This study has explored for the first time the role of CD28 in the induction of peripheral tolerance in CD8 T cells in vivo. Analyzing proliferation, cytokine secretion, and CTL generation of MHC class I–restricted T cells from CD28WT and CD28KO H-Y RAG-2 KO mice, we have found that CD28 costimulation is dispensible for in vitro activation. Activation of CD28KO T cells was not diminished even when using very low concentrations of peptide antigen, or male APCs that express a low but physiological concentration of antigen. It was therefore surprising that we found CD28 is required in vivo for both the induction of clonal deletion and for antigen-specific hyporesponsiveness of CD8 T cells. These data demonstrate a previously unanticipated role for CD28 in tolerance induction of CD8 cells.
We have previously used pregnancy as a model to study peripheral tolerance in the maternal immune system to the fetal antigen, H-Y. Tolerance to H-Y expressed by male fetuses appears to be mediated by a combination of decreased survival of H-Y–specific T cells in vivo and decreased responsiveness of the remaining T cells as determined by in vitro challenge with antigen (26
). That this is not a nonspecific effect of pregnancy itself but rather antigen-mediated is confirmed by the lack of deletion or hyporesponsiveness observed in pregnant females with litters consisting of only female pups. The reduction in clonotypic T cells could not be accounted for either by TCR down-regulation, or by migration to the sites of H-Y antigen expression, either the placenta itself or the draining lymph nodes of the placenta (our unpublished data). Therefore, it is likely that the decreased numbers of H-Y reactive T cells are due to cell death resulting from antigen encounter whether occurring in lymphoid tissues, or in nonlymphoid tissues after emigration.
The dependence on CD28 costimulation for induction of clonal deletion is unexpected. At least for CD4 T cells, it has been proposed that CD28 costimulation is required for clonal deletion of T cells to self-antigen because activation is required for activation-induced cell death to occur (31
). The absence of clonal deletion in CD28KO mice would therefore be consistent with this model. However, it is important to consider that while CD4 T cells have a requisite dependence on CD28 for activation, it appears that the H-Y–specific CD8 T cells studied here do not. Therefore, it is in this respect surprising that clonal deletion of these CD8 cells is CD28 dependent while activation is not. These results suggest a novel role for CD28 in determining the fate of CD8 T cells.
H-Y–specific T cells that escape deletion in pregnant CD28WT females exhibit a form of anergy defined by decreased proliferative responses that are not reversed by IL-2 (26
). The remaining CD28WT T cells in pregnant mice exhibit decreased levels of CD62L, confirmation that these cells have encountered antigen, and also have decreased CTL activity in response to H-Y antigen. In marked contrast, H-Y–specific T cells from pregnant CD28KO mice did not exhibit either decreased proliferation or decreased CTL activity, nor did they down-regulate CD62L raising the possibility that these T cells have not been activated by antigen. However, T cells from the d18 CD28KO pregnants had significantly lower levels of IFN-γ–producing cells compared with T cells from the CD28KO nonpregnant females, indicating that these cells are not “antigen ignorant” but have seen antigen in vivo and responded, albeit in an altered manner. The decrease in INF-γ–producing cells is a not a nonspecific effect of pregnancy since T cells from d18 CD28WT pregnants do not exhibit decreased IFN-γ–producing cells.
The phenomenon of “split tolerance” observed here in CD8 cells from pregnant CD28 WT mice, in which unresponsiveness was induced in some but not other parameters of T cell responses, has previously been reported in other experimental systems. Diminished proliferative response, but intact cytokine and/or CTL function, has been observed in vitro (32
) as well as in in vivo systems of peripheral tolerance (33
). However, the means by which this type of split tolerance occurs is unclear. Interestingly, the inverse functional pattern is observed in T cells from CD28KO H-Y pregnants, which have normal CTL activity and proliferation yet decreased numbers of IFN-γ–producing cells. Future studies will determine whether there are differences in TCR signaling in T cells from CD28WT and CD28KO H-Y pregnants that may account for functional differences in these populations, as has previously been suggested in circumstances of split tolerance (34
). One question that arises is whether the “split tolerance” that we as well as others have observed is truly “tolerance” since there is not a total abrogation of responsiveness in these systems. In the experimental system characterized here, the hyporesponsiveness observed by the residual T cells appears to be sufficient to prevent an immune response and rejection of the fetus, and therefore could be argued to be true functional tolerance.
The role of costimulation in the induction of anergy has previously been addressed in studies of CD4 T cells, and it has been proposed that CTLA-4/B7 interactions may be particularly important in the induction of anergy (14
). It has also been reported that two kinds of anergy can be induced in vitro, one that is induced by TCR stimulation in the absence of CD28 and is reversible by IL-2 and the other that is induced by CTLA-4 signals and is not reversible by IL-2 (35
). The antigen-specific hyporesponsiveness that we have observed in H-Y–specific transgenic CD8 T cells resembles the second of these anergic states, in that it is not reversible by IL-2 (26
). However, in our model of peripheral tolerance, anergy induction does not occur in CD28KO mice, despite the fact that CTLA-4/B7 interactions remain intact in these animals. This finding supports the observation by Frauwirth et al. that demonstrated CTLA-4 is not required for induction of in vivo anergy using CD8 TCR transgenic mice (36
), again suggesting different roles for the CD28/CTLA-4/B7 costimulatory pathway between CD4 and CD8 T cells. The importance of CD28/B7 interactions in tolerance induction is further confirmed by the lack of both deletion and anergy induced in the B7DKO H-Y pregnants.
In conclusion, our data demonstrate that CD8 cells can respond efficiently to even low concentrations of antigen in the absence of CD28 costimulation in vitro. However, peripheral tolerance in vivo does not occur in the absence of CD28. To our knowledge this is the first in vivo assessment of the role for CD28 in peripheral tolerance induction of CD8 cells and the first evidence that CD28 is critical for tolerance induction of CD8 cells. These findings suggest that the role of CD28/CTLA-4/B7 interactions in peripheral tolerance of CD8 T cells may differ significantly from that for CD4 T cells and demonstrate that CD28 is involved in both clonal deletion and induction of hyporesponsiveness by encounter with peripheral self-antigen.