Report

The Immunological Synapse Balances T Cell Receptor Signaling and Degradation

See allHide authors and affiliations

Science  14 Nov 2003:
Vol. 302, Issue 5648, pp. 1218-1222
DOI: 10.1126/science.1086507

Abstract

The immunological synapse is a specialized cell-cell junction between T cell and antigen-presenting cell surfaces. It is characterized by a central cluster of antigen receptors, a ring of integrin family adhesion molecules, and temporal stability over hours. The role of this specific organization in signaling for T cell activation has been controversial. We use in vitro and in silico experiments to determine that the immunological synapse acts as a type of adaptive controller that both boosts T cell receptor triggering and attenuates strong signals.

The mature immunological synapse is characterized by a reorganization of membrane proteins, resulting in a stable central cluster of T cell receptors (TCRs) (the C-SMAC) surrounded by a ring of adhesion molecules (the P-SMAC) (13). Although it was originally proposed that the immunological synapse serves to enhance and sustain signaling through the TCR for long periods of time (3), the paucity of active signaling molecules in the C-SMAC after a few minutes suggests that it may not be involved in signaling (4, 5). We combined in vitro experiments with CD2AP–/– T cells and simulations of a computational (or in silico) model to address this controversy. Our results demonstrate that the C-SMAC is a site for both strong receptor triggering and increased TCR degradation.

Because CD2AP is required for receptor segregation into the C-SMAC in a model immunological synapse (6), we compared synapse formation and signaling in CD2AP–/– cells with those in wild-type (WT) cells. Our attempts were initially complicated by lethal nephrotic syndrome in CD2AP–/– mice by 6 to 7 weeks of age (7). The location of the CD2AP gene on mouse chromosome 17, close to the major histocompatibility complex (MHC) class II locus, necessitated the use of a TCR transgenic mouse that is compatible with the b haplotype. Therefore, we crossed the CD2AP–/– animals to the AND TCR transgenic mouse (8). This TCR recognizes a cytochrome c peptide bound to the MHC class II molecule, I-Ek, but it can develop in the thymus of a b haplotype mouse. CD4 and CD8 profiles from young animals, as well as bone marrow chimeras, showed normal distributions of CD4- and CD8-positive cells, suggesting that thymocyte development is not grossly impaired in CD2AP–/– mice (9). Naïve T cells were isolated from mice at ∼3 weeks of age (the time of kidney disease initiation) and were used immediately or established as T cell lines.

Images of the interface between antigen-bearing APCs and T cells were reconstructed with confocal microscopy and staining with antibodies to the TCR and LFA-1. As reported previously (4), by 30 min, WT T cells form a “mature” synapse: a ring of LFA-1, marking the P-SMAC, surrounding a region enriched in TCR, marking the C-SMAC [fig. S9A (10)]. Consistent with previous work (6), CD2AP deficiency profoundly altered these morphological features [fig. S9B (10)]. At no time were clearly defined C- and P-SMACs visualized, and TCRs were homogeneously distributed with LFA-1 throughout the synapse.

To assess T cell activation, we incubated cells with peptide-pulsed splenocytes. Both naïve and cultured T cell lines from CD2AP–/– T cells showed an increased sensitivity to antigen and augmented cell proliferation compared with WT cells (Fig. 1, A and B). Some of the enhanced proliferation may be attributed to increased levels of secreted interleukin-2 (IL-2) because naïve CD2AP–/– T cells secreted increased IL-2 at all peptide concentrations tested [fig. S6A (9, 10)]. CD2AP–/– T cells also exhibited an increased number of cell divisions and increased apoptosis compared with WT cells [fig. S6B (9, 10)].

Fig. 1.

Hypersensitivity of CD2AP-deficient T cells to antigen. Proliferative responses of (A) freshly isolated naïve splenic T cells and (B) long-term cultured primary T cell lines from AND+/CD2AP+/– or AND+/CD2AP–/– mice were stimulated by antigenic peptide–pulsed APCs at the indicated doses of MCC peptide (88–108) for 48 hours. (A) A representative result from nine independent experiments. The T cell line in (B) was cultured for 6 weeks in vitro and rested for 2 weeks before antigenic stimulation. Mean 3H-thymidine uptakes ± SD are shown. (C) Delayed and sustained tyrosine phosphorylation in CD2AP-deficient T cells. T cells from AND+/CD2AP+/– or AND+/CD2AP–/– primary cell lines were stimulated with the B cell hybridoma TA3 prepulsed with 100 mM MCC peptide. At the indicated time points, cells were lysed and postnuclear lysates were immunoprecipitated with anti-ZAP70 or anti-TCRζ polyclonal rabbit sera. Samples were resolved by SDS–polyacrylamide gel electrophoresis and immunoblotted with anti-phosphotyrosine (4G10). As a protein-loading control, the same blots were stripped and reblotted with ZAP70- or TCRζ-specific antibodies.

The enhanced proliferation associated with CD2AP deficiency correlated with prolonged tyrosine phosphorylation of the TCRζ chain and of ZAP70 (Fig. 1C), parameters that reflect TCR signaling. TCRζ and ZAP70 immunoprecipitates, prepared from T cells stimulated with APC-peptide, were immunoblotted with an antibody to phosphotyrosine. In WT cells, tyrosine phosphorylation of ZAP70 and TCRζ was detectable at 10 min, peaked at 20 min, and returned to base line by 60 min. In contrast, CD2AP deficiency delayed and markedly prolonged the detectable response; tyrosine phosphorylation was barely detectable at 10 min, peaked at 20 min, and was maintained at this level for at least 60 min.

Thus, CD2AP–/– cells did not form a C-SMAC with APC and exhibited sustained tyrosine phosphorylation, which correlates with increased proliferation. These findings appear to support the view that the C-SMAC does not potentiate TCR signaling. However, the veracity of this interpretation is complicated by the many factors that affect amplification and termination of signaling (e.g., receptor-ligand binding, formation of intracellular signaling complexes, kinetic proofreading, serial triggering, receptor endocytosis and degradation, and receptor clustering). To delineate the roles played by each of these factors and to understand how their interplay and coordination affect signal transduction, we developed a computational model in which proteins are represented by particles on a lattice. A kinetic Monte Carlo (MC) algorithm (10) simulated the dynamics of receptor-ligand binding, signal transduction, and protein movement. Particles diffuse, form complexes, catalyze phosphorylation and nucleotide exchange, and undergo phosphotransfer. Each attempt of one such event corresponds to an increment in time of one MC step (the time unit in the simulations). The TCR signaling cascade (Fig. 2) was determined by a specific set of allowed reactions (1113). Signaling is initiated by binding of peptide MHC (pMHC) to the TCR. This allows Lck to sequentially phosphorylate two sites on the TCR (first TCRζ1 and then TCRζ2). These sites are a simplified representation (Fig. 2A, i) of the immunoreceptor tyrosine-based activation motifs (ITAMs). A TCR with only TCRζ1 phosphorylated corresponds to the p21 form, whereas a TCR phosphorylated at both TCRζ1 and TCRζ2 corresponds to the p23 form (14). The extent to which the TCR becomes phosphorylated is dependent on the half-life of the interaction between pMHC and TCR. The phosphorylated TCRζ1 and TCRζ2 can then be bound by one or more ZAP70s, which in turn can be phosphorylated and activated by Lck (Fig. 2A, ii). When ZAP70 is phosphorylated, it can recruit adapter molecules like LAT, GADS (or Grb2), Itk, SLP-76, and SLAP-130, as well as the signaling molecules that bind to them (Fig. 2A, iv to vi). The various downstream intracellular signaling molecules in the model (such as Ras) serve merely as “counters” that give us potential readouts and measure the signal strength. Lck interactions with SHP-1 and ERK (Fig. 2B) provided negative and positive feedback, respectively, to the cascade (13, 15, 16).

Fig. 2.

Schematic representation of the signaling network. (A) Basic network used in the simulations. Each box corresponds to a particle on the lattice; molecules composed of more than one particle are indicated by larger boxes, some of which are subdivided to indicate multiple internal-state variables (e.g., TCRαβ, TCRζ1, and TCRζ2). Colored boxes represent active states and gray boxes represent inactive ones; white boxes represent internal states that do not change. Yellow boxes represent phosphorylation, blue boxes represent conformational changes, and green boxes represent GTPase. Species pairs that form complexes are indicated by dashed lines. Thin single-headed arrows indicate activation (and deactivation) events, and wide double-headed arrows indicate equilibria between states. Thin T-shaped lines indicate inhibition. Two additional transfer reactions, corresponding to generic phosphatases (Pase1 and Pase2) that dephosphorylate TCRζ1 and TCRζ2, are not shown. These reactions were included to allow signaling TCR to revert in the absence of pMHC interactions. (B) Additional interactions and reactions included in the full network used in the simulations. The symbols are the same as in (A). LckY reflects the phosphorylation state of residue Y394 and is taken to be partially active even when not phosphorylated. We neglect regulation at the Y505 site of Lck (by CD45 and Csk), which corresponds to assuming that it is always dephosphorylated (not inhibitory). Ignoring CD45 prevents our model from exhibiting the brief reduction in tyrosine phosphorylation observed in the first few minutes of signaling (5). LckY activates its own inhibitor, the phosphatase SHP-1. When active, SHP-1 dephosphorylates LckY, as shown in (ii). This last reaction is prevented by phosphorylation of residue S59 (LckS) by ERK (a positive-feedback loop). SHP-1 also inhibits signaling by competing with ZAP70 to bind TCRζ1.

To model formation of the C-SMAC, we introduced a force that biases TCR motion toward the center of the interface with the APC. The central accumulation of TCR depends on cytoskeletal and membrane forces regulated by the Rho family of GTPases (17). This was modeled by linking the biasing force to active ZAP70-mediated recruitment of an adapter that can activate a heterotrimeric GTP-binding protein (G protein) (referred to in Fig. 2 as a GTPase that could be Rac). Specifically, receptor movement was stimulated when the number of these activated G proteins exceeded a threshold (10).

TCR internalization and degradation were also included in the model as these processes are thought to play critical roles in turning off TCR signaling. On the basis of existing data, our model constitutively internalizes TCRs at a fixed rate (18). If a receptor is not phosphorylated or only singly phosphorylated (p21), it is returned to the surface. If a TCR is fully phosphorylated (p23), it is degraded; i.e., it is removed from the system (18). Both bound and unbound TCRs in the p23 form were subject to degradation (19). The qualitative results of our simulations were insensitive to whether internalized TCRs were returned to the same spot on the membrane or whether they were randomly returned to any spot on the surface.

Although this model (Fig. 2) is clearly a simplified representation of signaling in T cells, it includes key features of TCR-mediated signaling, thus allowing us to assess how various factors interact to influence signal transduction. As an example of the time course of signaling obtained with our model (Fig. 3A), we used phosphorylation of ZAP70 as the readout because its activation is a critical intermediate in the signaling process. Other readouts downstream of active ZAP70 are also accessible in our simulations, and in all cases, exhibited the same qualitative behavior (10).

Fig. 3.

(A) Time course of ZAP70 activity from simulations of the full model: No C-SMAC formation (blue line; ps = 1.00, where ps is the probability of accepting a displacement away from the center of the junction) and C-SMAC formation (red line; ps = 0.85). Error bars indicate the standard error of the mean for 40 trials. The panels on the right show the spatial distribution of active ZAP70 at the approximate peak in signaling activity (20 × 109 MC steps). On average, TCRs were internalized every 1 × 109 MC steps. The basic results shown here are reproduced by a simpler field model (10). (B) The average number of active ZAP70 as a function of TCR-pMHC off rate at 10 × 109 MC steps. Blue line: basic network shown in Fig. 3B; red line: full network (includes the interactions in Fig. 3C). Error bars indicate the SEM for 10 trials. At the slowest off rate, each pMHC interacts with only 1.3 TCRs on average within the simulation time. As the off rate increases, each pMHC triggers a larger number of TCR until Lck can no longer phosphorylate TCRζ2 within the time a TCR is bound to a pMHC. On average, at the fastest off rate, the time for p23 TCR to form once pMHC is bound is about 13 times as long as the lifetime of the TCR-pMHC complex.

The simulations showed that the magnitude of signaling increases and then decreases over time. Analysis of these results demonstrated that the initial rise reflects the time it takes for the signal to propagate through the network of biochemical events, and the decline of signaling occurs because of receptor degradation. When the model was tested over a range of different TCR-pMHC half-lives, the magnitude of signaling was maximized at an intermediate TCR-pMHC half-life (Fig. 3B). Half-lives longer than the optimum value impaired the ability of pMHC to engage many TCRs (serial triggering) (20). In contrast, short half-lives did not allow sufficient time for receptor triggering (kinetic proofreading) (21). This competition results in an optimal half-life (1921). In addition, the simulations showed that a negative-feedback loop associated with SHP-1 (Fig. 2B) increased specificity by inhibiting TCR-based signaling for shorter TCR-pMHC half-lives (13, 15, 16).

Eliminating centrally biased TCR movement from the model allowed us to ask how the lack of an organized C-SMAC (as seen with conjugates of CD2AP–/– cells with APC) affected signaling. When the C-SMAC did not form, the strength of the signal at early times was somewhat lower than when the C-SMAC did form, but the signal was sustained for a much longer period of time (Fig. 3A). This qualitative difference resembles the difference in the time course of signaling between CD2AP–/– and WT cells (Fig. 1C), except that the in silico experiments without C-SMAC formation did not exhibit a marked delay in the onset of signaling. The qualitative results in Figure 3A are robust to 20-fold variations in the kinetics of the reactions that constitute the signaling pathways in our simulations.

Our analysis of the simulation results demonstrates that the C-SMAC enhances signaling by concentrating TCR, pMHC, and kinases like Lck into a small area. Clustering TCR and pMHC allows for more frequent TCR-pMHC binding. When the C-SMAC does not form, the median time for a pMHC to rebind a TCR after dissociation is 20.3 × 106 MC steps. In contrast, the corresponding time is only 2.7 × 106 MC steps when TCRs cluster in the C-SMAC. Receptor clustering in the C-SMAC should therefore directly enhance the frequency of TCR-pMHC complex formation without any change in the TCR-pMHC half-life. Our analysis also showed that the C-SMAC facilitates TCR phosphorylation by Lck because Lck can act on clustered TCRs rather than single TCRs. These factors combine to result in a much higher rate of production of fully phosphorylated TCRs (p23) when the C-SMAC does form. Why then do we observe a paucity of phosphorylated molecules in the C-SMAC over long times in the in vitro and in silico experiments (Figs. 1C and 3A)?

Because only fully phosphorylated receptors are subject to degradation, the higher rate of production of fully phosphorylated TCRs in the C-SMAC enhances receptor degradation. Thus, counterintuitively, enhanced receptor triggering in the C-SMAC also serves to limit sustained tyrosine kinase activity in the C-SMAC over long times. The higher the rate of receptor degradation, the shorter will be the time period over which phosphorylated molecules are observed in the C-SMAC. Indeed, the model suggests that the only way to sustain TCR signaling over longer periods would be to have TCR replenishment from new synthesis (22, 23). The persistent high level of signaling in the absence of CD2AP may therefore stem from defects that lower the rate of TCR triggering, and/or concomitant degradation of activated TCRs.

To directly assess whether CD2AP deficiency affects TCR down-regulation, we measured TCR expression levels before and after T cell activation. Antigen-pulsed APCs induced WT T cells to down-regulate TCR expression (Fig. 4A) (24). CD2AP–/– T cells only minimally down-regulated TCR expression (Fig. 4A). This was due directly to the absence of CD2AP because reconstitution of CD2AP–/– cells with a CD2AP-expressing retrovirus restored TCR down-regulation (Fig. 4B). To assess the effect of CD2AP deficiency on degradation of TCR (22), we measured TCRζ expression before and after stimulation with peptide-pulsed APCs. Whereas T cell activation greatly reduced the level of TCRζ in WT T cells, there was no effect on TCRζ chain levels in CD2AP–/– T cells (Fig. 4C). This could be due to a role for CD2AP in intracellular trafficking because CD2AP deficiency impaired delivery of the TCR to lysosomes (fig. S7) (10, 25, 26).

Fig. 4.

Defective TCR down-regulation and degradation in CD2AP-deficient T cells. (A) Naïve T cells from either AND/CD2AP+/– or AND/CD2AP–/– mice were stimulated with splenic APCs from B10.BR mice in the presence or absence of 10 μM MCC peptide. After 2 hours, cell conjugates were disrupted by treatment with EDTA-trypsin. The surface expression of AND TCRs was analyzed by flow cytometry with the TCR Vb3-specific antibody KJ25 and gating on Thy1.2-positive cells. (B) Retroviral transfection of CD2AP reconstitutes TCR down-regulation in CD2AP–/– T cells. (Left) CD2AP–/– T cell lines were transduced with CD2AP–green fluorescent protein (GFP) retrovirus and stimulated as described in (A). The surface expression of AND TCRs was analyzed in the GFP-positive (R1) or -negative (R2) T cell population. (Right) CD2AP–/– T cells were transduced with a retrovirus expressing GFP alone. (C) TCR degradation in CD2AP+/– and CD2AP–/– T cells. Lymph node T cells from either AND/CD2AP+/– or AND/CD2AP–/– mice were stimulated for 4 hours with the B cell hybridoma CH27 prepulsed with 100 μM MCC peptide. Cells were lysed in RIPA lysis buffer, and postnuclear lysates were immunoprecipitated and immunoblotted with anti-TCRζ polyclonal rabbit sera. To block synthesis of TCRs, T cells were pretreated with cycloheximide for 1 hour before antigenic stimulation.

To determine whether the defect in down-regulation also involved changes in TCR internalization, we measured the basal rate of TCR internalization using an inhibitor of anterograde transport, brefeldin A. In the presence of brefeldin A, TCRs were steadily lost from the plasma membrane because internalized TCRs are unable to recycle (18). CD2AP deficiency did not alter this basal rate of internalization (fig. S5) (9). Furthermore, antibody-mediated TCR internalization at early time points was similar between WT and CD2AP–/– T cells (9). Therefore, the failure to down-regulate TCR in the CD2AP–/– T cells appears to be due to a defect in TCR degradation, not internalization.

A counterintuitive prediction of the computational model is that the attenuation of phosphotyrosine levels in the C-SMAC is due to enhanced receptor triggering, which results in increased receptor degradation. Because CD2AP–/– cells cannot mediate receptor degradation, we reasoned that, if these cells could be induced to form a C-SMAC, we could directly assess whether there is enhanced receptor triggering in the C-SMAC. A mathematical model analyzing the thermodynamics of synapse formation predicts that a C-SMAC should form more readily with planar lipid bilayers containing intercellular adhesion molecule 1 (ICAM-1) and pMHC than with APCs because of higher ligand mobility in the planar bilayer and because only one deformable membrane is involved (27); both factors make reorganization of receptors and ligands easier. Furthermore, CD2AP may be required for C-SMAC formation only when there are large numbers of CD2-CD48 interactions that require organization, which are present in the cell-cell system but absent in the bilayer system. These reasons led us to examine whether CD2AP–/– cells could form a C-SMAC with planar bilayers.

Despite some defects in the fine structure, C- and P-SMACs were readily formed in the absence of CD2AP (Fig. 5A), demonstrating that CD2AP is not absolutely required for C-SMAC formation. This experimental system allowed us to measure signaling in a C-SMAC in the absence of receptor degradation. The in silico model was used to predict the outcome of signaling in the C-SMAC with (WT cells) and without (CD2AP–/– cells) receptor degradation. Based on the presence of phosphotyrosine as the readout, the model predicts sustained and strong phosphotyrosine staining in the C-SMAC, assuming formation of the synapse in the absence of receptor degradation (Fig. 5B). Consistent with this prediction, anti-phosphotyrosine stained the P-SMAC but not the C-SMAC in WT cells (Fig. 5C). In contrast, when CD2AP–/– T cells were stained for phosphotyrosine, the greatest staining was observed in the C-SMAC (Fig. 5C). These experiments confirm that concentrating receptors in the C-SMAC facilitates and enhances signaling by the TCR and that the absence of signaling intermediates at later time points is due to concomitantly higher receptor degradation. The observation that CD2AP–/– cells in the bilayer system exhibit discernable C-SMACs that are the site of the strongest signaling also provides evidence against a role for CD2AP only in formation of the C-SMAC. If CD2AP's only function was to promote C-SMAC formation, WT cells and CD2AP–/– cells should exhibit nearly identical behavior in the bilayer experiments. Indeed, CD2AP–/– cells exhibited the strongest phosphotyrosine levels, and WT cells, the weakest phosphotyrosine levels, in the C-SMAC (Fig. 5).

Fig. 5.

CD2AP–/– cells form a C-SMAC with planar bilayers and exhibit strong signaling. (A) Immune synapse formation with planar lipid bilayers. T cell blasts from either CD2AP+/– or CD2AP–/– mice were incubated on a supported planar lipid bilayer containing Oregon Green-I-Ek (prepulsed with 100 μM peptide) at 166 molecules/μm2 and Cy5–ICAM-1 at 266 molecules/μm2. Cells were imaged at 37°C with a Zeiss Confocal LSM510 microscope in real time. (B) Time course of phosphorylated TCR, LckY, and ZAP70 (pY) obtained from the computational model for cases that form the C-SMAC with (red curve) and without (blue curve) TCR degradation. Error bars indicate the SEM for 40 trials. The panels show the levels of phosphotyrosine at 60 ×109 MC steps in the C-SMAC only. Brighter shades of red correspond to higher phosphotyrosine levels. (C) Tyrosine phosphorylation patterns in CD2AP+/– and CD2AP–/– T cells on planar lipid bilayer. T cell blasts were plated on supported planar lipid bilayer containing I-EK (prepulsed with peptide) and ICAM-1, as in (A). After 1 hour, cell-bilayer conjugates were fixed and permeabilized. Tyrosine phosphorylation was visualized by staining with a phosphotyrosine-specific antibody (4G10). (Right) Quantitation of these patterns was obtained by calculating the ratio of C-SMAC fluorescence divided by the P-SMAC fluorescence in over 15 cells per experiment. Data are the average of three independent experiments. Error bars denote standard deviations.

Combining experiments with CD2AP–/– cells and simulations of a computational model reveals an unsuspected function of the C-SMAC and resolves the controversy regarding the role of the C-SMAC in propagating and attenuating signals. Concentrating antigen, TCR, and kinases in the C-SMAC enhances signaling by decreasing the amount of time required for antigenic ligand to search and find TCR and for subsequent receptor phosphorylation. Because of these factors, it is predicted that fully phosphorylated receptors (p23) form in the absence of C-SMAC formation only with high-affinity agonists; p23 generation is greatly facilitated by the C-SMAC when ligand quality is weaker. However, enhanced receptor triggering in the C-SMAC results in increased receptor degradation, which limits sustained tyrosine phosphorylation within the C-SMAC. The model therefore predicts that the C-SMAC is responsible for intense but self-limited signaling. Thus, the synapse functions as an adaptive controller. Limiting strong signaling over long times may serve to protect against cell death caused by overstimulation. This is consistent with our observation of enhanced apoptosis in CD2AP–/– cells (9).

The adaptive control function of the C-SMAC revealed by our work touches on many complex issues including ligand quality, TCR down-modulation, and partial TCR signaling. For example, our model has implications for the biology of signaling by altered peptide ligands (APLs) (28). Our model suggests that the critical link between partial ζ chain phosphorylation and inefficient TCR down-modulation of APLs (29, 30) is their inability to form the immunological synapse (3), which both enhances and limits signaling. These and other hypotheses emerging from the adaptive control function of the C-SMAC are testable and should motivate further experiments that will result in a deeper understanding of the complex issues underlying T cell activation.

The function of the immunological synapse is an emergent property involving many inextricably linked variables, and our work illustrates how the analysis of such complex biological systems benefits greatly from synergistic experimental and computational studies. Without the computational model, for example, we might have wrongly concluded that the C-SMAC is not involved in TCR triggering and that it functions only to attenuate signaling. The surprising interpretation that the C-SMAC balances TCR signaling and degradation emerged from the computational model and the cellular experiments it suggested.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1086507/DC1

SOM Text

Fig. S1 to S9

Table S1 to S4

References

References and Notes

View Abstract

Navigate This Article