Common Gamma Chain Cytokines Promote Rapid In Vitro Expansion of Allo-Specific Human CD8+ Suppressor T Cells

Human CD8+ regulatory T cells, particularly the CD8+CD28− T suppressor cells, have emerged as an important modulator of alloimmunity. Understanding the conditions under which these cells are induced and/or expanded would greatly facilitate their application in future clinical trials. In the current study, we develop a novel strategy that combines common gamma chain (γc) cytokines IL-2, IL-7 and IL-15 and donor antigen presenting cells (APCs) to stimulate full HLA-mismatched allogeneic human CD8+ T cells which results in significant expansions of donor-specific CD8+CD28− T suppressor cells in vitro. The expanded CD8+CD28− T cells exhibit increased expressions of CTLA-4, FoxP3, and CD25, while down-regulate expressions of CD56, CD57, CD127, and perforin. Furthermore, these cells suppress proliferation of CD4+ T cells in a contact-dependent and cytokine-independent manner. Interestingly, the specificity of suppression is restricted by the donor HLA class I antigens but promiscuous to HLA class II antigens, providing a potential mechanism for linked suppression. Taken together, our results demonstrate a novel role for common γc cytokines in combination with donor APCs in the expansion of donor-specific CD8+CD28− T suppressor cells, and represent a robust strategy for in vitro generation of such cells for adoptive cellular immunotherapy in transplantation.


Introduction
Allogeneic organ transplantation has emerged as the current best therapeutic option for selected patients with end stage organ failure. Tolerance induction is highly desirable in this population, because long-term immunosuppressive therapy is associated with significant risks for opportunistic infections, degenerative and metabolic diseases, as well as malignancies. Regulatory T cell (Treg) based therapy has emerged as a promising means for immunomodulation to achieve transplant tolerance. Great interest has been focused on CD4 + Tregs [1], and cell therapies using adoptive transfer of in vitro generated CD4 + Treg cells have demonstrated promising efficacy for immunomodulation in animal models of allogeneic transplantation as well as in clinical trials of human allogeneic bone marrow transplantation for the control of graft versus host disease [2,3,4]. However, evidence also suggests that CD8 + Tregs may play an important regulatory role in transplant tolerance [5,6,7,8], in addition to possible immunomodulatory roles in autoimmune disorders [9,10], cancers [11] and aging [12]. Several subsets of CD8 + Tregs have been observed. Natural CD8 + Tregs have been reported to be CD8 + CD25 + , CD8 + CD122 + , or CD8 + CXCR3 + in different systems [13,14,15]. Induced CD8 + Tregs have also been reported to bear various phenotypic characteristics, such as CD28 2 , CD56 + , CD57 + , CTLA4 + , CD103 + , CD25 + Foxp3 + or LAG3 + CCL4 + [5,16,17,18,19]. These natural or induced CD8 + Tregs exert their suppressive function by cell-cell contact [5] or by producing soluble cytokines [15,17], and in some circumstances by inducing tolerogenicity in antigen presenting cells (APCs) [20].
Among the various CD8 + Tregs, CD8 + CD28 2 T suppressor cells have emerged as an important modulator of alloimmunity. Human allo-specific CD8 + CD28 2 cells have been shown to emerge after repeated stimulation with allogeneic APCs with supplemental recombinant human IL-2 in vitro [5]. These cells have subsequently been termed ''Ts cells''. Similarly, autoantigen-specific CD8 + Tregs including CD8 + CD28 2 cells have been reported to be induced in vitro via stimulation of peripheral T cells obtained from patients with systemic lupus erythematosus (SLE) in the presence of a combination of common gamma chain (cc) cytokines IL-2, IL-7 and IL-15 [21]. These conditions under which CD8 + CD28 2 cells are induced are distinct from those in which CD4 + Tregs are induced, in that the latter frequently require the presence of tolerance promoting agents such as TGF-b, IL-10, or rapamycin in the culture [22,23,24,25]. While the mechanisms by which CD8 + CD28 2 cells are induced under such apparently nontolerance promoting conditions remain unclear, the induced CD8 + CD28 2 cells nevertheless exhibit potent antigen-specific suppressive capacities [5,21]. The cc cytokines have also been reported to induce stable loss of CD28 expression in actively dividing CD8 + CD28 + T cells [26,27], thereby promoting the generation of a preponderance of CD8 + CD28 2 T cells. We therefore questioned if cc cytokines in combination with donor APCs may be used to expand donor-specific CD8 + CD28 2 T suppressor cells in large numbers in vitro, and if so, what the mechanisms of donor-specific suppression are.
In this report, we utilized APCs and T cells variably matched for human leukocyte antigen (HLA) class I or class II to study the ability and mechanisms of suppression of the CD8 + CD28 2 T cells generated in vitro by this novel approach of combining allogeneic APCs plus cc cytokines. Our findings indicate that large numbers of CD8 + CD28 2 T cells with potent donor-specific suppressive capacity can be effectively generated in vitro using this approach. Furthermore, their suppressive capacity is contactdependent and is restricted only by the donor HLA class I antigens but promiscuous to HLA class II antigens. To our knowledge, this is the first report demonstrating a role of combination cc cytokines in the in vitro generation of large numbers of human donor-specific CD8 + CD28 2 suppressor cells. Findings from this study may have important implications for designing future Treg based therapies for prevention of rejection in human organ or tissue transplantation.

Results
Donor APCs plus cc cytokines induce rapid expansion of CD8 + CD28 2 T cells in culture The CD28 2 cell population accounts for a small fraction of freshly isolated human CD8 + T cells from PBMCs (5.5%, Figure 1A, top panel, Day 0). Fresh CD8 + T cells were stimulated with HLA-A, -B and -DR complete-mismatched APCs in the presence of IL-2 alone or a combination of cc cytokines IL-2, IL-7 and IL-15 as described in Materials and Methods [21]. As shown in Figure 1A bottom panels (Day 6, Day 9), culture in the presence of IL-2 alone allowed only minimal increase of the CD28 2 population over time. In contrast, culture in the presence of a combination of cc cytokines IL-2, IL-7 and IL-15 significantly increased this population from 5.5% of all CD8 + cells on day 0 to 45.7% (51.5+/28.5%) on day 6, and 55.5% (58.5+/27.2%) on day 9 ( Figure 1A), likely representing down-regulation of cell surface CD28 during the in vitro cell activation as previously described [26,27]. The total number of CD8 + CD28 2 recovered at the end of the 9 day culture increased by ,26.7+/27.5 fold ( Figure 1B). The most dramatic increase of this population was observed when IL-15 was added to the culture, with moderate effect observed with IL-7 on the expansion of this population (data not shown).
In vitro expanded CD8 + CD28 2 T cells suppress CD4 + T cells proliferation in a donor-specific manner CD8 + CD28 2 T cells were generated from individual A by culturing with allogeneic APCs from individual B (designated as B-APC ) in the presence of the cc cytokine combination as described above. After 9 days of culture, the resulting CD8 + CD28 2 T cells were isolated (purity.95%) and used as suppressors (S) to test for their suppressive capacity in mixed lymphocyte reactions (MLRs) using CD4 + T cells from individual A (designated as A-T4) as responders (R). Suppressor to responder ratios (S:R) of 0.5:1, 0.1:1 and 0.02:1 were tested (with the cell number of ''R'' kept constant).
As shown in Figure 2A, when APCs from the original priming stimulator (B-APC ) were used as stimulators of the MLRs, proliferation of A-T4 (measured by CFSE dilution) was markedly suppressed in a dose-dependent manner by the in vitro expanded CD8 + CD28 2 T cells. Suppression was also confirmed by using 3 H-thymidine incorporation as a read-out ( Figure 2B). In contrast, when APCs from an HLA-A, -B and -DR mismatched indifferent stimulator (I-APC ) were used as stimulators of the MLRs, CD8 + CD28 2 cells exhibited only marginal suppression at the highest S:R ratio tested (0.5:1, shown in Fig. 2A, middle panels). No suppression could be demonstrated when anti-CD3/CD28 coated Dynabeads were used for stimulating the CD4 + T cells (Figure 2A, bottom panels). Taken together, our results indicate that the CD8 + CD28 2 T cell generated from donor APCs plus cc cytokine-driven in vitro expansion suppress CD4 + T cells proliferation only in a donor-specific manner.
Suppression by the in vitro expanded CD8 + CD28 2 T cells is restricted by donor HLA class I antigens but promiscuous to HLA class II antigens To determine whether the donor specificity observed in the above suppression of proliferation was restricted via recognition of HLA class I antigens only, we examined the ability of the CD8 + CD28 2 T cells generated from culturing with B-APC to suppress proliferation stimulated by APCs from a third-party donor (C-APC ) who shared identical HLA class I but not class II antigens with individual B. The ability of suppressing C-APC driven proliferation was compared with that of suppressing I-APC driven proliferation (I-APC were from an indifferent individual who was mismatched with individual A for both HLA class I and class II antigens). Table 1 provides one example of the HLA class I and class II antigen typing of individuals A, B, C, and I used for one of three such experiments. As shown representatively in Figure 3, proliferation of A-T4 was profoundly suppressed when B-APC was used as the stimulator as expected. Interestingly, a significant suppression was also observed, albeit at a slightly reduced level, when C-APC was used as the stimulator. In contrast, suppression was almost completely lost when I-APC was used as the stimulator. These data support the hypothesis that suppression by the in vitro expanded CD8 + CD28 2 T cells is mediated via specific interactions between the T cell receptor (TCR) on the CD8 + cells and donor HLA class I molecules on the APCs, and such interactions can lead to suppression of CD4 + cells with TCRs specific for HLA class II molecules co-expressed on the same interacting APCs (the C-APC in this case) but different from those expressed on the original priming APCs (the B-APC in this case). This finding corroborates with the previous observation that CD8 suppressor cells are capable of exerting linked suppression via interactions with APCs [28].

Suppression by the in vitro expanded CD8 + CD28 2 T cells is contact dependent but IFN-c or TGF-b independent
The role of cell-cell contact in suppression was examined using transwell assays as shown in Figure 4A. The lower chamber was plated with CFSE-labeled responder cells (R) and stimulator cells (B-APC ), and the in vitro expanded CD8 + CD28 2 T cells (S) were added either in the lower chamber to allow cell-cell contact or in the upper chamber to prevent cell-cell contact. As shown in Figure 4A, suppression of proliferation by CD8 + CD28 2 T cells was abolished when these cells were plated in the upper chamber of the transwells, indicating that suppression by these cells requires cell-cell contact between the CD8 + CD28 2 T cells, stimulator APCs and/or the responder CD4 + T cells. This finding is consistent with the characteristic donor HLA class I restriction of the suppressive activity of the in vitro expanded CD8 + CD28 2 T cells observed in Figure 3.
Several reports have implicated TGF-b [13,29,30] and IFN-c [8,31,32] in the mechanisms of regulation by CD8 + suppressor cells. To test for a potential role of TGF-b or IFN-c in the  suppressive capacity of the in vitro expanded CD8 + CD28 2 T cells in our system, anti-human IFN-c or anti-human TGF-b antibody was added to the suppression assays described above (baseline suppression without the blocking antibodies is shown in Figure 4B). As shown in Figure 4C and Figure 4D, neither anti-IFN-c nor anti-TGF-b antibodies at increasing concentrations affected suppression by CD8 + CD28 2 T cells.

In vitro expanded CD8 + CD28 2 T cells do not exhibit cytotoxicity
To test whether cytotoxicity towards the stimulating APCs contributes to the suppression of donor-specific proliferation observed above, we assessed the cytotoxicity of CD8 + CD28 2 T cells using a CFSE-based assay [33]. We designed this assay to specifically determine the ability of the CD8 + CD28 2 T cells to lyse allogeneic donor APCs used as stimulators in the above MLRs. In this assay, ''effector cells'' were the CD8 + CD28 2 T cells generated by the donor APC plus cc cytokine-driven in vitro cultures as above. The CD8 + CD28 + T cells generated by the same cultures were used as control effector cells because of their known cytotoxicity [5,34,35]. Target cells were either the original priming APCs which were labeled with a high concentration of CFSE (B-APC -CFSE high ), or control APCs from an HLA-A, -B, and -DR mismatched indifferent individual which were labeled with a low concentration of CFSE (I-APC -CFSE low ). A 1:1 mixture of I-APC -CFSE low and B-APC -CFSE high cells were plated either by themselves or with the same number of the putative (CD8 + CD28 2 ) or control (CD8 + CD28 + ) effector cells. At 24, 72, and 120 hr of culture, cells were collected and analyzed by FACS to enumerate CFSE low and CFSE high cells. The ratio of cell numbers of I-APC -CFSE low over B-APC -CFSE high (shown as ''R'' in Figure 5) over time provided a measure of specific cytolytic effect of the putative effector cells on these two target populations. As shown in Figure 5 left panels, minimal spontaneous lysis was observed as the ratio R remained stable over time in the absence of any effector cells. When the CD8 + CD28 + T cells were added as effector cells ( Figure 5, middle panels), R increased to 1.5 at 72 hrs and 2.9 at 120 hrs, demonstrating that B-APC -CFSE high cells were specifically killed by the CD8 + CD28 + T cells. In contrast, when the CD8 + CD28 2 T cells were added as effector cells, R remained unchanged over time, indicating that the CD8 + CD28 2 T cells exhibit no specific cytolysis towards the B-APC cells ( Figure 5, right panels).   Figure 3. Phenotypic characteristics of the in vitro expanded CD8 + CD28 2 T cells Phenotypic characteristics of CD8 + CD28 2 T cells were examined before and after donor APCs plus cc cytokine-driven in vitro expansion to determine if reported markers of CD8 Tregs such as CTLA-4, FoxP3, CD25, CD56 and CD57 [5,16,17,18,19] were also expressed by the CD8 + CD28 2 T cells generated by our protocol. As shown in Figure 6A, among freshly isolated CD8 + T cells prior to in vitro stimulation, CD8 + CD28 2 T cells comprised a small fraction. At baseline, they expressed CD56 (35.0%), CD57 (76.5%), CD62L (28.0%), CD127 (13.6%), and perforin (36.6%), but exhibited minimal expressions of CTLA-4, FoxP3, CD25 ( Figure 6A). Following donor APCs plus cc cytokine-driven expansion, CD28 2 cells became the predominant population among CD8 + cells, and they significantly up-regulated expressions of CTLA-4 (78.7%), FoxP3 (16.0%), and CD25 (37.5%) compared with basline, while down-regulated expressions of CD56, CD57, CD62L, CD127, and perforin ( Figure 6B). While the exact percentages of these markers varied between the CD8 + CD28 2 cells from different individuals (data not shown), the trend of up-or down-regulations before and after stimulation in cultures remained the same (1 of 3 representative experiments is shown in Figure 6).

Discussion
In this study, we reported the novel finding that large numbers of donor-specific human CD8 + CD28 2 T suppressor cells can be  . Suppression by the in vitro expanded CD8 + CD28 2 T cells is contact dependent but IFN-c or TGF-b independent. Suppression assays were set up as described in Figure 2 at an S:R ratio of 0.5:1. Proliferation of the CD4 + cells was measured by CFSE dilution. A, Transwell assays: the lower chambers of 24-well transwell plates were plated with CFSE-labeled responder cells (3610 5 of R) and stimulator cells (3610 5 of B-APC ), and the in vitro expanded CD8 + CD28 2 T cells were added either in the lower chamber to allow cell-cell contact or in the upper chamber to prevent cellcell contact. Data shown are representative of three independent experiments. B, Control suppression without any blocking antibodies. C, Suppression in the presence of anti-human IFN-c antibody or its isotype control antibody at concentrations of 1 mg/ml, 10 mg/ml and 100 mg/ml. D, Suppression in the presence of anti-human TGF-b antibody or its isotype control antibody at concentrations of 1 mg/ml, 10 mg/ml and 100 mg/ml. Data shown for B and C are representative of three independent experiments. Abbreviations used for A-D: ''R'': CD4 + responder cells from donor A; ''B-APC '': APC stimulators from donor B; ''S'': the in vitro expanded CD8 + CD28 2 T cells. doi:10.1371/journal.pone.0028948.g004 generated in vitro using a combination of donor APCs and cc cytokines stimulated proliferation. Common cc cytokines IL-2, IL-7 and IL-15 play critical roles in T cell homeostasis [36]. Recent reports have demonstrated that IL-15 can promote the generation and proliferation of CD8 + memory T cells when stimulated by anti-CD3 and anti-CD28 antibodies [12,26]. In contrast, a combination of cc cytokines including IL-2, IL-7 and IL-15 has been used to expand CD8 + Foxp3 + Treg in vitro from autoimmune patients with SLE who have undergone autologous bone marrow transplant [21]. Human allo-specific CD8 + CD28 2 suppressor T cells have been reported to be generated by two rounds of stimulation by allogeneic APCs in which IL-2 was the only cytokine added during the last four days of cultures [20]. These cells have been termed as Ts cells. Here, we demonstrated that the combination of IL-2, IL-7 and IL-15, compared with IL-2 alone, was superior in inducing significant expansions of allo-specific CD8 + CD28 2 cells in cultures activated by donor APC stimulation ( Figure 1A, 1B), and that the expanded CD8 + CD28 2 cells were highly functionally suppressive. The role of each individual cytokine in the expansion versus inducing the ''suppressive'' characteristics of this cell population remains to be further elucidated.
The phenotypic and functional characteristics of the CD8 + CD28 2 T cells generated in our protocol with a combination of cc cytokines closely resemble those of previously described Ts cells [19,20]. They both up-regulate CTLA-4, FoxP3 and CD25, and have variable degrees of down-regulation of CD62L during expansion; possess minimal cytotoxicity; and exhibit suppressive capacity in a contact dependent fashion. We further demonstrated that our in vitro expanded CD8 + CD28 2 T cells exert linked suppression of CD4 T cells of a different HLA class II specificity as long as the HLA class II molecules are co-expressed with the priming HLA class I molecules on the same APCs. This has also been previously indirectly demonstrated as a characteristic of Ts cells [5]. The ability to generate large numbers of functionally suppressive donor-specific CD8 + CD28 2 T cells that closely mimic Ts cells with this novel approach of using common cc cytokines thus has significant potential for facilitating future clinical use of Treg cell-based immunotherapy for tolerance induction in transplantation.
Several mechanisms of suppression by CD8 suppressor cells have been described. These include direct cytotoxicity [37], secretion of inhibitory cytokines [32,38], down-regulation of costimulatory molecules [5,19], and generation of tolerogenic APCs [20]. In our suppression assays, the CD8 + CD28 2 T cells generated by donor APCs plus cc cytokines suppress proliferation of CD4 + T cells in a donor-specific and dose-dependent manner. Interestingly, distinct from CD4 Treg cells, the CD8 + CD28 2 T cells do not have the ability to suppress proliferation driven by non-specific stimulation by anti-CD3 and anti-CD28 coated Dynabeads. This finding indicates the importance of the interacting APCs in mediating the suppression by the CD8 + CD28 2 T cells. Further studies characterizing phenotypic and functional changes of the interacting APCs will provide definitive evidence for this hypothesis. Furthermore, the transwell assay showed that separation of the CD8 + CD28 2 T cells alone or CD8 + CD28 2 T cells together with stimulator APCs from the responder CD4 + T cells abolished the ability of the CD8 + CD28 2 T cells to suppress. These data further indicate that suppression of CD8 + CD28 2 T cells is mediated through cell-cell contact in a three-cell system, i.e. not only between the CD8 + CD28 2 T cells and the APCs, but also between the APCs (or the CD8 + CD28 2 T cells) and the responding CD4 + T cells.
The finding that suppression by the CD8 + CD28 2 T cells is restricted by HLA class I but is promiscuous towards HLA class II specificities with which the responding CD4 TCRs interact suggests a potential mechanism for linked suppression. Linked suppression classically depicts a situation in which co-expression of an indifferent antigen and a tolerant antigen on the same graft allows acceptance of the indifferent antigen in a host previously tolerized to the tolerant antigen [39]. Dendritic cells are thought to play a critical role in the process of linked suppression [39]. In our studies, C-APC cells co-express both matched HLA class I antigens with the priming donor B-APC and mismatched HLA class II antigens with the priming donor B-APC . Therefore linked suppression by the CD8 + CD28 2 T cells occurs via C-APC cells, and proliferation of CD4 + responder to the mismatched class II antigens is suppressed. These findings have significant implications for designing and understanding the therapeutic value of this CD8 + suppressor population.
The role of cytotoxicity in mechanisms of regulation by CD8 + suppressor cells has been debated. Suppressive CD8 + CD28 2 Ts cells generated by repeated stimulations with alloantigens have been shown to have no cytotoxicity towards the stimulating allogeneic APCs [5]. In contrast, Zheng et al [37] reported that human CD8 high T cells induced by allogeneic CD40-activated B cells have regulatory property that is at least in part mediated through cytotoxicity towards the allogeneic priming PBMC. In our in vitro CFSE-based cytotoxicity assay ( Figure 5), it appears that the CD8 + CD28 2 T cells generated by our protocol do not possess strong cytotoxicity towards their priming APCs. This is consistent with the down-regulation of perforin in these cells after culturing ( Figure 6). It should be noted, however, that cytotoxicity and other possible regulatory mechanisms, such as cytokine production or those mediated through high levels of expression of FoxP3 and CTLA-4, need not be mutually exclusive.
Both cytokine TGF-b [13,29,30] and IFN-c [8,31,32] have been implicated in the regulatory mechanisms of CD8 + suppressor cells in different studies. In our system, neither anti-IFN-c nor anti-TGF-b antibody reversed the suppressive effect of the expanded CD8 + CD28 2 T cells ( Figure 4C, D). CD56 and CD57 are two markers of natural killer cells previously reported to overexpress on CD8 + CD28 2 Ts cells [19]. However, in our culture system the expression of these markers on the CD8 + CD28 2 cells decreased after culture, suggesting that they may not play a major role in the functionality of expanded CD8 + CD28 2 cells in our system.
Two proteins, FoxP3 and CTLA-4, are discernibly up-regulated in the expanded CD8 + CD28 2 T cells. FoxP3 has been described as an important marker for CD8 + suppressor cells in humans [18,20,40,41]. However, unlike in mice, in humans non-regulatory T cells including effector cells can also up-regulate FoxP3 expression transiently upon activation [42,43]. Moreover, there are also CD8 + suppressor cells that are not characterized with FoxP3 expression [16,32,38]. Whether or not FoxP3 up-regulation in our in vitro expanded CD8 + CD28 2 T cells contribute to their suppressive function remains to be elucidated. CTLA-4 is another marker described in CD8 + suppressor cells [18,20,40,41]. CTLA-4 is a homologue of CD28, and is known as a negative costimulatory receptor which upon signaling from CD80 and CD86 delivers inhibitory signals to T cells to down-regulate T cell activation [44,45]. Recent data also suggest that signaling between CTLA-4 and CD80/CD86 can be bi-directional. CTLA-4 expressed on surface of CD4 + Tregs has been shown to interact with CD80/CD86 expressed on APCs and subsequently downregulate their expression, possibly via a process of transendocytosis [46], thereby decreasing the ability of APCs to stimulate T cells through CD28 [47,48]. CTLA-4 has also been reported to up-regulate indolamine 2,3 dioxygenase (IDO) in some subsets of APCs as yet another mechanism for regulating T cell responses [49]. Given the profound up-regulation of this molecule (from 0.2% to 78.7%, Figure 6) and simultaneous down-regulation of CD28 in our CD8 + CD28 2 T cells, CTLA-4 mediated negative signaling to the interacting APCs is a highly attractive potential mechanism of suppression by this cell population.
While robust suppression is observed from our in vitro expanded CD8 + CD28 2 T cells, an important issue which remains to be resolved is how these cells would behave in vivo upon adoptive transfer. Significant plasticity has been described of CD4 + CD25 + FoxP3 + T cells. Studies of in vivo viability, stability and functionality of the expanded human CD8 + CD28 2 T cells in our protocol will necessitate the use of functional humanized mouse models of allogeneic immune responses. Such studies are critical before this approach can be safely translated into clinical applications and are currently being actively pursued in our laboratory.
In conclusion, our findings show that donor APCs plus cc cytokines can be harnessed for efficient in vitro expansion of human donor-specific CD8 + CD28 2 T suppressor cells in a simple and robust culture system. This approach may hold promise for clinical application of using regulatory T cell-based immunotherapy for tolerance induction in transplantation in the future.

Ethics Statement
This study was approved by the Northwestern Institutional Review Board and written consent was obtained from the participants.

Human subjects and HLA typing
Peripheral blood mononuclear cells (PBMCs) were obtained from healthy volunteers. HLA typing was performed by the Northwestern Histocompatibility Laboratory using molecular methods. Donors for T cells and APCs were selected for their HLA-A, B and DR compatibility or incompatibility based on the specific requirements of individual experiments.

Isolation of PBMCs and cell subsets
PBMCs were isolated from fresh whole blood using Lymphocyte Separation Medium (Mediatech Inc., Manassas, VA) by density gradient centrifugation. CD8 + and CD4 + T cells were purified from PBMCs using CD8 or CD4 isolation kits (Miltenyi Biotec) through positive selection according to the manufacturer's protocols. Purities for both T cells were confirmed by flow cytometry and were routinely .98%. APCs were isolated from PBMCs by depletion of CD3 + cells using CD3 microbeads (Miltenyi Biotech). After culturing of the CD8 + cells for indicated number of days, CD28 + cells were removed by positive selection using human CD28 MicroBead Kit (Miltenyi Biotech, purity.99%), and the flow-through was collected as CD28 2 cells (purity.95%). Again, purity of all isolated cells was confirmed by flow cytometry.
In vitro generation and expansion of CD8 + CD28 2 cells with allogeneic APCs and cc cytokines 2610 6 purified CD8 + T cells from individual A were cultured with 1610 6 HLA-A, -B, -DR mismatched APCs from individual B in 2 ml culture medium (RPMI-1640 supplemented with 15% normal human blood group AB serum, 2 mM L-glutamine, 10 mM HEPES and 16antibiotic-antimycotic solution; all from Gibco-BRL, Gaithersburg, MD) supplemented with IL-2 (20 U/ ml), IL-7 (50 ng/ml) and IL-15 (50 ng/ml) (PeproTech Inc., Rocky Hill, NJ) in 24-well plates at 37uC in 5% CO 2 . Supplemented culture medium was changed on days 4, 7 and 8 (by replacing 1 ml of the culture medium with fresh medium containing cytokines). Cells in each well were split into two wells on day 6 and harvested on day 9, and the CD28 2 population was isolated as described above.
Suppression of donor-specific proliferation by in vitro generated CD8 + CD28 2 cells 5610 4 CFSE labeled purified CD4 + T cells from individual A (A-T4) were used as ''responders (R)'' and stimulated with 5610 4 APCs from the original priming donor (individual B; B-APC ). APCs from an HLA-A,-B,-DR complete-mismatched indifferent donor (I-APC ) or 1610 4 anti-CD3/anti-CD28 Dynabeads (Invitrogen, Carlsbad, CA) were used as third party or non-specific stimulation controls respectively. All cultures were prepared in triplicates and incubated in 96-well U-bottom plates in 37uC 5% CO 2 incubator. The CD8 + CD28 2 cells were added as ''suppressor (S)'' at S:R ratios of 0.5:1, 0.1:1 and 0.02:1 (the cell number of ''R'' was kept constant). CFSE dilution was assessed on day 7 to determine extent of proliferation. In selected experiments, anti-human IFN-c (eBioscience Inc.), anti-human TGF-b antibody (BioXcell, West Lebanon, NH), or respective isotype controls were added to cultures at concentrations of 1 mg/ml, 10 mg/ml or 100 mg/ml. Parallel cultures were analyzed using 3 H-thymidine incorporation. 1 mCi 3 H-thymidine (PerkinElmer, San Jose, CA) was added per well for the last 18 hrs of a 7-day culture, and incorporated radioactivity was measured as counts per minute (CPM) with a Perkin-Elmer scintillation counter.

Transwell experiments
The lower chambers of 24-well transwell plates were plated with either 3610 5 CFSE labeled naïve CD4 + T cells from individual A (A-T4), or with A-T4 and 3610 5 priming APCs from individual B (B-APC ) in the presence or absence of 3610 4 CD8 + CD28 2 cells (total volume 600 ml). The upper chambers were plated with medium only, CD8 + CD28 2 cells only, or CD8 + CD28 2 cells plus priming APCs (B-APC ). Cells collected from the lower chamber after 7 days of culture were assessed by FACS for CFSE dilution.
Cytotoxicity by CD8 + CD28 2 cells CFSE-based cytotoxic assay was set up according to published methods [33] as follows. APCs serving as target cells were labeled with two concentrations of CFSE: high concentration (2.0 mM) for APCs from the priming donor (B-APC -CFSE high ) and low concentration (0.2 mM) for APCs from an HLA-A,-B,-DR complete-mismatched indifferent donor (I-APC -CFSE low ). The in vitro generated CD8 + CD28 2 cells or their CD8 + CD28 + counterpart were used as putative or control effector cells respectively. 5610 4 effector cells were cultured with 5610 4 each of B-APC -CFSE high and I-APC -CFSE low together in 96-well U-bottom plates in triplicate wells. Cultured B-APC -CFSE high and I-APC -CFSE low in the absence of effector cells were used as controls for spontaneous cytolysis. B-APC -CFSE high and I-APC -CFSE low were enumerated by FACS after 0, 24, 72 and 120 hrs of culture. The ratios of cell numbers of I-APC -CFSE low over B-APC -CFSE high in the absence or the presence of effector cells were calculated over time to estimate specific killing of B-APC -CFSE high .

Statistical analysis
Comparisons were made by paired Student's t test. Differences were considered significant if the P value was ,0.05.