1 females, another HY-specific TCR-transgenic mouse line (30). To obtain a sufficient number of Tregs, we used Rag2+/− Marilyn females, which are referred to as Marilyn mice unless otherwise stated.
Compared with that of B6 mice, PBL from Marilyn mice displayed a marginal increase of frequency of CD4 T cells (top left, Fig. 1,A), a 60% reduction of percent of CD8 T cells (top middle, Fig. 1,A), and 30% reduction of Tregs (top right, Fig. 1,A). Importantly, >95% of Marilyn CD8+, CD4+, and Tregs expressed transgenic Vβ6 (bottom, Fig. 1 A). Similar data were also observed by another group (31). In addition, the mean fluorescence intensity (MFI) of Vβ6 on Marilyn CD4+Vβ6+ T cells was slightly higher that that on B6 T cells (103.3 ± 4.2 vs 81.6 ± 1.5, n = 3).
Expansion of Tregs by beads and by BMDC
When Marilyn naive and Tregs were compared for their response to beads plus rHuIL2, we observed that they expanded 40 and 20-fold, respectively (left, Fig. 1,B). Enhanced expansion was obtained when B6 naive and Tregs were used as responders (data not shown). The limited expansion of Marilyn Tregs was due to their delayed kinetics of proliferation over a wide range of IL2 concentrations (right, Fig. 1 B).
To generate mature BMDC, male B6 BM cells were cultured with GM-CSF and IL4 and matured by LPS. Fig. 1 C shows a typical phenotype of immature (top) vs mature BMDC (bottom). Maturation of BMDC was evidenced by the up-regulation of CD11c, H2Ab, and CD86.
For expanding Tregs, beads appear to be more efficient than BMDC (11, 12). However, relative efficiencies have not been directly compared using the same Tregs, nor has the capacity of high concentrations of IL2 to overcome the limited expansion of Tregs by BMDC been examined. When Marilyn Tregs were compared for expansion in response to beads or male BMDC under the same IL2 concentration (1000 IU/ml), we found the level of expansion obtained by beads was consistently higher (left, Fig. 1 D), indicating that 1) beads are superior to Ag-displaying, LPS-matured DCs for expanding Tregs, and 2) the limited capacity of expansion of Tregs by matured BMDC cannot be overcome by increasing IL2 dose.
It is possible that the different efficiency of expansion of Tregs by beads and BMDC reflects the fact that the beads induce polyclonal activation whereas DC are only capable of triggering T cells expressing transgenic TCR. To seek a clear-cut answer, we used monoclonal Tregs sorted from Rag2−/− Marilyn females and directly compared their expansion in response to beads or male BMDC. The fold of expansion of Tregs were very similar under these circumstances (right, Fig. 1 D), suggesting that one reason for the higher yield of expanded Tregs by beads is due to their ability to induce polyclonal T cell activation, regardless of Ag-specificity of T cells.
For suppressing in vivo anti-HY T cell responses, expanded Marilyn Tregs are more efficient than fresh Marilyn Tregs
To compare the in vivo regulatory activity between fresh and expanded Marilyn Tregs, we have used the adoptive cell transfer strategy to assess the ability of these two populations to protect syngeneic male skin grafts and hemopoietic cells from rejection by female recipients.
In the first series of adoptive transfer and skin transplantation experiments, we found that 1 × 105 fresh Marilyn Tregs, after adoptive transfer to immunocompetent B6 female recipients, prolonged graft survival but did not give indefinite protection, while HY nonspecific polyclonal Tregs isolated from B6 mice had no significant influence on skin graft survival (Fig. 2,A). In the second series of skin graft experiments, a single injection of 1 × 105 expanded Marilyn Tregs led to indefinite survival of male grafts in all B6 female recipients tested; in contrast, the same number of expanded B6 Tregs did not afford any protection (Fig. 2,B). Therefore, expanded Tregs are more effective than fresh Tregs with the same Ag-specificity (p = 0.025). To demonstrate a requirement for the specificity of HY-specific Tregs in this process, Tregs isolated from Rag2+/− OT-II females were expanded and adoptively transferred to B6 females, which were grafted with syngeneic male skin next day. To activate transferred OT-II Tregs, OVA protein was delivered to grafted mice via daily drinking water (32) until the experiment terminated. As shown in Fig. 2 C, expanded OT-II Tregs provided no protection for male graft survival, demonstrating that regulation-mediated by Tregs requires Ag-specificity in this model.
The in vivo cytotoxic assay was used to assess the ability of expanded Marilyn Tregs to protect male hemopoietic cells from rejection. Fig. 2,D represents a typical rejection response curve directed at male spleen cells by CD8+ T cells in B6 females (28), which received differentially CFSE-labeled male and female test spleen cells, and the relative survival of male cells was assessed in serial peripheral blood samples. In the first series of assays, we observed that adoptive transfer of 1 × 106 expanded Marilyn Tregs could significantly delay the course of rejection by prolonging the survival of male cells in B6 mice; there was no sign of rejection at day 17 (Fig. 2,E), and even at day 23 the survival of male cells in blood was >60% (Fig. 2,F). In contrast, administration of expanded B6 Tregs had no significant effect in altering the tempo of rejection. In the second series of assays, the capacity of fresh Tregs of Marilyn or B6 females to prevent male hemopoietic cells from rejection was directly compared and the results contrasted with those using the expanded populations. As shown in Fig. 2 G and H, fresh B6 Tregs offered no protection, with male cells being completely cleared from blood after 17 days. The same number of fresh Marilyn Tregs provided a limited protection; only ∼60% of male hemopoietic cells survived at day 17, and <20% at day 23. Because the adoptive transfer of the same number of expanded Marilyn Tregs provided a much better protection to male cells at both time points tested (100% at day 17; and >60% at day 23), we conclude that expanded Marilyn Tregs are more powerful suppressor cells than fresh Tregs with the same Ag-specificity for inhibiting the T cell response against male hemopoietic cells.
The location of adoptively transferred Treg cells in vivo
To understand how expanded Marilyn Tregs could protect male skin grafts from rejection by the female recipients, we tracked the fate of transferred Tregs at various time points over a relatively long period after transplantation. The use of Thy1.1 and Thy1.2 specific Abs allows us to reliably distinguish donor Tregs (Thy1.1+1.2+) from endogenous, host T cells (Thy1.1−1.2+). As shown in Fig. 3,A, 17 days after skin grafting, although Tregs were detectable in graft-draining and nondraining LN, they located predominantly in the graft-draining LN and expressed Foxp3. Evidence for preferential homing and/or expansion of Tregs at sites of Ag presentation was not only the finding of higher percentages (4-fold higher in the draining compared with the nondraining LN, at day 17 and 27) but also by the higher absolute cell numbers (3776 ± 315, n = 2) in the draining compared with the nondraining LN at day 27 (408 ± 32, n = 2). In addition, this pattern of selective location remained unchanged, although the proportion of Tregs in lymphoid tissues was reduced from day 17 to 27. More importantly, we could detect transferred Tregs (Vβ6+Thy1.1+CD25+) within individual accepted male skin grafts 21 days after grafting (top, Fig. 3,B) whereas these cells were virtually absent in host tail skin (bottom, Fig. 3 B). Finally, Tregs from graft-draining but not those from nondraining LN cells were capable of undergoing a strong clonal expansion following in vitro peptide stimulation, increasing from 0.04 to 1.8% representing a substantial 45-fold increase (data not shown).
To confirm the above results, we used an immunohistochemistry approach. Forty days after skin grafting, graft-draining LNs were examined under a microscope. The presence of transferred Tregs within graft-draining LN was evidenced by the detection of cells coexpressing Thy1.1 and Foxp3 (Fig. 3 C).
Expanded and fresh Tregs displayed similar phenotypes
Because Foxp3 is the most specific Treg marker (33, 34, 35), we compared Foxp3 expression between naive and expanded Tregs. Eighty percent of fresh Marilyn Tregs were positively stained by FJK-16s (top left, Fig. 4,A), which also stained 3% of naive cells (top right, Fig. 4,A), suggesting that Tregs are also present in CD4+CD25− fraction, although they are largely enriched in CD4+CD25+ fraction. Following expansion by beads plus IL2, Marilyn Tregs (left, Fig. 4,B) but not naive cells (right, Fig. 4 B) predominately expressed Foxp3, indicating that 1) Foxp3 appears to be stably expressed by Tregs following expansion, and 2) Foxp3, unlike CD25, is not a T cell activation maker.
CTLA4 is also expressed by Tregs (36, 37), and it is of interest to examine whether expansion would affect their CTLA4 expression. The expanded Marilyn Tregs (top left, Fig. 4,C) contained six-fold more CTLA4-expressing cells than expanded naive cells (top right, Fig. 4 C), suggesting that CTLA4 expression by Tregs was largely maintained after expansion.
Fresh Tregs are CD25highCD4low whereas naive cells are CD25negCD4high (30, 38). Although the expanded Marilyn naive cells greatly up-regulated CD25 expression, they still expressed a significantly lower level of CD25 than the expanded Marilyn Tregs (top left, Fig. 4,D). CD4 expression on expanded Marilyn naive cells was consistently slightly higher than that of the expanded Marilyn Tregs (top right, Fig. 4,D). Therefore, the differential expression pattern of CD4 and CD25 by Tregs and naive cells is conserved after expansion, and this is also evidenced by the cells from B6 mice (bottom, Fig. 4 D).
Expanded Tregs up-regulated TGFβ and IL10 but down-regulated IFN-γ, GM-CSF, IL5, and TNF-α
To understand how in vitro expansion can improve in vivo regulatory function of Tregs, we analyzed IL10 and TGFβ production. Compared with expanded Marilyn naive cells (right hand, Fig. 5,A), expanded Marilyn Tregs clearly up-regulated IL10 and TGFβ production in response to peptide (left hand, Fig. 5,A). Similar results were observed when expanded B6 Tregs were stimulated with PMA plus ionomycin (Fig. 5 B). Importantly, neither IL10 nor TGFβ was detectable in Marilyn fresh Tregs after stimulation by HY/Dby peptide presented by female BMDC (data not shown).
We also extended our analysis to other cytokines and found that the expanded Marilyn Tregs profoundly down-regulated production of IFN-γ, GM-CSF, IL5, and TNF-α. This was best illustrated by the finding that the proportion (for all cytokines tested) or the MFI (for TNF-α only) of cytokine-producing cells induced by HY/Dby peptide were much smaller in expanded Marilyn Tregs (left hand, Fig. 5,C) compared with expanded CD4+CD25− cells (right hand, Fig. 5,C). Again, down-regulation of these cytokines in expanded Tregs was not a unique feature of transgenic T cells, because a similar cytokine profile was observed when expanded Tregs from B6 mice were stimulated by PMA plus ionomycin (Fig. 5 D).
Repeated stimulation of expanded Tregs leads to loss of Foxp3 expression
We found that “green” Tregs (CD4+CD25+GFP+ cells isolated from Foxp3GFP reporter mice) (25) eventually decreased GFP expression upon adoptive transfer into Rag2−/− B6 mice (J. Chai and J. Dyson, unpublished data), indicating that Tregs can lose Foxp3 following homeostatic stimulation in vivo. Thus, it is of interest to assess whether loss of Foxp3 could also occur in vitro. To this end, FASC sorted “green” Tregs (99% purity) from Foxp3GFP mice were subjected to in vitro expansion by beads plus IL2. 7 days after the first round of stimulation, expanded green Tregs were still largely GFP-positive (left, Fig. 6,A), suggesting that Foxp3 expression in expanded Tregs is well maintained. In contrast, 7 days after the second round of culture, the majority of expanded Tregs became GFP negative (right, Fig. 6 A), indicating that repeated stimulation leads to the loss of Foxp3 in Tregs.
The decline in Foxp3 expression (Fig. 6) after the second round of stimulation could be the result of expansion of a minor subset of cells. To rule out this possibility, we have conducted an additional experiment in which the green cells were resorted to 99% purity (CD4+GFP+) at the end of the first round of expansion and then restimulated beads plus IL2. At the end of the first round of expansion (day 10), the percentage of GFP-expressing CD4+ cells was reduced from 99 to 85% (data not shown). Expanded Treg cells were resorted for 99% pure for CD4 and GFP at the beginning of the second round of expansion (data not shown). We observed that there is a significant reduction of proportion of GFP+CD4+ T cells (from 99 to 45%) at the end of the second round of expansion (data not shown). Therefore, it appears to be unlikely that the loss of GFP is due to selective expansion of a minor GFP-negative population.
TGFβ prevents down-regulation of Foxp3 in Tregs during the expansion
Above results raise the possibility that the expression of Foxp3 by Tregs has been already reduced following stimulation by beads plus IL2, and the down-regulation of Foxp3 in expanded Tregs is further accelerated by the second stimulation. To test this hypothesis, Foxp3 expression by expanded B6 Tregs was monitored during the first round of stimulation. In the absence of exogenous TGFβ, although expanded B6 Tregs were still Foxp3-positive, the expression level of Foxp3 was relatively low; the MFI of Foxp3 in the top 42% of Foxp3+ cells was only 141, while the MFI of the remaining 57% of cells was as little as 43 (top left, Fig. 6,B). In contrast, the presence of as little as 1 ng/ml exogenous TGFβ greatly improved Foxp3 expression by expanded Tregs, evidence by increased proportion (79%) and enhanced MFI (210) (bottom left, Fig. 6,B). Furthermore, the introduction of TGFβ only slightly reduced the level of CD25-expression on expanded Tregs (right hand, Fig. 6,B). Finally, addition of TGFβ to culture did not significantly reduce the yield of expanded Tregs (Fig. 6 C) especially at late stage of expansion, perhaps this was due to the capacity of TGFβ to protect activated T cells from apoptosis (16).
Expanded Tregs can produce IL2
Although the expanded Marilyn T cells of both types dramatically up-regulated IL2 production in response to HY/Ab/Dby peptide presented by female BMDC (bottom, Fig. 7,A), there was a remarkable difference in Foxp3 expression: IL2-producing expanded Marilyn CD4+CD25− cells did not express Foxp3, while up to 60% of IL2-producing-expanded Marilyn Tregs were Foxp3-positive (bottom, Fig. 7,B), demonstrating that the expanded Tregs were able to make IL2. When Foxp3 expression between IL2-producing and non-IL2-producing subsets within expanded Tregs was directly compared, we found that the proportion of Foxp3+ cells in non-IL2-producing fraction was higher than that in the IL2-producing fraction (85 vs 67%, data not shown). In other words, there were more Foxp3-negative cells in IL2-producing subset than those in non-IL2-producing subset (33 vs 15%, data not shown). However it is clear that the majority of IL2-producing cells do express Foxp3, indicating that they appear to be Tregs. Within expanded B6 CD4+CD25+ population, 18% of cells were Foxp3+IL2+ (bottom left, Fig. 7,C). In addition, a small proportion (1.7%) of Foxp3+ cells presenting in expanded B6 CD4+CD25− population produced IL2 upon PMA plus ionomycin stimulation (top right, Fig. 7 C). Therefore, resting Tregs unable to secrete IL2 appeared to be re-programmed into IL2-producing Tregs following strong stimulation via CD3, CD28, and IL2R.
Green Tregs are able to make IL2 after expansion
To conclusively demonstrate IL2 was produced by expanded Tregs rather than by contaminating memory/activated T cells, we have used Foxp3GFP reporter mice from which one can obtain the purest live green Tregs (25). CD4+GFP+CD25+ cells sorted from Foxp3GFP mice with 99% purity were cultured with beads and IL2. At day 5 and 8 of expansion, the cells were simultaneously examined for GFP expression and IL2 production. The proportion of IL2-producing Tregs was increased from 1.8% at day 5 to 11.3% at day 8 (right hand of Fig. 7,D), suggesting that expanded Tregs can up-regulate IL2 following the expansion. Interestingly, the progressive down-regulation of Foxp3 expression by expanded Tregs (was observed MFI of GFP reduced from 40 at day 5 to 35 at day 8; left hand of Fig. 7,D). Clearly, compared with those from WT and Marilyn mice (Fig. 7, A–C), Tregs from Foxp3GFP mice have a reduced capacity of producing IL2 following in vitro activation and expansion. The precise reasons for this are unclear, but we propose that the distinct MHC background (B6 for WT and Marilyn mice, and B6 × 129 for Foxp3GFP mice) and the expression of exogenous transgene would be potentially responsible for the difference observed.
The first major finding was that the immunoregulatory potential of HY-specific Tregs was significantly enhanced following a short period of in vitro expansion. Compared with fresh Tregs, expanded Marilyn Tregs are much more effective suppressor cells, evidenced in either male skin graft (Fig. 2, A–C) or male hemopoietic cell transplantation models (Fig. 2, E–H). There are several features in terms of prevention of graft rejection by expanded Marilyn Tregs. Firstly, a single injection of a small number of these cells is sufficient to prevent graft rejection. It has been estimated that infusion of 1 × 105 donor T cells will represent a frequency of 1:9000 of the peripheral CD4+ T cell pool in the recipient (39). Secondly, cell transfer was conducted in an immune-competent rather than a lymphopenic animal, and this is more relevant to clinical transplantation. Thirdly, the protection shows linked suppression, because the expanded Tregs are specific to a single HY T cell epitope and yet appeared to be able to inhibit T cell responses against skin graft mediated by at least two additional CD8+ T cell epitopes. Similar linked suppression has also been observed in a murine diabetes model (12). Finally, the Ag-specificity of Tregs was essential for efficient tolerance induction, because only the HY-specific Tregs protected the male graft from rejection; polyclonal or OVA323–339-specific Tregs were unable to do so (Fig. 2 C). Because immune regulation mediated by Tregs is generally found to be Ag specific in nonlymphopenic settings, this might reflect the fact that Tregs must be activated via their TCR before they are able to exert their suppressive activities (40). Polyclonal Tregs appeared to be efficient in treating GVHD (13, 41, 42, 43, 44), inflammatory bowel disease (45), and autoimmune gastritis (46) in mouse models, however, it is important to note these results were obtained when Tregs were cotransferred with effector T cells into lymphopenic recipients. Therefore, success of these therapies might be a result of a combination of extensive homeostatic proliferation properties of Tregs and/or a higher precursor frequency of alloantigen-specific Tregs within the polyclonal population (7).
There was clear up-regulation of “inhibitory” cytokines (IL-10 and TGFβ) and down-regulation of “inflammatory” cytokines (IFN-γ, TNF-α, etc.), but the magnitude of these changes was very small in most cases. In fact, there were still far more cells that produce the inflammatory cytokines than produce the inhibitory cytokines. We proposed that the potential contamination of non-Treg cells (activated/memory) within the CD4+CD25+ Tregs could be responsible for why there are more cells that produce the inflammatory cytokines than produce the inhibitory cytokines. It has been demonstrated that the IL10 protein in CD4+CD25+ Tregs is only detectable after 6 days of culture (47). This may help to explain why a low level of IL10 production was observed. This delayed kinetics of the production of IL10 by expanded Tregs may also apply to TGFβ. In addition, for measuring intracellular TGFβ, our direct staining method using conjugated anti-TGFβPE may be less sensitive than indirect staining used by others (48).
Expanded Marilyn Tregs could induce indefinite survival of male skin grafts, but they were clearly less effective in protecting male hemopoietic cells from rejection (Fig. 2, E–H). Similar results have been seen in our previous studies showing that male spleen cells are more sensitive to in vivo cytotoxicity than male skin grafts to rejection in B6 mice in which tolerance has been induced by intranasal peptide administration (28). This may be due to 1) target male cells introduced into the recipients by i.v. injection are not likely to be in sustained contact with CD4+CD25− cells in the circulation, 2) the lower density of HY Ag on the skin graft compared with male spleen cells, or 3) the environment of the skin graft is more conducive to Treg activity (49). In this regard, we found that expanded Marilyn Tregs are preferentially located within graft-draining LN as well as tolerated skin allografts by flow cytometric analysis (Fig. 3, A and B) which could be readily confirmed by immunohistochemistry studies (Fig. 3 C).
The transferred Treg cells were clearly shown to preferentially localize to grafted tissue and draining lymph nodes (Fig. 3), but the numbers with the graft-draining LN appeared to wane rather dramatically by day 27 (Fig. 3 A), yet the grafts were clearly maintained for >100 days. There are several speculations with respect to this observation. We speculate that by day 27, some host HY-specific CD4 T cells have been converted to adaptive Tregs within the graft draining LN by transferred Marilyn Tregs via poorly understood mechanisms of infectious tolerance, although we do not have direct evidence for this due to lack of HY-specific class II-restricted tetramers. Thus, the passive loss of donor Marilyn Tregs by day 27 can be potentially corrected by the peripheral generation of a new cohort of host’s Tregs of the same Ag-specificity. We also propose that Tregs are just one of the critical components responsible for maintaining tolerance. Interaction of between Tregs and APC within an accepted graft creates an inhibitory environment (i.e., tolerogenic Ag presentation and immunosuppressive cytokines), which further induces the expression of immunoregulatory genes (HO-1 and IDO) in the APC and/or tissue cells (endothelium and epithelium). Thus, reduced level of Tregs at day 27 can be potentially also compensated by the development of new local mechanisms of protection (acquired immune privilege) (50). The role of the expression of IDO by tolerogenic DC in maintaining tolerance is worth exploring in great detail in the future. It would be of interest to compare the regulatory activity of expanded Treg cells in WT and IDO knockout recipients-bearing male grafts from WT or knockout donors.
Ongoing work is exploring the precise mechanisms of action of expanded Tregs, and specifically we wish to differentiate the following possibilities: 1) prevention of the entry of naive T cells to the draining LN (7, 51), 2) suppression of the initiation of activation and expansion of naive T cells through inhibition of function of APC (52), 3) suppression of the differentiation of naive T cells into effector T cells (53), 4) creation of a microenvironment of immune privilege at sites of Ag (54), and 5) induction of apoptosis in T effector cells (55).
Although expanded Tregs up-regulated TGFβ and IL10 and maintained Foxp3 and CTLA4 expression, we have not explored the detailed parameters whereby expansion amplifies regulatory potentials of Tregs. The increased potency may be due to 1) increased individual Treg cell suppressive function following expansion, as demonstrated by others (11, 12) or 2) conversion by “infectious tolerance” or functional inactivation of contaminating CD4+CD25− cells during the expansion, a process requiring TGFβ (20) and which can be accelerated by IL2 (56). More experiments need to be performed in the future to establish the link between the up-regulation of inhibitory cytokines (IL10 and TGFβ) and enhanced regulatory activity by expanded Treg cells. For example, the comparison of in vivo regulatory activity between IL10-sufficient Marilyn Treg cells with IL10-defcient Marilyn Treg cells would be helpful to address this concern. Alternatively, anti-IL10 or anti-IL10R Ab can be coadministered with expanded Treg cells to examine the role of IL10 in the maintenance of tolerance. In vivo regulatory activity of Treg cells has been linked with IL10 and/or TGFβ (1, 2, 21, 50, 54, 57, 58, 59); IL10 produced by Treg cells is crucially important for the control of inflammatory bowel diseases (57) and wasting disease (60) induced by the transferred naive CD4 T cells. TGFβ is another important mediator of Treg cell function (1, 2). In addition, TGFβ acts as a growth factor for Treg cells and is needed for maintaining Foxp3 expression by Treg cells in vivo (18). Finally, TGFβ plays a key role in the conversion of naive CD4 T cells to Foxp3+ Treg cells in vivo (58, 59).
The second main finding was that green Tregs tended to lose Foxp3 following repeated expansion (Fig. 6,A). To exclude the possibility that the decline in Foxp3 expression after the second round of stimulation was due to a selective expansion of a minor subset of cells, expanded Treg cells were resorted for 99% pure at the beginning of the second round of expansion. We observed that there is a significant reduction of proportion of GFP+CD4+ T cells (from 99 to 45%) at the end of the second round of expansion (data not shown). Therefore, it appears to be unlikely that the loss of GFP is due to selective expansion of a minor GFP-negative population. In this regard, it is of interest to notice that two independent groups have recently demonstrated that during the 6 days of stimulation with anti-CD3, anti-CD28, and IL6, sorted CD4+GFP+ T cells from Foxp3GFP knockin mice gradually lost their Foxp3 expression (percentage of GFP+ cells reduced from 96% at day 2 to 73% at day 6) (48, 61). Therefore, at least in vitro during the course of activation and expansion, Foxp3 expression in Treg cells does not seem to be stable. Moreover, the loss of Foxp3 in expanded Tregs could be significantly prevented by adding exogenous TGFβ to the culture (Fig. 6 B). This finding not only represents a potential modification of existing Treg expansion protocols, but has several important implications for the clinical application of Tregs. For instance, due to the difficulty in obtaining human Tregs, there is a requirement for maintaining and preventing the loss of Foxp3 expression during in vitro expansion. Moreover, because the selection of Ag-specific Tregs with defined specificity from polyclonal Tregs (62, 63) needs several rounds of restimulation, it is critical to have a reliable procedure which is capable not only expanding Tregs, but also limiting their tendency to lose Foxp3.
Why is TGFβ able to prevent the loss of Foxp3 in expanded Tregs? It has been demonstrated that TGFβ is a growth factor for Tregs and also required for maintaining their Foxp3 expression (18). Furthermore, one more recent study uncovered a role for TGFβ in protection from activation-induced cell death (19). Moreover, TGFβ is needed to convert naive CD4 T cells to Foxp3-expressing Tregs in vitro (20), a process which also requires IL2 (56).
The third finding was that expanded Tregs were able to make IL2. The proportion of IL2-producing cells in expanded naive and Tregs was comparable, regardless of Ag specificity or the nature of stimuli (Fig. 7). We found that a significant proportion of IL2-producing expanded Tregs are Foxp3-positive (Fig. 7, B and C). This result was experimentally confirmed using green Tregs (Fig. 7 D). One recent report also showed that expanded Tregs indeed produced IL2 (64), although Foxp3 expression was not directly assessed. Our results are in contrast to others showing that expanded Tregs did not make IL2 (12). These discrepancies may be due to the use of different Ag-specific Tregs, distinct Treg subsets, and different APC and assays for IL-2 detection.
Our own data (Fig. 6 and our unpublished results) and those from others (16) indicate that Foxp3 expression by Tregs is less stable than predicted. The level of Foxp3 in Tregs has been directly linked to their efficiency of suppression (17). By taking these findings into account, one could interpret the finding of IL2 production by expanded Tregs (Fig. 7) as one of consequences of down-regulation of Foxp3 expression (Fig. 6). Further investigation is needed to determine whether IL2 production by expanded Tregs represents a “reprogramming” or a selection during in vitro expansion.
It remains difficult to discern whether the IL-2 producing Foxp3+ cells are due to a “re-expression” of the IL2 gene or due to the acquisition of Foxp3 by effector T cells. It is of important to distinguish above two possibilities, or a combination of both. We prefer the former possibility (i.e., IL2-producing Foxp3+ cells represent original input Tregs which turn on IL2 gene due to down-regulation of Foxp3 gene) due to following reasons. The small proportion of CD25+ nonregulatory cells which contaminated the sorted CD4+CD25+ cells are effector/memory CD4 T cells. Unlike naive CD4 T cells, these freshly isolated effector/memory CD4 T cells (63) as well as in vitro recently differentiated Th1 and Th2 cells (65) were resistant to in vitro TGFβ-mediated induction into Foxp3-expressing Tregs. Furthermore, for the induction of Foxp3 in naive CD4 T cells in vitro, a very high dose of exogenous TGFβ (2–5 ng/ml) (20, 63, 65) was added to the culture, such a high concentration of TGFβ might not be reached endogenously in our system, given the input of Tregs was very low (5–10 × 103 cells/well). Moreover, for expansion of Tregs, we have used beads coated with anti-CD3/CD28 which appear to be much less efficient than plate-bound anti-CD3/CD28 for the induction of Foxp3 (56, 65).
However, we cannot completely rule out the latter possibility (IL2+Foxp3+ cells come from effector T cells were induced to express Foxp3). Indeed, it has been suggested that the contamination of a minor population of effector/memory cells within CD4+CD25+ Tregs is responsible for IL2 production observed (66). A conclusive demonstration needs the cells from (Marilyn × Foxp3/GFP) F1 mice.
Unlike TCR transgenic mice, Ag-specific Tregs in humans are rare and clinical application represents a challenge. To generate large numbers, several approaches may be applicable. For example, Ag-specific Tregs can be expanded using Ag-pulsed DC cells followed by nonspecific stimulation (62, 63). At least in mice, nonregulatory CD4 T cells can be converted to Foxp3-expressing functional Tregs following Ag stimulation in the presence of TGFβ and IL2 (56, 65, 67). Furthermore, specificity can be altered by exogenous TCR expression and regulatory function can be induced by transfer of the Foxp3 gene (Refs. 33 , 68 and our unpublished observations).
In summary, we demonstrated that 1) HY-specific expanded Tregs are superior to fresh Tregs in suppressing in vivo anti-HY T cell responses, 2) expanded Tregs intended to reduce and even lose Foxp3, and 3) the down-regulation of Foxp3 can be prevented by adding exogenous TGFβ to culture. These findings not only provide new knowledge of Treg immunobiology but also are very informative for modifying current Treg expansion procedures. The successful treatment of GVHD, autoimmune diseases, and transplant rejections by immunotherapy requires a large number of high quality of Tregs. Because Foxp3 expression by Tregs is correlated with their suppressive function, but appears to be relatively unstable during in vitro expansion, it is worth paying more attention to maintain the high level of Foxp3 by expanded Tregs.
We thank Dr. O. Lantz and Prof. A. Rudensky (via Prof. R. Maizels) for providing Marilyn and Foxp3/GFP mice, respectively; and Dr. Gitta Stockinger for OT-II cells, and Eric O’Connor and Eugene Ng for their expert assistance with cell sorting.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the grants from Cancer Research U.K. (C11356/ A4033/A6994).
Abbreviations used in this paper: Treg, regulatory T cell; GVHD, graft-vs host disease; BMDC, bone marrow-derived dendritic cell; WT, wild type; LN, lymph node; DC, dendritic cell; MFI, mean fluorescence intensity.