CD4⁺ regulatory T cells require CTLA-4 for the maintenance of systemic tolerance

Ctla4^−/− mice develop massive lymphoproliferation and a multi-organ inflammatory response driven by self-antigen specific T cells (10, 11). Peripheral T cells are detectably activated by 4 d of age, and transfer of T cells from Ctla4^−/− mice to lymphopenic hosts such as Rag1^−/− mice recapitulates the disease (wasting and lymphoproliferation), highlighting the autoaggressive nature of Ctla4^−/− T cells (unpublished data). As initially shown by Bachmann et al. (22), BM chimaeras generated using a 1:1 mixture of wild-type and Ctla4^−/− (WT:Ctla4^−/−) cells are protected from disease and remained healthy for extended periods of time (Fig. 1 A). This contrasts with Ctla4^−/−-only BM chimaeras that succumb to a fatal lymphoproliferative and inflammatory disease. Similarly, mixed WT:Ctla4^−/− blastocyst chimaeras generated without any manipulation of the lymphoid microenvironment (e.g., γ-irradiation performed on BM recipients) were also disease resistant for their entire life span. (Fig. 1, A and B). Trans-regulation of Ctla4^−/− T cells appeared relatively insensitive to the ratio of WT:Ctla4^−/− cells as both blastocyst and BM chimaeras with WT T cells comprising only 10–20% of the peripheral T cell compartment, albeit rare, were completely healthy. As initial experiments did not reveal any differences between mixed blastocyst and BM chimaeras, all subsequent analyses were performed with BM chimaeras.

Phenotypically, WT and Ctla4^−/− T cells in the mixed BM chimaeras were indistinguishable, with a majority of the cells expressing low levels of CD44 and CD69, and high levels of CD62L, which is indicative of a naive/unactivated state (Fig. 1 C). In contrast, recipients of Ctla4^−/− BM alone developed a population of highly activated T cells, comparable to Ctla4^−/− mice. The frequency of cycling Ctla4^−/− T cells was dramatically reduced in the mixed BM chimaeras, as compared with chimaeras receiving Ctla4^−/− BM only (Fig. 1 C and not depicted), and the naive Ctla4^−/− CD4⁺ and CD8⁺ T cells were as quiescent as WT T cells. Additionally, whereas Ctla4^−/− only BM chimaeras developed lymphoproliferative disease and obvious tissue infiltration by activated T cells, WT:Ctla4^−/− mixed BM chimaeras remained healthy with no aberrant tissue infiltration or disruption in homeostasis (unpublished data).

The robustness of the trans-regulation in WT:Ctla4^−/− mixed BM chimaeras was tested in both acute viral (lymphocytic choriomeningitis virus [LCMV]) and chronic Mycobacterium tuberculosis (Mtb) models, which primarily elicit CD8⁺ or CD4⁺ T cell responses, respectively (33–36). The frequency and magnitude of antiviral or antibacterial T cell responses in challenged WT:Ctla4^−/− mixed BM chimaeras were essentially identical between WT and Ctla4^−/− T cells (Fig. S1, available, and not depicted). The results with LCMV infection confirmed previously published studies (37, 38). In the case of chronic Mtb infection, the WT:Ctla4^−/− ratio of T cells in the lung and spleen upon analysis was similar to preinfection ratios in the blood (Fig. S1 A). Further, the proportions of naive, activated, or CD25⁺CD4⁺ cells in the lung were comparable between WT and Ctla4^−/− T cell compartments (Fig. S1 B). Finally, in vitro IFN-γ responses to Mtb antigens by WT and Ctla4^−/− T cells from infected mixed BM chimaeras showed no significant differences (unpublished data). This remarkable stability of the Ctla4^−/− T cell compartment upon robust activation and chronic antigen presentation was also observed in mixed BM chimaeras challenged with skin allografts (unpublished data).

The WT BM-derived cell subset responsible for trans-regulating Ctla4^−/− T cells has not been formally demonstrated. To this end, we first generated Tcrb^−/−:Ctla4^−/− mixed BM chimaeras to confirm that αβ T cells are required. The absence of WT αβ T cells led to the expansion of activated Ctla4^−/− T cells and death of the animals, demonstrating their requirement for regulation (Table I). In addition, CTLA-4–sufficient αβ T cells expressing MHC class I– or class II–restricted TCR transgenes in a RAG-1/2–sufficient background (that allows for the generation of FOXP3⁺ T cells) were unable to prevent Ctla4^−/− T cell expansion in mixed chimaeras (Fig. 2), suggesting that a diverse TCR repertoire is required for the regulation. Further, Tcrd^−/−:Ctla4^−/− mixed chimaeras remained healthy, indicating that γδ T cells are not necessary for regulation.

We next addressed whether regulation required the activity of a specific αβ T cell subset. As both CD4⁺ (39–41) and CD8⁺ (42, 43) regulatory T cells have been reported, we examined whether the absence of either of these WT T cell subsets affected tolerogenic potential in mixed BM chimaeras. For these experiments, mixed chimaeras consisting of CD4- or CD8-deficient along with Ctla4^−/− BM-derived cells were generated. CTLA-4–sufficient Cd4^−/− or Cd8^−/− BM cells were able to regulate Ctla4^−/− T cells in mixed BM chimaeras, as most recipient chimaeras remained healthy for > 6 mo (Table I). Also, no significant phenotypic differences were detected between the Ctla4^−/− and CTLA-4⁺ T cells in the respective chimaeras (not depicted). This data initially suggested either that CD4⁺ or CD8⁺ T cells were sufficient for regulation or that CD4⁻CD8⁻ T cells may be responsible. Interpretation of these results was hampered by the fact that lack of CD8 or CD4 coreceptors does not preclude the development of T cells positively selected on MHC class I or class II molecules, respectively (44, 45) (see below).

Given the requirement for TCRβ-expressing cells and a possible involvement of CD4⁻8⁻ αβ TCR⁺ T cells, it was formally possible that NKT cells were responsible for regulation. To test this, WT:Ctla4^−/− mixed chimaeras were injected weekly with PK136 mAb that has been shown to deplete NK1.1-expressing cells (NKT and NK cells) in vivo (46). NK1.1⁺ cell-depleted WT:Ctla4^−/− chimaeras remained healthy for 2 mo, and analyses of lymphocytes in the treated mice confirmed that Ctla4^−/− T cells remained quiescent (unpublished data). Thus, NKT and NK cells are not required for trans-regulation.

Depending on the model system, induction of T cell tolerance has been demonstrated to have short- or long-lived consequences on subsequent T cell responses. The question therefore arises whether tolerance in this model system is irreversible such that, once established, regulation no longer requires continuous suppression by WT T cells. To address this we first tested whether the established in vivo regulation of Ctla4^−/− T cells in mixed BM chimaeras could be reversed by the selective depletion of WT T cells. To this end, WT:Ctla4^−/−mixed BM chimaeras distinguished by the CD90 (Thy1) allele were generated. Upon establishing a stable peripheral T cell repertoire (>10 wk after reconstitution), WT T cells (CD90.1) were selectively depleted using an anti-CD90.1–specific Ab. To confirm effective depletion of WT T cells, mice were bled 2 wk after the initiation of Ab administration. By this time, >90% of T cells in both the CD4⁺ and CD8⁺ subsets were CD90.2⁺, indicating effective depletion of CD90.1⁺ WT T cells (unpublished data).

3–4 wk after initiating depletion, mice began to show symptoms of sickness similar to Ctla4^−/−-only BM recipients (ruffled fur, inflammation around the ears and eyes, and loose stools indicative of enterocolitis). In contrast, CD90.1 WT: CD90.2 WT control chimaeras that were depleted of CD90.1⁺ WT cells showed no visible signs of illness and appeared healthy throughout the study. Animals were killed 8 wk after WT T cell depletion, and lymphocytes were analyzed for activation status as shown in Fig. 3 A. Most T cells from the WT (CD90.1)-depleted WT:Ctla4^−/− mixed BM chimaeras were activated (CD44^hiCD69⁺CD62L^lo) in contrast to the depleted WT:WT, which maintained a predominantly naive phenotype. Furthermore, almost a third of the Ctla4^−/− cells were now cycling, as indicated by the threefold increase in total LN and spleen cell numbers, primarily in the CD4⁺ T cell compartment. (Fig. 3 B). This contrasts with WT T cells from depleted control chimaeras that had a relatively low frequency of cycling cells. Histological analyses confirmed that Ctla4^−/− T cells had regained overt autoreactive characteristics as indicated by infiltration into various nonlymphoid organs (unpublished data). These data reveal that trans-regulation depends on the persistent presence of WT T cells and that the regulated Ctla4^−/− T cells are not permanently altered to prevent rampant autoreactivity. It is possible that disease induction is mediated solely by recent Ctla4^−/− thymic emigrants not subject to peripheral regulatory processes caused by the depletion of the WT T cells. However, this seems unlikely given the rapid kinetics of disease (∼3 wk) and the high proportion of activated peripheral T cells. Moreover, transfer of regulated Ctla4^−/− CD4⁺ T cells from mixed BM chimaeras into recipients without the appropriate WT T cells results in aberrant T cell activation as described below, which is consistent with a reversible autoreactive state of Ctla4^−/− T cells held in check by CTLA-4–sufficient T cells.

Although the BM chimaera system is useful in studying the overall development and/or regulation of hematopoetic cells, it is difficult to evaluate the importance of specific T cell subsets in the regulation of naive Ctla4^−/− T cells. Instead, adoptive transfer of sorted peripheral T cell subsets into lymphopenic hosts was chosen to characterize the importance of various T cell subsets. Previously, Tivol et al. (47) showed that WT T cells rapidly eliminate cotransferred activated Ctla4^−/− T cells in the transfer system, suggesting that distinct mechanisms of regulation operate on naive versus activated Ctla4^−/− T cells in vivo. To avoid this “deletional” mechanism of tolerance, we sought to isolate naive Ctla4^−/− T cells. Several potential in vivo sources of unactivated Ctla4^−/− T cells exist, such as day 2 or 3 Ctla4^−/− neonates or co-stimulation–impaired Ctla4^−/− mice (e.g., B7^−/−Ctla4^−/− or Cd28^−/−Ctla4^−/−). However, we deemed these sources problematic because of the difficulty in obtaining sufficient numbers of naive T cells from Ctla4^−/− neonates, and in the latter case, because of the possibility that T cells from these double-deficient mice may be abnormal because they are generated and maintained in a co-stimulation-deficient environment. We therefore chose to obtain naive Ctla4^−/− T cells from regulated WT:Ctla4^−/− mixed BM chimaeras.

We first tested the feasibility of the transfer assay for trans-regulation of naive T cells by performing bulk transfers of LN and/or spleen cells from healthy mixed BM chimaeras to Rag1^−/− (or Tcrb^−/−; similar data not depicted) recipients. Although the T cells rapidly up-regulated activation markers caused by homeostatic expansion, there was no significant alteration in the relative ratios of WT:Ctla4^−/− T cells and no phenotypic difference between the transferred WT or Ctla4^−/− T cells (Fig. S2, available). The recipient mice remained healthy for several months and, unlike transfers using activated Ctla4^−/− T cells, maintained Ctla4^−/− T cells in the periphery, demonstrating efficient regulation upon transfer. In contrast, transfer of LN/spleen cells depleted of WT T cells (using CD90.1 congenic marker) resulted in lymphoproliferation and wasting, confirming the requirement for the continuous presence of WT T cells for regulation (unpublished data).

To determine whether regulation could be established by either CD4⁺ or CD8⁺ αβ T cells, sorted CD4⁺ T cells from WT:Ctla4^−/− mixed BM chimaeras were transferred alone (Ctla4^−/− or WT) or together with either WT CD4⁺ or CD8⁺ T cells from C57BL/6 mice. Transfer of naive Ctla4^−/− (but not WT) CD4⁺ T cells alone resulted in lymphoproliferation and a wasting syndrome (data not shown). WT CD8⁺ T cells were unable to regulate Ctla4^−/− CD4⁺ T cells, as mice receiving sorted WT CD8⁺ T cells and naive Ctla4^−/− T cells became ill ∼4 wk after transfer. In contrast, mice receiving WT CD4⁺ T cells with Ctla4^−/− T cells remained healthy, with no detectable signs of wasting. In all conditions, T cells were CD44^hiCD62L^lo because of homeostatic proliferation (Fig. 4 A). Upon sacrifice, mice receiving Ctla4^−/− CD4⁺ and WT CD8⁺ T cells displayed a marked increase in T cell number (approximately sixfold) compared with mice receiving Ctla4^−/− CD4⁺ and WT CD4⁺ T cells (Fig. 4 B). These data demonstrate that WT CD4⁺, but not CD8⁺, T cells are necessary and sufficient to regulate Ctla4^−/− T cells. Our initial observation of regulation in Cd4^−/−:Ctla4^−/− mixed chimaeras suggested that non-CD4⁺ T cells can regulate Ctla4^−/− T cells. However, a more likely interpretation attributes regulation to MHC class-II restricted T cells or functional FOXP3⁺ T cells found in Cd4^−/− mice (unpublished data) (44, 48).

Among CD4⁺ T cells, CTLA-4⁺ FOXP3⁺ T reg cells were the most likely population necessary for regulating Ctla4^−/− T cells. In BM chimaeras consisting of Cd28^−/−:Ctla4^−/− cells, no trans-regulation was observed (Table I). Cd28^−/− mice have approximately fivefold reduced number of T reg cells (49), but they do not develop autoimmune disease, reflecting the necessity of CD28 signaling for naive T cell activation. T cell composition in Cd28^−/−:WT mixed chimaeras reflected the initial input ratios, demonstrating that Cd28^−/− T cells are not significantly disadvantaged during reconstitution in the presence of WT T cells (unpublished data). These results suggested that either normal T reg cell numbers or CD4⁺ T cell subsets other than T reg cells with normal CD28 signaling is required for the trans-regulation.

To directly test the functional relevance of T reg cells, we cotransferred WT CD4⁺25⁺ T reg cells (≥90% FOXP3⁺) with Ctla4^−/− T cells from healthy mixed BM chimaeras to Rag1^−/− recipients in an ∼1:10 ratio to mimic the normal frequency of T reg cells:conventional T cells. Mice receiving Ctla4^−/− cells alone became visibly sick by ∼3-4 wk of age despite the fact that ∼10% of the CD4⁺ T cells were FOXP3⁺; the same frequency as the starting donor population (Fig. 5 A). In contrast, mice infused with Ctla4^−/− T cells and WT T reg cells remained healthy for several months showing no signs of lymphoproliferation or aberrant T cell infiltration into tissues.

Comparative analyses of the regulated Ctla4^−/− T cells in the transfer system at 3–4 wk post-transfer showed that the T reg cells suppressed the expansion of both CD4⁺ and CD8⁺ Ctla4^−/− T cells. In the spleen and LNs, there was an approximately two- and sixfold increase in cell numbers, respectively, in mice receiving Ctla4^−/− T cells only compared with those reconstituted with Ctla4^−/− and WT T reg cells (Fig. 5 B). The suppression was more pronounced for CD4⁺ T cells, with ∼10-fold difference in number in the LNs. Given the varying degrees of T cell infiltration into nonlymphoid tissues in mice receiving only Ctla4^−/− T cells (unpublished data), the extent of expansion noted in lymphoid tissues is most likely an underestimate of the overall number of Ctla4^−/− T cells present in the mice. The activation status of Ctla4^−/− CD4⁺ T cells with our without T reg cells revealed negligible differences in CD4⁺ T cells, most likely as a result of homeostatic proliferation. In contrast, the CD8⁺ T cell compartment displayed a significant reduction in the frequency of “activated” cells (CD44^hiCD62L⁻) and an approximately twofold reduction in BrdU uptake when CTLA-4–sufficient T reg cells were cotransferred (Fig. 5 C). Collectively, these data demonstrate that CTLA-4–sufficient, but not CTLA-4–deficient, T reg cells can completely regulate autopathogenic Ctla4^−/− T cells in vivo.

We next investigated the potential mechanisms involved in regulating Ctla4^−/− T cells by CTLA-4–sufficient T reg cells in vivo. Based on the expression of B7.1/2 on activated T cells, it has been suggested that CTLA-4 on T reg cells can engage its ligand B7 on T cells and directly inhibit their activation (“inside-out” signal) (31, 32). Although initial reports indicated that B7-deficient T cells were resistant to T reg cell–mediated suppression in vivo, more recent studies using different in vivo models have generated inconsistent data (30).

To test whether CTLA-4 sufficient T reg cells control Ctla4^−/− T cells via B7, we generated mixed BM chimaeras using stem cells from WT and Cd80^−/−Cd86^−/−Ctla4^−/− (B7^−/−Ctla4^−/−) mice. 2 mo after reconstitution, both WT:Ctla4^−/− and WT:B7^−/−Ctla4^−/− cohorts were healthy with no signs of inflammation or wasting (Table I). B7^−/−Ctla4^−/− T cells in the mice were maintained in a naive state, indistinguishable from WT cells, as assessed by expression of CD44 and CD62L (unpublished data). In conjunction with published data showing that B7^−/− and B7^−/−Ctla4^−/− can be controlled in transfer models by WT T reg cells (30) (unpublished data), the regulation of Ctla4^−/− T cells by CTLA-4–sufficient T reg cells in vivo does not require B7 signaling in the T cells.

Although our data demonstrated that trans-regulation requires the continuous presence of CTLA-4–sufficient T reg cells, it was not known whether they were molecularly altering the target Ctla4^−/− T cells or inhibiting activation by preventing access to antigens and/or cytokines. To distinguish between these possibilities, we performed global gene expression profiling of naive Ctla4^−/− and WT T cells sorted from mixed BM chimaeras.

Overall, the differences were minimal (∼1% of expressed genes, ∼70/6000 assayed; unpublished data) and the gene expression signature of regulated naive Ctla4^−/− T cells closely resembled naive WT T cells from the same animal. At the transcriptional level, there was no evidence of active suppression (50, 51) or altered TCR/CD28 signaling on Ctla4^−/− T cells. Further analyses of select genes identified as altered in expression by microarray analyses (Fig. 6) revealed that Gadd45β (Myd118) was decreased in expression by approximately fourfold in Ctla4^−/− T cells compared with the WT counterpart, whereas Gadd45α and Gadd45γ expression were not significantly altered (not depicted). Gadd45β is required for normal responses to TCR and inflammatory cytokine stimulations as indicated by the development of autoimmunity in its absence (52). In contrast, expression of other genes previously implicated in T cell suppression were not altered in naive Ctla4^−/− T cells as illustrated by similar transcript levels for the representative genes in Fig. 6. These results suggest that CTLA-4–sufficient T reg cells indirectly prevent naive Ctla4^−/− T cells from becoming fully activated, possibly by regulating the immune stimulatory activities of APCs.

The two best-characterized effectors of T reg cells are the cytokines TGF-β and IL-10. Given the absence of alterations in TGF-β production by the T cells and candidate TGF-β target genes in regulated Ctla4^−/− T cells, trans-regulation appeared not to extensively involve suppressive cytokines acting specifically on Ctla4^−/− T cells. Consistent with this, WT:Ctla4^−/− mixed chimaeras injected with blocking Abs to TGF-β had minimal alterations at 8 wk of treatment, and even at 3 mo only some of the mice showed increased frequency of activated T cells in the both WT and Ctla4^−/− populations (unpublished data).

Similarly, regulation of Ctla4^−/− T cells also deficient in IL-10 receptor–mediated signaling appeared largely intact as mixed WT:Il10ra^−/−Ctla4^−/− BM chimaeras survived and maintained stable T cell subset frequencies (Fig. 7). Interestingly, the IL10ra^−/−Ctla4^−/− subset had a lower frequency of naive T cells in the CD4⁺ (Fig. 7 A), and especially in the CD8⁺ (Fig. 7 B) T cell populations, but over time the differences became statistically insignificant (data not shown). Thus, IL-10R signaling in Ctla4^−/− cells is dispensable for the establishment of trans-regulation. The functional redundancy of IL-10R signaling and TGF-β and the potential for long-term regulation of Ctla4^−/−Il10Ra^−/− T cells are currently under investigation.

Using three in vivo model systems, mixed genotype blastocyst chimaeras, mixed BM cell chimaeras, and adoptive T cell transfer into lymphopenic mice, we have conclusively demonstrated that naive Ctla4^−/− T cells are regulated in trans by CTLA-4–sufficient T reg cells. The first model system does not involve whole body irradiation, ruling out experimental artifacts such as a dysregulated cytokine environment in irradiated mice as a caveat to the physiological relevance of the observed Ctla4^−/− T cell homeostasis imposed in trans. Using the latter two model systems we have identified CTLA-4–sufficient CD4⁺CD25⁺ T reg cells as the subset necessary for trans-regulation of naive Ctla4^−/− T cells. Because CTLA-4–deficient FOXP3⁺ T reg cells in Ctla4^−/− mice cannot prevent the lymphoproliferative disease, the results definitively show that CTLA-4 is essential for in vivo T reg cell–mediated regulation of naive T cell activation. This function of CTLA-4 is not simply restricted to controlling Ctla4^−/− T cells, because a specific conditional loss of CTLA-4 in T reg cells in mice results in systemic lymphoproliferation whether or not naive T cells can express CTLA-4 (unpublished data).

Previous studies using mixed BM chimaeras had established that CTLA-4–sufficient cells can function in trans to maintain Ctla4^−/− T cells in a naive state (22). Using the colitis model system, it was shown that naive B7^−/−Ctla4^−/− effector T cells were prevented from causing colitis by B7^−/− CTLA-4–sufficient T reg cells, but not by B7^−/−Ctla4^−/− T reg cells (30). A major caveat to these studies was that the few T reg cells arising from B7-deficient mice may have altered properties caused by the loss of CD28 and CTLA-4 signaling during their development (53). More problematic was the observation that B7^−/−Ctla4^−/− T reg cells or sorted Ctla4^−/− T reg cells from WT:Ctla4^−/− mixed BM chimaeras were reported to be effective in regulating colitogenic CD4⁺ T cells, indicating that CTLA-4 is in fact dispensable for T reg cell function. To account for this discrepancy, it was suggested that Ctla4^−/− T reg cells can adapt to the loss of CTLA-4 and use alternate, compensatory mechanisms of immune suppression, such as the enhanced prominence of IL-10 as the suppressive factor (25, 30). The relevance of these findings to the regulation of Ctla4^−/− T cells in vivo was unclear because even the proposed “adapted” Ctla4^−/− T reg cells are incapable of regulating Ctla4^−/− T cells in unmanipulated Ctla4^−/− mice. Hence, the importance of T reg cells, or other immunoregulatory T cell subsets, in the control of Ctla4^−/− T cells in trans and the potential regulatory mechanisms involved remained ambiguous.

Here, we have presented a systematic characterization of effector T cell–extrinsic CTLA-4 function to maintain T cell tolerance. Only T reg cells expressing CTLA-4 can restrain the autoreactive T cells in vivo. Regulation is long-lasting, and depending on the persistent presence of T reg cells, is stable in the face of infection or immunity against alloantigens. The equal participation of Ctla4^−/− and WT T cells in immune responses against acute or chronic pathogen infection suggests that CTLA-4 is dispensable on effector T cells during the initiation and resolution of primary inflammatory T cell responses against foreign antigens.

The results, however, do not exclude T reg cell–independent functions of CTLA-4 and, conversely, CTLA-4–independent T reg cell effector mechanisms in the maintenance of T cell tolerance. There are three molecules whose absence in vivo causes early onset, lethal lymphoproliferative disease: CTLA-4, FOXP3, and TGF-β (54, 55). All three have immune-suppressive functions and, at this point, it is unclear whether they are components of a single pathway or parallel converging pathways. Current data would suggest that both hypotheses may be true. We have demonstrated that FOXP3⁺ T reg cells require CTLA-4 for in vivo regulatory functions. However, CTLA-4 also appears to play a role in regulating effector T cell activation independent of T reg cells because CTLA-4 expression targeted specifically in activated conventional T cells (but not naive or T reg cells) can delay fatality of Ctla4^−/− mice (unpublished data). TGF-β is required for the maintenance and function of T reg cells (25, 56, 57). Although T reg cells produce TGF-β, studies have also shown that T reg cell–derived TGF-β is not essential for T cell homeostasis as long as other cells are capable of its production (58). Further, TGF-β has T reg cell–independent functions in regulating T cell development and homeostasis (59–61). Hence, although FOXP3⁺ T reg cells use CTLA-4 and TGF-β as effectors of immune suppression, CTLA-4 and TGF-β also maintain T cell homeostasis and self-tolerance in a T reg cell–independent manner. Conversely, FOXP3⁺ T reg cells use immune suppressive pathways other than CTLA-4 and TGF-β to regulate T cell activation both in vitro and in vivo (62).

Data presented here do not provide a definitive identification of the mechanism(s) responsible for CTLA-4⁺ T reg cell–mediated trans-regulation of Ctla4^−/− T cells. However, the fact that, (a) TGF-β, IL-10, and B7 signaling are not singularly required in Ctla4^−/− T cells, (b) Il35 expression (63) was normal in Ctla4^−/− T reg cells (unpublished data), and (c) Ctla4^−/− and WT T cells displayed highly similar global gene expression patterns, strongly suggests that T regs do not act directly on Ctla4^−/− T cells. We thus favor the hypothesis that T reg cells control Ctla4^−/− T cells by limiting immunogenicity of APCs in a noninflamed setting, a well-documented mechanism of T reg cell function in vitro (64, 65) and in vivo (66). In particular, several studies have suggested that the immunosuppressive enzyme IDO is produced by DCs upon encounter with T reg cells expressing functional CTLA-4, although the in vivo requirement for IDO in T reg cell–mediated immunosuppression has not been established (67, 68).

There are other possible mechanisms by which Ctla4^−/− T cells could be maintained in a quiescent state by CTLA-4–sufficient T reg cells. It has been suggested that DCs can secrete exosomes containing MHC and co-stimulatory molecules that can be transferred intercellularly to other DCs to stimulate T cells (69). Hence, we have considered the possibility that CTLA-4–expressing T cells produce exosomes containing CTLA-4 that can be taken up by Ctla4^−/− T cells and used to intrinsically maintain regulation. We have been able to both detect exosomes containing CTLA-4 and transfer CTLA-4 protein to Ctla4^−/− CD4⁺ T cells in vitro, but have yet to obtain any direct in vivo evidence for this mechanism in the mixed chimaera model (unpublished data).

Alternatively, because CTLA-4 has been estimated to have 50–100-fold higher affinity to B7 than CD28, it is conceivable that in the mixed chimaeras CTLA-4–sufficient T cells, especially T reg cells, are competitively engaging available B7 molecules on the APCs, thereby preventing CD28 signaling, and subsequent T cell activation in Ctla4^−/− T cells. Physical competition alone appears unlikely given that as few as 10% of total T cells of WT origin can control the activation of the remaining 90% Ctla4^−/− T cells. Nevertheless, we have observed that in B7.2:Ctla4^−/− transgenic (overexpressing B7.2 on T and B cells) mixed BM chimaeras, tolerance was not established (unpublished data), consistent with a B7 “competition model” that predicts an increase in the probability of CD28 triggering in Ctla4^−/− T cells that could result in a breakdown in regulation. However, as B7.2 transgenic T cells are expanded and show signs of activation, the absence of regulation in these mixed chimaeras could alternatively be a direct consequence of aberrant T cell homeostasis in the tolerizing (Ctla4^+/+) T cell population rather than increased access to B7 ligands. Ongoing studies with more refined cell type–dependent, temporal manipulation of relevant co-stimulatory ligands and receptors will help delineate the mechanism(s) of trans-regulation involving non-T accessory cells.

All mice were bred and maintained in our specific pathogen–free animal facilities. All animal experiments were approved by the University of Massachusetts Medical School Institutional Care and Use Committee. Ctla4^−/− mice were backcrossed for 12 generations onto the C57BL/6 background. CD45.1⁺ or CD90.1⁺ congenic C57BL/6 mice were crossed with Ctla4^+/− mice to generate a source of allelically distinguished Ctla4^−/− cells in chimaeras and during transfer experiments. Rag1^−/−, Tcrb^−/−, Tcrd^−/−, Cd8a^−/−, and Cd28^−/− mice were obtained from The Jackson Laboratory. Cd80^−/−Cd86^−/− double-knockout mice (B7^−/−) were obtained from A. Sharpe (Harvard Medical School, Boston, MA). Cd4^−/− mice were obtained from N. Killeen (University of California, San Francisco, San Francisco, CA). 2CT, OTI, and OTII TCR transgenic mice were provided by K. Rock (University of Massachussetts Medical School, Worcester, MA). IL-10Rα mutant mice were generated by ENU mutagenesis and contain a single base mutation that results in a stop codon at amino acid 220 of the IL-10Rα chain (unpublished data). Similar to Il10^−/− mice, Il10ra^N/N mice (referred to as Il10ra^−/− here for simplicity) suffer from severe inflammatory bowel syndrome. The mice were backcrossed >10 generations to C57BL/6 and then crossed to Ctla4^−/− mice to generate Il10ra^−/−Ctla4^−/− double-deficient mice.

Blastocyst chimaeras were created by injecting Ctla4^−/− embryonic stem cells (12) into C57BL/6 blastocysts and reimplanting the embryos into surrogate mothers. Resulting progeny were periodically bled over 3 yr, and the frequency of Ctla4^−/− T cells monitored by enumerating the percentage of Ly9.1⁺ T cells (Ctla4^−/−) versus Ly9.1⁻ T cells (Ctla4^+/+) via flow cytometry.

The femurs and tibias of donor mice were flushed with RPMI-1640 supplemented with 5% FBS (HyClone; Invitrogen), sodium pyruvate, l-glutamine, β2-mercaptoethanol, nonessential amino acids, and 100 U/ml penicillin/streptomycin (R5) using a 25-gauge needle. BM was mechanically separated by gentle pipetting, and RBCs were removed by a 1-min lysis step (0.15 M NH₄Cl, 1 mM KHCO₃, and 0.1 mM EDTA, pH 7.3). After lysis and washing, BM cells were resuspended in 2.5 ml of R5 and depleted of T cells using anti-CD90.2 and/or a mixture of anti-CD4/CD8–coated magnetic beads (Invitrogen). BM was analyzed by flow cytometry for effective depletion of T cells (with anti–TCR-β-FITC), washed 2 times in PBS, and adjusted to 2 × 10⁷ cells/ml. Mice were sublethally irradiated (300 rads) and injected with ∼4 × 10⁶ BM cells i.v. 4 wk after transfer, peripheral blood samples were obtained and analyzed for T cell reconstitution by flow cytometry.

LNs (pooled from superficial, deep cervical, axillary, brachial, inguinal, mesenteric, and lumbar) and spleen were isolated and homogenized via glass slides (or using a tissue homogenizer for sorting experiments), and RBCs were lysed. Cells were washed, filtered, and resuspended in R5. Transfers involving sorted CD4⁺ or CD8⁺ T cells were performed as follows: LN and spleen cells were isolated from WT mice or mixed BM chimaeras, pooled, labeled with fluorescently conjugated CD4 and CD8 Abs plus a congenic marker (to differentiate Ctla4^−/− T cells in mixed chimaeras), and sorted on a MoFlo high-speed sorter (Dako). After sorting, cells were washed in PBS and ∼1 × 10⁷ total T cells were transferred to Rag1^−/− recipients (subsequent 1:1 mix of WT CD4⁺ or CD8⁺ T cells and CD4⁺ Ctla4^−/− T cells injected on consecutive days after sorting).

For T reg cell cotransfer studies, LN and spleen cells from healthy mixed BM chimaeras (Ctla4^−/−:CD90.1⁺WT) were labeled with CD90.1-biotin, followed by streptavidin magnetic beads, and then purified on an AutoMACS System (Miltenyi Biotec) to deplete WT T cells. After purification, T cells were typically >96% CD90.1⁻. To obtain T reg cells, LN and spleen cells were isolated from C57BL/6 mice, labeled with Abs against CD4 and CD25, and sorted on a MoFlo high-speed sorter. Sorted purity was routinely >95%, of which ∼90% were FOXP3⁺, as determined by intracellular staining. Cells were washed in PBS, counted, and injected i.v. (2 × 10⁷ cells/mouse) at approximately a 10:1 ratio of WT T cell–depleted cells from mixed chimaeras to T reg cells.

Abs were purified from supernatants of the following hybridomas using protein G–Sepharose columns: anti-CD90.1 (clone OX7), anti-NK1.1 (PK136), and anti–TGF-β (2G7). For WT T cell depletions, mice received three consecutive injections of OX7 mAb, followed by weekly injections to maintain depletion (0.25 mg i.p.). For NK depletions, mice received weekly injections of PK136 Ab (0.025 mg i.p.). For TGF-β blockade, mice received 0.25 mg of 2G7 Ab every 4 d.

Single-cell suspensions at 10⁶ cells/sample were preincubated in 2.4G2 hybridoma supernatant to block FcγR binding for 15 min, and then incubated with various combinations of the following conjugated Abs for 30 min: Ly9.1-FITC, CD25-FITC, FoxP3-FITC, CD69-FITC, CD8-PE, CD62L-PE, CD4-PECy5, CD4-PerCPCy5′5, CD25-PECy7, CD45.1-PECy7, CD8-APCCy7, CD44-APC, CD45.1-biotin, CD90.1-biotin, Streptavidin-ECD, Streptavidin-Pacific Blue. Abs used for flow cytometric analysis were purchased from BD or eBioscience. Cells were washed, fixed in 4% paraformaldehyde for 10 min, washed again, and then resuspended in FACS buffer (1X PBS/1.5% FCS/0.05% NaN₃). Samples were analyzed on either an Epics XL-MCL (Beckman Coulter) or LSRII flow cytometer (BD). For intracellular FOXP3 staining, the manufacturer's protocol was followed (eBioscience). For peripheral blood analyses, ∼50–100 µl of blood from the tail vein was mixed with heparin, stained with fluorochrome-conjugated antibodies, briefly incubated in 0.25% saponin, fixed in 4% paraformaldehyde for 10 min, and resuspended in FACS buffer for analysis. At least 100,000 events were collected (20,000–50,000 events for blood) for each sample, and data were analyzed using FlowJo software (Tree Star, Inc.).

Animals received twice daily injections of 1 mg of BrdU i.p. for 2 d before sacrifice. After cell surface marker staining, cells were incubated in 50 μl of 5 M NaCl, followed by the slow addition of 120 μl of 95% EtOH. Cells were then stored at –20°C overnight, washed, and incubated in 4% paraformaldehyde for 20 min on ice. Subsequently, cells were incubated in DNase solution (1X PBS/0.05% Tween-20/125 U DNase I/sample) for 15 min at room temperature, and then for 15 min on ice. Samples were washed once in PBS, once with labeling buffer (1X PBS/2% FBS/0.05% Tween-20) and incubated with anti–BrdU-FITC (BD) for 30 min. After being washed 2 times, cells were immediately analyzed by flow cytometry.

Healthy mixed BM chimaeras 12 wk or more after reconstitution were used for these studies. For LCMV infections, chimaeras were challenged i.p. with 5 × 10⁴ PFU of strain Armstrong virus. Activation status, T cell ratios, and antigen-specific splenic T cell responses were analyzed at 7–9 d after infection. For Mtb infection, sonicated Mtb Erdman stock was titrated to deliver ∼100 CFU per mouse. Mice were exposed to infectious aerosol for 30 min in a Glass-Col inhalation Exposure System (Glass-Col, LLC). Upon sacrifice at 6 or 14 wk after infection, mice were immediately perfused with PBS. Lungs were minced, digested with collagenase and DNase for 30 min at 37°C, and strained, and RBCs were lysed. Spleens were also removed for analysis. Cells were then analyzed by flow cytometry.

Naive (CD44^loCD62L^hi) Ctla4^−/− or WT CD4⁺ T cells (>95% pure) from healthy mixed BM chimaeras (pooled T cells from 3–5 mice/sort) were sorted on a MoFlo high-speed sorter using Abs to CD44, CD62L, and the congenic marker CD45.1. Cells were counted and resuspended in Trizol (Invitrogen). mRNA was isolated as per the manufacturer's protocol. cDNA preparation was performed with the OmniScript RT kit following the manufacturer's instructions (QIAGEN). Serial dilutions of cDNA were PCR amplified using the following primers: Actb, 5′-CTAGGCACCAGGGTGTGATGG-3′ and 5′-TCTCTTTGATGTCACGCACGA-3′; Cblb, 5′-TTCCAGATGGCAAATTCTATG-3′ and 5′-TACATCCTTTCCTTGCCTTCTTTA-3′; Gadd45b, 5′-TACGAGGCGGCCAAACTGATGAAT-3′ and 5′-ACCCATTGGTTATTGCCTCTGCTC-3′; Lklf, 5′-CTGGAGGCCAAGCCCAAACGCGGC-3′ and 5′-CGTTGGGGACAGTAAACTCAAAGGCA-3′; Nfatc2, 5′-TGCTGTTCTCATGGATGCCCTCAA-3′ and 5′-TCCTCTCATCTGCTGTCCCAATGA-3′; Pou2f1, 5′-ACGCACCAGCATAGAGACCAACAT-3′ and 5′-TCACAGCAGCACTGGTTAAAGGGA-3′; and Prkcz, 5′-TCCTTGATGCCGACGGACACATTA-3′ and 5′-GCTGGTGAACTGCGTGTCAAAGTT-3′.

Tissues (pancreas, small intestine, liver, lung, heart, tongue, and ear) were fixed in 10% buffered formalin, paraffin embedded, sectioned, and stained with hematoxylin and eosin. Two sections, 90 µm apart, were cut from each block, and slides were viewed by light microscopy.

Fig. S1 demonstrates that challenge with the bacterial pathogen Mtb does not abrogate regulation of Ctla4^−/− T cells in mixed BM chimaeras. Fig. S2 demonstrates that trans-regulation of Ctla4^−/− T cells from mixed BM chimaeras is maintained in the presence of WT T cells upon adoptive transfer to lymphopenic hosts.

We would like to thank the UMMS Flow Cytometry Core Facility staff for their dedicated efforts in cell sorting and the Genomics Core for microarray data acquisition and analysis. We thank Airiel Davis for excellent technical assistance. We thank the following investigators for providing essential mouse lines: Arlene Sharpe, Kenneth Rock and Nigel Killeen. The blastocyst chimaeras were generated in the laboratory of Dr. James Allison (at UC Berkeley, currently at Memorial Sloan-Kettering). We thank Dr. Tom Markees for performing skin graft experiments, Dr. Leslie Berg for discussion and critical reading of the manuscript, and Dr. Michael Brehm for performing LCMV infection studies. We also gratefully acknowledge the support of the UMMS Diabetes and Endocrinology Research Center and their staff, especially in providing purified Abs used for in vivo studies. This work was supported by grants from National Institutes of Health (AI054670, AI59880).

The authors declare that they have no conflicting financial interests.