Donor-specific CD8⁺Foxp3⁺ T cells protect skin allografts and facilitate induction of conventional CD4⁺Foxp3⁺ regulatory T cells

Abstract

CD4⁺ regulatory T cells play a critical role in tolerance induction in transplantation. CD8⁺ suppressor T cells have also been shown to control alloimmune responses in pre-clinical and clinical models. However, the exact nature of the CD8⁺ suppressor T cells, their induction and mechanism of function in allogeneic transplantation remain elusive. In this study, we show that functionally suppressive, alloantigen-specific CD8⁺Foxp3⁺ T cells can be induced and significantly expanded by stimulating naive CD8⁺ T cells with donor dendritic cells in the presence of IL-2, TGF-β1, and retinoic acid. These CD8⁺Foxp3⁺ T cells express enhanced levels of CTLA-4, CCR4 and CD103, inhibit the up-regulation of co-stimulatory molecules on dendritic cells, and suppress CD4 and CD8 T cell proliferation and cytokine production in a donor-specific and contact-depended manner. Importantly, upon adoptive transfer, the induced CD8⁺Foxp3⁺ T cells protect full MHC-mismatched skin allografts. In vivo, the CD8⁺Foxp3⁺ T cells preferentially traffic to the graft draining lymph node where they induce conventional CD4⁺Foxp3⁺ T cells and concurrently suppress effector T cell expansion. We conclude that donor-specific CD8⁺Foxp3⁺ suppressor T cells can be induced and exploited as an effective form of cell therapy for graft protection in transplantation.

INTRODUCTION

T regulatory cells (Tregs) capable of suppressing alloreactive T effector cells are thought to be a key element to long-term rejection free allograft survival (1). Several Treg populations have been described in both humans and mice, and the use of Tregs has recently emerged as a promising immunomodulating therapy in transplantation (2–7). Of the Treg populations described thus far, the CD4⁺CD25⁺Foxp3⁺ T cells are the most widely studied, while other populations such as CD8⁺ suppressor cells have been less extensively studied (8–10). CD8⁺ suppressor T cells encompass several populations bearing different phenotypes (11, 12), and have been identified both as naturally-occurring and induced populations (5, 7, 13, 14). These various populations of CD8⁺ suppressor T cells have been shown to suppress auto-reactive and alloreactive responses. In murine models, they can be induced by exposure to self-antigens expressed on neonatal parenchymal cells and subsequently contribute to the maintenance of self-tolerance independent of CD4⁺ T cells (15). Likewise, in response to allogeneic stimulations, CD8⁺ suppressive cells have been shown to be induced in mice treated with rapamycin and to home to skin grafts to confer long-term graft protection (13). Other studies demonstrate that CD8⁺ suppressor cells critical for graft protection can be induced in vivo by ICOS-B7h blockade or CD40Ig treatment (10, 16). In humans, CD8⁺ suppressor cells have been identified in recipients of kidney, heart and liver allografts (17–19). Interestingly, CD8 T cells expanded from rejecting human cardiac allografts could specifically inhibit anti-donor immune responses in vitro via a number of mechanisms (7, 18, 20). Collectively, these studies suggest that CD8⁺ suppressor cells may play an important role in the suppression of allograft rejection and induction of transplant tolerance.

In this study, we report that polyclonal naïve CD8⁺ T cells stimulated with allogeneic dendritic cells (DCs) in the presence of IL-2, TGF-β1 and retinoic acid expand robustly and convert to allo-suppressive CD8⁺Foxp3⁺ T cells capable of protecting full MHC mismatched allogeneic skin grafts. We further demonstrate that the CD8⁺Foxp3⁺ T cells act as a strong inducer for CD4⁺Foxp3⁺ cells, providing an important link between the CD8⁺ suppressor cells and the more conventional CD4⁺Foxp3⁺ Tregs.

MATERIALS AND METHODS

Mice

BALB/c (H-2^d), C57BL/6 (H-2^b), SJL (H-2^s), CD45.1, Thy1.1 congenic C57BL/6, C57BL/6.RAG^−/− and C57BL/6.GFP-Foxp3 mice were purchased from Jackson Laboratory. Mice were used according to protocols approved by the ACUC at NU.

Cell purifications and cultures

BALB/c bone marrow dendritic cells (BM-DCs) were generated (21). On day 6, harvested BM-DCs were pre-conditioned with 10 nM rapamycin (Sigma-Aldrich) and 2 ng/ml mouse TGF-β1 (R&D Systems) (22), followed by co-culturing with naive CD8⁺ T cells from B6 mice at a DC to T cell ratio of 1:3 with 2 ng/ml of mTGF-β1, 100 nM of retinoic acid (Sigma-Aldrich) and 1500 units/ml of rmIL-2 (R&D Systems) in RPMI-1640 with 10% FCS. The resulting CD8⁺CD25⁺(Foxp3⁺) T cells were purified by MACS. CD8⁺Foxp3⁺ T cell-APC secondary cultures were setup using splenic APCs from BALB/c mice purified by MACS depletion of Thy1.2⁺ cells. APCs were co-cultured with the induced CD8⁺Foxp3⁺ T cells or CD8⁺Foxp3⁻ T cells at an APC to T cell ratio of 5:1. For in vitro conversion experiments, naïve CD4⁺CD25⁻ T cells from B6 mice were added (5×10⁵) to the CD8⁺Foxp3⁺ T cell-APC secondary co-cultures and analyzed for Foxp3 induction 7 days later. Purity by MACS ranged from 60–75%.

Proliferation assay and cytokine detection

Details are provided in Supplementary Materials.

Skin Transplants and Adoptive Transfer

Details are provided in Supplementary Materials.

Antibodies, FACS analysis, and quantitative RT-PCR

Details are provided in Supplementary Materials.

Statistical Analysis

Statistical significance was determined by Wilcoxon nonparametric tests or by Student’s t-test with significance determined at P < 0.05. Statistical significance of graft survival was determined using a Log-Rank (Mantel-Cox) text. GraphPad PRISM 5 software was used.

RESULTS

Naïve CD8 T cells stimulated with allogeneic DCs and TGF-β1 convert to CD8⁺Foxp3⁺ T cells

We have previously developed an in vitro culture system that effectively differentiates naive CD4⁺ T cells to donor-specific CD4⁺CD25⁺Foxp3⁺ Tregs using allogeneic DCs preconditioned with rapamycin and TGF-β1 (22). We used the same culture system to test whether naive CD8⁺ T cells could also be differentiated to CD8⁺Foxp3⁺ suppressor cells. Naïve CD8⁺ T cells were co-cultured for 5–7 days with pre-conditioned BALB/c DCs in the presence of retinoic acid (100 nM), rmIL-2 (1500 U/ml), and mTGF-β1 (2 ng/ml) (22). Similar to CD4⁺ T cells, naïve CD8 T cells also differentiated into a CD8⁺Foxp3⁺ phenotype in a TGF-β1 dependent fashion (Fig. 1A), and the total number of CD8⁺Foxp3⁺ cells continued to expand over the course of the 7-day co-culture (Fig. 1B).

FIG. 1. Naïve CD8 T cells stimulated with allogeneic DCs and TGF-β1 convert to CD8⁺Foxp3⁺ T cells.

Freshly purified naïve B6 CD8 T cells (5×10⁴ cells/well) were cultured with rapamycin-conditioned BALB/c bone marrow derived CD11c⁺ DCs at a T:DC ratio of 3:1 in the presence of IL-2 (1500 U/ml) and retinoic acid (100 nM), with or without recombinant TGF-β1 (2 ng/ml), as described in Materials and Methods. Cells were analyzed on Day 5 of culture for A, or Days 3, 5, and 7 for B. A: Expression of Foxp3 and CD25 on CD8 T cells stimulated by allogeneic APCs with or without TGF-β1 for 5 days. Naïve CD8 T cells are shown as a control. B: The total number of CD8⁺Foxp3⁺ T cells (expressed as number per well of a 96-well plate) continues to increase over the 7-day culture period. Data shown in A and B are representative of eight independent experiments.

The induced CD8⁺Foxp3⁺ T cells express enhanced levels of CD103, CTLA-4 and CCR4

We next analyzed in-depth the phenotype of the induced CD8⁺Foxp3⁺ T cells. We first examined CD103, a molecule known to be induced by TGF-β and to be associated with potent suppressive capabilities of human CD8 T cells (7). As shown in Fig. 2, CD103 expression was significantly increased on CD8⁺Foxp3⁺ T cells when compared to that on CD8⁺Foxp3⁻ T cells (MFI=1360±31.9 vs. 6.6±1.4, P<0.0001). CTLA-4 is known to play an important role in Treg function (23). Notably, CD8⁺Foxp3⁺ T cells expressed an increased level of CTLA-4 in comparison to CD8⁺Foxp3⁻ T cells (MFI=357±14.8 vs. 56±4.3, P=0.0064). Another molecule found enriched on Tregs is GITR, a member of the tumor-necrosis family (TNF) receptor superfamily. The expression of GITR on both CD8⁺Foxp3⁺ and CD8⁺Foxp3⁻ T cells was elevated in comparison to naïve CD8 T cells (MFI=123.9±1.3 and 125±2.1 vs. 29±1.9, respectively). Surface latency-associated peptide (LAP) has been found to be expressed on both human and mouse Tregs (24, 25). In our cultures, both CD8⁺Foxp3⁺ and CD8⁺Foxp3⁻ T cells expressed elevated levels of LAP (MFI=23.65±8 and 12.5±1.3, P=NS) compared with naïve cells. In agreement with their putative regulatory phenotype, CD8⁺Foxp3⁺ cells did not express FasL or produce IFN-γ, unlike the CD8⁺Foxp3⁻ cells (FasL MFI=111.7± 38.7 vs. 354±11.5, P=0.073; and IFN-γ MFI=137.7±44.8 vs. 391.5±4.5, P=0.0021). Of note, neither the CD8⁺Foxp3⁺ nor the CD8⁺Foxp3⁻ T cells produced IL-10. Consistent with an activated phenotype following allogeneic stimulation (26), both CD8⁺Foxp3⁺ and CD8⁺Foxp3⁻ T cells exhibited increased expression of CD44 (MFI=6261±362 and 5850±257, respectively, vs. 70±23 on naïve T cells), and decreased expressions of CD62L (MFI = 34 ± 2 and 42 ± 4.6, respectively, vs. 349 ± 122 on naïve CD8 T cells) and CD28 (MFI=145.8±24.6 and 141±22.4, respectively, vs. 350±169 on naïve CD8 T cells). We further analyzed the expression of CD122, the β-chain of IL-2R, because a population of CD122⁺CD8⁺ Tregs has been described to play a role in maintaining immune homeostasis (27). Interestingly, we found the CD122 level to be elevated on CD8⁺Foxp3⁻ cells but not on CD8⁺Foxp3⁺ T cells (MFI=1393±54 vs. 398±11.5, P=0.0031). We also analyzed CCR4 and CD11a, two molecules associated with Treg migration (28, 29). Compared to their Foxp3⁻ counterpart, expression levels of CCR4 and CD11a were higher on CD8⁺Foxp3⁺ cells (CCR4 MFI=370±84.9 vs. 10±3.4, P=0.0133; and CD11a MFI=6945±763 vs. 2896±438, P=0.0441). Taken together, our results indicate that the induced CD8⁺Foxp3⁺ T cells display a CD103⁺CTLA-4⁺CCR4⁺ phenotype distinct from that of CD8⁺Foxp3⁻ T cells.

FIG. 2. Phenotypic characterization of the in vitro induced CD8⁺Foxp3⁺ T cells.

Freshly purified naïve B6 CD8 T cells were cultured with BALB/c BM-DCs as described as in FIG. 1 with or without recombinant TGF-β1 to generate CD8⁺CD25⁺Foxp3⁺ cells or CD8⁺CD25⁺Foxp3⁻ cells. On day 5 of DC-T cell culture, cells were harvested, stained for CD8, CD25, and respective markers shown in each histogram. Histograms were gated on CD8⁺CD25⁺ cells for activated cultures. Naïve CD8⁺ T cells prior to DC-T culture are shown as baseline for comparison. Shaded histogram, isotype control; red line, +TGF-β1 condition; blue line, −TGF-β1 condition; green line, naïve CD8⁺ T cells. Data shown are representative of at least 5 independent experiments. MFI values with standard deviations are listed in the text.

The induced CD8⁺Foxp3⁺ T cells suppress T cell proliferation and interferon-γ production in an alloantigen specific and contact-dependent manner

We next tested the suppressive function of the CD8⁺Foxp3⁺ T cells induced with allogeneic DCs and TGF-β1. As shown in Fig. 3A, when the in vitro induced B6 CD8⁺Foxp3⁺ T cells (putative suppressors, or “S”) were added to B6 naïve CD8 or CD4 responders (“R”) stimulated with BALB/c APCs in MLRs, proliferation of both CD8 and CD4 cells was suppressed. The suppression observed was alloantigen specific, as a decrease in T cell proliferation was not seen when APCs from 3^rd party SJL mice were used as stimulators (Fig. 3A), or when T cells from 3^rd party SJL mice were used as responders (Supplementary Fig. 1). IFN-γ production by responder T cells was also significantly decreased by the presence of the induced CD8⁺Foxp3⁺ T cells in a donor-specific fashion (Fig. 3A). Separating the CD8⁺Foxp3⁺ suppressor cells by transwells completely abolished their ability to suppress proliferation of both CD8 and CD4 responders (Fig. 3B), indicating that suppression exerted by the induced CD8⁺Foxp3⁺ T cells required cell-cell contact. Furthermore, the induced CD8⁺Foxp3⁺ T cells expressed minimal levels of granzyme B and IFN-γ, suggesting that they do not possess cytotoxicity (Fig. 3C). Consistent with this notion, using an in vitro cytotoxicity assay, we demonstrated that the induced CD8⁺Foxp3⁺ T cells indeed did not possess cytotoxicity towards donor splenocytes (Supplementary Fig. 2) confirming that cytotoxicity towards donor APCs was not the underlying reason for the suppression observed in Fig. 3A or 3B. Taken together, these results indicate that the induced CD8⁺Foxp3⁺ T cells possess alloantigen-specific suppressive activity in vitro, and the mechanism through which these cells suppress is contact dependent but cytotoxicity independent.

FIG. 3. Induced CD8⁺Foxp3⁺ T cells suppress T cell proliferation and interferon-γ production in an alloantigen-specific and contact-dependent manner.

B6 CD8⁺Foxp3⁺ T cells were generated using BALB/c BM-DCs as in FIG. 1. A: MLRs were set up using naïve B6 CD8 or CD4 responders (R) stimulated with either BALB/c (donor, black bars) or SJL (3^rd party, gray bars) APCs. The in vitro induced CD8⁺Foxp3⁺ cells (S) were added to indicated wells at a 5:1 R:S ratio. Proliferation was assessed by [³H] thymidine uptake during the last 18 hours of a 5-day MLR. IFN-γ was measured from culture supernatants collected immediately prior to thymidine pulse. B: The lower chamber of a 24-well transwell plate was plated with CFSE-labeled CD8 or CD4 responder cells (“R”, 5×10⁵/well) and BALB/c APCs (1×10⁶/well). The induced CD8⁺Foxp3⁺ T cells (5×10⁵/well) were added either in the lower chamber to allow cell-cell contact or in the upper chamber to prevent cell-cell contact, as indicated. Proliferation was assessed by CFSE dilution. C: Freshly purified naïve B6 CD8 T cells were cultured with BALB/c BM-DCs as described as in FIG. 1 with or without recombinant TGF-β1. On day 5 of DC-T cell culture, cells were harvested and enriched for CD8⁺CD25⁺ fraction: CD8⁺CD25⁺Foxp3⁺ cells were obtained from (+)TGF-β1 cultures; CD8⁺CD25⁺Foxp3⁻ cells were obtained from (−)TGF-β1 cultures. Relative expressions of granzyme B and IFN-γ mRNA were measured by real-time RT-PCR. Values were standardized to that of 18s RNA. Data shown for A to C are representative of at least 3 independent experiments.

CD8⁺Foxp3⁺ T cells inhibit the expression of co-stimulatory molecules on CD8α⁺CD11c⁺ DCs

CD4 Tregs have been shown to affect DC development by preventing maturation and inducing immunosuppressive molecules of the B7-H family (30). CD8⁺CD28⁻ regulatory cells have also been shown to interfere with CD40L signaling on T-helper cells thus preventing up-regulation of co-stimulatory molecules on APCs (31). We next sought to determine if the induced CD8⁺Foxp3⁺ T cells could directly alter the expression of co-stimulatory molecules on APCs. Splenic APCs from BALB/c mice were co-cultured with the induced CD8⁺Foxp3⁺ T cells or the CD8⁺Foxp3⁻ T cells. As shown in Fig. 4A, after just 1 day of co-culture, significant up-regulation of CD40, CD80 and CD86 was observed on the CD8α⁺CD11c⁺ DCs co-cultured with CD8⁺Foxp3⁻ T cells, and this trend continued at day 3 and day 5 of co-cultures (Fig 4B, C). In stark contrast, this up-regulation was completely inhibited by the CD8⁺Foxp3⁺ T cells (Fig. 4A–C). The inhibition of co-stimulatory molecules was not observed on CD8α⁻CD11c⁺ DCs (Fig. 4A–C) or on non-CD11c⁺ APCs (data not shown), demonstrating a specific effect of the induced CD8⁺Foxp3⁺ cells on the CD8α⁺CD11c⁺ DCs. These results highlight one potential mechanism by which the induced CD8⁺Foxp3⁺ cells may regulate effector T cell responses.

FIG. 4. CD8⁺Foxp3⁺ T cells down-regulate co-stimulatory molecules on CD8α⁺CD11c⁺ DCs.

B6 CD8⁺Foxp3⁺ and CD8⁺Foxp3⁻ T cells were generated using BALB/c BM-DCs as in FIG. 1. On day 5, the CD8⁺Foxp3⁺ and CD8⁺Foxp3⁻ T cells were harvested and secondarily cultured with fresh donor (BALB/c) splenic APCs as described in Materials and Methods. The phenotype of the resulting CD11c⁺ DCs was analyzed on days 1 (A), day 3 (B) and day 5 (C) of the secondary cultures. Shaded histogram, isotype control; black line, DCs cultured with CD8⁺Foxp3⁺cells; grey line, DCs cultured with CD8⁺Foxp3⁻ cells. Results are representative of two independent experiments.

CD8⁺Foxp3⁺ T cells induce naïve CD4⁺CD25⁻ T cells to differentiate to CD4⁺CD25⁺Foxp3⁺ Tregs in a TGF-β dependent fashion

Previous studies of CD4⁺ Tregs have demonstrated that their regulatory capacity may in part be due to their ability to further induce suppressive phenotypes (e.g. induction of Foxp3 expression) in naive CD4 T cells, a phenomenon known as infectious tolerance (22, 32, 33). Therefore, we next examined the effect of the induced CD8⁺Foxp3⁺ cells on CD4 helper T cell differentiation. Naive B6 CD4⁺CD25⁻ T cells were cultured with BALB/c APCs for 7 days in the presence or absence of the induced CD8⁺Foxp3⁺ T cells, and evaluated for their differentiation to CD4⁺CD25⁺Foxp3⁺ cells. As shown in Fig. 5A and 5B, in the absence of the induced CD8⁺Foxp3⁺ T cells, very few CD4⁺ cells differentiated into Foxp3⁺ cells. In contrast, in the presence of CD8⁺Foxp3⁺ T cells, a significant number of CD4⁺ cells acquired a Foxp3⁺ phenotype (Fig. 5A for percentages among total CD4⁺ cells, and Fig. 5B for total numbers of Foxp3⁺ CD4 cells per well). It has been previously shown that TGF-β1 is critical for the induction of Foxp3 expression in the periphery (34, 35). Consistent with this notion, anti-TGF-β mAb substantially decreased the induction of Foxp3 among CD4⁺ T cells by the CD8⁺Foxp3⁺ cells (Fig. 5A and 5B).

FIG. 5. CD8⁺Foxp3⁺ T cells induce naïve CD4⁺CD25⁻ T cells to differentiate to CD4⁺CD25⁺Foxp3⁺ Tregs in vitro.

B6 CD8⁺Foxp3⁺ T cells were generated using BALB/c BM-DCs as in FIG. 1. 1×10⁵ naïve B6 CD4⁺CD25⁻ T cells were cultured with 5×10⁵ BALB/c APCs with or without 1×10⁵ of the induced B6 CD8⁺Foxp3⁺ T cells. Anti-TGF-β (20 μg/mL) was added to corresponding wells on day 0 of co-culture. Cells were harvested and analyzed on day 7 of co-culture. A: Percentages of the resulting CD4⁺ T cells expressing Foxp3 at day 5 of co-culture. B: Total number (per well of 96-well plates) of the resulting CD4⁺ T cells expressing Foxp3 at day 5 of co-culture. Data in A are representative of 3 independent experiments. Data in B are average of 3 independent experiments. C: Induced CD8⁺Foxp3⁺ T cells produce TGF-β upon re-stimulation. Left panel: the induced B6 CD8⁺Foxp3⁺ cells (5×10⁵) were re-stimulated with 1×10⁶ BALB/c splenic APCs. Three days post co-culture, supernatants were harvested and active TGF-β in the supernatants was measured by ELISA. Right panel: LAP expression on B6 CD8⁺Foxp3⁺ T cells generated during primary culture as in FIG. 1 (1° Culture CD8⁺Foxp3⁺, grey line) or after the 3 day re-stimulation (2° Culture CD8⁺Foxp3⁺, black line). D: The induction of CD4⁺CD25⁺Foxp3⁺ Treg differentiation by the CD8⁺Foxp3⁺ T cells is contact-dependent. The lower chamber of a 24-well transwell plate was plated with naïve B6 CD4⁺CD25⁻ T cells (5×10⁵/well) and BALB/c APCs (1×10⁶/well). The induced CD8⁺Foxp3⁺ T cells (5×10⁵/well) were added either in the lower chamber to allow cell-cell contact or in the upper chamber to prevent cell-cell contact, as indicated. The resulting CD4⁺ cells on day 5 of culture were analyzed for Foxp3 expression. E: Total number (per well of 24-well plates) of CD4⁺Foxp3⁺ cells induced in the transwell experiments as shown in D. Data in D and E are representative or average of 2 independent experiments. F: Phenotype of the CD4⁺CD25⁺Foxp3⁺ T cells induced in the presence of CD8⁺Foxp3⁺ T cells. Histograms were gated on the induced CD4⁺CD25⁺Foxp3⁺ cells (in the presence of CD8⁺Foxp3⁺ T cells) or the resulting CD4⁺CD25⁻Foxp3⁻ (in the absence of CD8⁺Foxp3⁺ T cells). Shaded histogram, isotype control; black line, CD4⁺CD25⁺Foxp3⁺ cells; grey line, CD4⁺CD25⁻Foxp3⁻ cells. Data shown are representative of 3 independent experiments.

We next determined the source of TGF-β in our co-cultures needed to drive the differentiation of CD4⁺Foxp3⁺ cells. When the induced CD8⁺Foxp3⁺ T cells were re-stimulated with donor APCs, active TGF-β could be detected in culture supernatant (Fig. 5C). Additionally, re-stimulated CD8⁺Foxp3⁺ T cells further enhanced their cell surface LAP expression (Fig. 5C). These results indicate that the CD8⁺Foxp3⁺ T cells themselves are likely to be an important source of TGF-β critical for the induction of CD4⁺Foxp3⁺ cells.

We next examined whether the generation of CD4⁺Foxp3⁺ T cells required cell-cell contact with the CD8⁺Foxp3⁺ T cells. As shown in Fig. 5D and 5E, transwell separation resulted in a complete loss of induction of CD4⁺Foxp3⁺ T cells (Fig. 5D for percentages, and Fig. 5E for total numbers). Finally, the CD4⁺Foxp3⁺ T cells induced by CD8⁺Foxp3⁺ cells expressed high levels of CD103, GITR and LAP, but not IL-17 or IFN-γ (Fig. 5F). Collectively, these results indicate that CD8⁺Foxp3⁺ T cells are capable of inducing antigen-stimulated differentiation of bona fide CD4⁺Foxp3⁺ T cells in a contact and TGF-β-dependent fashion.

CD8⁺Foxp3⁺ T cells protect full MHC-mismatched skin allografts

We next examined the in vivo suppressive function of the induced CD8⁺Foxp3⁺ T cells. To do so, we performed adoptive transfers in B6.RAG^−/− recipients bearing full thickness BALB/c skin grafts. Ten to fourteen days post skin transplant, 2×10⁵ B6 total T cells with or without 1×10⁶ B6 CD8⁺Foxp3⁺ T cells (generated with BALB/c APCs) were adoptively transferred to the B6.RAG^−/− recipients. As shown in Fig. 6A, in the absence of CD8⁺Foxp3⁺ T cells, all skin allografts were rejected between days 13 and 17 (Group 1, graft median survival time (MST) = 14 days) post adoptive transfers. In contrast, co-injection of CD8⁺Foxp3⁺ T cells significantly prolonged the survival of the skin allografts (Group 2, MST = 33 days). Representative skin morphology of a rejected and a protected graft on day 14 post-adoptive transfers is shown in Fig. 6B. Graft protection seen with CD8⁺Foxp3⁺ T cells was donor-specific because these cells (generated with BALB/c APCs) did not protect 3^rd party SJL skin allografts (Group 5). To control for the possibility of in vivo competition due to disparate cell numbers transferred contributing to the observed protection, we performed an additional adoptive transfer group in which 2×10⁵ B6 total T cells were injected together with 1×10⁶ B6 CD8⁺Foxp3⁻ T cells generated without TGF-β1. As seen in Fig. 6A (Group 3), identical numbers of CD8⁺Foxp3⁻ T cells did not provide any protective effect, and in fact had a tendency to accelerate rejection. To examine whether the adoptively transferred CD8⁺Foxp3⁺ T cells themselves could revert to pathogenic cytotoxic T lymphocytes in vivo, we performed a further adoptive transfer group in which only 1×10⁶ B6 CD8⁺Foxp3⁺ T cells were injected. All mice transferred with only CD8⁺Foxp3⁺ T cells kept skin allografts for > 90 days (Group 4), confirming that these cells do not revert to a pathological phenotype in vivo. This is unlikely due to the CD8⁺Foxp3⁺ cells being blasted after the induction culture resulting in unresponsiveness to antigenic stimuli, because identical numbers of CD8⁺Foxp3⁻ T cells were fully capable of timely rejecting the BALB/c skin allograft (Group 6).

FIG. 6. CD8⁺Foxp3⁺ T cells protect full MHC-mismatched skin allografts.

B6.RAG^−/− mice received full-thickness BALB/c skin allografts. 10–14 days post skin-transplant, a total of 2×10⁵ naïve B6 T cells were adoptively transferred (i.v.) into the B6.RAG^−/− recipients with or without 1×10⁶ induced CD8⁺Foxp3^gfp+ T cells generated from (+)TGF-β1 cultures, or with 1×10⁶ CD8⁺Foxp3^gfp− T cells generated from (−)TGF-β1 cultures. Additional groups of control mice received 1×10⁶ induced CD8⁺Foxp3^gfp+ T cells only or 1×10⁶ CD8⁺Foxp3^gfp− T cells only. Third party control was provided by B6.RAG^−/− mice receiving SJL skin grafts. A: Graft survival. Day 0 designates the day of cell adoptive transfer. B: Skin grafts on day 14. Left: rejected skin graft from Group 1. Right: protected skin graft from Group 2.

CD8⁺Foxp3⁺ T cells accumulate in graft draining lymph nodes (DLNs) and induce CD4⁺Foxp3⁺ T cells in vivo

To study the in vivo behavior of the adoptively transferred CD8⁺Foxp3⁺ cells, we reconstituted B6.RAG^−/− recipient mice bearing BALB/c skin grafts with 2×10⁵ CD45.1⁺ B6 total T cells with or without 1×10⁶ B6 CD45.2⁺CD8⁺Foxp3^gfp+ T cells (again generated with BALB/c APCs). We first examined the relative prevalence of the CD45.2⁺CD8⁺Foxp3^gfp+ T cells in various secondary lymphoid organs (SLO) on day 14 and day 28 post adoptive transfer. As shown in Fig. 7A, there was a clear preference of these cells homing to the DLN of the skin graft over the non-draining lymph nodes (NDLNs). This preference was observed at day 14 and persisted at day 28.

FIG. 7. CD8⁺Foxp3⁺ T cells accumulate in graft draining lymph nodes (DLNs) and induce CD4⁺Foxp3⁺ T cells in vivo.

CD45.2.B6.RAG^−/− mice received full-thickness BALB/c skin allografts. 10–14 days post skin-transplant, a total of 2×10⁵ naïve CD45.1⁺ B6 T cells were adoptively transferred (i.v.) into the recipients with or without 1×10⁶ induced CD45.2⁺CD8⁺Foxp3^gfp+ T cells. Mice were sacrificed on day 14, day 28, and day 35 post adoptive transfers, and the spleen, DLN and NDLNs were harvested and analyzed. A: Total number of CD45.2⁺Foxp3^gfp+ in DLN and NDLN of mice 14 and 28 days after adoptive transfer of both naïve CD45.1⁺ B6 T cells and the induced CD45.2⁺CD8⁺Foxp3^gfp+ T cells. B: The CD45.1⁺ effector population diminished in the DLN, but not NDLN, at day 28 post-adoptive transfer in mice that received induced CD45.2⁺CD8⁺Foxp3^gfp+ T cells compared to those that did not. The total number of CD45.1⁺ cells in the DLN on day 14 and day 28 is shown in the bar graph. C and D: The percentages of CD25⁺Foxp3⁺ T cells among total CD45.1⁺CD4⁺ T cells (C) or the total number of CD45.1⁺CD4⁺CD25⁺Foxp3⁺ T cells (D) in the spleen, DLN and NDLN (Day 14) were increased in mice that received the induced CD45.2⁺CD8⁺Foxp3^gfp+ T cells compared to those that did not. E: The total number of CD45.2⁺CD8⁺Foxp3^gfp+ T cells (left panel) or CD45.1⁺CD4⁺Foxp3⁺ cells (right panel) in the graft draining lymph node on day 35 post adoptive transfers with or without the adoptive transfer of CD8⁺Foxp3⁺ cells. At this time point, ~50% of the recipients receiving CD8⁺Foxp3⁺ cells and 100% of the recipients not receiving CD8⁺Foxp3⁺ cells had rejected their skin allografts. All data are representative of 3 independent experiments with at least 3 mice in each group.

We next examined the population size of CD45.1⁺ effector cells in the DLN and NDLN in the presence or absence of co-adoptively transferred CD45.2⁺CD8⁺Foxp3^gfp+ T cells. As shown in Fig. 7B, there was a significant increase of the percentage of CD45.1⁺ effector cells in the DLN over time (from 39% on day 14 to 62% on day 28) in the absence of CD45.2⁺CD8⁺Foxp3^gfp+ T cells. In contrast, in the presence of CD45.2⁺CD8⁺Foxp3^gfp+ T cells, the percentage of CD45.1⁺ effector cells remained un-expanded over time (from 42% on day 14 to 39% on day 28). Enumerating the total numbers of CD45.1⁺ effector cells in the DLN at days 14 and 28 confirmed these findings (Fig. 7B). Furthermore, this decrease of CD45.1⁺ effector cell numbers over time was not observed in NDLN (Fig. 7B) or the spleen (data not shown), demonstrating that suppression of expansion of effector cells by the CD45.2⁺CD8⁺Foxp3^gfp+ T cells occur mainly in the DLN.

We next examined the in vivo effect of the adoptively transferred CD8⁺Foxp3⁺ on the differentiation of the co-injected CD4⁺ T cells. We injected CD25-depleted CD45.1⁺ B6 T cells with or without 1×10⁶ B6 CD45.2⁺CD8⁺Foxp3^gfp+ T cells, and examined the phenotype of CD45.1⁺CD4⁺ T cells in the spleen, the DLN and NDLN 14 days later. Complimentary to our in vitro findings, we found that co-injection of CD45.2⁺CD8⁺Foxp3^gfp+ cells resulted in a significant induction of the CD45.1⁺CD4⁺CD25⁺Foxp3⁺ population (Fig. 7C for percentages among total CD45.1⁺CD4⁺ cells, and Fig. 7D for total numbers of Foxp3⁺ CD4 cells). Interestingly, this induction was seen in all SLO, rather than limited to the graft DLN. These findings indicate that functional suppression of the CD8⁺Foxp3⁺ cells in vivo may indeed be mediated via induction of conventional CD4⁺CD25⁺Foxp3⁺ Tregs. Finally, we examined whether the CD45.2⁺CD8⁺Foxp3^gfp+ T cells or CD45.1⁺CD4⁺Foxp3⁺ cells declined in the graft DLN over time to account for the eventual rejection of the skin allograft observed. As shown in Fig. 7E, at a later time point (day 35 post adoptive transfers) when ~50% of the recipients in Group 2 had rejected their skin allografts, we indeed observed a decline of the number of CD45.2⁺CD8⁺Foxp3^gfp+ T cells in the DLN in comparison to earlier time points (Fig. 7E left panel compared with 7A), although a significant population of CD45.1⁺CD4⁺Foxp3⁺ cells still remained in the DLN at this time point (Fig. 7E right panel compared with 7D). The CD8⁺Foxp3⁺ cells are terminally differentiated and ultimately short-lived. While the CD8⁺Foxp3⁺ cells may exert suppression indirectly through induction of CD4⁺Foxp3⁺ cells, they are also capable of directly suppressing CD8 cells and inhibiting DC maturation as shown earlier. Therefore, a decline of the CD8⁺Foxp3⁺ cells may ultimately lead to a revert of a favorable Treg:Teff ratio, resulting in the eventual graft demise.

DISCUSSION

In this study, we show that naïve CD8 T cells can be induced to become suppressive CD8⁺Foxp3⁺ T cells when cultured with allogeneic DCs in the presence of IL-2, TGF-β1 and retinoic acid. While CD8⁺ suppressor cells have been reported to mediate transplant tolerance in several models (10, 13, 16, 36), we show for the first time that in vitro induced CD8⁺Foxp3⁺ suppressor cells exert donor-specific graft protection upon adoptive transfer in a full MHC-mismatched skin transplant model. Importantly, these CD8⁺Foxp3⁺ cells inhibit up-regulation of co-stimulatory molecules on CD8α⁺CD11c⁺ DCs and facilitate conversion of naïve CD4 T cells to CD25⁺Foxp3⁺ T cells.

Foxp3 has been described to confer functional importance in CD4⁺ Tregs in humans and in mice (37–39). Whether Foxp3 is critical for the suppressive function of CD8⁺Foxp3⁺ T cells here is currently unclear. Three additional markers, CTLA-4, CCR4, and CD103, also distinguish the CD8⁺Foxp3⁺ cells from the CD8⁺Foxp3⁻ cells resulting from our cultures. Interestingly, CCR4 is found on skin homing T cells (40), consistent with the ability of our CD8⁺Foxp3⁺ cells to protect skin allografts, likely via exerting local suppression of allogeneic T cell responses. Ongoing investigation is attempting to provide definitive evidence of functional importance of these markers to the observed in vitro and in vivo suppression.

One mechanism through which Tregs exert their suppressive function is by tolerizing APCs through direct cell-cell interactions (20, 41–43). In a rat model of heart allotransplantation, CD8⁺ Tregs were found to have the capacity to induce inhibitory receptors on DCs and endothelial cells (44, 45). Accordingly, we found that when donor APCs were co-cultured with the CD8⁺Foxp3⁺ T cells, the CD8α⁺CD11c⁺ DCs failed to up-regulate co-stimulatory molecules (Fig. 4). Interestingly, this effect was specific only to the CD8α⁺ DCs, a population that has been shown to produce higher amounts of TGF-β than the CD8α⁻ DCs, and hence is more efficient in inducing antigen-specific CD4 Tregs (46). Therefore, we suggest that the CD8⁺Foxp3⁺ T cells accumulate in recipient lymphoid organs where they tolerize CD8α⁺CD11c⁺ DCs, synergistically induce differentiation of CD4⁺CD25⁺Foxp3⁺ T cells and suppress effector T cell expansion and function.

An important characteristic of the CD8⁺Foxp3⁺ cells in our model is their ability to induce CD4⁺CD25⁺Foxp3⁺ cells from naïve CD4⁺ T cells. CD4⁺ Tregs have been shown in numerous models and in humans to induce regulatory functions in naïve CD4 T cells in a process described as infectious tolerance (32, 47–51). It is thought that infectious tolerance is initiated when Tregs and naïve T cells interact with the same APC (50), and that TGF-β plays an obligatory role (34, 35). The observed effect of the CD8⁺Foxp3⁺ cells on CD8α⁺CD11c⁺ DCs, and the contact- and TGF-β-dependent fashion by which the CD8⁺Foxp3⁺ cells induce CD4⁺Foxp3⁺ T cells are consistent with the infectious tolerance model. However, our findings do not exclude other mechanisms such as catabolism of essential amino acids (35, 52) or the generation of extracellular adenosine (53) as additional mediators by which our CD8⁺Foxp3⁺ mediate such a conversion.

In human organ and bone marrow transplantations, increased graft survival has been observed with a concurrent robust CD4⁺Foxp3⁺ Treg population (54–56). Tolerance strategies have taken the approach of establishing an initial optimal Treg: T effector cells ratio (57–59). Adoptive transfer of CD8⁺Foxp3⁺ cells represents a novel approach for altering this ratio in favor of tolerance, due both to their own suppressive capacity and their ability to further induce CD4⁺Foxp3⁺ cells. Consequently, reproducible prolongation of full-thickness skin graft survival was observed in our study after the adoptive transfer of CD8⁺Foxp3⁺ cells into B6.RAG^−/− recipients. While the ratio of suppressor: effector cells used in this study might be difficult to achieve clinically, we provide a proof of principle that CD8⁺Foxp3⁺ cells as a form of adoptive cell therapy have the potential to be further explored in the context of transplantation for inhibition of graft rejection. The B6.RAG^−/− utilized in the current study allows us to study the behavior of the CD8⁺Foxp3⁺ cells in a lymphopenic environment, which bears some resemblance to that created by lymphocyte depleting induction therapies. However, it has been shown that suppressive cells in a lymphopenic environment may preferentially divide and consequently regulate other cells around them (60). Furthermore, CD1d-recruited NKT cells may have altered ability to suppress immune effector mechanisms in lymphopenic hosts (61). Therefore, future studies using non-lymphopenic hosts will be more ideal to allow better understanding of the behavior of these cells in vivo.

In conclusion, we have demonstrated that naïve CD8 T cells co-cultured with allogeneic DCs can become potent regulatory cells capable of protecting full-thickness skin allografts in a donor-specific fashion. As therapies utilizing ex vivo generated Tregs have already met with some success in clinical trials (62–64), we propose that the CD8⁺Foxp3⁺ cells have the potential to be used as an adoptive transfer cell therapy for a variety of clinical applications including transplantation and autoimmunity.

ABBREVIATIONS

APC

antigen presenting cell

CFSE

carboxyfluorescein succinimidyl ester

CTLA-4

cytotoxic T-lymphocyte antigen 4

DC

dendritic cell

DLN

draining lymph node

Foxp3

forkhead box P3

GFP

green fluorescent protein

ICOS

inducible T-cell costimulator

IFN-γ

interferon gamma

LAP

latency-associated peptide

MFI

mean fluorescent intensity

MLR

mixed lymphocyte reaction

MST

median survival time

NDLN

non-draining lymph node

SLO

secondary lymphoid organs

NKT

natural killer T cell

Tregs

T regulatory cells