Research ArticleTransplantation

Functional Regulatory T Cells Produced by Inhibiting Cyclic Nucleotide Phosphodiesterase Type 3 Prevent Allograft Rejection

See allHide authors and affiliations

Science Translational Medicine  18 May 2011:
Vol. 3, Issue 83, pp. 83ra40
DOI: 10.1126/scitranslmed.3002099


Regulatory T cells (Tregs) manipulated ex vivo have potential as cellular therapeutics in autoimmunity and transplantation. Although it is possible to expand naturally occurring Tregs, an attractive alternative possibility, particularly suited to solid organ and bone marrow transplantation, is the stimulation of total T cell populations with defined allogeneic antigen-presenting cells (APCs) under conditions that lead to the generation or expansion of donor-reactive, adaptive Tregs. Here we demonstrate that stimulation of mouse CD4+ T cells by immature allogeneic dendritic cells combined with pharmacological inhibition of phosphodiesterase 3 (PDE) resulted in a functional enrichment of Foxp3+ T cells. Without further manipulation or selection, the resultant population delayed skin allograft rejection mediated by polyclonal CD4+ effectors or donor-reactive CD8+ T cell receptor transgenic T cells and inhibited both effector cell proliferation and T cell priming for interferon-γ production. Notably, PDE inhibition also enhanced the enrichment of human Foxp3+ CD4+ T cells driven by allogeneic APCs. These cells inhibited T cell proliferation in a standard in vitro mixed lymphocyte assay and, moreover, attenuated the development of vasculopathy mediated by autologous peripheral blood mononuclear cells in a functionally relevant humanized mouse transplant model. These data establish a method for the ex vivo generation of graft-reactive, functional mouse and human Tregs that uses a clinically approved agent, making pharmacological PDE inhibition a potential strategy for Treg-based therapies.


Compelling evidence obtained from experimental models and from specific groups of patients indicates that regulatory T cells (Tregs) play an important role in T cell homeostasis and may offer therapeutic opportunities in autoimmune disease, bone marrow (BM), and solid organ transplantation (15). A number of current collaborative efforts are aimed at the identification of immunosuppressive protocols that might lead to Treg development in organ transplant patients, but these may be frustrated by the heterogeneity of immunosuppressive regimens and the paucity of reliable phenotypic markers of functional regulatory cells. An alternative approach is to develop ex vivo protocols that generate or expand recipient-derived Tregs for administration to the patient as a cellular therapy, and several promising approaches depend on T cell receptor (TCR) signaling combined with the provision of specific cytokines or exogenous costimulation (610). However, both recent and historical data indicate that manipulation of specific metabolic pathways may also offer possibilities for promoting Treg development ex vivo.

The cyclic nucleoside monophosphates cAMP (adenosine 3′,5′-monophosphate) and cGMP (guanosine 3′,5′-monophosphate) contribute to cell growth, proliferation, differentiation, and survival, and phosphodiesterases (PDEs), which hydrolyze the 3′-5′ phosphodiester bond to yield free adenosine and guanosine, provide an important control point in these processes (11). Indeed, T cell responses in vitro can be inhibited by extracellular nucleosides such as adenosine (12), and membrane-permeable analogs have shown that this inhibition is mediated at least in part by cAMP (13). Furthermore, elevation of intracellular cAMP in vitro with PDE inhibitors such as theophylline and caffeine not only inhibits T cell proliferation and interleukin-2 (IL-2) secretion (14) but also generates populations of T cells that can inhibit proliferation of other T cells in secondary cultures (15, 16). However, these PDE inhibitors are broadly reactive against most PDE isoforms, making data interpretation difficult, particularly in the context of current knowledge of Treg differentiation and function.

Recent data have shown that the Treg-associated transcription factor Foxp3 binds to target sequences within the PDE3b gene (17, 18), and in a screen of genes regulated by Foxp3, pde3b has been identified as one of the most repressed (19). Foxp3+CD25+ T cells express ectoenzymes that yield free adenosine from ATP (adenosine 5′-triphosphate) and AMP (adenosine 5′-monophosphate) (20, 21), and it has been proposed that the local delivery of adenosine provides another mechanism by which Tregs can provide relatively specific control of T cell responses (20).

We have shown previously that stimulation of CD4+ T cells ex vivo with allogeneic BM dendritic cells (DCs) in the presence of interferon-γ (IFN-γ) enriches for Foxp3+ cells that regulate allograft rejection (9, 10). This effect appears to be dependent on nitric oxide (NO) because enrichment is repressed in the presence of N-methyl-L-arginine (L-NMMA), an inhibitor of both constitutive and inducible forms of NO synthase (NOS), and induced in the absence of exogenous IFN-γ by provision of the NO donor S-nitroso-N-acetylpenicillamine (SNAP) (9). The ability of exogenous NO to drive the development of functional Tregs has also been shown using the NO donor NOC-18 (22). The role of NO in the generation of adaptive Foxp3+ Tregs hints at a possible connection with soluble guanylate cyclase (sGC); this signal transduction enzyme catalyzes the conversion of GTP (guanosine 5′-triphosphate) to cyclic GMP (cGMP) and is activated by NO (23). cGMP is an important second messenger that mediates many physiological effects, but one of the consequences of elevated cGMP levels is an inhibition of phosphodiesterase 3 (PDE3), resulting in a net increase in the intracellular concentration of cAMP (24, 25).

The link between NO, PDEs, and Tregs prompted us to determine whether functional alloreactive Tregs could be induced by alloantigen stimulation in the presence of pharmacological PDE inhibition. Foxp3+ cells enriched by this protocol prevented skin allograft rejection and development of transplant vasculopathy, which suggests that this method of alloreactive Treg enrichment may have potential for use in clinical transplantation.


PDE3 inhibition enriches alloreactive Foxp3+ T cells

Elevated levels of cAMP have been linked to inhibition of T cell responses, prompting us to determine whether pharmacological inhibition of PDE3 might promote the enrichment of alloreactive Foxp3+ T cells. Naïve CBA (H-2k) CD25CD4+ T cells were cocultured with B6 (H-2b) BM DCs in the presence of the PDE3 inhibitor cilostamide. Cells were restimulated on day 7 and harvested on day 14 for analysis of Foxp3 expression. As shown in Fig. 1A, the addition of cilostamide resulted in a dose-dependent increase in the proportion of Foxp3+ cells, reaching an ~10-fold enrichment at a concentration of 20 μM (Fig. 1B, left panel). A similar dose-dependent effect was seen in the absolute number of Foxp3+ T cells recovered, with a decline above a concentration of 10 μM, which probably reflects toxicity. These data indicate a potential route for the enrichment of alloreactive Foxp3+ cells ex vivo using a pharmacological agent in widespread clinical use.

Fig. 1

Stimulation of CD4+ T cells with allogeneic APCs combined with PDE3 inhibition leads to an enrichment of Foxp3+ T cells. Naïve CBA (H-2k) CD25CD4+ T cells were cocultured with B6 (H-2b) BM DCs in the presence of the PDE3 inhibitor cilostamide at the concentrations shown. Cells gated on CD4 and TCR-β for analysis of intracellular Foxp3 expression. (A) Representative plots. (B) Quantification.

Cilostamide promotes the development of functional Tregs

To determine whether CD4+ T cells driven by alloantigen during PDE3 inhibition are functional in vivo, we used an established adoptive transfer major histocompatibility complex (MHC)–mismatched skin transplant model. Naïve CBA (H-2k) CD25CD4+ T cells were cocultured with granulocyte-macrophage colony-stimulating factor (GM-CSF)/transforming growth factor–β (TGF-β)–differentiated B6 (H-2b) BM DCs in the presence or absence of 10 μM cilostamide, restimulated under identical conditions on day 7, harvested on day 14, and transferred into CBA-Rag−/− mice with 1 × 105 naïve CBA CD25CD4+ T cells or 1 × 105 BM3 CD8+ T cells as effector populations followed by B6 skin grafting (day +1). As shown in Fig. 2A, mice reconstituted with polyclonal CD25CD4+ cells rejected their grafts acutely [median survival time (MST), 18 days; n = 6], but cotransfer of 2 × 105 PDEi (PDE inhibitor)–conditioned T cells (Tcilos) prevented rejection, with all animals accepting their grafts long term (MST, >100 days; n = 5). In contrast, grafts were not protected by cotransfer of T cells cultured without cilostamide (MST, 18.5 days; n = 7, respectively, P = <0.05). We also examined regulation in a more stringent setting with a model of rejection by H-2Kb–reactive TCR transgenic BM3 T cells. In this model, essentially 100% of the effector population is reactive against donor H-2Kb. Reconstitution with 1 × 105 CD8+ BM3 T cells resulted in acute B6 skin graft rejection (MST, 15 days; n = 6; Fig. 2B), but cotransfer of 2 × 105 Tcilos delayed rejection and resulted in long-term survival in two of five recipients (MST, 51 days; n = 5, P = 0.02). Thus, the effect of alloantigen stimulation combined with PDE inhibition is not restricted to the phenotypic acquisition of Treg markers but also results in a population capable of functional regulation in vivo.

Fig. 2

Ex vivo PDEi-conditioned CD4+ T cells prevent skin allograft rejection and inhibit T cell proliferation and priming in vivo. (A) B6 (H-2b) skin graft survival in CBA-Rag−/− (H-2k) mice reconstituted with CBA CD25CD4+ effector cells alone or with cultured CD4+ T cells. (B) B6 (H-2b) skin graft survival in CBA-Rag−/− mice reconstituted with naïve CD8+ TCR transgenic BM3 cells alone or with syngeneic CD4+ T cells. Syngeneic CD4+ T cells in (A) and (B) were stimulated with B6 bone marrow (BM)–derived DCs in the presence or absence of 10 μM cilostamide. (C) Number of CD8+ T cells recovered from CBA-Rag−/− mice reconstituted with CFSE-labeled BM3 CD8+ T cells ± ex vivo PDEi-conditioned CBA CD25CD4+ T cells (Tcilos) and transplanted with B6 (H-2b) skin grafts the next day. (D) Proliferation of CD8+ T cells in (C) as determined by CFSE labeling. Histogram shows means ± SD (five to six mice per group) of CD8+ cells remaining CFSE+ defined according to the dot plot gate shown.

To explore potential regulatory mechanisms, we examined the effect of PDEi-conditioned cells on the proliferation of BM3 T cells in vivo. CBA-Rag−/− mice were reconstituted with 1 × 105 carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled BM3 CD8+ T cells with or without 2 × 105 ex vivo PDEi-conditioned CBA CD25CD4+ T cells. The use of BM3 cells allowed unequivocal identification of the effector population. Reconstituted mice received a B6 (H-2b) skin allograft the next day, and 15 days later, splenocytes were stained for CD8 and TCR-β to determine both absolute numbers and in vivo proliferation history. BM3 cells were readily detected in mice reconstituted with this population alone, but cotransfer of PDEi-conditioned cells resulted in a 10-fold reduction in the number of BM3 cells recovered (Fig. 2C; 33.6 ± 13.0 × 103 versus 3.2 ± 0.8 × 103 CD8 cells per spleen, BM3 cells only and BM3 cells + Tcilos, respectively; P = 0.01), indicating that inhibition of expansion is one mechanism involved in regulation mediated by PDEi-conditioned cells. This is further supported by division history analysis (Fig. 2D). When transferred alone, almost 90% of this monoclonal alloreactive population was CFSE at day 15, reflecting extensive in vivo proliferation. In contrast, in the presence of PDEi-conditioned CD4+ T cells, a large proportion of BM3 cells remained CFSE+ (11.3 ± 2.4% versus 43.7 ± 4.2% BM3 only and BM3 plus Tcilos, respectively; P = 0.004), thus providing evidence of inhibition of expansion by this ex vivo–generated regulatory population.

To assess the impact of the PDEi population on defined T cell effector function, we reconstituted CBA-Rag−/− mice with either 1 × 105 naïve CBA CD25CD4 T cells or BM3 CD8 T cells as effector populations ± cotransfer of CBA Tcilos and transplanted with B6 (H-2b) skin allografts the following day. Spleens were harvested 15 days after skin grafting. CD4+ or CD8+ cells were purified by magnetic positive selection from pooled splenocytes and challenged with T cell–depleted splenocytes from B6 mice as stimulators in an IFN-γ enzyme-linked immunospot (ELISpot) assay. The presence of PDEi-conditioned cells at the time of skin grafting resulted in a marked reduction in the capacity of the responding polyclonal CD4+ population to secrete IFN-γ when rechallenged with alloantigen in vitro (Fig. 3A, left panel), which translates to a 97% decrease in the frequency of donor-reactive IFN-γ–producing cells (Fig. 3A, right panel, P = 0.01 with respect to CD25 cells only). PDEi-conditioned cells also had a marked effect on the ability of monoclonal BM3 cells to produce IFN-γ in response to alloantigen rechallenge, equating to a greater than 95% reduction when expressed as the number of IFN-γ–producing cells per spleen (Fig. 3B, P = 0.01). Overall, these data indicate that the effect on graft outcome shown in Fig. 2, A and B most likely reflects an inhibition of both cell proliferation and acquisition of direct effector function mediated by this ex vivo–stimulated regulatory population.

Fig. 3

Cilostamide-induced Tregs inhibit IFN-γ production by effector T cells. (A and B) Frequency of donor-reactive IFN-γ−secreting CD4+ (A) or BM3 (B) T cells per well (means ± SD; five to six mice per group) with representative ELISpot images (left panels) from CBA-Rag−/− mice reconstituted with naïve CBA CD25CD4+ T cells or BM3 CD8+ T cells as effectors with or without ex vivo PDEi-conditioned CBA CD25CD4+ T cells (Tcilos) transplanted with B6 (H-2b) skin allografts. Right panels show the absolute number of IFN-γ–positive cells per spleen.

PDE inhibition results in demethylation of the Foxp3 gene

Epigenetic modification leading to unmethylated CpGs in the Foxp3 Treg-specific demethylated region (TSDR) has recently been shown to be a hallmark of naturally occurring Tregs (nTregs) in both mice and humans and is thought to reflect stable, constitutive Foxp3 expression in this population (26). To examine epigenetic changes in the cilostamide-induced population that showed functional regulation in vivo (Fig. 2), naïve CBA CD4+ T cells were stimulated with alloantigen in the presence of cilostamide as described above then harvested on day 14 for DNA isolation followed by DNA bisulfite pyrosequencing. The TSDR region and polymerase chain reaction (PCR) primers used are shown in Fig. 4A. To provide a relevant comparator population in the same mouse strain, we gated naïve CBA T cells on CD4 and flow-sorted them on the basis of GITR and CD25 expression. As shown in Fig. 4B, this strategy provided a population depleted of nTregs, and previous direct comparisons performed in this laboratory have shown that for naïve, unstimulated CD4+ T cells, this approach is as effective as sorting cells on the basis of green fluorescent protein (GFP) expression from GFP-Foxp3 reporter mice (27). Naïve CD25GITR cells (containing <0.5% nTregs) showed high levels of methylation in the TSDR region (Fig. 4C, upper panel), with levels at some CpG sites exceeding 90%, consistent with their Foxp3 protein signature. In contrast, the reciprocal CD25+GITR+ population showed extensive demethylation within the same region, in agreement with data reported previously for BALB/c mouse nTregs (26).

Fig. 4

PDE3 inhibition leads to CpG demethylation in the Foxp3 5′UTR. (A) CpG map of the Treg-specific demethylated region (TSDR) and indication of the positions of the eight primer sets. Replicate populations from two independent experiments were analyzed with four technical replicates per population. (B) CBA CD4+ T cells were stained for CD25 and GITR and flow-sorted on the basis of the gate shown (left panel). This results in a population of sorted cells essentially devoid of naturally occurring Tregs (nTregs) (right panel). (C) Percent CpG methylation in the TSDR for each of the cell populations shown. (D) Percent CpG methylation for CD4+CD25 cells from CBA Foxp3 GFP reporter mice that were stimulated with B6 BM-derived DCs in the presence of 10 μM cilostamide. (E) CD4+CD25 T cells from CBA Foxp3 GFP reporter mice were stimulated with B6 BM DCs in the presence of 10 μM cilostamide, and 14 days later, GFP+ cells were isolated by flow sorting and restimulated with B6 BM DCs under identical conditions without cilostamide. Cells were harvested on days 7 and 14 and stained for intracellular Foxp3. (F) TSDR CpG methylation profiles of cells in (E) and that of sorted GFP+ cells restimulated under neutral conditions after initial simulation in the presence of TGF-β plus IL-2. (G) Representative dot plots show Foxp3 expression in the gated CD4+ TCR+ input population (day 0; CBA CD4+ T cells stimulated with B6 BM DCs for 14 days in the presence of cilostamide) and in the gated population in spleen and draining lymph node (dLN) 14 days after transfer to CBA-Rag−/− mice together with BM3 CD8+ T cells followed by B6 skin grafting. Histogram shows means ± SD of data from five individual mice.

To examine the effect of PDE3 inhibition on TSDR methylation, we stimulated naïve CBA CD4+ T cells with B6 BM-derived DCs in the absence or presence of 10 μM cilostamide. Cells were restimulated under identical conditions on day 7 and then harvested on day 14 for DNA isolation. Cells were not sorted or enriched before DNA extraction because the intention was to determine the TSDR methylation status of the population that without further manipulation regulates rejection responses in vivo (Fig. 2). Cells stimulated with alloantigen in the absence of cilostamide remained heavily methylated in the TSDR region (Fig. 4C, lower left) and, in this respect, were similar to naïve non-Tregs. In contrast, T cells activated in the presence of cilostamide showed reduced TSDR methylation. Although the TSDR region in these cells is considerably more methylated than in nTregs, the levels are similar to those seen using the prototypic IL-2 + TGF-β protocol (Fig. 4C, upper right) that has been reported elsewhere (6, 28, 29) and appear to be similar to those reported by Floess et al. (26). Therefore, although T cells driven by allogeneic antigen-presenting cells (APCs) in the presence of cilostamide are more highly methylated in the critical Foxp3 locus than their naturally occurring Treg counterparts, they show levels of demethylation that are comparable to those seen in other adaptive or inducible Tregs.

Although the cilostamide-treated cells showed less TSDR demethylation than CD25+GITR+ nTregs, by design, the TSDR data for these cells were obtained from the total population that showed functional regulation in vivo (Fig. 2). As shown in Fig. 1, these populations typically contain ~60% Foxp3 cells. To determine whether demethylation of the TSDR in cilostamide-treated cells corresponded with Foxp3 expression, we stimulated CD4+CD25 cells from CBA Foxp3.GFP reporter mice with B6 BM-derived DCs in the presence of 10 μM cilostamide, restimulated them under identical conditions on day 7, and then flow-sorted them on day 14 on the basis of GFP (Foxp3) expression. Bisulfite sequencing was then performed on the isolated GFP and GFP+ populations. As shown in Fig. 4D, TSDR demethylation was more pronounced in GFP+ than GFP cells, confirming that alloantigen stimulation combined with PDE inhibition induces TSDR epigenetic modifications preferentially in Foxp3+ cells.

One of the observations made by Floess et al. (26) was that although nTregs stimulated in the absence of cytokines retained Foxp3 expression (~99.4% Foxp3+ at day 0, 92% after 6 days), restimulation of TGF-β + IL-2–induced Tregs in the absence of TGF-β resulted in a loss of Foxp3 expression (~98% Foxp3+ at day 0, <10% after 6 days). To examine the stability of cilostamide-induced alloreactive Tregs, we stimulated CD4+CD25 T cells from CBA-Foxp3 GFP reporter mice with B6 BM DCs in the presence of 10 μM cilostamide as described above, flow-sorted them on the basis of GFP (Foxp3) expression (>96% purity), and then restimulated them with B6 BM DCs under identical conditions without cilostamide. This population retained Foxp3 expression 7 and 14 days after restimulation (Fig. 4E). To determine whether this stability reflects epigenetic modifications, we prepared and isolated cells as described for Fig. 4E, and isolated DNA 14 days after restimulation for TSDR bisulfite sequencing. To provide a comparator population, we stimulated cells from CBA-Foxp3 GFP reporter mice in the presence of IL-2 + TGF-β (Fig. 2C) then isolated them on the basis of GFP expression. These cells were restimulated without cytokine modification and then harvested 14 days later for DNA isolation. As shown in Fig. 4F, PDEi cells restimulated in the absence of further cilostamide treatment retained substantial levels of TSDR demethylation, and the overall TSDR profile was similar to that seen in restimulated TGF-β cells. Although the TSDR in the restimulated PDEi cells was slightly less demethylated than that in equivalent cells at the time of initial harvest (Fig. 4D), when seen in combination with the Foxp3 expression data (Fig. 4E), the data suggest that cilostamide-induced Tregs maintain a relatively stable phenotype under conditions of antigen reexposure.

To assess the functional stability of this population under conditions of alloantigen challenge in vivo, we transferred B6-reactive CBA Tcilos to CBA-Rag−/− mice (day 0) together with Kb-specific BM3 CD8+ TCR transgenic T cells as described in Fig. 2B. One day later, these mice were transplanted with B6 skin grafts. On day 14, spleen and draining lymph node cells (left axillary) were analyzed phenotypically. Foxp3+ cells were readily detected in both compartments (Fig. 4G) but were especially enriched in the draining lymph nodes (P = 0.01 with respect to spleen), which are known to be sites of early Treg recruitment and function in a comparable skin graft model (30). Thus, under conditions of active in vivo regulation (Fig. 2, A and B), alloreactive CD4+ cilostamide-induced Tregs retain Foxp3 stability during the critical period where their presence leads to long-term skin graft survival instead of acute transplant rejection.

PDE3 inhibition increases Foxp3 expression in vivo

The demonstration that PDE3 inhibition in vitro leads to an enrichment of alloreactive Foxp3+ cells prompted us to determine whether a similar phenomenon might occur in vivo. To explore this possibility, we adoptively transferred CBA CD4+CD25 cells (input day 0 population, Fig. 5B) to normal CBA×B.10 F1 mice, which were then treated with cilostamide or vehicle only. On day 5, spleen cells were harvested and analyzed for Foxp3 expression. Transfer of CBA cells into these F1 mice established an in vivo mixed lymphocyte reaction (MLR), where the transferred T cells respond to H-2b alloantigens without NK (natural killer) killing of the input population. Foxp3+ cells were gated out by staining for Kb, allowing Foxp3 expression to be determined in the transferred T cell population alone. Relative to vehicle-only controls, the administration of cilostamide resulted in a twofold increase in both the percentage and the absolute number of Foxp3+ T cells in the input population (Fig. 5, C and D), providing in vivo support for the effects of cilostamide demonstrated in vitro. Transfer of the same population to cilostamide-treated syngeneic CBA mice in which transgenic T cell expression of hCD52 allowed host hCD52+Foxp3+ cells to be excluded from the analysis did not result in an increase in Foxp3+ cells (Fig. 5D), indicating that this effect is dependent on direct T cell activation.

Fig. 5

Phosphodiesterase inhibition in vivo enhances Foxp3 expression. (A and B) CBA (H-2k) CD4+CD25 cells (A) were adoptively transferred on day 0 (B) to normal CBA×B.10 F1 (H-2k+b) mice, which were then treated with cilostamide (6.4 mg/kg per day) or vehicle [phosphate-buffered saline (PBS)] for 5 consecutive days. On day 5, spleen cells were analyzed for Foxp3 expression. (C) Representative dot plots. (D) Absolute numbers of Foxp3+ T cells in the spleen. Means ± SD (n = 3 per group).

PDE3 inhibition ex vivo results in functional human Foxp3+ Tregs

To determine whether inhibition of PDE3 would also be an effective means for the generation ex vivo of human Tregs, we cocultured CD45ROCD25CD4+ T cells with GM-CSF/IL-4/TGF-β–differentiated allogeneic monocyte-derived DCs in the presence of cilostamide, restimulated them on day 7 under identical conditions, and harvested them on day 14 for phenotypic analysis and for use in a regulation assay based on a standard MLR. As in the mouse, between 5 and 10% of resting human CD4+ T cells are positive for Foxp3, but unlike the mouse, Foxp3 is up-regulated nearly threefold after activation (Fig. 6A). However, as with mouse cells, the addition of cilostamide resulted in a significant increase in Foxp3 expression such that most cells became Foxp3+ (Fig. 6A). This was also reflected by the up-regulation of cell surface markers previously associated with a regulatory phenotype, including CD25, CD62L, and CD152 (solid curves, Fig. 6B).

Fig. 6

Phosphodiesterase inhibition generates functional human regulatory cells. (A) Frequency of Foxp3+ cells derived from CD25CD45ROCD4+ T cells from human PBMCs cultured with irradiated allogeneic APCs in the absence (control) or presence of 20 μM cilostamide. (B) The same populations were examined for expression of CD25, CD127, CD62L, and CD152 (bold curves). Data are representative of two independent experiments. (C) PBMCs were stimulated by allogeneic APCs ± cilostamide as in (A), reisolated (conditioned T cells), and cocultured with autologous PBMCs (responders) driven by allogeneic APCs. Responder proliferation was determined by thymidine incorporation. Results expressed as mean cpm ± SD of triplicate wells. (D) BALB/c Rag2−/−/cγ−/− mice were transplanted with human arterial interposition grafts on day 0 and reconstituted intraperitoneally the next day with 107 PBMCs alone or together with 107 cilostamide-induced Tcilos driven by APCs from the same PBMC donor. Grafts were harvested 30 days later. Representative EvG-stained sections (left) and intimal expansion analysis from two independent experiments (right) are shown. Results are means ± SEM.

These ex vivo–conditioned CD4+ T cells were then tested in an MLR to determine their ability to suppress the proliferation of autologous peripheral blood mononuclear cells (PBMCs) against allogeneic stimulation. As shown in Fig. 6C, when autologous responder PBMCs were stimulated by irradiated allogeneic PBMCs, the presence of cells conditioned in the primary culture with cilostamide resulted in a marked inhibition of proliferation in the secondary suppression assay. This inhibition could be detected at a conditioned T cell/responder ratio of 1:64 but increased in a dose-dependent manner such that, at a ratio of 1:2, the proliferation was about 15-fold lower than that of responders alone (P < 0.05, ratios 1:32 and above). By contrast, cells conditioned in the absence of cilostamide had no impact on proliferation except at a ratio of 1:1, and even here, proliferation was reduced by less than a factor of two (P < 0.05 at 1:1, P > 0.05 all other groups). To determine whether these conditioned human cells have the capacity for regulation in vivo, we took advantage of a human artery transplantation model established in our laboratory. BALB/c Rag2−/−/cγ−/− mice were transplanted with human arterial interposition grafts and then reconstituted the next day with 107 PBMCs allogeneic to the vessel with or without 107 cilostamide-induced Tregs from the same cell donor. Thirty days after transplant, the grafts were harvested, sectioned, stained, and evaluated for transplant vasculopathy. As shown in Fig. 6D, although neither the procurement process nor transplantation itself resulted in vasculopathy, reconstitution of these immunodeficient recipients with PBMCs alone resulted in a significant degree of intimal hyperplasia as indicated by the open arrows in the upper representative section. In clear contrast, cotransfer of cilostamide-treated T cells resulted in a significant reduction in vasculopathy, which on average was about twofold but, in some cases, reduced intimal expansion almost to that seen in unreconstituted animals (Fig. 6D, right panel). Thus, as in the mouse, allogeneic stimulation of human CD4+ T cells in the presence of cilostamide promotes the enrichment of functional Tregs.


The observation that Tregs can control destructive T cell responses in numerous transplantation and autoimmunity models has recently been given new focus by the demonstration that nTregs play an essential role in peripheral self-tolerance in normal, unmanipulated animals (5, 31). The fact that selective depletion of these cells leads to widespread autoimmune disease underlines the potential of such populations and has rekindled attempts to identify protocols in which Tregs could be manipulated for therapeutic benefit. Although it is possible that some forms of current immunotherapy may induce the selection or development of human Tregs in vivo (32, 33), an alternative possibility is to develop protocols that expand and/or generate Tregs ex vivo for use as a cellular therapy.

Much of the current research focus is on the expansion of nTregs using polyclonal stimulation that can result in a several thousand-fold expansion of the starting population. These cells can regulate alloreactivity in vivo, but more importantly, the initial observations made with mouse T cells (6, 34) have now been replicated in our laboratory with expanded human nTregs in a clinically relevant model of human chronic allograft vasculopathy (35). Thus, expanded nTregs clearly have considerable potential for regulating alloreactivity, and several protocols now exist for the use of such populations in hematopoietic stem cell transplantation for the control of graft-versus-host disease (GVHD) and encouraging preliminary results have recently been published (36, 37). However, the impact on normal immune responses and immune homeostasis of administering large numbers of polyclonally activated Tregs is not known, and the uncertainty is compounded by the fact that methods do not yet exist to eliminate potential effector T cells from the expanded population. An alternative approach is to promote the enrichment, expansion, or generation of alloreactive-induced Tregs by stimulation of recipient CD4+ T cells with donor APCs under defined culture conditions. Here, we have demonstrated that stimulation of both mouse and human CD4+ T cells in the presence of a PDE3 inhibitor resulted in a population with the phenotypic characteristics of Tregs and the functional ability to prevent allograft rejection in vivo. We also demonstrated that the administration of cilostamide can promote the development of Tregs in vivo. The importance of T cell activation for the emergence of Foxp3+ cells is highlighted by the fact that although transfer of CD4+ T cells to F1 recipients results in an expanded Foxp3+ population, this did not occur on transfer to syngeneic recipients (Fig. 5, C and D). This is consistent with results of previously published work examining the role of TCR ligation and in vivo development of nTregs (3840) and with our own experiments where alloreactive adaptive Tregs develop in vivo after tolerance induction (27).

In an attempt to understand the origin and development of alloreactive Tregs demonstrably responsible for allograft protection after tolerance induction in vivo (41), we have adoptively transferred mouse CD4+ T cells purged of nTregs by stringent flow sorting to syngeneic Rag−/− recipients. These mice were then either tolerized with an established induction protocol based on donor alloantigen challenge or left untreated before delivery of an effector T cell population followed by transplantation with donor strain skin grafts. Mice transplanted without tolerance induction rejected their grafts acutely, but in contrast, mice that had received the donor alloantigen induction protocol all accepted their grafts long term. Critically, this engraftment was associated with the emergence of a Foxp3+ T cell population (27). Thus, as with the development of nTregs, the development of adaptive Tregs in the periphery is critically dependent on TCR signaling. Although this is an important mechanistic observation, its relevance to the cilostamide data shown in Fig. 5 is that an increase in the proportion of Foxp3+ Tregs is unlikely to be a clinical consequence in patients receiving PDE inhibitors, for example, for the treatment of claudication. However, the possibility exists that when combined with established alloantigen-dependent tolerance induction protocols, cilostamide treatment may be a useful adjunctive strategy for the development of functional Tregs in vivo.

In the protocol described, both CD4+ T cells and allogeneic DCs are exposed to cilostamide in the 14-day coculture period (Fig. 1), and so either population could be potentially affected by the inhibitor. The reduced PDE3b expression in nTregs (17, 18) suggests that T cells rather than APCs are the likely primary targets of cilostamide in the culture system described; however, further studies will need to be performed to confirm this hypothesis. The observation that cilostamide-induced enrichment of Foxp3+ Tregs is dependent on T cell activation suggests that stimulation of bulk CD4+ T cells with CD3/CD28 beads in the presence of cilostamide might be an alternative route to promote the development of Tregs in vitro. However, polyclonal stimulation in the presence of cilostamide results in an outgrowth of CD25+Foxp3 cells. Although it might be possible to modify this outcome by using alternative or weaker agonists, the potential risk of expanding non-Tregs probably means that polyclonal stimulation of total CD4+ cells is less attractive than polyclonal expansion of isolated Treg populations.

In an attempt to determine whether the enrichment of Tregs extends beyond alloreactive T cells in our system, we have asked whether the in vitro cilostamide protocol can promote Treg development in a polyclonal response to human γ globulins, which are known to be capable of inducing tolerance in vivo (42, 43). However, attempts to detect an enrichment of Foxp3+ cells in this setting have been unsuccessful, possibly because of the low frequency of T cells responding to nominal antigens. To overcome this problem, we have used ovalbumin-reactive DO11.10 TCR transgenic T cells. These preliminary experiments have been performed with total CD4+ T cells from DO11.10 mice on a Rag−/− background, and even though positive controls proliferate vigorously as anticipated, the addition of cilostamide does not result in the development of Foxp3+ cells. It is known that TCR transgenic Rag−/− mice frequently have low numbers of endogenous nTregs, probably because of a failure of positive selection (39, 40), and this is also a feature of DO11.10 TCR transgenics (44). We have previously shown that the enrichment of alloreactive Foxp3+ T cells in vitro depends, at least in part, on the presence of preexisting nTregs in the input T cell population (10), perhaps representing an in vitro correlate of in vivo infectious tolerance (45); these data suggest that our inability to drive the enrichment of Foxp3+ DO11.10 T cells may be a direct reflection of this nTreg deficit. Although the contribution that preexisting nTregs make to this process is unknown, one intriguing possibility is that the conversion of non-Tregs depends on direct delivery of cAMP through gap junctions as has been shown elsewhere (46).

A growing understanding of how Foxp3 influences regulatory cell differentiation combined with the wide use of PDE inhibitors to treat clinical conditions such as claudication, pulmonary hypertension, and asthma highlights the relevance of targeting PDEs as a potential route toward Treg cellular therapy. Indeed, although the focus of this investigation was the inhibition of PDE3, additional PDE isoforms exist, some of which show differential expression in specific cell populations (11). For example, although both PDE3 and PDE4 are expressed in human and mouse T cells, human T cells also express PDE7 (47, 48), and the availability of selective inhibitors for defined isoforms suggests further options for manipulating T cell responses. Inhibition of PDEs may contribute to the development of a number of therapeutic approaches for the generation of Tregs for use in autoimmune disease and transplantation.

Materials and Methods


CBA.Ca (CBA, H-2k), C57BL/6 (B6, H-2b), and BM3 (H-2k) mice were bred and maintained at our institution. CBA recombination–activating gene 1 knockout (CBA-Rag−/−) H-2k mice were provided by D. Kioussis (National Institute for Medical Research, London, UK). CBA transgenic mice expressing human CD52 were a gift from H. Waldmann (Sir William Dunn School of Pathology, Oxford, UK). BALB/c recombination–activating gene 2 knockout and common γ chain knockout (BALB/c Rag2−/−/cγ−/−) mice were obtained from Charles River Laboratories. All mice were maintained in the Biomedical Service Unit, John Radcliffe Hospital.

Reagents and monoclonal antibodies

The antibodies to Foxp3, FJK-16s, and PCH101 were obtained from eBioscience. Other conjugated antibodies were from BD Pharmingen. Cilostamide was from Sigma-Aldrich.

Cell purification

CD25CD4+ T cells and CD14+ monocytes were isolated with CD4, CD25, and CD14 MicroBeads (Miltenyi). CD45ROCD25CD4+ T cells were isolated with DynaBeads (Invitrogen). On reanalysis, all populations were >95% pure.

In vitro generation of DCs

Mouse BM-derived DCs were generated with recombinant mouse GM-CSF (rmGM-CSF) and recombinant human TGF-β1 (rhTGF-β1) as described (10). Human monocyte-derived DCs were generated from peripheral blood CD14+ cells by culture for 6 days with rhGM-CSF, rhIL-4, and rhTGF-β1.

Ex vivo conditioning protocol

Purified naïve CD25CD4+ T cells or CD45ROCD25CD4+ T cells were cocultured with allogeneic DCs (5 × 105 T cells plus 5 × 104 DCs per 2-ml well) in complete medium in the presence of additions specified. On day 7, the T cells were restimulated under identical conditions and harvested 7 days later.

Adoptive transfer and skin transplantation

CBA-Rag−/− mice were reconstituted intravenously with 1 × 105 CD25CD4+ cells from naïve CBA or BM3 CD8+ cells with or without 2 × 105 ex vivo–conditioned cells and transplanted the next day with full-thickness H-2b skin grafts.

CpG methylation analysis

DNA methylation in the TSDR of the Foxp3 5′ untranslated region (UTR) was assessed with bisulfite DNA pyrosequencing. A nested PCR approach was applied with two overlapping larger [<400 base pairs (bp)] and six smaller (>200 bp) amplicons for PCR1. Pyrosequencing amplicons covered all 14 CpG sites in the TSDR for PCR2. Genomic DNA (20 ng) was bisulfate-converted with the EZ DNA Methylation Kit (Zymo Research). DNA thermocycler settings were as follows: 95°C for 30 s, 50°C for 15 min for 20 cycles. PCR1 on two oligo sets was performed according to standard protocols (49) with minor changes to the cycling conditions: 95°C for 5 min (94°C for 30 s, 57°C for 1 min, 72°C for 1 min) for 40 cycles, 72°C for 5 min.

PCR1 primers


Six oligo sets were designed for pyrosequencing with Biotage primer design software in conjunction with published protocols for DNA methylation pyrosequencing assays (49). A universal M13 tag was added to the opposite oligo of the sequencing primer, as described (50). Two primer sets were designed for FoxP3_1 and four for FoxP3_2. Cycling conditions were as follows: 95°C for 5 min (95°C for 30 s, 58°C for 30 s, 72°C for 30 s) for 45 cycles, 72°C for 5 min.

Pyrosequencing primers


Pyrosequencing reactions were performed with the manufacturer’s protocol (PSQ96, Biotage), and CpG methylation was analyzed with Q-CpG (Biotage) software.

Mixed lymphocyte reaction

PBMCs (1 × 105) and syngeneic irradiated CD45ROCD25CD4+ T cells conditioned ex vivo in the absence or presence of cilostamide were cocultured with irradiated allogeneic PBMCs (1 × 105) for 7 days, with 0.5 mCi [3H]thymidine per well added for the last 16 hours. Thymidine incorporation was measured by scintillation counting, and results were expressed as the mean of triplicate wells ± SD.

Transplant vasculopathy in human vessels

Human internal mammary artery side branches retrieved with informed consent (REC Ref No. 07/H0605/130) were transplanted into BALB/c Rag2−/−/cγ−/− mice as aortic interposition grafts. The day after transplantation, mice were reconstituted intraperitoneally with human PBMCs with or without PDEi-conditioned T cells. Grafts were harvested on postoperative day 30. Data were analyzed only from mice displaying >1% human CD45+ cells as judged by phenotypic analysis of the spleen. Intimal expansion on elastic van Gieson (EvG)–stained sections was measured with Adobe Photoshop and calculated as follows: (Arealamina − Areaintima)/Arealamina) × 100.

Statistical analysis

Two-tailed comparisons were made with the Mann-Whitney test except for transplant outcome data (log-rank test) and ELISpot analyses (unpaired t test).


  • Citation: G. Feng, S. N. Nadig, L. Bäckdahl, S. Beck, R. S. Francis, A. Schiopu, A. Whatcott, K. J. Wood, A. Bushell, Functional Regulatory T Cells Produced by Inhibiting Cyclic Nucleotide Phosphodiesterase Type 3 Prevent Allograft Rejection. Sci. Transl. Med. 3, 83ra40 (2011).

References and Notes

  1. Acknowledgments: We are indebted to D. Taggart and the cardiac surgery team at the Nuffield Department of Surgical Sciences for providing human vessels from patients undergoing coronary artery bypass surgery; D. Kioussis for providing CBA-Rag−/−mice; H. Waldman for providing hCD25 transgenic mice; the staff of the Biomedical Service Unit, John Radcliffe Hospital, for animal care; N. Jones for advice and discussion of the data obtained using BM3 TCR transgenic mice; and J. Huehn for critical appraisal of the manuscript. Funding: This work was supported by The Wellcome Trust, the European Union Framework 6 Integrated Project, RISET, and the British Heart Foundation (PG/06/050). G.F. received a Dorothy Hodgkin Post-Graduate Award and support from the China-Oxford Scholarship Fund. R.S.F. received a Kidney Research UK Training Fellowship. S.N.N. received an American Society of Transplantation Research Fellowship Award. S.B. was funded by the Wellcome Trust (084071) and a Royal Society Wolfson Research Merit Award. A.S. is supported by the Swedish Heart and Lung Foundation and the Swedish Research Council. K.J.W. holds a Royal Society Wolfson Research Merit Award. Author contributions: G.F. designed and conducted the experiments, analyzed the data, and contributed original ideas. R.S.F. designed and conducted key in vivo experiments. S.N.N. and A.S. performed and analyzed the human arterial transplants. K.J.W. obtained funding and provided important ideas, comments, and leadership. A.W. performed key repeat experiments. S.B. and L.B. designed and performed the bisulfite sequencing experiments and interpreted the data. A.B. designed the experiments, interpreted the data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article