Hierarchical Regulation of CTLA-4 Dimer-Based Lattice Formation and Its Biological Relevance for T Cell Inactivation1
http://www.100md.com
免疫学杂志 2005年第14期
Abstract
CTLA-4 is an activation-induced, homodimeric inhibitory receptor in T cells. Recent crystallographic reports have suggested that it may form lattice-like arrays on the cell surface upon binding B7.1/B7.2 (CD80, CD86) molecules. To test the biological relevance of these CTLA-4-B7 lattices, we introduced a C122A point mutation in human CTLA-4, because this residue was shown to be essential for dimerization in solution. Surprisingly, we found that up to 35% of C122A CTLA-4 dimerized in human T lymphocytes. Moreover, C122A CTLA-4 partitioned within lipid rafts, colocalized with the TCR in the immunological synapse, and inhibited T cell activation. C122-independent dimerization of CTLA-4 involved N-glycosylation, because further mutation of the N78 and N110 glycosylation sites abrogated dimerization. Despite being monomeric, the N78A/N110A/C122A triple mutant CTLA-4 localized in the immunological synapse and inhibited T cell activation. Such functionality correlated with B7-induced dimerization of these mutant molecules. Based on these data, we propose a model of hierarchical regulation of CTLA-4 oligomerization by which B7 binding ultimately determines the formation of dimer-dependent CTLA-4 lattices that may be necessary for triggering B7-dependent T cell inactivation.
Introduction
T cell activation is regulated by a complex balance of costimulatory and inhibitory signals (1). One of the latter type of signals is delivered by CTLA-4 (CD152), an activation-induced transmembrane receptor expressed by T cells (2, 3). The ligands for CTLA-4 are B7.1 (CD80) and B7.2 (CD86), the same ligands as those for the costimulatory receptor CD28 (4). However, CTLA-4 binds to B7 molecules with a higher affinity and avidity compared with CD28. This allows CTLA-4 to sequester B7 ligands from CD28 and antagonize CD28-dependent costimulation (5, 6). In addition, CTLA-4 can inhibit T cell activation by initiating a signal transduction pathway that leads to inhibition of TCR-dependent signaling (reviewed in Ref. 7). This latter mechanism involves accumulation of CTLA-4 in lipid rafts, where it may exert its down-regulatory function on signaling initiated from TCR subunits (8, 9, 10). CTLA-4 also colocalizes with the TCR in the immunological synapse (IS),6 where it may inhibit TCR-proximal signaling (8, 9, 11).
An unresolved question regarding the biology of CTLA-4 is how ligation of this receptor by B7 molecules triggers inhibitory signals for T cell activation. Classical receptor signaling models claim that signaling either results from ligand-induced conformational changes, as in the case of the insulin receptor (12), or, alternatively, results from ligand-induced dimerization, as in the case of human growth hormone (13). However, CTLA-4 does not undergo any detectable conformation change upon B7 binding (14), and before ligation, it exists as a nonfunctional covalent homodimer (15). Therefore, alternative models are currently being explored to explain CTLA-4 signaling.
Based on the recent cocrystal structure of CTLA-4 and B7, it has been suggested that CTLA-4 homodimers form periodic arrays, or lattices, with homodimeric B7 (14, 16). Such lattices are postulated to provide the appropriate oligomerization conditions for CTLA-4 signaling to occur (17, 18). Because CTLA-4 coclusters with the TCR in the IS (8, 11), such lattices would be poised to inhibit TCR/CD28-mediated T cell activation. To test the hypothesis that dimer-dependent CTLA-4-B7 lattices facilitate the function of CTLA-4, we analyzed the function of CTLA-4 molecules containing a cysteine to alanine point mutation at position 122 (C122A). This cysteine residue has been shown to mediate intermolecular disulfide bonding in the CTLA-4 dimer and is critical for CTLA-4 dimerization in solution (15). Surprisingly, when a C122A mutation was introduced, up to 35% of CTLA-4 was dimeric. Furthermore, C122A CTLA-4 compartmentalized in lipid rafts, migrated to the IS upon stimulation where it colocalized with the TCR, and inhibited T cell activation. C122A CTLA-4, which underwent aberrant N-glycosylation, failed to dimerize when N78 and N110 glycosylation sites were mutated. Interestingly, this triple-mutant N78A/N110A/C122A (NNC) CTLA-4 was still functional, localized in the IS, and inhibited T cell activation. Such functionality correlated with NNC CTLA-4 becoming predominantly dimeric upon ligation with B7. Our findings demonstrate that CTLA-4 dimerization/oligomerization is hierarchically regulated by intermolecular disulfide bonding, N-linked glycosylation, and B7 ligand-driven dimerization. This provides a structural framework for CTLA-4 lattice formation leading to B7-dependent functions of CTLA-4.
Materials and Methods
Plasmids and T cell transfectants
CTLA-4 mutants were generated using the QuickChange site-directed mutagenesis kit (Stratagene) on the previously described doxycycline-inducible pBIG2i vector containing wild-type (WT) CTLA-4 as a template (19, 20). The NA mutations were made on pBluescript vector containing WT CTLA-4 or C122A CTLA-4 and subcloned into pBIG2i using the Gateway kit with pENTR/d/topo as a shuttle vector (Invitrogen Life Technologies). Stable transfectants were generated by electroporating linearized plasmids into Jurkat E6.1 T cells and were screened for stable expression as described previously (20). Individual clones with high levels of CD3, CD28, and doxycycline-inducible expression of CTLA-4 were selected for study. Jurkat T cells and B lymphoblast cell lines were cultured under standard conditions in R-10 medium (21). Doxycycline (Sigma-Aldrich) was added in culture overnight at 100 ng/ml for inhibition assays or 1000 ng/ml for biochemical experiments (8).
Antibodies
Mouse anti-human CTLA-4 mAbs 11 and 24 were supplied by Dr. B. M. Carreno (Wyeth Research, Cambridge, MA) and used for immunoblotting and immunoprecipitation respectively. For flow cytometry and confocal microscopy, FITC-conjugated anti-human IgG, Fc fragment-specific Ab (Jackson ImmunoResearch Laboratories) and FITC-labeled anti-CD3 (UCHT-I), FITC-labeled anti-CD28, and PE-labeled anti-CTLA-4 (BNI3) mAbs (BD Pharmingen) were used. Immunoblotting for ERK-1/2 was performed using a rabbit polyclonal immunoaffinity-purified antiserum (Stressgen Biotechnologies). Linker for activation of T cells (LAT) was immunoblotted using an anti-LAT rabbit polyclonal antiserum (Upstate Biotechnology). Ezrin was immunoblotted with a goat polyclonal antiserum against ezrin (Santa Cruz Biotechnology). Actin was immunoblotted using an affinity-purified goat polyclonal antiserum (Santa Cruz Biotechnology). Isotype control immunoprecipitations were performed with a mouse IgG1 (eBioscience). Immunoprecipitation of CTLA-4 with B7 was performed with B7.2-Ig (Genetics Institute/Wyeth Research).
Confocal microscopy
For confocal microscopic analysis of IS formation, T cells (1.0 x 106 cells/group) were stimulated with LG2 B lymphoblastoid cells (1.0 x 106 cells/group) and staphylococcal enterotoxin E (SEE; 100 ng/ml; Toxin Technology) on poly-L-lysine-coated confocal dishes (MatTek) as described previously (8). Cells were first fixed with 4% paraformaldehyde, then stained with PE-labeled anti-CTLA-4 mAb and FITC-labeled anti-CD3 mAb for 30 min on ice. Samples were analyzed on a confocal microscope (Zeiss) and LSM 510 software (Zeiss).
T cell functional assays
T cells (0.2 x 106 cells/group), with or without preincubation with doxycycline (100 ng/ml) to induce CTLA-4 expression, were plated in triplicate on 96-well plates with the B lymphoblast LG2 (0.1 x 106 cells/group) with or without 10 ng/ml SEE or with or without 0.1 ng/ml SEE in the presence or the absence of 100 ng/ml doxycycline at 37°C for 48 h. Supernatants were collected and diluted 1/10 before measurement of IL-2 by ELISA following the manufacturer’s specifications (BD Biosciences).
Results
CTLA-4 dimerizes in vivo in the absence of intermolecular disulfide bonding at C122
A previous report had indicated that C122 is required for dimerization of soluble CTLA-4 (15). To study the functional implications of CTLA-4 dimerization in T lymphocytes, we introduced a cysteine to alanine point mutation in that position (C122A), subcloned the mutant CTLA-4 cDNA into a doxycycline-inducible plasmid, and generated stably transfected Jurkat T cells. These cells are an appropriate cell line for structural analysis of CTLA-4 because they do not express endogenous CTLA-4 even upon activation (22), and they preserve a glycosylation machinery comparable to that of peripheral blood T lymphocytes (23). Disruption of intermolecular disulfide bonds by mutating cysteine residues to alanine residues has been validated in a variety of dimeric or oligomeric protein complexes (24, 25, 26). Interestingly, we found that C122A CTLA-4 could still dimerize in significant amounts, as illustrated by the presence of a mixture of monomers and reduction-sensitive dimers in Western blots under nonreducing conditions (Fig. 1A). Quantification of monomeric and dimeric forms from multiple stably transfected T cell clones indicated that, on the average, 35% of C122A CTLA-4 were dimeric, although the percentage of dimeric C122A CTLA-4 was as high as 45%, compared with 90% or more for WT CTLA-4 (Fig. 1B). We also noticed a difference in the gel mobility of C122A CTLA-4 that was due to aberrant N-linked glycosylation of this mutant form of CTLA-4 (see below). The observed dimerization of C122A CTLA-4 was not due to intragel dimerization, because it was still present when the T cell lysates were prepared in the presence of excess N-ethylmaleimide, a reagent that chemically modifies free cysteine residues and thus prevents de novo formation of disulfide bonds (data not shown). Moreover, C122A CTLA-4 dimers were preserved under reducing conditions if DSS, a membrane-permeable, 11.4-? arm length cross-linker, was added before lysing the cells, indicating that C122A CTLA-4 dimers are present in intact cells (data not shown). We also noted the presence of a lower band in the C122A CTLA-4 lysates under reducing and nonreducing conditions representing nonglycosylated and/or partially glycosylated CTLA-4.
To explain the mechanistic basis for the function of NNC CTLA-4, we hypothesized that the presence of a bivalent CTLA-4 ligand on the surface of an APC could induce dimerization of NNC CTLA-4. To test this hypothesis, whole cell lysates from stable WT CTLA-4, C122A CTLA-4, and NNC CTLA-4 T cell transfectants were immunoprecipitated with recombinant B7.2-Ig and analyzed for dimer formation under nonreducing conditions. WT CTLA-4, C122A CTLA-4, and NNC CTLA-4 were able to bind to B7.2-Ig to a similar extent on the cell surface (Fig. 7A). We found that B7.2-Ig pulldown caused a significant increase in the proportion of C122A and NNC dimers compared with that seen in whole cell lysates (Fig. 7B). This was particularly apparent for the NNC CTLA-4, for which dimers were not detected in whole cell lysates, but constituted the majority of the CTLA-4 signal in B7.2-Ig pulldown experiments. The C122A CTLA-4 transfectant selected for this experiment expressed a lesser amount of dimer than the average C122A CTLA-4 transfectant to better illustrate the B7.2-Ig-induced dimerization. Therefore, B7 ligands can drive dimerization of monomeric CTLA-4 on the cell surface, and this is associated with normal function of monomeric CTLA-4.
Discussion
Recent crystallographic studies have suggested that CTLA-4-B7 lattices provide the mechanistic basis for triggering the inhibitory function of CTLA-4 (14, 16). To test the biological relevance of CTLA-4-B7 lattices in vivo, we generated a monomeric version of CTLA-4, reasoning that monomeric CTLA-4 would not be able to drive lattice formation with B7. To do this, we mutated residue C122 of CTLA-4 to alanine, because a previous report had shown that C122 forms an intermolecular disulfide bond that is essential for homodimerization in solution (15). These studies provide new insights on the regulation of CTLA-4 oligomerization.
The first remarkable finding is that, contrary to what was expected (15), CTLA-4 in T cells can still dimerize in the absence of C122, with up to 35% of C122A CTLA-4 being dimeric. Such a form can reach the T cell surface, is fully functional, and does not misfold based on its ability to be detected by anti-CTLA-4 Abs and by the fact that it binds B7. The apparent discrepancy between this finding and that of Linsley et al. (15) may be due to the fact that we used intact CTLA-4 expressed in T cells, whereas Linsley et al. used soluble CTLA-4 cleaved from the surface of COS cells. In addition, as shown in this study, CTLA-4 dimers are rescued by aberrant N-glycosylation in T cells, which may not be operational or may be different in COS cells. The second key finding of our studies is that N-glycosylation of CTLA-4 plays a key role in facilitating CTLA-4 dimerization, particularly in the absence of C122-dependent covalent bonding. If it is N-glycosylated, then the mutant C122A CTLA-4 can still partition within lipid rafts, is expressed on the T cell surface, colocalizes with the TCR in the IS, and is fully functional. Our data with PNGase-F and brefeldin A indicate that the rescuing of C122A CTLA-4 by N-glycosylation occurs in the N-glycan maturation process within the Golgi as opposed to the initial addition of immature N-glycans in the ER. It is important to note that the lack of C122 and N-glycosylation sites did not lead to misfolding based on the evidence that the NNC mutant CTLA-4 molecules trafficked to the cell surface, were still recognized by anti-CTLA-4 Abs, relocalized to the IS, were able to bind B7, and inhibited T cell activation. These findings would not be possible if CTLA-4 was grossly misfolded.
The relationship between glycosylation, receptor dimerization, and receptor function has been examined in a few studies, although no single paradigm has emerged. Glycosylation can either enhance or interfere with receptor dimerization. For example, glycosylation at residue N15 is required for 1-adrenergic receptor dimerization and surface expression (31), and deglycosylation of the oncogenic truncated epidermal growth factor receptor impairs its homodimerization and kinase activity (32). In addition, single N-linked glycosylation has been shown to contribute to disulfide-dependent dimer formation of mucins (33). In contrast, glycosylation (sialylation and O-glycosylation) negatively modulates CD45 dimerization (34, 35, 36, 37). Nevertheless, in some cases, glycosylation has no effect on dimerization, as seen with CD26, a serine protease with nine glycosylation sites that retains its ability to dimerize and function in the absence of these N residues (38). Differential glycosylation also has functional consequences, some of them through its effects on receptor dimerization. For example, on CD45, glycosylation decreases dimerization, and this is associated with increased activity of this phosphatase (34, 35, 36, 37). In this sense, it is of interest to point out that T cells express a hypoglycosylated form of CD86 with reduced binding affinity for CTLA-4 and CD28 and thus may act as a regulator of T cell activation (39). N-glycosylation of CD28 can also negatively regulate its binding to CD80 and thus inhibit its costimulatory function (40).
There are several possible mechanisms to account for the contribution of N-glycosylation to CTLA-4 dimerization. The crystal structure of B7.1-engaged CTLA-4 showed that the N-linked oligosaccharide at residue N110 is oriented toward the CTLA-4 dimerization interface, suggesting a role in facilitating dimerization (16). However, we have shown that N110A/C122A mutants do not have a diminished ability to dimerize, implying that dual glycosylation is necessary to drive dimerization. The fact that the NNC CTLA-4 fails to dimerize suggests a role for N78 in this process. The N78 oligosaccharide is thought to hide a hydrophobic patch that can lead to aggregation of CTLA-4 when the N78 residue is mutated (28). Another possibility is that N-glycosylation stabilizes the small noncovalent binding surface at the dimerization interface seen in crystal structures (14, 16). However, this interface, which is composed of 10 highly conserved residues that interact with the same surface on the other CTLA-4 monomer, has a small buried surface area compared with other protein-protein binding surfaces and is thought to be relatively weak (16). We think it unlikely that this weak noncovalent interaction could protect up to 35% of the dimers from the boiling conditions used for Western blotting.
The sensitivity to reduction would imply that a disulfide bond is involved in this process. There are two intramolecular disulfide bonds per CTLA-4 monomer: one between C21 and C92, which is involved in forming the classical Ig fold and is required for maintaining the conformation of the B7 binding motif, and the second between C48 and C66, which is unique to CTLA-4 and has no known function (15). These intramolecular disulfide bonds may indirectly stabilize C122A dimers in a manner analogous to what has been reported for mucins, which can form intramolecular disulfide-dependent cysteine knots (33). Although no such cysteine knot was seen in WT CTLA-4 crystal structures, we cannot rule out such an effect in the C122A mutant. Another possibility is that the C48-C66 bond, which is found close to the dimerization interface (16), could form a novel intermolecular disulfide bond during biosynthesis. We ruled out the possibility of intragel dimerization through de novo disulfide bond formation by treating the cells with N-ethylmaleimide before lysis, which blocks exposed cysteine resides from bonding. Also, the NNC CTLA-4 mutant has all the other cysteine residues intact and still does not dimerize, thus making it unlikely that these other cysteine residues contribute to the formation of C122A dimers. The contribution of glycosylation in this respect could be to slow the rate of biosynthesis and facilitate interactions with chaperones in the ER, as was reported for mucin (33). However, we have shown that dimerization of C122A and NNC CTLA-4 could occur upon ligation with monomeric B7-Ig during immunoprecipitation, suggesting that mature molecules are also capable of forming stable dimers.
Based on the data presented in this study, we propose a hierarchical model of CTLA-4 oligomerization leading to B7:CTLA-4 lattice formation and the triggering of B7-dependent inhibitory functions of CTLA-4. In this model there are three separate steps that supersede each other in a sequential fashion. These are C122-dependent covalent bonding, N-glycosylation, and binding to B7. In the ER, disulfide bonds at C122 form and generate CTLA-4 dimers. In the event that this covalent bond cannot form, oligosaccharide modifications completed in the Golgi allow for dimerization of CTLA-4. Nevertheless, in the absence of N-glycosylation of the molecule (e.g., similar to what has been reported to occur in vivo for B7 (39)), monomeric CTLA-4 can reach the T cell surface, where it will dimerize after binding to B7. This allows for monomeric CTLA-4 to form lattices and inhibit T cell activation in a B7-dependent fashion. It remains to be determined whether the fine architecture of CTLA-4 oligomers induced by B7 in the presence or the absence of disulfide bond/N-glycosylation-dependent dimers is the same, and whether the redistribution of these oligomers changes during T cell activation (41).
Although spontaneous mutations of the C122 residue in CTLA-4 have not been reported in a clinical setting, conditions of cellular stress (e.g., oxidative stress) could cause defects in glycosylation and cysteine-based dimerization/oligomerization of CTLA-4 leading to the generation of monomeric CTLA-4. This possibility has been documented for other molecules. In the case of receptor-like protein phosphatase-, oxidative stress leads to a conformational change in the cytoplasmic tail of the molecule that is subsequently transmitted to the extracellular portion of the molecule and modulates dimer formation and the function of the protein (42, 43). The tumor suppressor ARF also reversibly forms homo-oligomers under oxidizing conditions, and such oligomers can form even after the three sole cysteine residues in this protein are deleted (44). Disulfide bond-independent oligomerization has also been shown for the tumor suppressor B23, which forms oligomers under reducing conditions (45). Finally, the DNA-binding activity of NtcA was perturbed by the reducing agent DTT, and this effect was preserved in NtcA cysteine mutants (46). Thus, the results presented in this study have direct biological implications on CTLA-4 function. By analogy to these reported cases, it is plausible to conclude that the capacity of B7 binding to determine the formation of functional dimers of CTLA-4 is a safe mechanism to avoid loss of function of CTLA-4 under conditions of redox stress after T cell activation.
In closing, it is important to note that the model proposed in this study for CTLA-4 oligomerization is operational for those CTLA-4 functions that are B7 dependent. Recently, there has been increasing interest in B7-independent functions of CTLA-4 (47). The different biological implications for B7-dependent and B7-independent CTLA-4 functions and their differential structural requirements may have contributed to the selective advantage for such elaborate regulation of CTLA-4 dimerization/oligomerization.
Acknowledgments
We thank Drs. B. M. Carreno (Wyeth Research, Cambridge, MA), M. Dustin (Skirball Institute, New York, NY), and the members of the Madrenas laboratory for helpful discussions.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by the Canadian Institutes of Health Research, the Kidney Foundation of Canada, and the London Health Sciences Center Multi-Organ Transplant Program. M.G.K. is the recipient of a Canadian Institutes of Health Research M.D./Ph.D. studentship, and J.M. holds a Canada Research Chair in Transplantation and Immunobiology.
2 P.J.D. and M.G.K. share first authorship.
3 Current address: Montreal Neurological Institute and Hospital, Montreal, Quebec, Canada.
4 Current address: Kennedy Institute of Rheumatology, Imperial College London, London, U.K.
5 Address correspondence and reprint requests to Dr. Joaquín Madrenas, Robarts Research Institute, Room 2.05, P.O. Box 5015, 100 Perth Drive, London, Ontario, Canada N6A 5K8. E-mail address: madrenas@robarts.ca
6 Abbreviations used in this paper: IS, immunological synapse; ER, endoplasmic reticulum; LAT, linker for activation of T cell; NNC, N78A/N110A/C122A; PNGase, peptide:N-glycosidase; SEE, staphylococcal enterotoxin E; WT, wild type; DSS, disuccinimidyl subarate.
Received for publication October 29, 2004. Accepted for publication May 5, 2005.
References
Rothstein, D. M., M. H. Sayegh. 2003. T-cell costimulatory pathways in allograft rejection and tolerance. Immunol. Rev. 196: 85-108.
Waterhouse, P., J. M. Penninger, E. Timms, A. Wakeham, A. Shahinian, K. P. Lee, C. B. Thompson, H. Griesser, T. W. Mak. 1995. Lymphoproliferative disorders with early lethality in mice deficient in Ctla-4. Science 270: 985-988.
Tivol, E. A., F. Borriello, A. N. Schweitzer, W. P. Lynch, J. A. Bluestone, A. H. Sharpe. 1995. Loss of CTLA-4 leads to massive lymphoproliferation and fatal multiorgan tissue destruction, revealing a critical negative regulatory role of CTLA-4. Immunity 3: 541-547.
van der Merwe, P. A., D. L. Bodian, S. Daenke, P. Linsley, S. J. Davis. 1997. CD80 (B7-1) binds both CD28 and CTLA-4 with a low affinity and very fast kinetics. J. Exp. Med. 185: 393-403.
Thompson, C. B., J. P. Allison. 1997. The emerging role of CTLA-4 as an immune attenuator. Immunity 7: 445-450.
Carreno, B. M., F. Bennett, T. A. Chau, V. Ling, D. Luxenberg, J. Jussif, M. L. Baroja, J. Madrenas. 2000. CTLA-4 (CD152) can inhibit T cell activation by two different mechanisms depending on its level of cell surface expression. J. Immunol. 165: 1352-1356.[(Peter J. Darlington2,3, M)
CTLA-4 is an activation-induced, homodimeric inhibitory receptor in T cells. Recent crystallographic reports have suggested that it may form lattice-like arrays on the cell surface upon binding B7.1/B7.2 (CD80, CD86) molecules. To test the biological relevance of these CTLA-4-B7 lattices, we introduced a C122A point mutation in human CTLA-4, because this residue was shown to be essential for dimerization in solution. Surprisingly, we found that up to 35% of C122A CTLA-4 dimerized in human T lymphocytes. Moreover, C122A CTLA-4 partitioned within lipid rafts, colocalized with the TCR in the immunological synapse, and inhibited T cell activation. C122-independent dimerization of CTLA-4 involved N-glycosylation, because further mutation of the N78 and N110 glycosylation sites abrogated dimerization. Despite being monomeric, the N78A/N110A/C122A triple mutant CTLA-4 localized in the immunological synapse and inhibited T cell activation. Such functionality correlated with B7-induced dimerization of these mutant molecules. Based on these data, we propose a model of hierarchical regulation of CTLA-4 oligomerization by which B7 binding ultimately determines the formation of dimer-dependent CTLA-4 lattices that may be necessary for triggering B7-dependent T cell inactivation.
Introduction
T cell activation is regulated by a complex balance of costimulatory and inhibitory signals (1). One of the latter type of signals is delivered by CTLA-4 (CD152), an activation-induced transmembrane receptor expressed by T cells (2, 3). The ligands for CTLA-4 are B7.1 (CD80) and B7.2 (CD86), the same ligands as those for the costimulatory receptor CD28 (4). However, CTLA-4 binds to B7 molecules with a higher affinity and avidity compared with CD28. This allows CTLA-4 to sequester B7 ligands from CD28 and antagonize CD28-dependent costimulation (5, 6). In addition, CTLA-4 can inhibit T cell activation by initiating a signal transduction pathway that leads to inhibition of TCR-dependent signaling (reviewed in Ref. 7). This latter mechanism involves accumulation of CTLA-4 in lipid rafts, where it may exert its down-regulatory function on signaling initiated from TCR subunits (8, 9, 10). CTLA-4 also colocalizes with the TCR in the immunological synapse (IS),6 where it may inhibit TCR-proximal signaling (8, 9, 11).
An unresolved question regarding the biology of CTLA-4 is how ligation of this receptor by B7 molecules triggers inhibitory signals for T cell activation. Classical receptor signaling models claim that signaling either results from ligand-induced conformational changes, as in the case of the insulin receptor (12), or, alternatively, results from ligand-induced dimerization, as in the case of human growth hormone (13). However, CTLA-4 does not undergo any detectable conformation change upon B7 binding (14), and before ligation, it exists as a nonfunctional covalent homodimer (15). Therefore, alternative models are currently being explored to explain CTLA-4 signaling.
Based on the recent cocrystal structure of CTLA-4 and B7, it has been suggested that CTLA-4 homodimers form periodic arrays, or lattices, with homodimeric B7 (14, 16). Such lattices are postulated to provide the appropriate oligomerization conditions for CTLA-4 signaling to occur (17, 18). Because CTLA-4 coclusters with the TCR in the IS (8, 11), such lattices would be poised to inhibit TCR/CD28-mediated T cell activation. To test the hypothesis that dimer-dependent CTLA-4-B7 lattices facilitate the function of CTLA-4, we analyzed the function of CTLA-4 molecules containing a cysteine to alanine point mutation at position 122 (C122A). This cysteine residue has been shown to mediate intermolecular disulfide bonding in the CTLA-4 dimer and is critical for CTLA-4 dimerization in solution (15). Surprisingly, when a C122A mutation was introduced, up to 35% of CTLA-4 was dimeric. Furthermore, C122A CTLA-4 compartmentalized in lipid rafts, migrated to the IS upon stimulation where it colocalized with the TCR, and inhibited T cell activation. C122A CTLA-4, which underwent aberrant N-glycosylation, failed to dimerize when N78 and N110 glycosylation sites were mutated. Interestingly, this triple-mutant N78A/N110A/C122A (NNC) CTLA-4 was still functional, localized in the IS, and inhibited T cell activation. Such functionality correlated with NNC CTLA-4 becoming predominantly dimeric upon ligation with B7. Our findings demonstrate that CTLA-4 dimerization/oligomerization is hierarchically regulated by intermolecular disulfide bonding, N-linked glycosylation, and B7 ligand-driven dimerization. This provides a structural framework for CTLA-4 lattice formation leading to B7-dependent functions of CTLA-4.
Materials and Methods
Plasmids and T cell transfectants
CTLA-4 mutants were generated using the QuickChange site-directed mutagenesis kit (Stratagene) on the previously described doxycycline-inducible pBIG2i vector containing wild-type (WT) CTLA-4 as a template (19, 20). The NA mutations were made on pBluescript vector containing WT CTLA-4 or C122A CTLA-4 and subcloned into pBIG2i using the Gateway kit with pENTR/d/topo as a shuttle vector (Invitrogen Life Technologies). Stable transfectants were generated by electroporating linearized plasmids into Jurkat E6.1 T cells and were screened for stable expression as described previously (20). Individual clones with high levels of CD3, CD28, and doxycycline-inducible expression of CTLA-4 were selected for study. Jurkat T cells and B lymphoblast cell lines were cultured under standard conditions in R-10 medium (21). Doxycycline (Sigma-Aldrich) was added in culture overnight at 100 ng/ml for inhibition assays or 1000 ng/ml for biochemical experiments (8).
Antibodies
Mouse anti-human CTLA-4 mAbs 11 and 24 were supplied by Dr. B. M. Carreno (Wyeth Research, Cambridge, MA) and used for immunoblotting and immunoprecipitation respectively. For flow cytometry and confocal microscopy, FITC-conjugated anti-human IgG, Fc fragment-specific Ab (Jackson ImmunoResearch Laboratories) and FITC-labeled anti-CD3 (UCHT-I), FITC-labeled anti-CD28, and PE-labeled anti-CTLA-4 (BNI3) mAbs (BD Pharmingen) were used. Immunoblotting for ERK-1/2 was performed using a rabbit polyclonal immunoaffinity-purified antiserum (Stressgen Biotechnologies). Linker for activation of T cells (LAT) was immunoblotted using an anti-LAT rabbit polyclonal antiserum (Upstate Biotechnology). Ezrin was immunoblotted with a goat polyclonal antiserum against ezrin (Santa Cruz Biotechnology). Actin was immunoblotted using an affinity-purified goat polyclonal antiserum (Santa Cruz Biotechnology). Isotype control immunoprecipitations were performed with a mouse IgG1 (eBioscience). Immunoprecipitation of CTLA-4 with B7 was performed with B7.2-Ig (Genetics Institute/Wyeth Research).
Confocal microscopy
For confocal microscopic analysis of IS formation, T cells (1.0 x 106 cells/group) were stimulated with LG2 B lymphoblastoid cells (1.0 x 106 cells/group) and staphylococcal enterotoxin E (SEE; 100 ng/ml; Toxin Technology) on poly-L-lysine-coated confocal dishes (MatTek) as described previously (8). Cells were first fixed with 4% paraformaldehyde, then stained with PE-labeled anti-CTLA-4 mAb and FITC-labeled anti-CD3 mAb for 30 min on ice. Samples were analyzed on a confocal microscope (Zeiss) and LSM 510 software (Zeiss).
T cell functional assays
T cells (0.2 x 106 cells/group), with or without preincubation with doxycycline (100 ng/ml) to induce CTLA-4 expression, were plated in triplicate on 96-well plates with the B lymphoblast LG2 (0.1 x 106 cells/group) with or without 10 ng/ml SEE or with or without 0.1 ng/ml SEE in the presence or the absence of 100 ng/ml doxycycline at 37°C for 48 h. Supernatants were collected and diluted 1/10 before measurement of IL-2 by ELISA following the manufacturer’s specifications (BD Biosciences).
Results
CTLA-4 dimerizes in vivo in the absence of intermolecular disulfide bonding at C122
A previous report had indicated that C122 is required for dimerization of soluble CTLA-4 (15). To study the functional implications of CTLA-4 dimerization in T lymphocytes, we introduced a cysteine to alanine point mutation in that position (C122A), subcloned the mutant CTLA-4 cDNA into a doxycycline-inducible plasmid, and generated stably transfected Jurkat T cells. These cells are an appropriate cell line for structural analysis of CTLA-4 because they do not express endogenous CTLA-4 even upon activation (22), and they preserve a glycosylation machinery comparable to that of peripheral blood T lymphocytes (23). Disruption of intermolecular disulfide bonds by mutating cysteine residues to alanine residues has been validated in a variety of dimeric or oligomeric protein complexes (24, 25, 26). Interestingly, we found that C122A CTLA-4 could still dimerize in significant amounts, as illustrated by the presence of a mixture of monomers and reduction-sensitive dimers in Western blots under nonreducing conditions (Fig. 1A). Quantification of monomeric and dimeric forms from multiple stably transfected T cell clones indicated that, on the average, 35% of C122A CTLA-4 were dimeric, although the percentage of dimeric C122A CTLA-4 was as high as 45%, compared with 90% or more for WT CTLA-4 (Fig. 1B). We also noticed a difference in the gel mobility of C122A CTLA-4 that was due to aberrant N-linked glycosylation of this mutant form of CTLA-4 (see below). The observed dimerization of C122A CTLA-4 was not due to intragel dimerization, because it was still present when the T cell lysates were prepared in the presence of excess N-ethylmaleimide, a reagent that chemically modifies free cysteine residues and thus prevents de novo formation of disulfide bonds (data not shown). Moreover, C122A CTLA-4 dimers were preserved under reducing conditions if DSS, a membrane-permeable, 11.4-? arm length cross-linker, was added before lysing the cells, indicating that C122A CTLA-4 dimers are present in intact cells (data not shown). We also noted the presence of a lower band in the C122A CTLA-4 lysates under reducing and nonreducing conditions representing nonglycosylated and/or partially glycosylated CTLA-4.
To explain the mechanistic basis for the function of NNC CTLA-4, we hypothesized that the presence of a bivalent CTLA-4 ligand on the surface of an APC could induce dimerization of NNC CTLA-4. To test this hypothesis, whole cell lysates from stable WT CTLA-4, C122A CTLA-4, and NNC CTLA-4 T cell transfectants were immunoprecipitated with recombinant B7.2-Ig and analyzed for dimer formation under nonreducing conditions. WT CTLA-4, C122A CTLA-4, and NNC CTLA-4 were able to bind to B7.2-Ig to a similar extent on the cell surface (Fig. 7A). We found that B7.2-Ig pulldown caused a significant increase in the proportion of C122A and NNC dimers compared with that seen in whole cell lysates (Fig. 7B). This was particularly apparent for the NNC CTLA-4, for which dimers were not detected in whole cell lysates, but constituted the majority of the CTLA-4 signal in B7.2-Ig pulldown experiments. The C122A CTLA-4 transfectant selected for this experiment expressed a lesser amount of dimer than the average C122A CTLA-4 transfectant to better illustrate the B7.2-Ig-induced dimerization. Therefore, B7 ligands can drive dimerization of monomeric CTLA-4 on the cell surface, and this is associated with normal function of monomeric CTLA-4.
Discussion
Recent crystallographic studies have suggested that CTLA-4-B7 lattices provide the mechanistic basis for triggering the inhibitory function of CTLA-4 (14, 16). To test the biological relevance of CTLA-4-B7 lattices in vivo, we generated a monomeric version of CTLA-4, reasoning that monomeric CTLA-4 would not be able to drive lattice formation with B7. To do this, we mutated residue C122 of CTLA-4 to alanine, because a previous report had shown that C122 forms an intermolecular disulfide bond that is essential for homodimerization in solution (15). These studies provide new insights on the regulation of CTLA-4 oligomerization.
The first remarkable finding is that, contrary to what was expected (15), CTLA-4 in T cells can still dimerize in the absence of C122, with up to 35% of C122A CTLA-4 being dimeric. Such a form can reach the T cell surface, is fully functional, and does not misfold based on its ability to be detected by anti-CTLA-4 Abs and by the fact that it binds B7. The apparent discrepancy between this finding and that of Linsley et al. (15) may be due to the fact that we used intact CTLA-4 expressed in T cells, whereas Linsley et al. used soluble CTLA-4 cleaved from the surface of COS cells. In addition, as shown in this study, CTLA-4 dimers are rescued by aberrant N-glycosylation in T cells, which may not be operational or may be different in COS cells. The second key finding of our studies is that N-glycosylation of CTLA-4 plays a key role in facilitating CTLA-4 dimerization, particularly in the absence of C122-dependent covalent bonding. If it is N-glycosylated, then the mutant C122A CTLA-4 can still partition within lipid rafts, is expressed on the T cell surface, colocalizes with the TCR in the IS, and is fully functional. Our data with PNGase-F and brefeldin A indicate that the rescuing of C122A CTLA-4 by N-glycosylation occurs in the N-glycan maturation process within the Golgi as opposed to the initial addition of immature N-glycans in the ER. It is important to note that the lack of C122 and N-glycosylation sites did not lead to misfolding based on the evidence that the NNC mutant CTLA-4 molecules trafficked to the cell surface, were still recognized by anti-CTLA-4 Abs, relocalized to the IS, were able to bind B7, and inhibited T cell activation. These findings would not be possible if CTLA-4 was grossly misfolded.
The relationship between glycosylation, receptor dimerization, and receptor function has been examined in a few studies, although no single paradigm has emerged. Glycosylation can either enhance or interfere with receptor dimerization. For example, glycosylation at residue N15 is required for 1-adrenergic receptor dimerization and surface expression (31), and deglycosylation of the oncogenic truncated epidermal growth factor receptor impairs its homodimerization and kinase activity (32). In addition, single N-linked glycosylation has been shown to contribute to disulfide-dependent dimer formation of mucins (33). In contrast, glycosylation (sialylation and O-glycosylation) negatively modulates CD45 dimerization (34, 35, 36, 37). Nevertheless, in some cases, glycosylation has no effect on dimerization, as seen with CD26, a serine protease with nine glycosylation sites that retains its ability to dimerize and function in the absence of these N residues (38). Differential glycosylation also has functional consequences, some of them through its effects on receptor dimerization. For example, on CD45, glycosylation decreases dimerization, and this is associated with increased activity of this phosphatase (34, 35, 36, 37). In this sense, it is of interest to point out that T cells express a hypoglycosylated form of CD86 with reduced binding affinity for CTLA-4 and CD28 and thus may act as a regulator of T cell activation (39). N-glycosylation of CD28 can also negatively regulate its binding to CD80 and thus inhibit its costimulatory function (40).
There are several possible mechanisms to account for the contribution of N-glycosylation to CTLA-4 dimerization. The crystal structure of B7.1-engaged CTLA-4 showed that the N-linked oligosaccharide at residue N110 is oriented toward the CTLA-4 dimerization interface, suggesting a role in facilitating dimerization (16). However, we have shown that N110A/C122A mutants do not have a diminished ability to dimerize, implying that dual glycosylation is necessary to drive dimerization. The fact that the NNC CTLA-4 fails to dimerize suggests a role for N78 in this process. The N78 oligosaccharide is thought to hide a hydrophobic patch that can lead to aggregation of CTLA-4 when the N78 residue is mutated (28). Another possibility is that N-glycosylation stabilizes the small noncovalent binding surface at the dimerization interface seen in crystal structures (14, 16). However, this interface, which is composed of 10 highly conserved residues that interact with the same surface on the other CTLA-4 monomer, has a small buried surface area compared with other protein-protein binding surfaces and is thought to be relatively weak (16). We think it unlikely that this weak noncovalent interaction could protect up to 35% of the dimers from the boiling conditions used for Western blotting.
The sensitivity to reduction would imply that a disulfide bond is involved in this process. There are two intramolecular disulfide bonds per CTLA-4 monomer: one between C21 and C92, which is involved in forming the classical Ig fold and is required for maintaining the conformation of the B7 binding motif, and the second between C48 and C66, which is unique to CTLA-4 and has no known function (15). These intramolecular disulfide bonds may indirectly stabilize C122A dimers in a manner analogous to what has been reported for mucins, which can form intramolecular disulfide-dependent cysteine knots (33). Although no such cysteine knot was seen in WT CTLA-4 crystal structures, we cannot rule out such an effect in the C122A mutant. Another possibility is that the C48-C66 bond, which is found close to the dimerization interface (16), could form a novel intermolecular disulfide bond during biosynthesis. We ruled out the possibility of intragel dimerization through de novo disulfide bond formation by treating the cells with N-ethylmaleimide before lysis, which blocks exposed cysteine resides from bonding. Also, the NNC CTLA-4 mutant has all the other cysteine residues intact and still does not dimerize, thus making it unlikely that these other cysteine residues contribute to the formation of C122A dimers. The contribution of glycosylation in this respect could be to slow the rate of biosynthesis and facilitate interactions with chaperones in the ER, as was reported for mucin (33). However, we have shown that dimerization of C122A and NNC CTLA-4 could occur upon ligation with monomeric B7-Ig during immunoprecipitation, suggesting that mature molecules are also capable of forming stable dimers.
Based on the data presented in this study, we propose a hierarchical model of CTLA-4 oligomerization leading to B7:CTLA-4 lattice formation and the triggering of B7-dependent inhibitory functions of CTLA-4. In this model there are three separate steps that supersede each other in a sequential fashion. These are C122-dependent covalent bonding, N-glycosylation, and binding to B7. In the ER, disulfide bonds at C122 form and generate CTLA-4 dimers. In the event that this covalent bond cannot form, oligosaccharide modifications completed in the Golgi allow for dimerization of CTLA-4. Nevertheless, in the absence of N-glycosylation of the molecule (e.g., similar to what has been reported to occur in vivo for B7 (39)), monomeric CTLA-4 can reach the T cell surface, where it will dimerize after binding to B7. This allows for monomeric CTLA-4 to form lattices and inhibit T cell activation in a B7-dependent fashion. It remains to be determined whether the fine architecture of CTLA-4 oligomers induced by B7 in the presence or the absence of disulfide bond/N-glycosylation-dependent dimers is the same, and whether the redistribution of these oligomers changes during T cell activation (41).
Although spontaneous mutations of the C122 residue in CTLA-4 have not been reported in a clinical setting, conditions of cellular stress (e.g., oxidative stress) could cause defects in glycosylation and cysteine-based dimerization/oligomerization of CTLA-4 leading to the generation of monomeric CTLA-4. This possibility has been documented for other molecules. In the case of receptor-like protein phosphatase-, oxidative stress leads to a conformational change in the cytoplasmic tail of the molecule that is subsequently transmitted to the extracellular portion of the molecule and modulates dimer formation and the function of the protein (42, 43). The tumor suppressor ARF also reversibly forms homo-oligomers under oxidizing conditions, and such oligomers can form even after the three sole cysteine residues in this protein are deleted (44). Disulfide bond-independent oligomerization has also been shown for the tumor suppressor B23, which forms oligomers under reducing conditions (45). Finally, the DNA-binding activity of NtcA was perturbed by the reducing agent DTT, and this effect was preserved in NtcA cysteine mutants (46). Thus, the results presented in this study have direct biological implications on CTLA-4 function. By analogy to these reported cases, it is plausible to conclude that the capacity of B7 binding to determine the formation of functional dimers of CTLA-4 is a safe mechanism to avoid loss of function of CTLA-4 under conditions of redox stress after T cell activation.
In closing, it is important to note that the model proposed in this study for CTLA-4 oligomerization is operational for those CTLA-4 functions that are B7 dependent. Recently, there has been increasing interest in B7-independent functions of CTLA-4 (47). The different biological implications for B7-dependent and B7-independent CTLA-4 functions and their differential structural requirements may have contributed to the selective advantage for such elaborate regulation of CTLA-4 dimerization/oligomerization.
Acknowledgments
We thank Drs. B. M. Carreno (Wyeth Research, Cambridge, MA), M. Dustin (Skirball Institute, New York, NY), and the members of the Madrenas laboratory for helpful discussions.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by the Canadian Institutes of Health Research, the Kidney Foundation of Canada, and the London Health Sciences Center Multi-Organ Transplant Program. M.G.K. is the recipient of a Canadian Institutes of Health Research M.D./Ph.D. studentship, and J.M. holds a Canada Research Chair in Transplantation and Immunobiology.
2 P.J.D. and M.G.K. share first authorship.
3 Current address: Montreal Neurological Institute and Hospital, Montreal, Quebec, Canada.
4 Current address: Kennedy Institute of Rheumatology, Imperial College London, London, U.K.
5 Address correspondence and reprint requests to Dr. Joaquín Madrenas, Robarts Research Institute, Room 2.05, P.O. Box 5015, 100 Perth Drive, London, Ontario, Canada N6A 5K8. E-mail address: madrenas@robarts.ca
6 Abbreviations used in this paper: IS, immunological synapse; ER, endoplasmic reticulum; LAT, linker for activation of T cell; NNC, N78A/N110A/C122A; PNGase, peptide:N-glycosidase; SEE, staphylococcal enterotoxin E; WT, wild type; DSS, disuccinimidyl subarate.
Received for publication October 29, 2004. Accepted for publication May 5, 2005.
References
Rothstein, D. M., M. H. Sayegh. 2003. T-cell costimulatory pathways in allograft rejection and tolerance. Immunol. Rev. 196: 85-108.
Waterhouse, P., J. M. Penninger, E. Timms, A. Wakeham, A. Shahinian, K. P. Lee, C. B. Thompson, H. Griesser, T. W. Mak. 1995. Lymphoproliferative disorders with early lethality in mice deficient in Ctla-4. Science 270: 985-988.
Tivol, E. A., F. Borriello, A. N. Schweitzer, W. P. Lynch, J. A. Bluestone, A. H. Sharpe. 1995. Loss of CTLA-4 leads to massive lymphoproliferation and fatal multiorgan tissue destruction, revealing a critical negative regulatory role of CTLA-4. Immunity 3: 541-547.
van der Merwe, P. A., D. L. Bodian, S. Daenke, P. Linsley, S. J. Davis. 1997. CD80 (B7-1) binds both CD28 and CTLA-4 with a low affinity and very fast kinetics. J. Exp. Med. 185: 393-403.
Thompson, C. B., J. P. Allison. 1997. The emerging role of CTLA-4 as an immune attenuator. Immunity 7: 445-450.
Carreno, B. M., F. Bennett, T. A. Chau, V. Ling, D. Luxenberg, J. Jussif, M. L. Baroja, J. Madrenas. 2000. CTLA-4 (CD152) can inhibit T cell activation by two different mechanisms depending on its level of cell surface expression. J. Immunol. 165: 1352-1356.[(Peter J. Darlington2,3, M)