Multifocal structure of the T cell – dendritic cell synapse



The structure of immunological synapses formed between murine naive T cells and mature dendritic cells has been subjected to a quantitative analysis. Immunofluorescence images of synapses formed in the absence of antigen show a diffuse synaptic accumulation of CD3 and LFA-1. In electron microscopy, these antigen-free synapses present a number of tight appositions (cleft size ∼15 nm), all along the synapse. These tight appositions cover a significantly larger surface fraction of antigen-dependent synapses. In immunofluorescence, antigen-dependent synapses show multiple patches of CD3 and LFA-1 with a variable overlap. A similar distribution is observed for PKCθ and talin. A concentric organization characteristic of prototypical synapses is rarely observed, even when dendritic cells are paralyzed by cytoskeletal poisons. In T–DC synapses, the interaction surface is composed of several tens of submicronic contact spots, with no large-scale segregation of CD3 and LFA-1. As a comparison, in T–B synapses, a central cluster of CD3 is frequently observed by immunofluorescence, and electron microscopy reveals a central tight apposition. Our data show that it is inappropriate to consider the concentric structure as a “mature synapse” and multifocal structures as immature.


Central supramolecular activation cluster




Immunological synapse


Peripheral SMAC


The structural organization of immunological synapses (IS) is expected to have important consequences on their functioning. In particular, large-scale molecular segregation in the synapse plane, as determined on immunofluorescence (IF) images, puts precise constraints on the molecular interactions that are likely to take place or not. These interactions are further constrained by the third dimension of the synapse, i.e. the precise size of the synaptic cleft that can be measured by EM.

The prototypical IS is considered as being organized around a c-SMAC (central supramolecular activation cluster) containing the TCR, surrounded by a p-SMAC (peripheral SMAC) 14; recently the idea of a d-SMAC (distal SMAC) has been added 5. The distance between the pre- and post-synaptic membranes is expected to be around 15 nm at the c-SMAC, to allow the interactions of 7-nm-high molecules such as the TCR or MHC, and 40–45 nm at the p-SMAC, to fit with the expected space required for integrin–ICAM-1 interactions 6.

The concentric, c-SMAC–p-SMAC organization has been well described in a number of IS including those formed between T cells and B cell lines, lipid bilayers containing ICAM-1 and MHC molecules, and macrophages 1, 2, 7. Cytotoxic T cells interacting with target cells show a minor variation on this theme, with a double c-SMAC where the TCR and secretory apparatus are juxtaposed 8. The synapses formed between NK cells and target cells under conditions where killing is prevented by inhibitory receptors has been described as being able to adopt several topologies: a concentric structure with a center of adhesion molecules surrounded by a ring of receptors; a multifocal structure; or a diffuse organization without apparent structure 9. A multifocal IS has also been observed at the synapse formed between thymocytes and thymic epithelial cells 10. The existence of well-defined synaptic regions with precise and distinct cleft widths of 15 and 40 nm has been predicted 6 and quoted in numerous reviews. However, despite two studies on NK–target-cell synapse structure 11, 12, this question has not yet been addressed in detail.

DC have the unique ability to activate naive T cells in vivo, and are thus at the origin of primary immune responses. Despite its exceptional physiological importance, the IS formed between T cells and DC has been little studied, except in a few papers where the existence of T–DC synapses in the absence of exogenous Ag was demonstrated 13, 14. These Ag-free synapses are characterized by the accumulation of CD3, CD4 or CD8, talin, CD45 and phosphotyrosine, and by the relocation of the centrosome towards the IS. In one previous study on the IS formed between T cells and DC, the EM data showed no indication in favor of the existence of a c-SMAC 13, but this point was not completed by IF data, or by a thorough ultrastructural analysis. Given the physiological importance of this synapse, we decided to examine more closely its structure, both in the presence and in the absence of exogenous Ag.

IF and EM were used to analyze the IS formed between mature DC and naive T cells from P14 TCR-transgenic mice. A quantitative analysis of the data allows several conclusions to be drawn. In the presence of Ag, a concentric organization is observed in a minority of the IS examined, whereas a multifocal organization is seen in most cases. In EM, a unique central apposition of 15 nm between the T cell and DC membranes was never observed, whereas such appositions were frequently observed at T–B synapses. In T–DC synapses, close appositions of 15 nm always appear to be multifocal, and cleft sizes of 40 nm for a significant distance are not observed. Thus, the structure observed at the T–DC synapse appears quite different from that of the prototypical IS.


IF analysis of T–DC synapses

The structure of T–DC synapses was first examined with the spatial resolution (∼0.2 µm in XY and ∼0.4 µm in XZ) permitted by fluorescence imaging with a high numerical aperture objective after image deconvolution. P14 T cells were added to adherent DC pulsed with 1 µM Ag peptide or left unpulsed. Fifteen or thirty minutes (as specified) after adding the T cells, the cells were fixed and labeled with specific Ab. In the conditions under which conjugates were formed, when the cells were not fixed and left in contact for 48 or 72 h, a standard functional outcome was observed: T cell survival without proliferation in the absence of Ag, massive proliferation in the presence of Ag (data not shown).

In the absence of added peptide, as previously reported, many Ag-free IS can be observed 14. They never exhibited a central accumulation of CD3. The structure of Ag-specific IS was more variable (Fig. 1). In a majority of the IS, CD3 distribution was uniform or multifocal, whereas a unique central accumulation (a c-SMAC-like structure) could be observed (Fig. 1) in approximately 30% of the T–DC conjugates (n=101). LFA-1 staining was increased in the contact zone in more than 80% of the conjugates.

Figure 1.

Synaptic CD3 clustering after 15 min of T–DC interaction. An Ag-free synapse (Ag-) shows a typical synaptic CD3 enrichment, without central accumulation. A central CD3 accumulation can be observed in some but not all Ag-dependent synapses (Ag+). Bar=5 µm.

When Ag-specific cells were simultaneously labeled with anti-CD3 and anti-LFA-1 Ab (Fig. 2A, B), en face views reconstructed from stacks of 30–40 XY planes revealed that a majority of synapses display a multifocal distribution of CD3 and LFA-1. The two molecules could either show some co-localization in the same patches (Fig. 2A) or appear distributed in distinct patches (Fig. 2B). Only 17.6±3.5% (n=4 experiments) of T–DC conjugates showed a molecular organization that resembled the c-SMAC–p-SMAC model, with a unique and central accumulation of CD3 and a peripheral localization of LFA-1. Even in these cases, LFA-1 was usually found on two sides of the CD3 cluster rather than all around it (Fig. 2B), and it was not always obvious if LFA-1 accumulation was in the T and/or in the DC membrane. In order to evaluate the amount of synaptic recruitment of CD3, the fraction of CD3 found in Ag-dependent synapses, relative to the amount found in the whole cell, was quantified either in equatorial planes or in three dimensions (see “Materials and methods”). The two measurements gave exactly the same result: 69±7% (n=18) of CD3 molecules from the whole cell.

Figure 2.

Structural diversity of Ag-dependent T–DC synapses. Double-labeling of CD3 and LFA-1 after 15 min of T–DC interaction, illustrated in two examples (A, B). First column: CD3 (deconvoluted XY and XZ views). Second column: LFA-1 (XY and XZ). Third column: DIC and overlay of the two XZ views (CD3 green and LFA-1 red). In the first two columns, the signal intensity was pseudo-colored with hues ranging from blue (low) to red or white (high). Z step: 250 nm. Bar=5 µm. (C) quantitative analysis of CD3 and LFA-1 localization in Ag-dependent T–DC synapses (single labeling). Analysis of labeling of CD3 or LFA-1 after 15 or 30 min of T–DC interaction (central, peripheral or multifocal distributions). Mean±SD, n=4 experiments (10–15 cells analyzed per experiment). (D, E) Double-labeling of PKCθ (red) and talin (green) after 30 min of T–DC interaction, illustrated in two examples. Bar=5 µm.

The multifocal pattern is not a transient state of T–DC synapse formation

A multifocal patterning at the IS has been suggested to represent an early transient state before distinct patches coalesce into a unique concentric c-SMAC–p-SMAC structure 4. To examine if the structure observed after 15 min of T–DC interaction was a transient one, we also analyzed IS after 30 min of T–DC interaction. Given the high efficiency of DC as APC and the short delay between T–DC contact and the triggering of a Ca2+ responses 13, 30-min-old T–DC conjugates cannot be considered as being in a transient state of synapse formation. In these older synapses, the distribution of CD3 and LFA-1 was strikingly similar to the one observed after 15 min of interaction, i.e. essentially multifocal (Fig. 2C). The percentage of conjugates containing both a unique central CD3 localization and a peripheral LFA-1 distribution was 17.6±3.5% (n=4 experiments, with 10–15 cells in each experiment) in 15-min-old synapses and 23.9±5.1% (n=3) in 30-min-old synapses; it was not significantly different in the two situations. Synapses statistically examined at earlier times points did not show different structures (not shown).

Next, we analyzed the distribution of PKCθ and talin, which are two intracellular molecules recruited at IS, after 30 min of T–DC interaction. As shown in Fig. 2D and E, the synaptic distribution of these molecules was also multifocal. Contrary to what has been described at the canonical “mature synapse” (where PKCθ is central and talin at the periphery, with no overlap between the two), there was no systematic exclusion or overlap of PKCθ and talin at T–DC synapses.

Dynamics of the T–DC synapse

The fact that a majority of the T–DC synapses appear multifocal at different times after their formation does not imply necessarily that these structures are stable. To further examine this point, we analyzed the dynamics of individual T–DC synapses. T cells expressing an HA-specific TCR and a GFP-coupled TCRζ chain were added to HA-loaded DC, and stacks of images were taken usually every 3 min in order to reconstruct en face views of the synapse. To examine the degree of functionality of a given type of structure, T cell Ca2+ responses were monitored in parallel every 20 s, together with transmitted-light images.

Fig. 3 illustrates two examples of the dynamics of the T–DC synapse. In the first example (top), 3 min after the beginning of the Ca2+ response, ζ–GFP clustering was observed in a central part of the synapse. This central accumulation was then disrupted into several distinct patches. Note the important mobility of the cell, both in the XY plane and along the Z axis.

Figure 3.

Dynamics of the T–DC synaptic structure in parallel with the Ca2+ response. Two examples of T cells expressing an HA-specific TCR and a GFP-coupled TCRζ chain, interacting with HA-loaded DC typical of five recordings (>10 min). Three series of images were acquired: Ca2+; transmitted light (with the contact zone indicated by an arrow); and stacks of ζ−GFP images used to reconstruct en face views. Each CD3ζ–GFP picture is numbered in white and its position reported on the calcium curve in bold. The dotted line gives the Ca2+ level in a typical unactivated cell. The bar in the top row is 5 µm.

The second example (Fig. 3, bottom) illustrates a case of a more sustained Ca2+ response that had started even before the beginning of the recording. In the first two en face views, most ζ–GFP appeared as a unique and mobile cluster. This cluster then split into several parts, and the Ca2+ level remained high during all these changes of the synaptic structure. In other cases (not shown), reversible coalescence of several clusters could also be observed. In general, there was no trend towards a progressive coalescence or towards a multiplication of the clusters, but just a systematic ability to go in either direction.

In the absence of antigen, the increased mobility of the T cell and the weaker synaptic accumulation of ζ–GFP did not allow the collection of complete stacks of images suitable for three-dimensional reconstruction of the IS.

Disruption of DC cytoskeleton does not increase the occurrence of c-SMAC–p-SMAC structures

Thus, T cells form multifocal synapses with DC, which are highly motile cells and require an intact cytoskeleton to be fully active as APC 15. On the contrary, T cells form concentric synapses with immobile APC such as B cell lines or lipid bilayers (see 16 for a review). It was thus of interest to examine if the DC motility was influencing the type of synaptic structure formed with T cells. To address this issue, DC were incubated with 20 µM cytochalasin D plus 5 µM nocodazole for 1 h prior to being added to T cells. Visual inspection of the cells confirmed the efficacy of the treatment: treated DC were immobile, with few dendrites. A similar treatment has been shown to reduce the number of T–DC conjugates 15. After 15 or 30 min of interaction, the most frequently observed synaptic structure was a multifocal one. The frequency of c-SMAC–p-SMAC structures was 16.0±5.2% (n=3 experiments) when DC were fixed or paralyzed with cytoskeletal poisons, and was almost identical, i.e. 16.2±5.3% (n=2) in the corresponding controls. Thus, the multifocal structure of the T–DC synapse is not related to the high cytoskeletal dynamics of DC.

T–DC synapses in EM: a quantification method

To get further insight on the synaptic structure with a better spatial resolution, T–DC synapses were next examined by transmission EM. Fig. 4A (top) shows typical examples of T–DC conjugates. DC and T cells can be distinguished by a number of features. T cells are smaller than DC, more spherical, their nucleus is more electron-dense and is surrounded by a thinner cytoplasm. Organelles such as Golgi apparatus, and lysosomes, are usually concentrated in one area of the cytoplasm. DC are larger, the nucleus is less electron-dense, and the abundant cytoplasm contains numerous organelles, in particular mitochondria, endoplasmic reticulum, endosomes and lysosomes.

Figure 4.

Quantification of synapse dimensions. Two typical examples of Ag-independent (left) and Ag-dependent (right) synapses are shown. (A) Original images (top) zoomed on the synapse (bottom). (B) Linearization, and drawing of regularly spaced lines spanning the synaptic cleft. (C) Plot of the cleft size as a function of the distance from the synapse edge. Dotted lines indicate cleft widths of 15 nm and 45 nm. (D) Histograms giving the relative frequencies of cleft widths measured in (C). The histograms corresponding to the close appositions (between 0 and 18 nm) and intermediate appositions (45±9 nm) are highlighted in black.

To examine membrane appositions more precisely, a systematic quantification of the size of the synaptic cleft was performed (see “Materials and methods”). Briefly, the original synapse image (Fig. 4A, bottom) was first linearized (Fig. 4B, top). Then, bars spanning the synaptic cleft were superimposed on the image. This series of bars (Fig. 4B, bottom) gives a faithful representation of the variations of the size of the synaptic cleft all along the synapse. A quantified profile of the IS was then automatically generated (Fig. 4C). To better visualize the expected zones of narrow (∼15 nm) and intermediate (∼40–45 nm) cleft widths, the 15 and 45 nm distances are indicated by dotted lines. A useful representation of the data is obtained by switching from the spatial to the frequency domain: each synapse profile was transformed into a histogram giving the relative frequencies of cleft widths measured in this particular IS (Fig. 4D). In the example shown in Fig. 4D, close-apposition frequency is higher in the Ag-dependent IS than in the Ag-independent one.

Quantitative analysis of T–DC synapse dimensions

This quantification method was then applied to the analysis of 13 synapses formed in the presence of Ag and 11 synapses formed in its absence, both after 15 min of interaction. As summarized in Table 1, in the absence of Ag, the average contact length of the synapses was 3.4 µm. Close appositions (cleft size <18 nm) were observed in several discrete regions, but they never formed a conspicuous central structure. The cumulative length of these close appositions was <20% of the IS length. Between the close appositions, the membranes were several tens of nm apart, presenting large undulations. The standard deviation (SD) of the distance profile, quantified in each IS, is an index of the amplitude of these undulations.

Table 1. Statistical analysis of T–DC immunological synapse parametersa)
FeatureAg–n=11 conjugatesAg+ (15 min) n=13 conjugatesAg+ (30 min) n=20 conjugatespb)–/+15 15/30
  1. a) Summary of the quantification performed on 11 Ag-free (Ag–) IS, 13 Ag-dependent (Ag+) synapses analyzed on conjugates fixed 15 min after adding T cells to DC, and 20 Ag+ synapses formed after 30 min of interaction.b) The left p-value accounts for the comparison between Ag– at 15 min and Ag+ at 15 min (–/+15), the right one for the comparison between Ag+ at 15 min and Ag+ at 30 min (15/30). ns, not significant.

Size of the synapse3.37±1.03 µm4.20±1.23 µm3.9±1.2 µm<0.01 ns
Mean cleft size52.3±17.9 nm24.3±18.0 nm20.2±16.0 nm<0.01 ns
Mean variations within an IS (SD measured for each IS and averaged)37.2±16.9 nm18.1±7.5 nm16.0±6.3 nm<0.01 ns
Cumulative length of close apposition per IS0.63±0.43 µm2.20±0.77 µm2.50±0.90 µm<0.01 ns
Fraction of the synapse occupied by close appositions (15-nm cleft)19.4±12.8%54.0±15.0%65.9±11.3%<0.01 <0.01
Fraction of the synapse occupied by intermediate appositions (45-nm cleft)20.7±8.0%13.6±7.0%8.5±4.2%<0.05 <0.01
Longest close apposition200.4±120.0 nm601±249 nm713±277nm<0.01 ns
Fraction of IS showing one close apposition >400 nm9.0%92.3%90.0%
Mean size of continuous region of close apposition (nm)86.5±38.0 nm222.3±120.0 nm312±156 nm<0.01 <0.10
Mean number of close appositions/image7.0±2.311.4±4.79.5±3.7<0.01 <0.10

These parameters were significantly modified when Ag was present at the DC surface. Under those conditions, the contact length was increased by 25%, and close appositions covered more than 50% of the total synaptic surface. In addition, membrane undulations were less marked, as measured by a 50% decrease of the SD of synaptic cleft size.

T–DC synapses do not show a central tight apposition circled by a 40-nm-wide crown

To establish if (despite the absence of clear-cut central appositions corresponding to a c-SMAC), there was a tendency for tight appositions to cluster in the IS center, we measured the frequency of occurrence of tight appositions as a function of the distance from the center of the synapse. In all situations, the probability of observing tight appositions (<18 nm clefts) was significantly larger in the presence than in the absence of Ag, and was widely distributed over the whole synapse (data not shown). This homogenous distribution of tight appositions all along the IS underlines that at the electron-microscopic level, there is no c-SMAC-like structure in T–DC synapses.

Fig. 5A, B and C show the histograms giving the relative frequencies of cleft sizes, similar to the ones shown in Fig. 4D, but averaged this time on all IS analyzed. A striking result is the absence of a peak around 40–45 nm. Fig. 5D shows the profile of a prototypical theoretical IS including a well-defined p-SMAC with a cleft size of 40–45 nm. Fig. 5E shows the quantified profile of such an IS that would have been generated by our quantification approach (extra noise has been added to the measurements, to mimic an experimental quantification). The resulting histogram giving the relative frequencies of cleft sizes is shown in Fig. 5F. It shows two prominent peaks at 15 and 45 nm. The comparison with experimental average histograms shows without ambiguity that in T–DC synapses, distinct close appositions of ∼15 nm do exist, whereas one never observes significant p-SMAC-like sections of the IS, i.e. 40–45 nm apart.

Figure 5.

Experimental and theoretical relative frequencies of cleft widths. Experimental (top): data averaged in 11 Ag-free synapses (A), 13 Ag-dependent synapses after 15 min of interaction (B), and 20 Ag-dependent synapses after 30 min of interaction (C). Highlighted in A, B, C and F are the close appositions and those in the 35–55-nm range. Theoretical (bottom): data expected for a prototypical IS. (D) Scheme of a prototypical IS with a 15-nm cleft at the c-SMAC and a 45-nm cleft at the p-SMAC. (E) Synaptic profile, corresponding to (D), with 15-nm and 45-nm widths indicated by dotted lines. (F) Histogram giving the relative frequencies of cleft widths corresponding to (E).

To analyze in more depth the multifocal nature of the T–DC synapse we performed 13 serial sections of one T–DC conjugate in order to analyze the surface of cell–cell contact. Fig. 6 (top) shows one of the sections of the T–DC synapse. The size of the synaptic cleft was measured in each section, and the information contained in the series of 13 profiles allowed the construction of a pseudo-color-coded en face view of the size of the synaptic cleft for a synaptic surface of 3000×900 nm (Fig. 6, bottom). This reconstructed view further illustrates the multifocal distribution of close membrane appositions.

Figure 6.

Multifocal distribution of close appositions at the single-cell level. Synaptic profiles were measured in 13 EM serial sections of one T–DC conjugate, as previously described, but without the linearization step. Each section being 70 nm thick, reconstruction covers approximately 900 nm in the Z- axis. Top: one section used to construct the en face view of the synapse. Bottom: the pseudo-colored reconstruction of the en face view (close appositions shown in red, as indicated in the color scale).

LFA-1 distribution in T–DC synapses estimated by EM

The possibility that the distance of interaction between LFA-1 and ICAM-1 could be smaller than previously expected has recently been questioned, on the basis of crystallographic data 17 (see “Discussion”). As our data show that in T–DC synapses, the cleft size is rarely close to 40–45 nm, we asked whether LFA-1–ICAM-1 interactions could take place when the synaptic cleft was narrower than that. Such a possibility should give rise to an accumulation of LFA-1 in tight appositions.

This question was addressed by performing immunogold labeling with anti-LFA-1 Ab and protein-A–gold, after fixation, and before embedding in epon. When LFA-1 distribution was examined at T–DC synapses after 30 min of cell–cell interaction (Fig. 7A), particles were observed all along the synapse, sometimes near or inside tight appositions (arrows). Occasional patches of 2–8 particles were also observed at the periphery of the synapse. The cleft size in which particles were observed shows a wide distribution with a maximum around 24–30 nm (Fig. 7B), with no sign of preferential localization of LFA-1 at a ∼40-nm distance.

Figure 7.

LFA-1 immunolabeling of Ag-dependent T–DC conjugates. The T–DC conjugates were labeled with an anti-LFA-1 Ab and 5-nm gold particles coupled to protein A. The cleft size was measured at the level of each particle. In (A) are shown examples of particles in synaptic clefts. Arrows indicate particles located near or within close appositions. (B) Fraction of particles found at the different cleft sizes (mean of two experiments, n=179 and 154 particles), with corresponding typical examples.

Is a multifocal structure typical of T–DC synapses?

To examine how specific were the synaptic structures observed with T–DC conjugates, we examined synapses made by exactly the same T cells, with primary B cells and macrophages. It is well established that such synapses do not form in the absence of Ag 13, so B cells or macrophages were loaded overnight with 1 µM antigenic peptide.

In IF experiments, a multifocal distribution of LFA-1 and CD3 can be observed in a majority of T–B synapses observed after 30 min of contact (Fig. 8A). Synaptic CD3 accumulation was less marked at T–B synapses than at T–DC synapses (52% at T–B synapses versus 69% at T–DC synapses). In approximately one-third of the conjugates, CD3 was strikingly clustered in the center of the synapse (Fig. 8B). In Fig. 8B, the CD3 cluster appears surrounded by a peripheral LFA-1 ring in the en face view. However, in the equatorial XY plane, it is clear that the origin of the ring of LFA-1 was essentially in the B cell; this ring cannot be considered as a T cell p-SMAC.

Figure 8.

Structures of T–B synapses. (A, B) Double-labeling of CD3 and LFA-1 after 15 min of T–B interaction, illustrated in two examples. First column: CD3 (deconvoluted XY and XZ views). Second column: LFA-1 (XY and XZ). Third column: overlay of the XY and XZ views (CD3 green and LFA-1 red). Bar=5 µm. (C) Two electron micrographs of T–B synapses. Bar=1 µm. The T cell is the cell on the right in each pair (B220 immunolabeling clearly visible at the surface of the left cell of the right micrograph).

T–B synapses were further examined by EM. Sometimes, multifocal structures were observed. More frequently (Fig. 8C), there was a large central apposition, with smaller lateral contacts. Strikingly, in all cases where a central apposition was observed, it was not surrounded by an apposition of ∼40 nm.

The structure of synapses between T cells and macrophages was examined by IF. P14 T cells formed conjugates less easily with macrophages than with B cells. However, almost all reconstructed synapses were again multifocal. Like in T–B synapses, CD3 and LFA-1 labeling did not show a systematic colocalization or exclusion (data not shown). Although CD3 was sometimes concentrated in the central part of the synapse, this central clustering was less frequent and less marked than with B cells. This feature is consistent with the fact that in electron micrographs, we never observed T–macrophage synapses with a dominant, central tight apposition similar to that frequently observed at T–B synapses.


We have performed a thorough quantitative analysis of IS images obtained by EM, and compared this information with those obtained by fluorescence microscopy. The data, centered on the T–DC synapse, and compared with the T–B synapse, confirms and extends previous findings already suggesting that the T–DC synapse does not share several features of the prototypical IS (see 16 for review), in particular concerning the question of large-scale molecular segregation.

It has been previously reported that in Ag-free T–DC conjugates, several molecules (CD3, CD4 or CD8, PKCθ and talin) were accumulated at the T–DC interface, forming the only example of Ag-free IS formed by naive T cells 14. Note that CD8+ clones can also form antigen-independent ‘presynapses’ or ‘ring junctions’, in 15–30% of the contacts with lipid bilayers expressing a high ICAM-1 density (at least 4-times larger than that required for T cell adhesion) 18.

In Ag-free T–DC synapses, the distribution of CD3 and LFA-1 was essentially diffuse. Ag-dependent T–DC synapses were generally multifocal, CD3 and LFA-1 being recruited in the whole synaptic surface, with additional hot-spots or patches showing a partial overlap. It is extremely rare to find CD3 only in the periphery of the synapse, or to find LFA-1 only in its center.

The hypothesis that the multifocal structure could have been related to DC motility was rejected, since DC paralyzed with a combination of cytochalasin D and nocodazole formed synapses with a similarly rare occurrence of c-SMAC–p-SMAC-like structures. The same result was obtained by using DC lightly fixed with 1% paraformaldehyde (not shown). In EM of T–DC synapses, regions of tight membrane apposition are always multiple. The mean size of these foci is 3-times smaller than previously reported 13. The difference may either be due to the cellular difference (CD8+ versus CD4+ T cells) or more probably to the lower spatial resolution used in the former study, as suggested by the fact that the cumulative length of tight appositions was similar in the two studies. In more than 60 Ag-dependent T–DC synapses, we never observed a major central close apposition.

Another striking feature of the different synapses examined in the present study is the absence in EM images of p-SMAC-like structures, i.e. of well-defined peripheral regions with a cleft width of 40–45 nm. Such a structure is absent in T–B synapses, even those that show a central tight apposition of the c-SMAC type. Where then could LFA-1 molecules interact with ICAM-1? Given the size of these molecules, certainly not in the regions where membranes are >55 nm apart. The possibility remains that the LFA-1–ICAM-1 interaction also takes place where the cleft width is <40 nm. Recent crystallographic data showing the existence in the αvβ3 molecule of a hinge allowing this integrin to bend 17 and to be able to interact with ICAM-1 in a narrower synaptic cleft would be in favor of this possibility, if LFA-1 can fold in a similar way, even though, in this crystallographic study, the folded molecule was related to a low-affinity state of the integrin.

The analysis of LFA-1 immunogold labeling at the T–DC synapse allows us to draw the following conclusions. More than half of LFA-1 labeling was found in appositions <35 nm. However, LFA-1 was very rarely found in the tight appositions (<20 nm cleft). This absence could have different explanations. One is that LFA-1 was concentrated in tight appositions, but this clustering could not be seen, because the local concentration of Ab within narrow clefts was too low. Alternatively, LFA-1–ICAM-1 interaction may not take place in the narrow clefts but only at their border. As we occasionally observed LFA-1 accumulation at the rim of tight appositions, we favor this latter explanation, implying that each tight apposition would then constitute a micro-c-SMAC–p-SMAC of a few hundred nm (range 60–600 nm) in diameter.

Do CD3 or LFA-1 patches detected by IF correspond to tight appositions seen in EM? The correspondence between the two cannot be one-to-one. Indeed, if the number of molecules in a tight apposition is small, it will not appear as a fluorescent patch. This is certainly what happens in Ag-free synapses, in which numerous tight appositions are observed, but no fluorescent patch can be detected. In addition, most tight membrane appositions are too small to be resolved by fluorescence microscopy. Several receptor-enriched appositions close enough to each other will appear as a single patch in IF. This is probably why the number of discrete tight membrane appositions (a mean of 8 in one section can be extrapolated to several tens in the whole synapse) is much larger than the average number of fluorescent patches detected in en face views of the synapse (2–4).

A tentative model of the different steps in the formation of a T–DC synapse is illustrated in Fig. 9. As T cells are bristling with filopodia, the contact is likely to be initiated by one or several filopodia, before these structures resolve and the contact zone flattens. After 15 min of cell–cell interaction, approximately half of the synapse surface corresponds to foci of tight membrane appositions alternating with regions where the membranes are too far apart to allow the interaction of molecules like the TCR or even LFA-1. After 30 min of interaction, there is no major change in the synaptic structure, except that the amplitude of the membrane undulations is decreased and that the foci of tight apposition (still several tens per synapse) now cover two-thirds of the IS surface. Our data suggest that LFA-1 is probably located at the rim of these submicronic appositions. Thus, there is no large-scale segregation between CD3 and LFA-1 in these synapses and the SMAC associated with the TCR and with integrins should not be analyzed as two separate entities but as a unique signaling complex.

Figure 9.

Model of T–DC synapse formation. Hypothetical sequence of events taking into account the variety of images of conjugates observed and the evolution of different parameters measured in the presence or absence of Ag. Middle row: electron micrographs, a partial zoom of which is shown in the top row. Bottom: schematic view of the T–DC interactions.

The ability of naive CD8+ T cells to form IS presenting a c-SMAC–p-SMAC structure is still a matter of debate. Lymph node naive OT1 cells are able to form c-SMAC-containing IS 19, whereas splenic 2C T cells do not 20. In addition to the differences in the T cells used, APC were also different (M12 B lymphoma in one case, P815 tumor cells in the other). Here, we have shown that the notion of mature synapses must be better defined. The experimental and explicit definition of a mature synapse is a c-SMAC–p-SMAC organization of CD3, PKCθ and/or LFA-1. The hypothetical and implicit definition is a central apposition at 15 nm surrounded by a 40-nm-wide cleft. We have never observed a structure corresponding to both definitions. T–B synapses frequently show c-SMAC structures (small CD3 cluster in IF, tight central apposition in EM), sometimes show p-SMAC-like structures in IF (but sometimes due to LFA-1 accumulation on the B cell), but never p-SMAC-like structures in EM. T–DC synapses sometimes show c-SMAC-like structures in IF (central CD3 cluster in IF, in a minority of the synapses), but the structure revealed in EM is always multifocal.

Depending on the molecules available in various types of T cells and APC, autoassembly phenomena may lead to different molecular organizations. Some conditions may favor a concentric organization (e.g. CD3 focusing is only observed in the presence of very large Ag concentrations, LFA-1 rings formed in the absence of Ag require a high ICAM density), others do not. The ability of T cells to be fully activated after forming synapses that do not present a concentric structure (this study and 10, 20) shows that it is inappropriate to consider the concentric structure as a “mature synapse” and multifocal structures as immature.

Materials and methods

Antibodies and cell lines

The following monoclonal Ab were used: mouse anti-human-CD3 (UCHT1, ascites); hamster PE-, FITC- or biotin-conjugated anti-murine-CD3ϵ, 145-2C11; rat IgG2a anti-LFA-1; and anti-MAC1, anti-CD4, anti-CD8 (Pharmingen, San Diego, CA, USA) anti-CD45RB (Cedarlane, Hornby, Canada). Conjugated goat secondary F(ab′)2 included: rhodamine-Red-X™-conjugated anti-mouse- or anti-Rat-IgG (Jackson ImmunoResearch Laboratories, West Grove, PA, USA); PE-conjugated anti-mouse-IgG (Immunotech, Marseille, France); and streptavidin Alexa Fluor® 488 conjugate (Molecular Probes, Eugene, OR, USA). The T cell line Jurkat J77cl20 (J77) and the lymphoblastoid B cell line Raji were cultured in RPMI-1640 medium with 10% FCS.

Cell preparation

C57BL/6 splenic DC were purified as described previously 13 and P14 splenic T cells (which recognize a gp33 peptide from LCMV 21) were purified by magnetic depletion with a cocktail of appropriate rat anti-mouse-receptor Ab followed by incubation with Dynabeads (Dynal, Lake Success, NY, USA) coupled to anti-rat-IgG. P14 cells were used in all experiments, except in the experiment illustrated in Fig. 3, performed with CD4+ HA-specific T cells (a kind gift from Benedita Rocha, INSERM U345, Institut Necker, Paris). These cells were transfected with a ζ–GFP construct (a kind gift from Claire Hivroz, Institut Curie, France). For transfection, freshly prepared CD4+ HA-specific lymph node T cells were transfected with the Amaxa Nucleofector technology (Köln, Germany) according to the manufacturer's instructions, with 5 µg of CD3ζ–GFP construct, and used 18 h after transfection. For this set of experiments, APC were Balb/c DC loaded with 1 µg/ml HA overnight.

Cell stimulation

Splenic DC were incubated with 1 µM LCMV gp33 peptide (Neosystem, Strasbourg, France) overnight at 37°C, or left unpulsed. Ten minutes before the experiment, DC were plated on poly(L)lysine-coated slides at RT. T cells (5×105) were added to DC. IF staining was performed as described previously 14. The efficiency of Ag stimulation was checked in parallel experiments by monitoring Ca2+ responses by video-imaging immediately after adding fura-2-loaded T cells to a monolayer of DC. During the initial 15-min period, a Ca2+ response was observed in practically every T cell that interacted with a DC.

Image acquisition and analysis

IF and transmission-light images were acquired with a 60× objective on a Eclipse TE2000 microscope (Nikon, Badhoevedorp, The Netherlands) equipped with a cooled CCD camera (CoolSNAPHq, Photometrics, Huntington Beach, CA, USA) and a Z-axis stepper motor. Image capture, quantitative analysis and deconvolution were done with Metamorph (Universal Imaging Corporation, Downington, PA, USA). Deconvolution was based on experimental point-spread functions determined with stacks of images of submicronic beads for each fluorescence wavelength. The fraction of CD3 concentrated at the synapse (whether central or more diffuse) was estimated either in a single plane or in three dimensions as follows. For single equatorial XY planes, two polygonal regions were drawn, one around the synapse, the other one including the rest of the membrane. After appropriate background subtraction, the ratio between the integrated intensities in both regions was calculated. For three dimensional studies, measurements of synaptic and extrasynaptic amounts of CD3 were measured and summed in 12 planes encompassing most of the synapse.

To monitor the live dynamics of the T–DC synapse, we used HA-specific fura-2-loaded T cells expressing a GFP-coupled TCRζ chain, interacting with HA-loaded DC. A series of three images was acquired: (1) transmitted-light images every 20 s; (2) Ca images every 20 s; and (3) stacks of GFP images usually every 3 min.

Ca2+ measurements were performed as previously reported 14. The GFP stacks were first subjected to a true point-spread function-based deconvolution (as above), before being used to reconstruct en face views of the T–DC synapse.


Ag-unpulsed or -pulsed APC were prepared as previously described and T cells were similarly added to the APC monolayer and left for 15 or 30 min to interact with APC at 37°C. Cell preparation for EM (Philips, CM120; FEI Company, Eindhoven, The Netherlands) was performed as described previously 13. For immunolabeling, cells were fixed with 2% paraformaldehyde, incubated with PBS plus 50 mM glycine for 10 min, incubated for 1 h with primary Ab in PBS plus 5% FCS, and further incubated with secondary Ab for 40 min. Ab were visualized with Protein-A–gold conjugates (Department of Cell Biology, Utrecht University, Netherlands). After a final glutaraldehyde fixation, cells were dehydrated and embedded in epon. The background staining corresponding to a control Ab was <1 bead/image. For reconstruction of a wide membrane apposition, serial sections of epon-embedded T–DC conjugates were prepared and photographed sequentially. Digitized images were analyzed using IMOD software (University of Boulder, Colorado, USA). For pseudo-color distance representation of en face views, intermembrane spaces were measured as described below and results were put in a Microsoft®Excel surface chart.

Quantitative analysis of electron micrographs

Measurement of the synaptic cleft size was a multi-step process. First, the original synapse image was linearized by applying appropriate rotations to one or several sections of the image. Then, bars spanning the synaptic cleft (starting in the middle of the plasma membranes) were drawn approximately every 30 nm on the image. A quantified profile of the IS was then automatically generated with NIH Image J. To do so, images plus bars were first thresholded and binarized to keep only the bars, then the size of these bars was plotted as a function of their distance to the edge of the synapse.

The accuracy of such measurements depends on the fact that the section is exactly orthogonal to the membrane planes. If the section plane makes an angle α with this optimal plane, membranes appear enlarged and less sharp. For a typical ultrathin section (70 nm thick), the enlargement, in nm, is given by 70 × tg α. With α =5°, the membrane thickness is doubled and its sharpness greatly reduced. However, if a suboptimal angle of section decreases the quality of the images, it does not introduce a significant error in the distance between two membranes, measured between their respective centers. If d is the true distance between the membranes and d’ is the apparent distance, one has d′ = d / cosα. If α=5°, d′ = d * 1.004 , i.e. the error is 0.4%. This is an upper limit for the error made in tilted sections, since membranes in images with angles >∼5° would appear too blurred to be usable, and have not been included in our measurements.


Data are expressed as mean±SD, and the significance of differences between two series of results assessed using Student's t-test.


We thank G. Bismuth and J. Delon for critical reading of the manuscript and suggestions, Benedita Rocha for the gift of HA-specific TCR-transgenic mice, Claire Hivroz for the TCRζ–GFP construct, and Meriem Garfa for data analysis. This study has been part of the PhD thesis of C. B. This work was supported by INSERM, CNRS, Ligue Nationale contre le Cancer, the Ministère de l'Education Nationale et de la Recherche (C. B.) and Région Ile-de-France (Sésame).


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5


  6. 6


  7. 7


  8. 8


  9. 9