Background
Cytotoxic T lymphocyte associated antigen-4 (CTLA-4, CD152) is an activation-induced glycoprotein of the Immunoglobulin superfamily, whose primary function is to down-regulate T cell responses [
1‐
4]. CTLA-4 shares its two known endogenous ligands, the B7 molecules B7.1 (CD80) and B7.2 (CD86), with the costimulatory receptor CD28 [
5‐
7]. Several mechanisms, including antagonism of CD28-dependent costimulation and direct negative signaling have been documented to explain the inhibitory capacity of CTLA-4 [
8]. Since the cytoplasmic tail of CTLA-4 lacks intrinsic enzymatic activity, the delivery of such a negative signal is likely provided through the association of CTLA-4 with key signaling molecules [
4].
CTLA-4 has been shown independently by two groups to associate with the serine/threonine phosphatase PP2A [
9,
10]. PP2A is a heterotrimeric holoenzyme which is comprised of a regulatory B subunit associated with a core dimer of a scaffolding A subunit (PP2AA) and a catalytic C subunit (PP2AC) [
11]. PP2A accounts for close to 1% of all cellular proteins and provides the majority of serine/threonine phosphate activity within eukaryotic cells [
12]. Using recombinant proteins, it has been reported that PP2AA interacts with the lysine rich motif located in the juxtamembrane region of the cytoplasmic tail of human CTLA-4, while the C subunit is thought to interact with the tyrosine residue in the YVKM motif located at position 165 [
9,
10]. However, it is currently unknown whether some of these associations occur
in vivo in T cells and if so what the functional consequences are.
We have previously reported that PP2A may regulate the ability of CTLA-4 to act as an inhibitor. Newly synthesized CTLA-4 becomes associated with PP2AA and remains associated when expressed on the cell surface, effectively blocking its inhibitory function [
10]. Following TCR:CTLA-4 co-ligation, where CTLA-4 engages B7 molecules expressed on antigen-presenting cells (APCs), PP2A is phosphorylated and dissociates from CTLA-4, and this dissociation correlates with the attenuation of T cell activation [
10]. Additionally, CTLA-4-dependent inhibition of Akt, a downstream target of PP2A, is sensitive to the PP2A inhibitor okadaic acid, implying that PP2A plays an important role in CTLA-4-mediated T cell inactivation [
13].
Under unique circumstances, some recombinant ligands of CTLA-4 can act as inverse agonists making CTLA-4 capable of activating T cells by itself, independent of TCR or CD28 ligation [
14,
15]. We have recently shown that soluble B7.1 Ig or 24:26, a bispecific, in-tandem single-chain Fv (ScFv) against human CTLA-4, function as inverse agonists of CTLA-4 resulting in the activation of primary human T cells and T cell lines. Such an inverse agonist activity correlates with the ability to induce the formation of a unique dimer-based CTLA-4 oligomer that signals through its cytoplasmic tail [
15]. Under these conditions of ligation, we have observed an increased association between PP2A and CTLA-4 suggesting that CTLA-4 may also induce T cell activation in a PP2A-dependent manner [
14].
As suggested by Rudd, the role of PP2A in CTLA-4 function needs clarification [
16]. Here, we started to address this issue by showing for the first time that the association between CTLA-4 and PP2A occurs in primary human T cells, suggesting that this interaction is physiologically relevant. Furthermore, we characterized the CTLA-4 interface interacting with PP2A using a panel of stably transfected Jurkat T cells expressing either wildtype (WT) CTLA-4 or CTLA-4 molecules mutated at various residues within the cytoplasmic domain. In this way, we eliminated any confounding effects as Jurkat T cells do not express endogenous CTLA-4 [
17]. Our results confirm the importance of the lysine rich motif for the association of PP2AA. However, contrary to previous studies, we report that not the first but the second tyrosine residue located at position 182 of human CTLA-4 is important for the binding of PP2AC to CTLA-4. Functionally, an increase in the association of PP2A to CTLA-4 was observed under conditions of inverse agonist ligation of CTLA-4 molecules with the exception of those mutated at the lysine residues. Such an increase correlated with the ability of CTLA-4 to induce T cell activation, and was dependent on the enzymatic activity of PP2A.
Discussion and conclusion
Understanding the mechanism of CTLA-4 function has proved to be remarkably puzzling over the past two decades. The ability of CTLA-4 to down-regulate T cell activation has been well established in multiple experimental systems including knock-out mouse models and T cell lines [
4]. Both extrinsic and intrinsic factors contribute to the inhibitory mechanism of CTLA-4
in vivo [
16]. Antagonism of CD28-dependent costimulation provides a plausible explanation for CTLA-4-mediated inhibition since CTLA-4 has a higher affinity and avidity for their shared ligands. However, the competition with CD28 for ligands only occurs when CTLA-4 is expressed at very high levels on the cell surface, indicating that an alternate mechanism lends to CTLA-4-dependent T cell inactivation [
8]. The direct delivery of a negative signal provides a more likely explanation for the inhibitory function of CTLA-4 at early stages of T cell down-regulation. This mechanism is functional at low levels of CTLA-4 surface expression and requires an intact cytoplasmic domain. The precise signaling pathway initialized by CTLA-4 is still undefined although it has been linked to down-regulation of CD28-dependent events [
23]. Many proteins have been shown to associate with CTLA-4. Among these, the serine/threonine phosphatase PP2A stands out as a candidate that can affect key molecules downstream of CD28, such as Akt, thereby affecting essential cellular events [
24].
In this study, we dissected the interaction between PP2A and CTLA-4 both from a structural point of view, to identify the areas of interaction, as well as from a functional point of view, to establish the requirement of such an interaction for the inhibitory and activating effects of CTLA-4 ligation. This was done using a panel of Jurkat T cells stably transfected with WT CTLA-4 or CTLA-4 molecules mutated at various locations throughout the intracellular domain. Previous data from yeast two hybrid studies suggested that the cytoplasmic domain of mouse CTLA-4 interacted with two subunits of the core dimer of PP2A [
9,
10]. The core dimer is comprised of a scaffolding A subunit and a catalytic C subunit, each existing as α and β isoforms. The association of the dimer to a third regulatory B subunit provides the cellular localization and target specificity of PP2A [
11]. Recent evidence has determined that post-translational modification of PP2AC plays an important role in the B subunit selection [
11]. The requirements for the interaction of PP2A with CTLA-4
in vivo in human T cells were not identified, justifying the current study. We confirm here that the K-rich motif (located at lysine residues 152, 155 and 156) is required for the interaction of CTLA-4 with PP2A. Mutation of these residues to alanine (KLESS CTLA-4) abrogated CTLA-4:PP2A co-precipitation. This confirmed our previous observation under conditions of equalized expression of WT CTLA-4 and KLESS CTLA-4, suggesting that the lower expression of KLESS CTLA-4 in this study is not likely contributing to its lack of association with PP2A [
10]. This observation is consistent with previous data pointing to the A subunit as the part of PP2A interacting with the K-rich motif [
10].
The C subunit of PP2A was shown to associate biochemically with murine CTLA-4 in HEK293 cells transfected with the cytoplasmic domain of CTLA-4 fused to GST, and the interaction site was suggested to be the tyrosine 165 of the YVKM motif by yeast two hybrid analysis [
9]. However, confirmation of the Y165 residue as the interaction site for PP2AC interaction was not examined in mouse or human T cells. An unexpected finding of our study here is that in human T cells it is the second tyrosine in the cytoplasmic tail of human CTLA-4 and not the first tyrosine (Y165) that is important for the interaction with PP2AC. We observed that co-precipitation of PP2A and CTLA-4 in human T cells was not affected when Y165 was mutated to phenylalanine. Since the A and C subunit of PP2A are almost always found associated with each other [
12], this result suggested that either mutating the PP2AC binding site was not enough to break the CTLA-4:PP2AA interaction or that Y165 was not the essential residue. We found that both PP2AA and PP2AC were able to co-precipitate Y165F CTLA-4 molecules, implying that another residue was likely responsible for interacting with the C subunit of PP2A. Our data indicate that the key residue for the second site of the CTLA-4:PP2AA interaction in human T cells is Y182. CTLA-4 molecules mutated at this second tyrosine (Y182F CTLA-4) had severely diminished interaction with PP2A. We cannot exclude that the Y165 may contribute to this interaction when Y182 is not available because a small level of association is observed between Y182F CTLA-4 and PP2A. However, the same level of co-precipitation is noted when both tyrosine residues are mutated (Y165F/Y182F CTLA-4) indicating that Y182 is the important residue for binding PP2AC and that the small amount of association observed is likely due to the intact PP2AA binding motif. Based on our results we propose a model in which the CTLA-4:PP2A interaction occurs at two distinct binding motifs: one is the lysine-rich motif binding to the A subunit of PP2A and the other is the Y182 residue of CTLA-4 binding to the C subunit of PP2A. This model predicts that the lysine-rich motif is the primary site responsible for stabilizing the CTLA-4:PP2A interaction and the tyrosine residues may be less important since they may be redundant in their ability to interact with PP2AC. This prediction correlates with the functional data presented in this study.
From a functional point of view, CTLA-4 displays a remarkable plasticity as it can inhibit or even activate T cells depending on the ligand it engages and the conditions in which this engagement occurs. The primary physiological function of CTLA-4 is to down-regulate T cell activation. We have previously reported that, under conditions of TCR and CTLA-4 co-ligation, PP2A is phosphorylated and dissociates from CTLA-4 [
10]. This correlates with the ability of CTLA-4 to inhibit T cell activation. This suggested that PP2A when bound to CTLA-4 prevents rather than mediates the inhibitory function of CTLA-4. In contrast, when PP2A is dissociated from CTLA-4, it likely inactivates downstream targets including Akt, consistent with the observation that CTLA-4-dependent inhibition of Akt phosphorylation is sensitive to OA [
13]. This model is consistent with the findings reported here that all the CTLA-4 mutants, independently of their ability to bind PP2A, inhibited IL-2 production when co-ligated with the TCR. The magnitude of inhibition through CTLA-4 in different in vitro models, including our own, is relatively modest (50–70% on average) compared to the striking phenotype of CTLA-4 knockout mice. This may be due to the use of cell lines rather than primary cells, to more intense activation conditions used in the in vitro systems, or other factors. Still, such an inhibition is reproducible and statistically significant. It remains to be determined how such co-ligation of CTLA-4 and TCR triggers the activity of PP2A.
The other aspect of CTLA-4 function is its ability to activate T cells when binding recombinant inverse agonist ligands, such as soluble B7.1 Ig and 24:26. Under these conditions, PP2A also stands as a key player. We show here that all CTLA-4 variants capable of interacting with PP2A showed enhanced association with this phosphatase following CTLA-4 engagement with 24:26. Such an enhanced association is in contrast to the PP2A dissociation observed when CTLA-4 acts as an inhibitory receptor. This enhanced association between CTLA-4 and PP2A is likely the result of stabilization of the interaction between these two molecules. We have shown that 24:26 induces the formation of dimer-based CTLA-4 oligomers that are tightly associated with each other on the T cell surface [
15]. The formation of such oligomers may provide a unique structure to facilitate the interaction between PP2A and CTLA-4. The enhanced CTLA-4:PP2A interaction upon inverse agonist ligation correlated with the ability of CTLA-4 to induce T cell activation. Moreover, the inverse agonist response was sensitive to the protein phosphatase inhibitor OA as IL-2 production induced upon 24:26 engagement of CTLA-4 was diminished its presence. Although OA is best known as an inhibitor of PP2A we can not rule out the inhibition of other phosphatases which may contribute to CTLA-4-mediated T cell activation. However, the effect of OA did not completely abolish the ability of CTLA-4 to induce IL-2 production, likely owing to the constitutive Akt activation in Jurkat T cells [
25].
Delineation of the interaction between CTLA-4 and PP2A provides mechanistic insights into the signaling pathways targeted by CTLA-4. The costimulatory molecule CD28 has also been shown to associate with PP2A [
9,
10]. Microarray analysis of genes regulated upon B7 ligation of CTLA-4 suggests that CTLA-4 inhibits T cells by inhibiting CD28-dependent genes and not TCR-dependent genes [
23]. Furthermore, we have shown that CD28 expression is essential for the inhibitory and activating function of CTLA-4 [
15]. Therefore, it is plausible that CTLA-4 may function through PP2A as an inhibitor by blocking CD28 signaling and as an activator by triggering the CD28 signaling pathway. PP2A activity is dependent on its phosphorylation state, with unphosphorylated PP2A being active and phosphorylated PP2A rendered inactive. PP2A is able to dephosphorylate itself to regain activity [
26].
One distinction between the inhibition of T cell activation by CTLA-4 and the activation of T cells by inverse agonists of CTLA-4 is that the former does not require the association of PP2A to CTLA-4 whereas the latter does. However, both responses are inhibited by okadaic acid [
13], implying that PP2A is a critical mediator of them. We propose that, under conditions of T cell inhibition, PP2A is phosphorylated and dissociates from CTLA-4, becoming available to block CD28 signaling. Such a blockade may affect upstream events (eg., inhibition of lck, blockade of CD28-PI3K interaction), or downstream events (eg., direct inhibition of Akt). In this context, CTLA-4 would act as a shuttle of PP2A to the immunological synapse where, upon release, PP2A could act on TCR-ζ and CD28 signaling events [
27]. It is unclear, how in our experimental system, CTLA-4 molecules that cannot bind PP2A can still inhibit T cell activation. Perhaps these molecules function to inhibit T cell responses primarily by sequestering B7 molecules from CD28.
Under conditions of T cell activation through CTLA-4, the pool of PP2A bound to CTLA-4 may trigger TCR-ζ and CD28 signaling by activating lck [
14]. Lck activation, occurs through dephosphorylation of the negative regulatory tyrosine (residue 505) which induces autophosphorylation at tyrosine 394, initiating its kinase activity [
28‐
30]. In addition to serine/threonine phosphatase acitivity, it has been reported that PP2A may also have tyrosine phosphatase activity [
24]. Since both the phosphatase activity of PP2A and lck expression are required for 24:26-induced T cell activation [
14], it is plausible to propose that PP2A could activate lck and initiate CD28 signaling.
Authors' contributions
WAT carried out the majority of the experimental work, participated in the experimental design and organization and drafted the manuscript. TAC performed some immunoprecipitation experiments. JM participated in the experimental design and organization and helped draft the manuscript. All authors have read and approved the manuscript for submission.