|Home | About | Journals | Submit | Contact Us | Français|
CTLA-4 is constitutively expressed by CD4+CD25+Foxp3+ regulatory T cells (Treg) but its precise role in Treg function is not clear. Although blockade of CTLA-4 interferes with Treg function, studies using CTLA-4 deficient Treg have failed to reveal an essential requirement for CTLA-4 in Treg suppression in vivo. Conditional deletion of CTLA-4 in Foxp3+ T cells disrupts immune homeostasis in vivo but the immune processes disrupted by CTLA-4 deletion have not been determined. We demonstrate that Treg expression of CTLA-4 is essential for Treg control of lymphopenia-induced CD4 T cell expansion. Despite IL-10 expression, CTLA-4 deficient Treg were unable to control the expansion of CD4+ target cells in a lymphopenic environment. Moreover, unlike their WT counterparts, CTLA-4-deficient Treg failed to inhibit cytokine production associated with homeostatic expansion and were unable to prevent colitis. Thus, while Treg developing in the absence of CTLA-4 appear to acquire some compensatory suppressive mechanisms in vitro, we identify a non-redundant role for CTLA-4 in Treg function in vivo.
Naturally occurring Foxp3+ regulatory T cells (Treg) constitutively express CTLA-4. Antibody blockade of CTLA-4 abrogates Treg activity in some in vitro and in vivo settings [1–3]. Moreover the lymphoproliferative disease associated with CTLA-4 deficiency [4, 5] can be controlled by provision of CTLA-4-sufficient T cells [6, 7], suggesting functional defects in Treg in the absence of CTLA-4. Despite these observations, purified CTLA-4-deficient CD4+CD25+(Foxp3+) Treg retain potent suppressive activity in vitro and prevent the development of colitis in vivo [2, 8], supporting the idea that CTLA-4 expression is not an absolute requirement for suppression. To reconcile these differences it is postulated that Treg developing in a CTLA-4-deficient setting develop compensatory suppressive mechanisms; more TGF-β-dependent in vitro  and more heavily dependent on IL-10 in vivo . Recently, the conditional deletion of CTLA-4 in Foxp3+ T cells has strongly implicated the requirement for CTLA-4 expression by Treg in the control of immune homeostasis .
The degree to which Treg control homeostatic proliferation and/or lymphopenia-induced proliferation (LIP) is controversial. Treg have been shown to control entry into cell cycle or control survival of expanded cells depending on the TCR affinity of the target cell . The exact mechanism(s) of Treg control of LIP is poorly understood but may require Treg production of IL-10 . No single molecule has been identified that accounts for all Treg actions therefore the mechanisms used to control LIP, antigen-induced proliferation and effector function are likely to be distinct yet complementary [12, 13].
It has been difficult to directly test the contribution of the CTLA-4 molecule to Treg suppressive function as CTLA-4 KO mice die of lymphoproliferative disease at 3 weeks of age. CTLA-4−/− mice have been crossed to CTLA-4-Ig transgenics or to B7−1/B7−2 knock-out mice to study KO cells in the absence of lymphoproliferation [2, 8]. Both strategies limit lymphoproliferation but also manipulate molecules that may themselves alter T cell (Treg or effector) development and/or function. While the conditional deletion of CTLA-4 in Foxp3-expressing cells is a strong indicator of an absolute requirement for CTLA-4 in Treg function in vivo, the possibility that the pathology observed in these mice is due in part to the of loss of CTLA-4 on non-regulatory T cells transiently expressing Foxp3 following activation, as seen in human T cells , cannot be ruled out. We used young, un-manipulated, CTLA-4-deficient mice and isolated Treg based on CD62L expression to exclude activated non-Treg. We show that CTLA-4 KO Treg prevent CD4 T cell proliferation and cytokine production in vitro but, unlike the WT Treg, are unable to control lymphopenia-induced expansion and cytokine production of CD4+ T cells in vivo. This breakdown in CTLA-4 KO Treg function impaired their ability to prevent the development of colitis. Thus, although Treg developing in the absence of CTLA-4 may acquire compensatory suppressive mechanisms for in vitro suppression, CTLA-4's requirement for Treg control of lymphopenia-induced expansion is absolute.
CTLA-4 KO mice contained a high proportion of recently activated effector cells within the CD4+CD25+ compartment: 44% were Foxp3-negative (Fig. 1A, pre-sort). To exclude this population we used CD62L expression . The sorted CD4+CD25+CD62Lhigh population was highly enriched for Foxp3+ cells (Fig 1A, post-sort) and both WT and CTLA-4 −/− Treg were routinely >95% Foxp3+. In standard in vitro suppression assays, the CTLA-4 KO Treg suppressed proliferation similarly to WT Treg at multiple target/Treg ratios (Fig. 1B, 1C), as described by others [8, 15]. In addition, CTLA-4 KO Treg potently suppressed IFN-γ and IL-4 cytokine production in vitro (Fig 1D). Thus CD62Lhigh Treg enriched from a CTLA-4 KO environment retain functional activity in vitro.
Lympho-proliferative disease in CTLA-4 KO mice is likely due to both a lack of intrinsic control of T cell activation and a defect in Treg [6, 7, 9]. To directly test the ability of CTLA-4 KO Treg to control lymphopenia-induced expansion, WT Thy1.2+ CD4+ CFSE-labeled targets were transferred to lymphopenic hosts (Rag2−/−) with an equal number of Thy1.1+ WT or CTLA-4 KO Treg. To control for cell number, the `no Treg' group received an equal number of (non-regulatory) WT Thy1.1+ CD4+CD25−CD62Lhigh cells. Day 13 after cell transfer, cells from the lymph nodes and spleen of individual mice were counted and assessed for proliferation by CFSE dilution. Similar to results reported by others [10, 11], WT Treg (from either 2 week of 6 week donors) did not prevent the majority of targets cells from entering cell division (Fig 2A). However, WT Treg, but not KO Treg, did modulate the extent of division/survival with a reduced frequency of highly proliferated cells (Fig 2A, gate 1) and a relative increase in cells that had divided less than four times (Fig 2A, gate 2). Correspondingly, there was a marked reduction in the total number of target cells recovered from co-transfers with WT Treg (Fig 2B). Strikingly, CTLA-4 deficient Treg failed to limit target T cell expansion/survival (Fig 2B). Proliferation in lymphopenic animals can be divided into two mechanistically distinct phases ; a rapid “spontaneous” proliferation where cells completely dilute their CFSE (Fig 2A, gate 1), that is thought to be antigen-driven, and a slower cytokine-dependent phase (Fig 2A, gate 2). Based on the cell numbers recovered, WT Treg specifically affected the rapid spontaneous phase but not the slower cytokine-dependent phase (Fig 2C) . In the absence of CTLA-4, Treg failed to regulate the purported antigen-driven rapid proliferation (Fig 2C).
To determine the functional fate of cells regulated by Treg during LIP, we isolated the target cells from reconstituted mice and re-stimulated in vitro, in the absence of Treg. Targets cells that had expanded in vivo in the absence of Treg proliferated and made IL-2 while the response was markedly attenuated if co-transferred with WT Treg (Fig. 2D). In contrast, CTLA-4 KO Treg failed to alter subsequent target T cell function (Fig. 2D). Thus the CTLA-4 KO Treg, unlike their WT counterparts, were unable to render target T cells non-responsive. The difference in WT and CTLA-4 KO Treg function was not due to differences in their own survival/expansion (Fig 2E) or to differences in Foxp3 expression (Fig 2F). These data point to a non-redundant requirement for CTLA-4 expression by Treg for the in vivo suppression of lymphopenia-induced expansion of CD4+ T cells. The unchecked lympho-proliferation observed in CTLA4-deficient mice may therefore be in part due to a failure of CTLA-4-deficient Treg to control lymphocyte homeostasis. That CTLA-4 KO Treg retain the ability to suppress proliferation in vitro but not in vivo suggests that the signals for target and/or Treg activation must be qualitatively or quantitatively different in the two settings.
A number of cell surface molecules have been implicated in the functional activity of Treg including CD25, PD1, GITR and LAG3 . CTLA-4-deficient Treg expressed all of the molecules tested and in some cases (CD25, GITR, LAG3) to a much higher degree than WT Treg (Fig 3A). Thus we find no evidence for the loss of other molecules that might indirectly abrogate suppression by CTLA-4 KO Treg. There was however a marked increase in the expression of LAG3 by KO Treg. Although LAG3 negatively regulates T cell function, in its soluble form it can have stimulatory properties on dendritic cells . Thus over-expression of LAG3 by KO Treg may activate APC to enhance, rather than suppress, expansion of CD4+ T cells in the lymphopenic environment.
A number of mechanisms have been proposed for Treg function . For control of LIP in particular, IL-10 and IL-35 have been shown to play a role [11, 19]. Ex vivo, WT and CTLA-4 KO Treg expressed similar levels of mRNA for IL-10 and for both of the IL-35 chains, Ebi3 and IL-12α (Fig 3B), suggesting that the failure to control LIP by KO Treg does not lie at the level of differences in these immunosuppressive cytokines. Indeed, CTLA-4 KO Foxp3+ population contained a slightly higher frequency of IL-10 producers (Fig 3C) as previously described . CTLA-4 engagement can also enhance TGF-β production and/or activity [20, 21] so CTLA-4-deficient Treg may have attenuated TGF-β-mediated functions. TGF-β plays a key role in immune homeostasis but its role in modulating LIP is not known. To formally test this, we used target T cells unable to respond to TGF-β from the TGF-β dominant negative receptor transgenic mice (dnTGF-β RII) . The dnTGF-β RII targets proliferated better than WT targets on transfer to lymphopenic hosts but were still potently suppressed by WT Tregs (Fig 3D), demonstrating that suppression of LIP is independent of target cell TGF-β signaling. Thus potential changes in TGF-β production by CTLA-4 KO Tregs are unlikely to account for the loss of suppressive activity during LIP. This is in contrast to the dependency on target T cell responsiveness to TGF-β in Treg control of colitis , highlighting potential differences in the mechanisms for control of LIP and gut inflammation by Treg.
Treg CTLA-4 expression could also modify target T cell functions during LIP through the induction in dendritic cells of the tryptophan-degrading enzyme indoleamine 2,3-dioxygenase (IDO) . Indeed, CTLA-4-Ig treatment potently suppressed clonal expansion of a large number of alloreactive T cells in an IDO-dependent manner . To determine the role of this pathway in Treg control of LIP, lymphopenic hosts receiving WT CD4+ T cells with and without WT Treg were treated with the IDO-inhibitor D-1-methyl tryptophan throughout the 2 weeks LIP period . As shown in Fig 3E, the IDO-inhibitor did not disrupt LIP and failed to block CTLA-sufficient Treg control of LIP. Thus Treg control of LIP is IDO independent.
Many experimental systems of Treg control of immune pathology contain an element of lymphopenia-induced proliferation raising the possibility that the primary function of Treg in these models is to limit the degree of initial expansion . While a number of studies suggest that Treg control of homeostatic expansion and disease pathology can be independently regulated [11, 28, 29], the relationship between LIP and immune pathology is still not fully understood. Given the striking loss of Treg control of LIP in the absence of CTLA-4, we examined the effect of this loss of function on cytokine production and immune pathology, namely the development of colitis.
During LIP CD4 T cells acquire the ability to secrete effector cytokines that may impact on immune homeostasis. Indeed, both IFN-γ and IL-4 cytokine production by the CD4+ targets was potently inhibited by co-transfer of WT Treg (Fig 4A). In contrast, CTLA-4 deficient Treg failed to control cytokine production (Fig 4A). There were small but significant changes in the quality of the cytokine response by CD4+ T cells with the transfer of CTLA-4 KO Treg compared to no Treg transfer, with an increase in IL-4 and decrease in IFN-γ producers (Fig 4A). Nonetheless, the predominant cytokine producers remained IFN-γ positive on co-transfer of CD4 targets alone or with CTLA-4 KO Treg. We followed the long-term consequences of the brake-down in Treg function in the absence of CTLA-4 by analysis of subsequent development of colitis 8–10 weeks after cell transfer. Correlating with the loss of inhibition of proliferation and cytokine production in the lymphopenic host at 2 weeks, CTLA-4 KO Treg failed to prevent the development of colitis (Fig 4B and 4C). The pathology was accompanied by a 10–30% loss in body weight in both CD25– and CD25– + KO Treg groups. We did not observe pathology in other tissues such as the liver and lung. Thus, while CTLA-4 KO Treg can be shown to regulate immune function in vitro (Fig 1), they are significantly impaired in their ability to control both LIP and colitis in vivo.
The role of CTLA-4 in Treg function has remained elusive. Previous reports of the function of CTLA-4-deficient Treg have suggested that the Treg that develop in the absence of CTLA-4 have acquired compensatory suppressive mechanisms through enhanced TGF-β or IL-10-dependent pathways [2, 8]. We reveal a non-redundant requirement for CTLA-4 expression by Treg to limit lymphopenia-induced CD4 T cell expansion in vivo, which appears independent of the immuno-suppressive cytokines IL-10, IL-35 and TGF-β and independent of IDO induction.
Loss of control of LIP was accompanied by an inability of CTLA-4 KO Treg to prevent colitis. Our results are in contrast to the retention of Treg function for control of colitis in Treg isolated from B7–1/B7–2/CTLA-4 KO hosts or mixed bone marrow chimeras . Differences in both the genetic and experimental systems likely account for the difference in Treg function. For instance, while we isolated CD62LhighCD25+ cells from young CTLA-4 −/− mice we cannot rule out that changes in the early inflammatory environment in the absence of CTLA-4 (not present in the B7–1/B7–2/CTLA-4 KO mice) may modulate Treg function. In addition, one noticeable difference between our study and that of the Powrie group is in the phenotype of CD25+CD4+ T cell subset used. We purified CD62LhighCD25+ cells while the Powrie group isolated CD45RBlowCD25+ T cells. The CD4+CD25+Foxp3+ T cell population is phenotypically and functionally heterogeneous and CD45RBlowCD25+ T cells can be subdivided based on the expression of CD62L, indeed we find at least 30% of CD45RBlow cells are CD62Llow. Thus distinct Treg subsets may be differentially dependent on CTLA-4 for their action. It is also possible that, unlike the CD62Lhigh population in this current study, the CD45RBlow population could contain recently activated effector/memory T cells. Interestingly, the Stockinger group has reported that both memory and Treg populations can prevent LIP  and they too may be differentially dependent on CTLA-4 for their action. Our results are in agreement with the recent study of conditional deletion of CTLA-4 in Foxp3+ T cells that revealed an essential role for CTLA-4 in Foxp3+ T cells for maintenance of immune homeostasis . We extend these studies in vivo to pinpoint an essential role for Treg expression of CTLA-4 in control of lymphopenia induced proliferation and associated effector cytokine production.
We suggest that CTLA-4-deficiency interferes with suppression through a direct action of Treg-expressed CTLA-4 with its ligand, presumably B7-family molecules on DC and/or T cells. Although CD4+ T cell LIP is not dependent on costimulatory signals, B7-CD28 costimulation provides a competitive advantage for T cell expansion . Indeed, the pattern of suppression by CD4+CD25+ Treg is quite similar to that of B7 blockade by CTLA-4-Ig . Thus Treg may effectively compete for limiting activation signals and restrain expansion of autoreactive clones through their constitutive expression of CTLA-4. Treg CTLA-4 expression could also modify target T cell functions by modulating the expression of costimulatory molecules on APC. The downregulation of B7 molecules on DC by Treg was found to be CTLA-4 dependent in vitro  but whether the same is true in vivo is not known. Additional experiments are needed to determine if the mechanism of Treg action is via B7 downregulation in vivo but such experiments are complicated by the fact that any attempt to modulate B7-binding will impact activation signals for both Treg and target T cell. Another possibility is that expression of CTLA-4 by Treg may prolong the contact time between Treg and DC via LFA-1 activation [32,33] increasing the efficiency of Treg suppression in the lymphopenic environment. This is an interesting model but will need to be further investigated given the seemingly conflicting data that CTLA-4 can also reverse the TCR-stop signal and reduce T-DC contact time .
Our studies once again highlight differences between Treg action in vitro and in vivo. We speculate that in vitro the requirement for Treg effector molecules is highly redundant, possibly due to non-physiological density of cellular interactions, antigenic stimulus and/or the type of APC used. In LIP, dendritic cells are sufficient to drive T cell proliferation [35–37] but whether other APC play a role is not clear. Interestingly, the Sakaguchi group found that CTLA-4-deficient Treg failed to downregulate B7 molecules on DC in vitro and this correlated with a defect in suppression of T cell proliferation in vitro similar to what we see in vivo. The requirement for key Treg effector molecules in vivo is likely to be highly context dependent[12, 13]. For the control of homeostatic proliferation this appears to be CTLA-4 but not IL-10 or TGF-β dependent while the control of CD8 effector functions looks to be TGF-β-dependent . Although Treg appear to have multiple mechanisms for regulating immune function, each component part may disable a very specific facet of a given immune response.
CTLA4+/+ and CTLA4−/− C57BL/6.CD90.1 mice were generated from CTLA-4+/− breeders (kindly provided by Jim Allison). CTLA-4−/− mice were used 2 weeks after birth. WT C57BL/6.CD90.2 mice were used as donors of WT cells and Rag2−/− mice (Taconic) were used as recipients. dnTGF-βRII transgenic mice were obtained from Jackson Labs. All mouse work was approved by the University of Rochester Medical School.
CD4+ T cells were enriched from spleen and LN by C' lysis of CD8, MHC class II, and heat-stable Ag-bearing cells. CD25+ and CD25− CD4+ T cells were further isolated by MACs (Miltenyi Biotec). The CD25− and CD25+ subsets were FACS sorted for CD62L, CD4 and CD25 expression (FACSAria, BD Biosciences). T depleted splenocytes (APCs) were prepared by C' lysis of Thy1+ cells and irradiated, 2500 rad.
Thy1.2+ CD4+CD25−CD62Lhigh sorted target cells (107/ml) were incubated with 5μM CFSE (Molecular Probes) for 5min at room temperature and washed three times. 0.5×106 CFSE-labeled targets cells were injected alone or at a 1:1 ratio with Thy1.1+ WT or CTLA-4 KO CD4+CD25+CD62Lhigh cells or control cells Thy1.1+ CD4+CD25−CD62Lhigh.
Colons were removed from mice 8–10 wks post cell transfer and fixed in 10% formalin. 4μm paraffin-embedded sections were cut and stained with hematoxylin and eosin. Inflammation in the tissues was scored from 0 to 5 in a blinded fashion: 0, normal colon; 1, slight epithelial cell hyperplasia and increased leukocytes in mucosa; 2, pronounced epithelial cell hyperplasia (2–3 fold increase in crypt length, significant leukocytic infiltration; 3, marked hyperplasia (up to 10-fold increase in crypt length), extensive leukocytic infiltrate mucosa and submucosa; 4, marked hyperplasia (up to 10-fold increase in crypt length) extensive transmural leukocytic infiltrate from submucosa to serosa, crypt abscesses.
1×105 CD4+CD25−CD62Lhigh target T cells were incubated in triplicate with 1×105 APC and 1μg/ml anti-CD3 mAb (2C11) with or without WT or CTLA-4 KO CD4+CD25+CD62Lhigh cells in a 96 well U bottom plate in RPMI-1640/10% heat-inactivated FCS. In some experiments cells were primed under Th1 (10ng IL-12, anti-IL-4 Ab 11B11, 10units/ml of rhIL-2) or Th2 (10ng IL4, anti-IFN-γ Ab XMG1.2, 10units/ml of rhIL-2) conditions. Cells were incubated for 72h at 37°C and proliferation measured by 3[H] thymidine incorporation (1μCi added for the last 6h).
Lymph nodes and spleen cells were pooled from mice in each group and the Thy1.2+ target cell population isolated by FACS. The sorted target cells were restimulated using anti-TCRβ plate bound mAb (clone H57) and the frequency of IL-2, IL-4, IFN-γ assessed by ELISPOT at 16h. In parallel supernatants were collected at 48h for IL-4 and IFN-γ production by ELISA. For IL-10 production, cells were stimulated with anti-TCRβ Ab for 8h, with Brefledin A added to final 6h of stimulation. Cells were surfaced labeled for CD4 and Thy1.1 before fixation and permeabilization. Cells were then labeled for both IL-10 (JES4−16E3) and Foxp3 (FJK-16s).
The cytokine secretion assay was preformed as described .
Real-time (RT) PCR for IL-10, Ebi3 and IL-12α was performed on Treg d13 after injection. Primers from ABI systems and detected by ABI prism 7700 Sequence BioDetector. HPRT as endogenous control and CD4+CD25− T cells used as the calibrator.
1-methyl-[D]-tryptophan (D-1MT) was dissolved in animal grade water to a final concentration of 2mg/ml (Sigma Aldrich) with aspartame for better taste as previously described . The pH of the D-1MT solution was adjusted to 7.0, filter sterilized and put into light sensitive water bottles. The mice were injected with the cells and placed into cages with the water bottles that contained the D-1MT solution or H20 (with aspartame). The oral treatment was continuous for two weeks and changed with fresh solutions 1 week into the experiment.
Statistical analyses used a 2 tailed Student's t test and are stated in each figure legend. A p value of less than or equal to 0.05 was considered significant: *p<0.05, **p<0.005, ***p<0.0005.
We thank Jim Allison for CTLA-4+/− mice, Megan Enos for Rag−/− mice and Andrew Mellor and David Munn for advice on the IDO-inhibitor experiments. A special thanks goes to Nathan Laniewski for cell sorting. This work was supported by the NIH grants R01-AI070826, U19 AI056390 and the JDRF (to DJF) and NIH T32-A107285 (to DKS).
Conflict of interest: The authors declare no financial or commercial conflict of interest.