Preventing NK Cell Activation by Donor Dendritic Cells Enhances Allospecific CD4 T Cell Priming and Promotes Th Type 2 Responses to Transplantation Antigens1

Although much progress has been made in understanding the role of NK cells in bone marrow transplantation, little is known about their function in CD4 T cell-mediated allograft rejection. We have previously shown that in the absence of CD8 T lymphocyte priming, the in vivo default development pathway of alloreactive CD4 T cells was strongly biased toward Th2 phenotype acquisition. In this study, we investigate the impact of NK cells on the activation and differentiation of alloreactive CD4 T cells in various donor/recipient combinations. Our data demonstrate that defective inhibition of host NK cells by donor APCs including dendritic cells (DCs) results in diminished allospecific Th cell responses associated with the development of effector Th cells producing IFN-γ rather than type 2 cytokines. Turning host NK cells off was sufficient to restore strong alloreactive CD4 T cell priming and Th2 cell development. Similar results were obtained by analyzing the effect of NK cell activation on CD4 T cell responses to skin allografts. However, despite the dramatic effect of NK cells on alloreactive Th1/Th2 cell development, the kinetics of skin graft rejection were not affected. Thus, Th2 differentiation is a major pathway of alloreactive CD4 T cell development during solid organ transplant rejection, as long as host NK and CD8 T cells are not activated. We propose the hypothesis that MHC class I-driven interactions between donor DCs and host NK cells or CD8 T cells might result in DC-carried signals controlling the dynamics of alloreactive CD4 T cell priming and polarization.

N atural killer cells are large granular lymphocytes capable of killing cells without prior immunization (1). The recognition of tumors or pathogen-infected cells by NK cells is controlled by the engagement of multiple cell surface inhibitory or stimulatory receptors that bind to MHC class I or non-MHC ligands. Such interactions result in a cascade of events that transduces either inhibitory or activatory signals, with their balance regulating NK cell activation. Inhibitory receptors encompass several families of cell surface molecules such as Ly-49 and CD94/NKG2 receptors in mouse that recognize self MHC class Ia molecules (2). The activatory receptors recognize MHC class Ilike molecules and other undefined ligands and trigger lysis and cytokine production (2). NK cells can distinguish allogeneic from syngeneic cells, and the expression of self MHC class I molecules protects target cells from NK cell lysis by interacting with NK cell inhibitory receptors. According to the missing self hypothesis, lack of expression or down-regulation of class I molecules on target cells render them susceptible to NK cell lysis (3).
Although much progress has been made in understanding the role of NK cells in bone marrow transplantation, little is known about their function in CD4 T cell-mediated allogeneic transplant rejection (4 -7). For instance, in fully allogeneic combinations, it is likely that in addition to CD4 and CD8 T lymphocytes, NK cells can be activated by donor-derived APCs. Among donor APCs, tissue-resident dendritic cells (DCs) 3 are likely to migrate from the graft to secondary lymphoid organs where priming of alloreactive T cells can occur initiating transplant rejection (8,9). How NK cells can interfere with donor-derived DC and whether these interactions can modulate allospecific T cell development in vivo is presently not known. We have recently analyzed the development of alloreactive CD4 T cells in the absence of CD8 T cell activation in vivo. We showed that while immunization of adult mice with semiallogeneic splenocytes induces the differentiation of donorspecific CD4 T cells toward the Th1 phenotype, responses strongly polarized toward the Th2 type occurred in CD8-deficient mice. This led us to conclude that in the absence of CD8 T cell priming and whatever the genetic background of the host, the default response of alloreactive CD4 T cells is a sustained production of Th2-type cytokines (10).
In this study, we investigate the role of NK cell activation on the development of alloreactive CD4 T cells in vivo in the absence of CD8 T cell activation. The experimental model we have chosen involved priming of the CD8-deficient parental C57BL/6 (B6) or BALB/c strains with either semiallogeneic (BALB/c ϫ B6) F 1 (CB6 F 1 ) or fully allogeneic APCs. Semiallogeneic APCs that expressed both parental MHC products were used instead of fully allogeneic cells to avoid NK cell activation (10). Although immunization of CD8-deficient mice with semiallogeneic APCs induced strong alloreactive CD4 T cell priming and Th2 cell development, injection of allogeneic APCs resulted in reduced allospecific CD4 T cell responses and in the selective expansion of IFN-␥-producing cells. Ab-mediated NK cell depletion before immunization resulted in dramatic expansion of alloreactive Th cells secreting high levels of IL-4 and IL-10 similar to the responses observed in ␤ 2microglobulin (␤ 2 m) and CD8 double knockout mice where NK cell activity is impaired (11)(12)(13). Interestingly, grafting CD8-deficient mice with semiallogeneic skin also induced marked allospecific Th2 cell priming; whereas rejection of allogeneic grafts was associated with the selective development of Th1 cells. In this latter combination, turning NK cells off resulted in a skewing of the alloreactive CD4 T cell response to the Th2 pathway without affecting the kinetics of skin graft rejection. Taken together, our data demonstrate that recognition of donor APCs by host NK cells strongly affects the magnitude of allospecific Th cell responses as well as their cytokine secretion profiles in various donor/recipient strain combinations.

In vivo mAb treatment
For in vivo NK cell depletion in B6 mice, the anti-NK cells PK136 mouse IgG2a mAb (HB 191; American Type Culture Collection, Manassas, VA) was purified from ascites fluid by caprylic acid precipitation. For in vivo NK depletion, mice were injected i.v. with 200 g PK136 mAb or with 25 l anti-asialo GM-1 in 200 l PBS 2 days prior to immunization, followed by two injections of 100 g PK136 mAb or 25 l anti-asialo-GM-1 in 200 l PBS the day of immunization (day 0) and 2 days later. Anti-rat transferrin receptor OX26 mouse IgG2a was used as isotype control for PK136 anti-NK treatment in B6 mice and was kindly provided by Dr A. Saoudi (Institut National de la Santé et de la Recherche Médicale, Toulouse, France). In BALB/c mice, NK depletion was obtained by the administration of anti-asialo-GM-1 rabbit serum (Wako Chemicals, Richmond, VA). Rabbit serum was precipitated with 50% ammonium sulfate, dialyzed in PBS then adjusted at the same concentration as anti-asialo GM-1, and used as control for anti-NK treatment in BALB/c mice.

T cell assays
For cytokine production analysis, CD8-depleted lymph node cells (LNCs) from mice immunized with allogeneic SCs were cultured at 3 ϫ 10 5 cells/ well in 96-well culture plates (Costar, Cambridge, MA) in the presence of various concentrations of irradiated SCs. For removal of CD8 cells, LNC were incubated with KT1.5 mAb (16) culture supernatant, washed, and subsequently incubated with sheep anti-rat IgG M-450 Dynabeads (Dynal Biotech, Great Neck, NY) previously adsorbed with 10% normal mouse serum. CD8-positive cells were then selectively depleted with a magnet (BioSource International, Camarillo, CA). A similar procedure was used for CD4 T cell enrichment by incubating LNC with anti-B220 RA3-3A1 (TIB 146; ATCC), anti-CD8 KT1.5, anti-class II M5/114 (TIB 120; ATCC), and anti-CD11b (TIB 128; ATCC) mAb. Cells were cultured in HL-1 synthetic medium (Hycor, Irvine, CA) supplemented with 2 mM L-glutamine and 50 g/ml gentamicin (Sigma-Aldrich). Cultures were incubated for 3 days in a humidified atmosphere of 5% CO 2 in air. Supernatants from replicate cultures were collected after 72 h and pooled for cytokine analysis. IFN-␥, IL-4, IL-5, and IL-10 were quantified by two sites of sandwich ELISA as previously described (10). For T cell proliferation assays, cell cultures were pulsed 16

Flow cytometric analysis of intracellular cytokine synthesis
LNC were cultured with allogeneic-irradiated (2400 rad) SCs from B6 or BALB/c or CB6 F 1 mice as indicated above. After 72 h of culture, cells were harvested, washed, and cultured for an additional 72 h in complete medium. After Ficoll separation, living cells were collected, resuspended at 10 6 /ml, and stimulated with PMA (50 ng/ml; Sigma-Aldrich) plus ionomycin (0.5 g/ml; Sigma-Aldrich) for 4 h. Two hours before cell harvest, brefeldin A (Sigma-Aldrich) was added at a concentration of 10 g/ml. Cells were harvested, washed in the presence of brefeldin A, and stained using biotinylated anti-CD4 mAb (BD PharMingen), followed by streptavidin-CyChrome (BD PharMingen). Labeled cells were then fixed with 2% paraformaldehyde (Fluka Chemie, Buchs, Switzerland). Intracytoplasmic staining for IFN-␥, IL-4, IL-5, and IL-10 were performed as described previously (10). Data were collected on 20,000 CD4 ϩ cells on a XL Coulter cytometer (Beckman Coulter France S.A., Roissy, France) and analyzed using the CellQuest software (BD Biosciences, Mountain View, CA).

Skin grafting
Skin grafts ϳ1 cm in diameter were prepared from tails of female CB6 F 1 , B6, or BALB/c mice and grafted onto the flank of recipient mice. Petroleum gauze was placed over the graft and sticking plaster was applied around the trunk. After 7 days, bandages were removed and the grafts monitored daily. Grafts were considered rejected when complete epithelial breakdown had occurred.

Statistical analysis
Results are expressed as the mean Ϯ SD, and statistical differences between variables were evaluated by the Mann-Whitney U test.

Effect of MHC class I deficiency of allogeneic APCs on allospecific CD4 T cell development in vivo
We have previously shown that priming of alloreactive CD4 T cells with semiallogeneic APCs results in the development of effector Th cells producing large amounts of type 2 cytokines in the absence of CD8 T cell activation (10). This is illustrated in Fig. 1, where immunization of ␤ 2 m Ϫ/Ϫ BALB/c mice with semiallogeneic ␤ 2 m Ϫ/Ϫ -irradiated splenocytes resulted in the generation of I-A breactive CD4 T cells producing high levels of IL-4 and IL-10 in agreement with our previous work (17). This was confirmed by the analysis of intracellular cytokine synthesis in the same T cell populations after 6 days of in vitro stimulation with CB6 F 1 APCs (Fig. 2C). Strong expansion of IL-4-or IL-10-producing cells was observed, and accounted for the majority of effector CD4 T cells in this combination. In contrast, the frequency of CD4 T cells that produced IFN-␥ was Ͻ10%. Conversely, immunization of BALB/c mice with CB6 F 1 APCs induced the development of IA b -reactive Th1 cells, producing mainly IFN-␥ (38%) but little type 2 cytokines ( Fig. 2A). We have previously shown that the difference in alloreactive CD4 T cell phenotype acquisition between these two combinations was due to the in vitro activation of MHC class I-specific CD8 T cells (10). Because MHC class I deficiency in hemopoietic cells has been shown to promote NK cell activation, we next evaluated the effect of priming BALB/c mice with ␤ 2 m-deficient CB6 F 1 APCs. As shown in Fig. 1, priming BALB/c mice with MHC class I-deficient CB6 F 1 APCs induces H-2 b -reactive CD4 T cells able to proliferate and to secrete IFN-␥, but little if any type 2 cytokines upon in vitro restimulation. Indeed, most of the alloreactive effector CD4 T cells that develop in this combination exhibited a Th1-phenotype as shown by intracellular cytokine staining (Fig. 2B). To demonstrate directly that NK cell activation was responsible for this effect, recipient mice were pretreated with asialo-GM-1 antiserum. As shown in Fig. 3, NK cell depletion resulted in the restoration of IL-4 and IL-10 production by H-2 b -reactive CD4 T cells. The T cell proliferative response, as well as IFN-␥ synthesis were also up-regulated in anti-asialo-GM-1-treated mice (Fig. 3). Thus, impaired MHC class I expression by donor APCs dramatically affects the magnitude and the phenotype of the allospecific CD4 T cell response due to the absence of negative regulation of host NK cells.

NK cell-mediated regulation of allospecific CD4 T cell development due to lack of expression of self MHC class I molecules by allogeneic APCs
NK cell-mediated bone marrow rejection can be caused by the complete lack of expression of MHC class I molecules, but also by the lack of self MHC class I molecules on donor cells (2, 18) albeit less efficiently (11,12). To evaluate the effect of NK cell activation on alloreactive CD4 T cell development, BM-DC from either fully allogeneic B6 or semiallogeneic CB6 F 1 mice were used to immunize BALB/c CD8 Ϫ/Ϫ recipients. Draining lymph nodes were harvested 6 days later and LNC restimulated in vitro with titrated amounts of semiallogeneic irradiated splenocytes (Fig. 4). In agreement with our previous experiments (10), priming of CD8deficient BALB/c mice with semiallogeneic CB6 F 1 BM-DCs induced a vigorous expansion of memory/effector CD4 T cells secreting large amounts of type 2 cytokines in response to allogeneic MHC class II products (Fig. 4, A and B). Similar results were obtained using allogeneic B6 SC as APCs (data not shown). In contrast, although a proliferative response could be induced in CD8-deficient BALB/c mice primed with B6 BM-DCs (Fig. 4A), these B6-specific CD4 T cells secreted less IFN-␥ in primary culture and no type 2 cytokines (Fig. 4B). To test whether this effect was due to NK cell activation as a consequence of the lack of expression of self MHC class I molecules, mice were pretreated with asialo-GM-1 antiserum. The B6-specific proliferative response was slightly up-regulated in NK-depleted mice as compared with control serum-treated and untreated CD8-deficient BALB/c recipients (Fig. 4A). Interestingly, in vivo elimination of  Immunization of BALB/c mice with MHC class I-deficient semiallogeneic APCs impairs Th2 cell development. WT (n ϭ 3 per group) or ␤ 2 m Ϫ/Ϫ (n ϭ 4) BALB/c mice were immunized into the hind footpads with 50 ϫ 10 6 irradiated SC from WT or ␤ 2 m Ϫ/Ϫ CB6 F 1 mice. Six days later, draining lymph nodes were harvested. Immune LNC were cultured (1.5 ϫ 10 6 cells/ml) in the presence of irradiated CB6 F 1 SC (1.5 ϫ 10 6 cells/ml). After 6 days of culture, LNC were stimulated using PMA and ionomycin for 4 h and stained intracellularly for IFN-␥ and IL-4. Analysis was performed on CD4 T cells. Results expressed as mean of three to four individual mice per group Ϯ SD are from one representative experiment of four performed. NK cells led to the expansion of allospecific CD4 T cells producing high amounts of IL-4 and IL-10 in BALB/c CD8 Ϫ/Ϫ mice immunized with fully allogeneic B6 APCs (Fig. 4B). IFN-␥ synthesis was also up-regulated in mice treated with anti-asialo GM-1 Abs (Fig. 4B). Because both type 1 and 2 cytokines were increased by turning NK cells off, we analyzed the frequency of IL-4-and IFN-␥-producing cells. As shown in Fig. 4C, NK cell depletion resulted in a dramatic increase in effector/memory Th2 cell expansion producing IL-4 but no IFN-␥ (40 -60%). The remaining effector T cells were composed of Th1 (2-6% IFN-␥ pos ) and Th0 (3-7% IL-4 pos IFN-␥ pos ) lymphocytes. This pattern of cytokineproducing cells was similar to the one observed in CD8-deficient BALB/c mice primed with CB6 F 1 APCs (Fig. 4C). Although T cell proliferative responses when measured in the fully allogeneic combination were associated with IFN-␥ synthesis (Fig. 4, A and  B), the yield of allospecific CD4 T cells after 6 days of primary culture was very low (Ͻ0.2 ϫ 10 6 cells for 10 6 input cells) pre-cluding further analysis (data not shown). However, in some experiments where a sufficient number of T cells could be obtained from this combination, the response was dominated by IFN-␥producing cells with a frequency of 20% (data not shown).

NK cell-mediated regulation of allohelper T cell activation and differentiation is not restricted to the BALB/c genetic background
Experiments described so far were performed in BALB/c mice using anti-asialo GM-1 Abs for NK cell depletion, which can also damage macrophages (19) or T cells (20). Therefore, we tested whether our observations were also valid in B6 mice in which the method of NK cell depletion using the anti-NK1.1 mAb PK136 is well established (21). To examine the allogeneic CD4 T cell response in the absence of CD8 T cell activation, we used CD8deficient B6 mice as recipients. We compared two combinations where mice were injected with BM-DCs from either semiallogeneic CB6 F 1 or allogeneic BALB/c mice. Unlike BALB/c BM-  DCs, CB6 F 1 BM-DCs express MHC class I molecules of H2 b haplotype; and therefore, are able to inhibit NK cell activation. As shown in Fig. 5, immunization of CD8-deficient B6 mice with allogeneic BALB/c BM-DCs resulted in a low lymphocyte recruitment in the draining lymph nodes (Fig. 5A). Upon allogeneic stimulation with APCs expressing H-2 d class II molecules, LNC failed to proliferate (data not shown) and did not produce cytokines (Fig.  5B). In contrast, immunization of CD8-deficient B6 mice with semiallogeneic CB6 F 1 APCs resulted in a strong expansion of H-2 d -specific CD4 T cells that produce, in addition to IFN-␥, high levels of type 2 cytokines (Fig. 5, B and C), in agreement with our previous observations (10).
Using the PK16 mAb, we next assessed whether NK cell depletion in B6 CD8 Ϫ/Ϫ mice before their immunization with fully allogeneic APCs could restore efficient CD4 T lymphocyte priming. Furthermore, because ␤ 2 m-deficient mice have been shown to have an impaired NK activity due to the lack of expression of MHC class I molecules (11,12), CD4 T cell priming in response to fully allogeneic APCs was also tested in CD8 and ␤ 2 m double knockout B6 mice. As shown in Fig. 6A in agreement with data in Fig. 5, CD8-deficient B6 mice injected with BALB/c BM-DCs failed to develop a significant allospecific CD4 T cell response. Lack of T cell priming was also observed in B6 CD8 Ϫ/Ϫ mice treated with an irrelevant isotype-matched control mAb (data not shown). In striking contrast, in NK-cell depleted B6 CD8 Ϫ/Ϫ mice, a strong T cell proliferation and cytokine production following in vitro allogeneic restimulation was observed. Similar CD4 T cell responses were recorded in NK-depleted B6 CD8 Ϫ/Ϫ mice and in ␤ 2 m-deficient B6 CD8 Ϫ/Ϫ mice (Fig. 6A) consistent with an inhibitory effect of NK cells on alloreactive T cell priming. To define the cytokine production profile of allospecific CD4 T lymphocytes, expanded T cells were stained intracellularly for IL-4 and IFN-␥ after PMA/ionomycin stimulation. Data shown in Fig. 6B illustrate that allospecific CD4 T cells primed in the absence of NK cell activation, either following Ab depletion (Fig. 6B, upper panel) or in mice with impaired NK cell activity (Fig. 6B, lower panel), exhibited strikingly similar profiles of cytokine-producing cells. The response was dominated by IL-4-producing cells (up to 30% of CD4 T cells), of which one-third also secreted IFN-␥. The frequency of IFN-␥ pos T cells was ϳ10% of CD4 T cells. Similar results were obtained when B6 CD8 Ϫ/Ϫ mice were primed with semiallogeneic F 1 APCs (Fig. 5C). Because it has been reported that syngeneic DC could be killed efficiently by autologous NK cells (22), and that there is a bidirectional cross-talk between DC and NK cells (23), we tested the effect of NK cell depletion in mice immunized with semiallogeneic DCs. Data in Fig. 7 show that in this combination where NK cells are silenced by the expression of syngeneic MHC class I molecules, NK cell depletion had no significant effect on alloreactive CD4 T cell activation and differentiation. Thus, in agreement with our previous observations in the BALB/c genetic background, inhibition of NK cell activation in B6 recipients immunized with fully allogeneic DCs enhances allospecific CD4 T cell priming and promotes Th2 cell development.  Inhibition of host NK cells in mice immunized with semiallogeneic DCs does not affect alloreactive CD4 T cell activation and differentiation. CD8 Ϫ/Ϫ B6 mice were immunized into the hind footpads with 5 ϫ 10 5 BM-DC from semiallogeneic CB6 F 1 mice. Mice were i.v. injected with anti-NK PK136 mAb at days Ϫ2, 0, and ϩ2 of the experiment. Six days after immunization, immune LNC were cultured (3 ϫ 10 5 cells/well) in the presence of irradiated BALB/c SC (3 ϫ 10 5 cells/well). Proliferation and cytokine production were measured as in Fig. 6. Results are expressed as mean (bar) of three mice per group and are from one representative experiment of two performed.

NK cell-mediated regulation of allospecific CD4 T cell response following fully allogeneic skin graft transplantation
We next determined whether NK cell activation in the absence of self-MHC class I expression can also control allospecific CD4 T cell responses induced by tissue transplantation. To this end, we performed skin graft on CD8-deficient BALB/c recipients in combinations in which NK cells were activated (B6 donor) or not (CB6 F 1 donor). As shown in Fig. 8A, the outcome of CD4 T cell-mediated graft rejection was similar in CD8-deficient recipient grafted with either a semi or fully allogeneic skin, and occurred between days 8 and 11. We then analyzed the polarization of allospecific CD4 T cells from lymph nodes draining the graft before and during the rejection process. As shown in Fig. 8B, CD4 T cells from CD8-deficient BALB/c mice grafted with CB6 F 1 skin produced large amounts of IL-4 in addition to IFN-␥ upon restimulation with CB6 F 1 APCs at both time points tested. In individual mice, the level of IL-4 synthesis was inversely correlated to IFN-␥ secretion, indicating that strong Th2 polarization had occurred in CB6 F 1 -grafted recipients. In contrast, while IL-4 production was low or undetectable in recipient mice bearing fully allogeneic graft, the allospecific CD4 T cell response was characterized by the selective development of IFN-␥-producing cells.
We next analyzed the impact of NK cell activation on the development of CD4 T cells in fully allogeneic skin graft recipients. CD4 T cell-mediated rejection of BALB/c skin by CD8-deficient B6 mice was characterized by the selective development of IFN-␥-producing T cells (Fig. 9A). In contrast, NK cell depletion in vivo resulted in enhanced IL-4 and decreased IFN-␥ production by T cells from the lymph nodes draining the BALB/c graft (Fig. 9A). Interestingly, a similar skewing of the H-2 d -specific T cell response to the Th2 phenotype was observed in ␤ 2 m-and CD8deficient B6 mice. In contrast to PK136-treated mice, those injected with control mAb exhibited a Th1-polarized allospecific T cell response. Finally, although the character of alloreactive CD4 T cell responses was dramatically affected by NK cell depletion, the kinetics of allogeneic BALB/c skin graft rejection in PK136treated B6 CD8 Ϫ/Ϫ mice was not modified, demonstrating that acute graft rejection can occur in the context of either a Th1-or Th2-polarized immune response (Fig. 9B). Taken together, these data demonstrate that donor DC-driven host NK cell activation controls alloreactive Th cell development during skin graft transplantation without affecting the kinetics of rejection.

Discussion
NK cells represent a key component of innate immune response and can mediate graft rejection following bone marrow transplantation (6,18). However, little is known about the role of NK cells in allorejection of solid organs (4 -7). In the present paper, we have evaluated whether NK cells can influence the response of alloreactive CD4 T lymphocytes, presumably through their interaction with donor-derived APCs. We have previously shown that allospecific CD4 T lymphocytes differentiate into type 2 cytokine-producing cells in the absence of CD8 T lymphocyte activation (10). FIGURE 8. CD4 T cell-mediated semi-identical skin graft rejection is associated with Th2 cell response. A, BALB/c CD8 Ϫ/Ϫ mice were grafted with skin from fully allogeneic B6 mice (n ϭ 6) or from semiallogeneic CB6 F 1 mice (n ϭ 5) and surveyed for graft rejection. B, BALB/c CD8 Ϫ/Ϫ mice were grafted with skin from either syngeneic BALB/c mice, allogeneic B6 mice, or from semiallogeneic CB6 F 1 mice (3-4 mice per group). The draining lymph nodes were harvested 7-9 days after grafting and LNC were cultured (3 ϫ 10 5 cells/well) in the presence of irradiated SC (3 ϫ 10 5 cells/well) from a CB6 F 1 mouse. Cytokine production in 72-h culture supernatant was measured by ELISA. Results expressed as mean (bar) of three to four individual mice (symbols) per group, are from one representative experiment of three performed. FIGURE 9. NK cell depletion in mice grafted with allogeneic skin restores Th2 cell development without affecting the kinetics of rejection. A, B6 CD8 Ϫ/Ϫ mice were treated or not with anti-NK mAb or irrelevant mAb from the same isotype as described in Materials and Methods. CD8-deficient B6 mice were grafted with skin from either allogeneic BALB/c mice or from syngeneic B6 mice (3-4 mice per group). Seven days after grafting, immune LNC were cultured (3 ϫ 10 5 cells/well) in the presence of irradiated CB6 F 1 SC (3 ϫ 10 5 cells/well). Cytokine production in 72-h culture supernatant was measured by ELISA. Results expressed as mean (bar) Ϯ SD of three to four individual mice per group are from one representative experiment of two performed. Statistical significance between untreated B6 CD8 Ϫ/Ϫ mice grafted with allogeneic skin and other groups is indicated ‫,ء (‬ p Ͻ 0.05). B, B6 CD8 Ϫ/Ϫ mice that have been either treated or not with anti-NK mAb were grafted with skin from allogeneic BALB/c mice or from semiallogeneic CB6 F 1 mice (4 -5 mice per group) and surveyed for graft rejection. Strong Th2 responses were first observed in a MHC class I-deficient donor/recipient combination where ␤ 2 m-deficient BALB/c mice were immunized with ␤ 2 m-deficient semiallogeneic CB6 F 1 (H-2 bxd ) APCs. In contrast, immunization of wild-type (WT) BALB/c mice with CB6 F 1 APCs resulted in the selective development of H-2 b -specific Th1 cells (10). The absence of Th2 cell development was controlled by CD8 T cells, because immunization of CD8-deficient parental strains with CB6 F 1 APCs induced allospecific CD4 T cells producing a large amount of Th2-type cytokines (10). We initially thought of using CB6 F 1 APCs to prime alloreactive host CD4 T cells on the assumption that F 1 APCs expressed, in addition to allogeneic MHC products, self MHC class I molecules that bind to inhibitory receptors on host NK cells, thereby silencing inhibitory receptor-bearing cells. We now provide direct evidence that NK cells, through their interaction with allogeneic APCs, can quantitatively and qualitatively control allospecific CD4 T responses in vivo. This was tested in different combinations: 1) by injecting ␤ 2 m-deficient allogeneic APCs into normal mice, or 2) by immunizing CD8-deficient mice with fully allogeneic APCs. In the first combination, host NK cells were likely to be activated as a consequence of the lack of expression of self MHC class I molecules on donor-derived APCs, thus resulting in NK cell activation. In this combination, the allospecific CD4 T cell response was impaired and led to the development of a Th1-dominated immune response. Depletion of asialo-GM-1positive cells before immunization with ␤ 2 m-deficient APCs restored a strong alloreactive CD4 T cell response characterized by the emergence of type 2 cytokine-producing cells. These data indicate that NK cells can regulate CD4 T cell activation and differentiation in vivo. This was further demonstrated in the second type of combination where inhibitory receptors on NK cells cannot be engaged by allogenic donor APCs that unlike CB6 F 1 APCs do not express syngeneic MHC class I molecules. Again, NK cell depletion was found to enhance CD4 T cell priming and to unmask Th2 cell development in both CD8-deficient B6 and BALB/c-recipient mice. Furthermore, we showed that H-2 d -specific Th2 responses of comparable magnitude were generated in NK-depleted B6 CD8 Ϫ/Ϫ mice and in B6 CD8 Ϫ/Ϫ ␤ 2 m Ϫ/Ϫ double-deficient mice that have an impaired NK cell activity due to the lack of MHC class I molecules (11)(12)(13). Thus, our data demonstrate that in various donor/recipient combinations where NK cells are activated, allospecific Th2 cell responses are abrogated or strongly diminished with the development of effector Th cells producing selectively IFN-␥ but no type 2 cytokines. Inhibition of host NK cells was sufficient to induce alloreactive CD4 T cell responses dominated by type 2 cytokine-producing effector Th cells. This mechanism was not restricted to the BALB/c background and occurred also in B6 mice. However, in this latter strain, NK cell activation resulted in a complete inhibition of alloreactive CD4 T cell priming when DCs were used for immunization. In contrast, when skin grafts were applied as stimulus, unipolar Th1 cell responses were selectively observed in combinations where NK cells were activated in both BALB/c and B6 strains. For instance, in CD8-deficient BALB/c recipients, Th1 responsiveness was not significantly different between mice grafted with either allogeneic or semiallogeneic skin. Conversely, strong priming of IL-4-producing T cells was exclusively observed in the semiallogeneic combination. The impact of NK cell activation on Th1/Th2 differentiation was further demonstrated by grafting CD8-deficient B6 mice with allogeneic BALB/c skin. In this combination, turning NK cells off resulted in a decreased production of IFN-␥ and an increased synthesis of IL-4 by alloreactive CD4 T cells from the lymph nodes draining the graft. Taken together, our data support the conclusion that NK cell activation reduces alloreactive CD4 T cell priming and thus, the subsequent development of both subsets of memory/effector T lymphocytes with a major impact on Th2 cells. Indeed, IL-4-producing cells expand exclusively when NK cells were turned off. In contrast, when CD4 T cell responses develop in combinations where NK cells are activated by donorderived APCs, these Th cells exhibited a unipolar Th1 profile.
Our study unveils a novel mechanism by which NK may regulate the adaptive immune response directed against transplantation Ags in the absence of appropriate interaction between inhibitory receptors on NK cells and MHC class I molecules on donor APCs. These observations are relevant to the clinical transplant situation because similar conclusions could be drawn by analyzing CD4 T cell responses to allograft transplantation. Allospecific CD4 T cells producing IL-4 could readily be detected only when CD8-deficient mice were grafted with semi-identical skin. In this combination, NK cells were not activated by donor-derived DCs because they could receive a silencing signal via self MHC molecules. In contrast, rejection of fully allogeneic skin grafts was associated with the selective development of Th1 cells due to NK cell activation by donor-derived DCs. Indeed, we showed that inhibition of NK cells during CD4 T cell-mediated allogeneic skin graft rejection induced a skewing of the alloreactive T cell response to the Th2 pathway. Therefore, the impact of NK cell activation on the development of Th2 effector functions was similar when skin grafts or BM-DCs were used as stimulus. This conclusion is in agreement with the hypothesis that donor-derived APCs, e.g., DCs, play a central role in inducing CD4 T cell-mediated allograft rejection through direct presentation of donor allogeneic MHC class II molecules to host Th lymphocytes in the draining lymph nodes (8,9,24). However, despite the dramatic effect of NK cells on alloreactive Th1/Th2 cell development, there was no effect on the kinetics of skin graft rejection. This is not surprising since it has been shown by others that both Th1 and Th2 responses are capable of causing acute transplant rejection with identical kinetics in recipient mice receiving either cardiac (25) or skin (26) grafts. Furthermore, subsequent studies have established that rejections mediated by Th2 cells were characterized by marked eosinophilic infiltration of skin and heart transplants (27,28). Finally, it has been reported by Le Moine et al. (29) in a model of MHC class II disparate skin grafts (B6 bm12 3 B6) that IL-4, IL-5, and eosinophils were critically involved not only in chronic but also in acute (30,31) skin graft rejection. These observations are in agreement with our present study and support the conclusion that alloreactive CD4 T cell development during solid organ transplant rejection is strongly biased toward Th2 phenotype as long as host NK and CD8 T cells are not activated.
DCs have been shown to play a crucial role in allorejection by migrating from the transplanted tissue to secondary lymphoid organs of the host where they can prime allospecific T cells and initiate graft rejection (8,9). Accumulating evidence indicates that DCs are phenotypicaly heterogeneous, and represent a multilineage system of leukocytes with variable functions (32). These DC subsets appear to play a role in determining the specific cytokines secreted by Th cells. It has been hypothesized that DCs recruited in immune lymph node can be instructed by environmental stimuli to perform different functions (33,34). It has been shown in humans that IL-12 is produced by DCs within a narrow time window so that only recently activated DCs can promote Th1 cell development (33,35). At later time points, DCs become exhausted in their capacity to secrete various cytokines including IL-12, thereby favoring conditions for priming of Th2 responses, which are dependent on IL-4 production by responding T cells (33)(34)(35). Transient IL-12 production by DCs in vivo have also been documented in mice following systemic stimulation with microbial products (36). To explain our data, we hypothesize that in a situation where both the NK and CD8 T cell pathways are not operative, donor DCs may accumulate in immune lymph nodes, thereby favoring strong CD4 T cell priming and Th2 type responses. Thus, tissue resident NK cells and possibly also CD8 T cells present in the secondary lymphoid organs would inhibit the default Th2 differentiation of allospecific CD4 T cells by limiting the flux of incoming allogeneic DCs into the draining lymph nodes and/or by shortening DC half-life in situ. This may lead to incomplete kinetics of DC differentiation preventing allospecific CD4 T lymphocyte activation by a particular DC subset with Th2-prone capacity. This would explain why in these conditions, CD4 T lymphocytes selectively differentiate toward the Th1 phenotype.
It has recently been shown that inhibition of NK cells combined with CD28 costimulation blockade induced long-term survival of allogeneic vascularized cardiac grafts. In contrast, none of these treatments alone resulted in the acceptance of cardiac allografts. Thus, it has been suggested that NK cells may have a critical role in allorejection by providing help to T cells, such function being essential in the absence of CD28-mediated costimulation (37). Our data are contradictory to this hypothesis, and several reasons could explain this discrepancy. First, CD28-deficient mice have reduced humoral responses but normal cell-mediated immunity (38), indicating that CD4 T cells are more profoundly affected by CD28-B7 blockade than CD8 T lymphocytes. Indeed, it has been shown that blockade of CD28/B7 costimulation inhibits intestinal allograft rejection mediated by CD4 ϩ but not CD8 ϩ T cells (39). Furthermore, it has been shown that CD28-B7 costimulation plays an important role in generating the Th2 compartment (40 -42). Altogether, this could explain why no Th2 cell priming was observed by turning NK cells off in CD28-deficient mice (37). Second, the CD4/ CD8 ratio among graft infiltrating cells was strongly skewed toward CD8 T cells in CD28-deficient mice (37), suggesting that CD8 T cells appear to be the major effector T cell subset responsible for graft rejection. In contrast, it has been clearly established that rejection of skin and cardiac allografts in mice with functional CD28/B7 costimulation pathways was dependent upon CD4 T cells, whereas CD8 T cells were never necessary nor sufficient (43,44). Therefore, the observation that NK cells in allogeneic graft might preferentially provide help to CD8 T cells might be a particularity of the CD28 Ϫ/Ϫ model and is most likely not relevant to the normal recipient situation where rejection mainly involves CD4 T cells (43,44).
In conclusion, our data strongly support the hypothesis that NK cells rather than delivering help to allospecific CD4 T cells may actually limit their activation and differentiation in vivo through their interaction with allogeneic APCs. Indeed, it has been recently shown that donor NK cells from transplanted allogeneic bone marrow are able to eliminate host APCs, preventing donor T cell activation and the consequent graft-vs-host reaction (45). Therefore, our present observations should be taken into consideration in situations where NK cell-inactivating strategies are proposed to improve transplantation tolerance of allogeneic organs (37). According to our data, inactivation of both CD8 T cells and NK cells would promote allospecific CD4 T cell priming and type 2 cytokine production, resulting in an alternative pathway of solid organ rejection involving eosinophils (27,28,31) rather than transplantation tolerance. We hypothesize that NK cells and/or CD8 T cells may limit the flux of graft-derived DCs and/or their kinetics of differentiation in immune lymph nodes, thereby preventing alloreactive CD4 T cell activation by a subset of terminally differentiated DCs displaying Th2-prone capacity. Current experiments are in progress to address this issue.