Identification of a previously unknown antigen-specific ... - Nature

© 2000 Nature America Inc. • http://medicine.nature.com

ARTICLES

Identification of a previously unknown antigen-specific regulatory T cell and its mechanism of suppression ZHU-XU ZHANG, LIMING YANG, KEVIN J. YOUNG, BARB DUTEMPLE & LI ZHANG


Department of Laboratory of Medicine and Pathobiology, Multi Organ Transplantation Program, Toronto General Hospital Research Institute, UHN, University of Toronto, Toronto, M5G, 2C4, Canada Z.-X.Z. and L.Y. contributed equally to this study. Correspondence should be addressed to L.Z.; email: [email protected]

Despite increasing evidence for the existence of antigen-specific regulatory T cells, the mechanisms underlying suppression remain unclear. In this study we have identified and cloned a novel subset of antigen-specific regulatory T cells and demonstrated that these T cells possess a unique combination of cell surface markers and array of cytokines. The regulatory T cells are able to inhibit the function of T cells carrying the same T-cell receptor specificity and prevent skin allograft rejection in an antigen-specific, dose-dependent manner. The regulatory T cells are able to acquire alloantigen from antigen-presenting cells, present the alloantigen to activated syngeneic CD8+ T cells and then send death signals to CD8+ T cells. These findings provide a novel mechanism of regulatory T-cell-mediated, antigen-specific suppression.

Evidence in support of the existence of antigen-specific regulatory T cells is accumulating1–12, and downregulation of immune responses to self or allogeneic antigens in the periphery by regulatory T cells has been demonstrated in a number of in vivo models of autoimmunity2–4,11 and transplantation1,5,7,10,13,14. Nonetheless, the underlying mechanisms remain obscure due to the inability to isolate and clone antigen-specific regulatory T cells. We have recently demonstrated that pre-transplant donorspecific transfusion (DST) of lymphocytes from (B6xBALB/c)F1 mice (H-2 Kb/d, Db/d, Ld+) into either (B6xdm2)F1 (dm2 is BALB/c Ld loss mutant) non-transgenic or (2Cxdm2)F1 anti-Ld (MHC class I) T-cell receptor (TCR)- transgenic recipient mice (both are H-2 Kb/d, Db/d, Ld–, and thus mismatched for Ld with the donor) leads to permanent acceptance of skin allografts of lymphocyte donor origin. The rejection of third-party skin allografts is normal15,16. Because none of the recipients was thymectomized, the function of newly released anti-Ld T cells from the thymus must have been inhibited in order to maintain Ld+ skin grafts. Consistent with this, we found that the spleen cells from recipient mice that permanently accepted the Ld+ skin grafts (defined as tolerant mice) are able to inhibit the anti-Ld response mediated by syngeneic naïve T cells in vitro15. These findings indicate the presence of regulatory cells in the spleen of tolerant mice. The objectives of this study were to identify and characterize these antigen-specific regulatory cells and delineate their mechanism of suppression. 1 B2+ CD4– CD8– T cells inhibit anti-Ld responses in vitro and in vivo To confirm the existence of antigen-specific regulatory cells in tolerant animals, we first investigated whether the spleen cells from tolerant mice can inhibit anti-Ld responses in vivo. We adoptively transferred naïve (B6xdm2)F1 mice with syngeneic splenocytes from either DST-treated tolerant mice, or mice that did not receive DST and rejected Ld+ skin grafts within 2 weeks as control, and subsequently gave them skin grafts from both Ld+ ((B6xBALB/c)F1, anti782

gen-specific) and third-party (C3H, H-2K) mice. All animals receiving spleen cells from control mice acutely eliminated the Ld+ skin allografts (median survival time (MST) = 11 d, Fig. 1a). However, the survival of Ld+ skin allografts was prolonged significantly in the recipients infused with splenocytes from DST-treated mice (MST = 33 d, n = 8, P < 0.05). The third-party skin allografts were rejected with similar kinetics (MST = 11 d) by all of the recipients. These data demonstrate that spleen cells from DST-treated mice are able to specifically enhance Ld+ skin allograft survival when infused into naïve syngeneic animals. The vast majority of peripheral T cells in (2Cxdm2)F1 mice express transgenic TCR specific for MHC class I Ld 16,17. The transgenic TCR can be detected by the clonotypic monoclonal antibody 1B218. To determine which subset of T-cells was responsible for the inhibition of the anti-Ld response, we purified 1B2+CD8+, 1B2+CD4+ and 1B2+CD4–CD8– (double negative, DN) T-cells from the spleen of DST-treated (2Cxdm2)F1 mice 120 d after skin grafting, and tested for their regulatory function in vitro. A dose-dependent inhibition of proliferation of naïve antiLd T-cells in mixed lymphocyte reactions (MLR) was observed only in cultures to which 1B2+DN T-cells from DST-treated mice were added (Fig. 1b). No inhibition was seen when 1B2+CD8+ or 1B2+CD4+ T-cells were added to the MLR. These results demonstrate that the mature DN, but not CD4+ or CD8+ T cells, are responsible for inhibiting the anti-Ld response mediated by naïve T cells in this model. DN T cells inhibit CD8+ T cells of the same TCR specificity In order to characterize the DN regulatory T cells and understand the mechanism of suppression, we generated panels of 1B2+DN and 1B2+CD8+ T-cell clones from the spleens of both naïve and DST-treated (2Cxdm2)F1 mice. Regardless of the origin, the DN T cells grew only when stimulated by Ld+ cells in the presence of exogenous interleukin (IL)-2 and IL-4. All 8 1B2+DN T-cell clones NATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000


ARTICLES


a

b

Fig. 1 DN T cells from the spleen of tolerant mice are able to inhibit anti-Ld responses in vivo and in vitro. a, Spleen cells from DST-treated mice enhance Ld+ skin allograft survival. (2Cxdm2)F1 mice were given DST from (B6xBALB/c)F1 (DST) or left untreated (control) and skin grafts from both (B6xBALB/c)F1 and C3H. Three weeks after skin grafting, all DSTtreated mice accepted (B6xBALB/c)F1 grafts but rejected C3H grafts, whereas control mice rejected both grafts15,17. Naïve (B6xdm2)F1 mice were adoptively transferred with viable splenocytes collected from either DST-treated (, n = 8) or control (, n = 5) mice. The next day, each mouse received skin grafts from both (B6xBALB/c)F1 (dashed lines) and

C3H (solid lines) mice. b, Inhibition of anti-Ld MLR by 1B2+DN T cells. 1B2+CD8+ cells from naïve (2Cxdm2)F1 mice were used as responder cells and stimulated by irradiated splenocytes from (B6xBALB/c)F1 mice. Purified 1B2+CD8+ (), 1B2+CD4+ () or 1B2+DN () cells from DSTtreated mice 120 d after skin grafting were added into the MLR cultures as putative regulatory cells at various ratios as indicated. Proliferation was measured by [3H] TdR incorporation. The cultures to which no putative regulatory cells were added were used as controls. The results represent three independent experiments each with five replicates. cpm, counts per minute.

that grew successfully showed a dose-dependent inhibition of proliferation of naïve anti-Ld T cells, whereas three of the four 1B2+CD8+ T-cell clones showed no suppression. The representative results are shown in Fig. 2a. In addition to suppression of proliferation, the cytotoxicity mediated by naïve anti-Ld T cells was also significantly impaired (Fig. 2b). To confirm that the DN regulatory T cells obtained from the TCR transgenic mice have a physiological relevance, we first investigated whether the DN T-cell clones generated from transgenic mice were able to enhance Ld-specific skin allograft survival in non-transgenic mice. We infused naïve unmanipu-

lated (B6xdm2)F1 mice with the 1B2+DN-suppressive clone TN12 and then gave them Ld+ and third-party (C3H) skin allografts. As a control, we also infused some mice with the non-suppressive 1B2+CD8+ T-cell clone C02. All the recipients infused with the non-suppressive C02 cells rejected Ld+ skin allografts within 14 d as seen in non-injected mice16. In contrast, the survival of Ld+ skin grafts was prolonged in the recipients of suppressive 1B2+DN T-cell clones in a dose-dependent manner (MST = 43.5 d, P < 0.01). Two out of 6 recipients accepted the Ld+ skin allografts indefinitely (>120 d) (Fig.2c). The third-party skin allografts were all acutely rejected (MST = 12 d) regardless of the types of cells in-

Fig. 2 Dose-dependent inhibition of syngeneic CD8+ T cells in vitro by DN T cells. a, 1B2+DN T-cell clones inhibit proliferation of syngeneic CD8+ T cells. Purified naïve CD8+ T cells from (2Cxdm2)F1 mice were used as responders and stimulated by irradiated (B6xBALB/c)F1 splenocytes. Varying numbers of 1B2+DN clones generated from DST-treated (TN02, " and TN12, #) and naïve (CN04, X) mice were added to the MLR cultures as putative regulatory cells, and a 1B2+CD8+ T-cell clone (C02, ) was used as a control. Cell proliferation was measured by 3H-TdR incorporation. The data are expressed as percent inhibition of proliferation compared to the controls to which no putative regulatory T cells were added. The experiments were repeated four times and the results represent the other five DN clones and two 1B2+CD8+ Tcell clones (not shown). b, Inhibition of 1B2+CD8+ T-cell-mediated cytotoxicity by 1B2+DN cells. Varying numbers (as indicated) of 1B2+CD8+ cells from the spleen of naïve (2Cxdm2)F1 mice were used as effector cells and stimulated by irradiated Ld+ spleen cells from (B6xBALB/c)F1 mice. 1B2+DN T-cell clone (TN12) were added at a 5:1 ratio 24 h later. Killing of P815 target cells by 1B2+CD8+ T cells (), 1B2+DN T-cell clones (), and 1B2+CD8+ plus 1B2+DN T cells () was measured by using standard 51Cr release assay at 90 h after the addition of the DN T-cell clones. The data are expressed as percent specific killing of P815 target cells, and represents six replicates. c, Infusion of 1B2+DN T cells prolongs Ld+ skin survival in non-transgenic mice. Naïve (B6xdm2)F1 mice were infused with either the 1B2+DN clone TN12 (1 × 107, dotted line, n = 6; or 2 × 107 solid line, n = 6) or with cells from the 1B2+CD8+ clone C02 (2 × 107 dashed line, n = 3) as a control. Each mouse was given skin grafts from both (B6xBALB/c)F1 and C3H (not shown) 1 d after injection of T-cell clones. d, Suppression of syngeneic CD8+ T cells by CD3+DN T cells in normal mice. CD8+ and DN T cells were purified from spleens of (B6xdm2)F1 () and MRL/+ (, HNATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000

a

b

c

d

2K) mice. (B6xdm2) T cells were stimulated with irradiated C3H splenocytes and MRL/+ T cells were stimulated with irradiated (B6xBALB/c)F1 splenocytes. CD8+ T cells were used as responders, and varying numbers of syngeneic DN T cells were added to the corresponding MLR cultures as putative regulatory cells. Cell proliferation was measured as described in Methods. The data represent two experiments each with five replicate cultures 783



ARTICLES Fig. 3 a, DN regulatory T cells express a unique combination of cell surface markers and require direct cell contact for suppression. Regulatory and non-regulatory T-cell clones (i–vi) as well as fresh spleen cells from DST-treated mice 120 d after skin grafting (vii–ix) were stained with various monoclonal antibodies as described in Methods. The relative levels of expression of 1B2 and CD8, CD25, CD28, CD30, and CD44 on DN regulatory T cells (grey) and 1B2+CD8+ non-regulatory T cells (white) are shown. The negative control is shown in black (iii). The data represent eight regulatory and three non-regulatory T-cell clones and spleen T cells from five DST-treated animals. The observed pattern of expression remained the same before and after stimulation. b, Cytokines, Fas and FasL mRNA expression in 1B2+DN and 1B2+CD8+ T cells. 1B2+CD8+ clone (C02, lane 1), 1B2+DN clones (TN12, TN01 and TN11 in lanes 2–4, respectively) and purified splenic 1B2+CD8+ T cells (lane 5) were activated by irradiated Ld+ spleen cells. Ten hours after activation, RNA samples were collected from viable cells. Semi-quantitative analysis of the mRNA levels of IL-2, IL-4, IL-10, IL-13, interferon-γ, tumor necrosis factor-α, TGFβ, Fas, FasL and GAPDH was accomplished by RTPCR17. Similar results were obtained for other time points (not shown). C, DN T-cell-mediated suppression requires direct cell contact. Naïve 1B2+CD8+ cells were stimulated with irradiated Ld+ spleen cells and 1B2+DN T cells were either added directly to the culture (left) or to the top chamber of a trans-well culture system (right) as described in Methods. The data are shown as percent inhibition of proliferation compared with controls to which no DN clones were added. Two different DN clones tested: TN12, ; TN02, . The results represent two independent experiments each with five replicate cultures.

a

b

fused (not shown). These results indicate that 1B2+DN T-cell clones can specifically enhance Ld+ skin graft survival in normal mice. Next, we studied whether DN T cells in the spleen of normal (non-transgenic) mice also have regulatory function. We purified αβ-TCR+CD3+DN T cells from normal (B6xdm2)F1 and MRL/MpJ-+/+ (MRL/+, H-2k) mice, and examined their ability to suppress syngeneic CD8+ T cells. As seen in 1B2+DN T-cell clones, a dose-dependent inhibition of proliferation of syngeneic CD8+ T cells by activated CD3+DN T cells was also observed when the same culture conditions (that is, alloantigen plus IL-2 and IL-4) were used (Fig. 2d). This finding indicates that peripheral DN T cells from both transgenic and normal mice are able to function as regulatory T cells, obviating concerns that the DN regulatory T cells observed in the TCR transgenic animals are a non-physiologically relevant oddity. Moreover, we can generate the DN regulatory T-cell clones from both DST-treated and naïve animals (Fig. 2a), indicating that the DN regulatory T-cell precursors may exist in the spleen of normal mice. The involvement of DST may be to promote the activation/function of DN T cells. DN T cells express a unique set of cell surface markers To further characterize the antigen-specific regulatory T cells, we compared the expression of cell surface markers between regulatory and non-regulatory T cells. Regulatory clones obtained from both DST-treated and naïve animals express equivalent levels of αβ-TCR, CD25, LFA-1, CD69, CD45, CD62L and CTLA-4 when compared with non-regulatory 1B2+CD8+ T-cell clones or primary activated 1B2+CD8+ T cells. They are negative for TCR and NK1.1 (natural killer). Interestingly, unlike 1B2+CD8+ T cells or clones, none of the DN regulatory T-cell clones express CD28 or CD44 at any time point after activation. The only molecule that 784

c

is expressed on 1B2+DN but not on 1B2+CD8+ T-cell clones observed to date is CD30 (Fig. 3a, i–vi). A similar pattern of cell surface markers was observed when DN T cells collected from DST-treated mice 120 d after skin grafting were studied (Fig. 3a,vii–ix). These results demonstrate that DN regulatory T cells do express a unique combination of cell surface markers (that is, αβ-TCR+CD4–CD8–-CD25+CD28–CD30+CD44–) that makes them distinguishable from any previously described T-cell subset (for example, activated helper, cytotoxic or memory T cells)19,20. These mature, peripheral DN T cells also differ from bone marrow-derived DN natural suppressor T cells that express NK1.121–24 and DN T-cell clones described by others25. These findings may explain why earlier attempts to identify a single unique marker on CD4+ or CD8+ suppressor T cells failed. Suppression mediated by DN T cells requires cell–cell contact To delineate the mechanisms of suppression, we first examined whether suppression was mediated by secreting soluble-suppressive factors such as transforming growth factor-β or IL-10 as reported by others2,4,26,27. Semi-quantitative reverse-transcription polymerase chain reaction (RT-PCR) showed that both regulatory and non-regulatory T-cell clones express equivalent levels of interferon-γ, transforming growth factor-β and tumor necrosis factor-α transcripts, and none of them express IL-2, IL-4 or IL-13 mRNA at any time after activation. Interestingly, IL-10 was only NATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000



ARTICLES Fig. 4 DN T cells specifically lyse activated CD8+ T cells through the Fas/FasL pathway. a, Regulatory and non-regulatory T-cell clones mediate cytotoxicity through different pathways. The 1B2+CD8+ or 1B2+DN T cells were stimulated independently by Ld+ spleen cells and used as effector cells. Concanavalin A-activated (B6x2C)F1 syngeneic (Ld-, Fas+; ) and (B6xBALB/c)F1 allogeneic (Ld+, Fas+; ) T cells (>95% CD8+) as well as allogeneic P815 tumor cells (Ld+, Fas-; ) were used as target cells. After co-culture with the effector cells, % specific lysis of target cells was calculated. The data are expressed as mean percent killing of five replicate cultures from three independent experiments. b, Fas Fc fusion protein (Fas-Fc) blocks DN Tcell-mediated cytotoxicity. DN T-cell clones (TN12, and CN04, ) were activated by Ld+ splenocytes and used as effector cells. The 1B2+CD8+ T cells from (2Cxdm2)F1 mice were stimulated by (B6xBALB/c)F1 spleen cells and used as targets. Fresh irradiated (B6xBALB/c)F1 splenocytes were added during the cytotoxicity assay. Lysis of primary activated 1B2+CD8+Fas+ targets was measured at E:T ratio 5:1 in the presence of varying doses of Fas-Fc as indicated. Mean percent inhibition of specific lysis of 1B2+CD8+Fas+ target cells compared to killing detected in the absence of Fas-Fc is shown. The data represent three replicate cultures from two independent experiments. c, DN T cells lyse activated Fas+ but not Fas-mutant CD8+ cells. Primary CD8+ T cells were purified from the spleen of MRL/+ (), MRL/lpr () (B6xdm2)F1 (), and B6/lpr () mice. Primary αβTCR+CD3+DN T cells were purified from both (B6xdm2)F1 (dashed lines) and MRL/lpr mice (solid lines). The CD8+ and DN T cells from (B6xdm2)F1 and B6/lpr mice were stimulated by irradiated C3H (H-2K) splenocytes, and CD8+ and DN T cells from MRL/+ and MRL/lpr mice were stimulated by irradiated (B6xBALB/c)F1 cells. Activated CD3+DN T cells were used as effector cells at E:T ratios as indicated. Fresh irradiated (B6xBALB/c)F1 splenocytes were added during the cytotoxicity assay. Specific lysis of target cells was determined as described in Methods. The data are expressed as mean percent killing of three replicate cultures from two independent experiments. d, DN T cells do not lyse CD8+ T cells with a different TCR specificity. Spleen cells from mice expressing an

a

b

c

d

expressed in primary activated 1B2+CD8+ T cells and 1B2+CD8+ clones, but not in any DN regulatory T-cell clones (Fig. 3b). These findings indicate that our DN regulatory T cells possess a unique array of cytokines that differ from T-helper (Th)1, Th2 or Th3/T regulatory-1 cells2,4, and that IL-10 is unlikely a suppressive factor produced by DN regulatory T cells as seen in CD4+ regulatory cells2,4. The findings also imply that suppression may be mediated by means other than secretion of suppressive cytokines. Consistent with this idea, supernatant collected from regulatory T-cell clones after stimulation with irradiated Ld+ spleen cells was not able to inhibit the naïve anti-Ld response in vitro (not shown). When naïve anti-Ld T cells were co-cultured with regulatory T cells in a trans-well system to prevent direct cell-cell contact but maintain diffusion of secreted soluble factors, no suppression was observed (Fig. 3c). These results support the conclusion that suppression by DN regulatory T cells requires cell contact. In addition, suppression is not likely due to competition for either antigen-presenting cells (APC) or growth factors, as reported by others11,28. Specifically, increasing the number of APC or the concentrations of IL-2/IL-4 did not reverse the suppression (not shown). DN T cells to kill CD8+ T cells through the Fas-dependent pathway Next, we addressed whether suppression is mediated by directly NATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000

αβTCR-transgene specific for the LCMV-gp were stimulated in vitro with peptide p33 as described40. Female T3.70+CD8+ anti-male HY transgenic T cells (anti-HY) were stimulated with irradiated male B6 spleen cells. C3H (H-2K) T cells were stimulated with irradiated SJL (H-2S) spleen cells. 1B2+CD8+ T cells were stimulated by irradiated (B6xBALB/c)F1 spleen cells. Activated CD8+ T cells from these cultures were used as targets. The 1B2+DN T cells were stimulated independently by (B6xBALB/c)F1 spleen cells, then co-cultured with various target cells. Fresh, appropriate irradiated spleen cells were added during cytotoxicity assays. Percentage specific lysis of anti-LCMV p33 (), anti-HY ( ), anti-H-2S () and anti-Ld () target cells was determined. The data are expressed as mean percent killing of three replicate cultures from two to three independent experiments.

killing 1B2+CD8+ T cells, and if so, which pathway DN T cells use. Two major pathways are involved in T-cell-mediated cytotoxicity: one perforin-dependent, the other Fas-dependent29. To study the ability of DN T cells to kill anti-Ld T cells through these pathways, we used three different target cells in cytotoxicity assays: Concanavalin A-activated spleen cells from both (B6xBALB/c)F1 (Ld+, Fas+), 1B2+CD8+ T cells (Ld–, Fas+), and a mastocytoma cell line P815 (Ld+, Fas–). Interestingly, although both activated regulatory and non-regulatory T cells express similar levels of TCR (Fig. 3a) and Fas ligand (FasL)30, they lysed target cells by using different cytolytic pathways. As seen in most cytotoxic CD8+ T cells, the non-regulatory 1B2+CD8+ T cells lysed Ld+ allogeneic, but not Ld– syngeneic target cells, regardless of Fas expression through the perforin-dependent pathway (Fig. 4a, left). In contrast, 1B2+DN regulatory T cells killed neither Ld+ allogeneic nor syngeneic T cells through the perforin-dependent pathway (Fig. 4a, middle). Furthermore, these regulatory T cells did not kill Fasallogeneic cells, but they were able to kill both Fas+ anti-Ld and Fas+ Ld+ CD8+ T cells in an 18-h killing assay (Fig. 4a, right). These results indicate the possibility that Fas–FasL interactions were involved in DN T-cell-mediated killing, which is consistent with the findings of others34. To verify the involvement of Fas–FasL interaction in DN T-cellmediated killing, we incubated the DN T cells with varying doses 785


ARTICLES


of Fas–Fc fusion protein before and during the cytotoxicity assays. The ability of DN T cells to kill activated 1B2+CD8+ T cells was blocked in a dose-dependent manner by Fas–Fc fusion protein (Fig 4b). This indicates that blocking FasL on the DN T-cell abolishes DN T-cell-mediated cytotoxicity. To further determine the importance of Fas in DN T-cell-mediated cytotoxicity, we studied the ability of purified αβTCR+CD3+DN T cells from (B6xdm2)F1 and MRL/MpJ-lpr/lpr (MRL/lpr) mice to kill CD8+ cells from wild-type and Fas mutant lpr mice. DN T cells from both normal and lpr mice were able to kill activated CD8+ T cells that express wild-type Fas, and the cytotoxicity of DN T cells collected from lpr mice was more pronounced than those from normal mice. However, we observed a significant reduction of cytotoxicity in both strains of mice when activated CD8+ cells from Fas mutant lpr mice were used as target T cells (Fig. 4c). Taken together, these results demonstrate that DN regulatory T clones and cells are able to kill activated CD8+ T cells, and that Fas–FasL interactions are important for the cytotoxicity to take place. Killing by 1B2+DN T-cells is not due to bystander cytotoxicity Our skin grafting results indicate that DN T-cell suppression is antigen specific in vivo. The in vitro finding that both activated Fas+ Ld+ T cells and syngeneic Ld– 1B2+CD8+ T cells are killed by anti-Ld DN T cells indicates the possibility that DN T cells may mediate a non-specific bystander killing through Fas–FasL interactions, as seen in some CD4+ and CD8+ T cells31–33. In order to determine the antigen-specificity of DN T-cell-mediated cytotoxicity, 1B2+DN T cells were stimulated by Ld+ cells and used as effector cells. Activated Ld– Fas+CD8+ T cells with different antigen-specificities including a) 1B2+CD8+ T cells (anti-Ld); b) female anti-male HY TCR transgenic T cells (anti-HY); c) anti-lymphocytic choriomeningitis virus glycoprotein TCR transgenic T cells (anti-LCMV-gp); d) C3H anti-SJL non-transgenic T cells (anti-H-2S) were used as targets. Although the target cells are all Ld–, and express a similar level of Fas after activation, only those 1B2+CD8+ T cells that carry the same TCR specificity as the DN T cells were killed. None of the CD8+Fas+ T cells that express a TCR with different antigen-specificity from DN T cells were lysed (Fig. 4d). These data demonstrate that 1) 1B2+DN T

Fig. 5 TCR -Ld interaction is critical for DN T-cell-mediated suppression. a, Kinetics of Ld expression on the surface of 1B2+DN and 1B2+CD8+ T cells. Primary 1B2+CD8+ T cells and 1B2+DN clones were cultured separately with irradiated (B6xBALB/c)F1 splenocytes. At various time points after culture, cells were harvested and triple stained with monoclonal antibodies 1B2, anti-CD8 and anti-Ld. The expression of Ld on 1B2+ and 1B2+CD8+ T cells was analyzed by using a flow cytometer. Changes in Ld mean fluorescence intensity (MFI) over time after activation of 1B2+DN (dashed line) and 1B2+CD8+ T cells (solid line) are shown. The results represent two independent experiments. b, Histograms show Ld expression on 1B2+DN (left) and 1B2+CD8+ (right) T cells after 2 d stimulation by Ld+ spleen cells (). Markers indicate % cells that express Ld in each subset of cells compared to corresponding controls (). c, Blocking either Ld on DN T cells or 1B2 on CD8+ T cells abrogates DN T-cell-mediated cytotoxicity. The 1B2+DN T cells were purified after 2 d of stimulation with irradiated Ld+ splenocytes, and used as effector cells. 1B2+CD8+ T cells were activated with irradiated Ld+ splenocytes for 4 d, labeled and used as target cells. The DN cells were either left untreated (solid line) or pre-incubated with anti-Ld monoclonal antibody before addition to the cultures. Labeled CD8+ T cells were either left untreated (), pre-incubated with 1B2 () or irrelevant isotype-matched control monoclonal antibody () and then washed before being used as targets. Specific lysis of target cells was determined as described in Methods. The results represent five replicate cultures. 786

cells do not lyse Fas+ T cells through bystander killing. 2) Sharing TCR-specificity between DN and target T cells is required for cytotoxicity to take place. TCR-Ld interaction is critical for 1B2+DN T-cell-mediated killing The finding that 1B2+DN T cells kill both Ld+ and anti-Ld T cells, but not other activated Ld– CD8+ T cells, indicates that a specific TCR-Ld interaction may be involved in the mechanism of suppression. In order for suppression to be mediated by the anti-Ld TCR, either the target or effector cell must express Ld. Recent studies have indicated that T cells are able to acquire MHC from APC34–36. To study whether either Ld- target or regulatory T cells could acquire Ld during their activation, we incubated 1B2+DN and 1B2+CD8+ cells independently with irradiated Ld+ APC, and monitored their ability to pick-up and express Ld at different time points by triple staining of cultured cells with 1B2, antiCD8, and anti-Ld monoclonal antibodies. Both 1B2+DN and 1B2+CD8+ T cells express Ld within minutes after incubation with Ld+ spleen cells (Fig. 5a). By 12 h, there was no detectable Ld on the surface of 1B2+CD8+ T cells. In contrast, 24% of the 1B2+DN T cells expressed Ld even at 48 h (Fig. 5b). These results demonstrate that both 1B2+DN and 1B2+CD8+ T cells are able to acquire Ld from APC and express it on their surface. The expression of Ld on 1B2+CD8+ T cells, however, was transient as seen by others34, whereas, the Ld expression on DN T cells persisted for at least 2 d. Next, we investigated whether Ld expressed on DN T cells was critical for lysis of syngeneic anti-Ld CD8+ T cells by blocking studies. When Ld molecules on DN T cells were neutralized by pre-incubation with anti-Ld monoclonal antibody before being

a

b

c

NATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000


ARTICLES


Fig. 6

1B2+DN T-cell suppression of 1B2+CD8+ T cells.

used as suppressor cells, their ability to kill 1B2+CD8+ T cells was eradicated. Similarly, when the TCR on 1B2+CD8+ cells was blocked by pre-incubation with 1B2 monoclonal antibody before being used as target cells, DN T-cell-mediated cytotoxicity was also abolished (Fig. 5c). Thus, blocking either Ld on the DN T cells or TCR on anti-Ld CD8+ T cells abrogates DN T-cell-mediated lysis of syngeneic CD8+ T cells. Together, these data clearly demonstrate that lysis of syngeneic CD8+ T cells by DN T cells requires a specific interaction between TCR and Ld alloantigen along with Fas–FasL interaction. These findings not only explain why suppression seen in our model is antigen-specific in vitro and in vivo, but also provide a new model for understanding the mechanism of regulatory T-cell-mediated suppression. Discussion This article identified and characterized a new subset of antigenspecific regulatory T cells. We demonstrate that 1B2+DN T cells (Fig. 1) and clones (Fig. 2) can specifically suppress anti-Ld responses in vitro and enhance donor-specific skin allograft survival. As DN regulatory T cells can be obtained from either pre-transplantation DST or in vitro stimulation of naïve cells with alloantigen in the presence of exogenous IL-2/IL-4, it indicates that the DN regulatory T-cell precursors may pre-exist in normal individuals. The involvement of DST in the induction of donor-specific tolerance may be to promote the activation and expansion of DN regulatory T cells in vivo by providing antigen stimulation and IL-2/IL-4. The results support this notion: 1) Both antigen and exogenous IL-2 and IL-4 are required for the DN T cells to survive and proliferate in vitro30,37. 2) The level of IL4 in the sera of DST recipients increased significantly15. 3) The number of DN T cells in the peripheral lymphoid organs and accepted Ld+ skin allografts of DST-treated animals was significantly increased (Young et al., manuscript submitted). Downregulation of specific immune responses by regulatory T cells in vitro and in vivo has been observed in numerous cases1–15. However, the mechanism by which regulatory T cells mediate antigen-specific suppression remains unclear. Various mechanisms, such as competition with antigen-specific T cells for APC or growth factors11,28 and the production of suppressive cytokines2,4, have been proposed. We demonstrate that DN regulatory T-cell mediated suppression requires direct contact with activated CD8+ T cells, indicating that suppression is not simply due to secretion of inhibitory cytokines or soluble factors (Fig. NATURE MEDICINE • VOLUME 6 • NUMBER 7 • JULY 2000

3). As we cannot abolish DN T-cell-mediated suppression by adding an excessive number of APC or IL-2/IL-4, it indicates that suppression is neither due to competition for the surface area on APC nor growth factors with CD8+ cells. Moreover, DN T cells are able to kill activated CD8+ T cells through the Fas–FasL pathway (Fig. 4a). This killing can be inhibited by Fas-Fc fusion protein (Fig. 4b), and the ability of DN T cells to kill syngeneic CD8+ T cells from Fas-mutant lpr mice was significantly impaired (Fig. 4c). Together, these results demonstrate that Fas–FasL interactions are involved, at least partially, in DN T-cellmediated suppression. The fact that DN T cells are able to kill both activated Ld+ allogeneic and anti-Ld syngeneic T cells, but not other CD8+ T cells activated by 3rd party antigens (Fig. 4), indicates the involvement of specific antigen recognition during suppression. Huang et al.34 recently reported that naïve 1B2+CD8+ T cells were able to acquire Ld peptide-MHC clusters from APC through TCRmediated endocytosis. During this process, the 1B2+CD8+ T cells were sensitized to peptide-specific lysis by neighboring 1B2+CD8+ T cells and therefore killed each other (fratricide), leading to downregulation of immune response34. In our study, Ld expression on 1B2+CD8+ cells was also detected within the first few hours after interacting with Ld+ APC. If suppression was due to fratricide between CD8+ T cells, a significant reduction of CD8+ T cells would be observed within the first 24 h of encountering the antigen. However, our results from kinetic studies indicated no significant death of 1B2+CD8+ T cells within the first 66 h after encountering Ld+ spleen cells (not shown). In addition, the finding that 1B2+CD8+ T cells did not suppress an anti-Ld response (Fig. 1b) does not support the hypothesis that suppression of CD8+ T cells in our model is due to fratricide. Alternatively, we hypothesized that DN T cells may acquire Ld from APC and present it to anti-Ld CD8+ T cells. The following observations support this hypothesis. First, we demonstrate that 1B2+DN T cells are able to acquire Ld from APC and express it on their surface for at least 48 h, in contrast to its transient (