Increased Mobility of Major Histocompatibility Complex I-Peptide Complexes Decreases the Sensitivity of Antigen Recognition*

CD8+ cytotoxic T lymphocytes (CTL) can recognize and kill target cells expressing only a few cognate major histocompatibility complex (MHC) I-peptide complexes. This high sensitivity requires efficient scanning of a vast number of highly diverse MHC I-peptide complexes by the T cell receptor in the contact site of transient conjugates formed mainly by nonspecific interactions of ICAM-1 and LFA-1. Tracking of single H-2Kd molecules loaded with fluorescent peptides on target cells and nascent conjugates with CTL showed dynamic transitions between states of free diffusion and immobility. The immobilizations were explained by association of MHC I-peptide complexes with ICAM-1 and strongly increased their local concentration in cell adhesion sites and hence their scanning by T cell receptor. In nascent immunological synapses cognate complexes became immobile, whereas noncognate ones diffused out again. Interfering with this mobility modulation-based concentration and sorting of MHC I-peptide complexes strongly impaired the sensitivity of antigen recognition by CTL, demonstrating that it constitutes a new basic aspect of antigen presentation by MHC I molecules.

plexes (1,2). The initial encounter of CTL with target cells is a nonspecific adhesion, mainly based on intercellular interactions between LFA-1 (CD18,CD11a) and ICAM-1 (CD54) (3)(4)(5)(6). This nonspecific adhesion starts at an initial contact site, and after about 2.5 min LFA-1-ICAM-1 ring junctions are formed (4). In such antigen-nonspecific conjugates LFA-1 triggers reorganization of the actin cytoskeleton and phosphorylation of molecules involved in cell adhesion on CD8 ϩ T cells, which lowers the threshold of T cell activation (5). At this stage CTL scan target cells for expression of cognate MHC I-peptide complexes. Effective eradication of transformed or virus-infected cells depends heavily on the efficiency of this scanning; to recognize target cells expressing only a few cognate MHC I-peptide complexes, CTL need to be able to recognize these among high numbers (Ͼ10 6 ) of vastly diverse MHC I-peptide complexes during the brief nonspecific CTL target cell encounter. Interaction of cognate MHC I-peptide complexes with TCR induces elevation of intracellular calcium, diverse phosphorylation events, activation of integrins, rearrangement of the cytoskeleton, translocation of adhesion and auxiliary molecules and their ligands to the contact site, resulting in the formation of an immunological synapse (IS) (2)(3)(4)(5)(6). As the IS matures, antigen recognition molecules (e.g. TCR/CD3, MHC molecules, and coreceptor) localize in the central supramolecular activation complex (cSMAC) and adhesion molecules (e.g. LFA-1/ICAM-1) in peripheral SMAC (pSMAC) (6 -10). This molecular segregation occurs by differential actin cytoskeletonmediated translocations and allows avid polyvalent interactions of TCR and coreceptor with MHC-peptide complexes, respectively, and integrins with their ligands, as both sets of molecules are of different size (6,11).
The mobility of human and murine MHC-I molecules has been examined on different cell types using fluorescence recovery after photobleaching (FRAP), indicating that a fraction of MHC I molecules was immobile within the measured time scale of 50 ms to 30 s (12)(13)(14). Studies using single particle tracking (SPT) showed mobile MHC I molecules following anomalous diffusion (i.e. nonlinear time dependence) over time scales of 100 ms to 300 s or transient confinements within domains of various sizes (13)(14)(15)(16). The lateral diffusion of MHC-I molecules in plasma membranes was increased upon truncation of their cytoplasmic tail or replacement of their transmembrane and intracellular portions with a glycosylphosphatidylinositol (GPI) anchor (12,16). There is compelling evidence that the anomalous diffusion of MHC I molecules depends on the actin cytoskeleton. Actin plays a key role in membrane compartmentalization and determines the barriers of a random walk of MHC I molecules (12,14,17). An attractive possibility is that MHC I molecules transiently interact with the actin cytoskeleton via ICAM-1, which has been shown by coimmunoprecipitation and fluorescence resonance energy transfer experiments to associate with MHC I molecules (18 -20) and to interact with the actin cytoskeleton (21).
It is not clear what role, if any, the anomalous diffusion of MHC I-peptide complexes has for antigen recognition by CTL. Biological studies on GPI-linked MHC I molecules have reached vastly diverging conclusions. Although some studies showed that GPI-linked MHC I molecules exhibit defective thymic selection and alloantigen recognition by CD8 T cells (22)(23)(24), others reported that they are well recognized by antigen-specific CTL (25,26).
The objective of this study was to investigate the relation between the movements of normal and GPI-linked MHC I-peptide complexes on target cells and their recognition by CTL and to elucidate how cognate and noncognate MHC I-peptide complexes diffuse in nascent IS. The dynamics of MHC II-peptide complexes containing fluorescent peptides have been assessed by single molecule microscopy, a technique that provides unprecedented detailed and accurate information on molecular movements on living cells (11,(27)(28)(29)(30). For tracking experiments, we used fluorescent derivatives of the Plasmodium berghei circumsporozoite (PbCS) peptide 252-260 (SYIPSAEKI). This peptide, conjugated with photoreactive 4-azidobenzoic acid (ABA) on Lys-259 (PbCS(ABA)), is recognized by S14 and related CTL clones in the context of K d (31). These clones were derived from mice immunized with PbCS-(ABA) and hence the ABA group is an essential part of the epitope they recognize (31). Fluorescent ATTO 647 dye was introduced in PbCS(ABA) by replacing PbCS Ser-252 with N-␥ ATTO 647 diaminopropionic acid (Dap(ATTO 647 )-YIPSAEK-(ABA)I). Alternatively, ATTO 647 was introduced in PbCS-(252-260) on Lys-259, i.e. SYIPSAEK(ATTO 647 )I. For simplicity, the former peptide derivative henceforth is referred to as peptide P1 and the latter as peptide P2. This system has the significant advantage that although both modifications preserved the binding of the peptide to K d , peptide P1 but not peptide P2 was recognized by S14 CTL (31,32).
We report that K d -peptide complexes on fibroblast L-K d and RMA-S-K d thymoma cells exhibit dynamic transitions between states of free diffusion and immobility. The immobile periods greatly increased in cell adhesion sites, thus increasing the local concentration of complexes. In the nascent synapse with CTL, cognate but not noncognate K d -peptide complexes became immobile, resulting in their selective local trapping. The mobility of K d -peptide complexes was substantially increased upon disruption of the cytoskeleton or for GPI-linked complexes. In both cases the recognition by S14 CTL was greatly reduced, demonstrating for the first time that mobility modulationbased sorting and trapping of MHC I-peptide complexes in the immunological synapse increases the sensitivity of antigen recognition by CTL.

EXPERIMENTAL PROCEDURES
Cells and K d Transfectants-ICAM-1 and K d -transfected L cells (L-K d ) and S14 CTL were propagated as described (31,33). DNA of K d GPI was obtained by truncating C-terminal K d after Thr-280 and fusion with the GPI anchor of CD55 using the insertion sequence (SRPNKGSGTTSGTTRLLSGHTCFTLT-GLLGTLVTMGLLT; GPI anchoring is at the S) as described (34). Both K d and K d GPI DNA were inserted in the pRRLsin. PPT.hPGK lentiviral vector (35) and viruses produced by transfecting HEK293T cells with envelope (pMD2.VSVG), packaging (pCMV⌬R8.91), and lentiviral plasmids. RMA-S cells (10 5 /ml) in DMEM supplemented with Polybrene (4 g/ml) were cultured for 24 h at 37°C with viruses incubated for 2.5 h at 37°C with PbCS(ABA) peptide (10 Ϫ6 M), stained with anti-K d monoclonal antibody SF1-1.1.1-FITC, and sorted on an Aria sorter (BD Biosciences). After PbCS(ABA) peptide loading, the K d surface expression, as assessed upon staining with SF1-1.1.1-FITC, was 81 (MFI) for RMA-S-K d , 171 for RMA-S-K d GPI , and 279 for L-K d cells. The ICAM-1 expression, as determined by staining with monoclonal antibody YN1/ 1.7.4-PE was 58 (MFI) for RMA-S-K d , 54 for RMA-S-K d GPI , and 82 for L-K d cells.
Single Molecule Microscopy-L-K d cell monolayers on glass coverslips (Assistant, Sondheim, Germany) were incubated in measurement buffer (Hanks' buffered salt solution supplemented with HEPES (10 mM) and ␤ 2 -microglobulin (2 g/ml)) with 1.5 nM peptide P1 for 1 h at 32°C. Alternatively, L-K d cells were pulsed likewise with 1.5 nM peptide P2 followed by addition of 1 M PbCS(ABA) peptide after 30 min. RMA-S-K d cells were incubated with 10 nM of peptide P1 in DMEM supplemented with 2.5% dialyzed fetal calf serum and ␤ 2 -microglobulin (2 g/ml) for 90 min at 32°C. RMA-S-K d GPI cells were incubated with 1 nM of the ATTO peptides at 37°C for 5 min. Single molecule images were recorded at 37°C on a modified wide field microscope (Axiovert 200 M, Zeiss, Feldbach, Switzerland) as described (29). Samples were illuminated with 632.8 nm light of a He-Ne laser for 10, 20, or 50 ms per image using average illumination intensities of around 1.2, 0.5, and 0.3 kilowatt/cm 2 , respectively, and imaged with an intensified CCD camera (I-Pentamax 512 EFT, Roper scientific, Trenton, NJ) or with a back-illuminated electron multiplying CCD camera (Ixon DV 887, Andor Technology, Belfast, UK). Using white light illumination and the settings for single molecule microscopy, we focused on the contact site of newly formed conjugates, just exhibiting high and stable calcium flux and recorded series of single molecule images.
Data Analysis of Single Molecule Trajectories-Fluorescence spots of single K d -peptide complexes were discriminated from cellular fluorescence background by their near-diffraction-limited sizes and intensities. Analysis of single molecule images was performed with a program written in IGOR PRO (30). Trajectories were evaluated from complexes displaying more than six steps and single-step photobleaching. The diffusion coefficients (D) were determined from a linear fit of the first four data points of the plot of the mean square displacement (MSD) versus the time lag, t lag , using MSD ϭ 4 D t lag . The confinement length (L) was evaluated from a fit of the plot of MSD versus t lag using a confined diffusion model MSD ϭ noise ϩ A 2 (1 Ϫ exp(Ϫ 4 Dt lag /A)), where noise is the measurement noise and A the confinement size. To discriminate between immobile K d -peptide complexes and those diffusing in confined domains, the distribution of L was compared with those of ATTO647 molecules immobilized on a glass surface (30). Average domain sizes ͗S͘ were evaluated from ͗S͘ ϭ ͗2A͘ ϭ ((͗L͘/2) 2 Ϫ ͗noise͘) 1/2 using (͗L͘/2) 2 measured on glass-immobile ATTO molecules to assess ͗noise͘. Mobility transitions within trajectories were detected by screening for immobile/confined periods (38). The probabilities of transitions from mobile to immobile, P mob-Ͼimmob , and from immobile to mobile, P immob-Ͼmob , were calculated by dividing the number of detected transitions by the total number of steps summed over all mobile or immobile molecules, respectively. The lifetimes of the mobile, mob , and immobile periods, immo , were calculated using mob ϭ 1/(⅐P mob-Ͼimmob ) and immo ϭ 1/(⅐P immob-Ͼmob ), where is the recording frequency of 10 Hz. Values are given as (mean Ϯ 95% confidence intervals). Error bars in graphs and histograms represent standard deviations and standard errors, respectively.

Single Molecule Microscopy of K d -Peptide P1 Complexes on
Living L-K d Cells-Single molecule images of individual K dpeptide complexes on K d -transfected fibroblasts (L-K d ) were acquired using the fluorescent peptide P1 (ATTO 647 -PbCS-(ABA)), which was recognized by cloned S14 CTL nearly as efficiently as the parental PbCS(ABA) peptide (Fig. 1A). For tracking experiments L-K d cells were loaded for 1 h at 32°C with 1.5 nM peptide P1, corresponding to about maximal lysis by S14 CTL. As determined from images, such L-K d cells displayed 65 Ϯ 35 (n ϭ 8) K d -peptide P1 complexes per cell. Individual complexes were discernible as discrete, diffraction limited spots, distinct from diffuse cellular autofluorescence (Fig.  1B). As assessed by confocal microscopy, peptide internalization was barely detectable under these conditions, and unspecific binding on untransfected L cells was negligibly small (data not shown). These spots exhibited single step photo-bleaching, proving that they were single K d -peptide complexes (Fig. 1C). The diffusion of the complexes was assessed by recording sequences of images at the rate of 10 Hz (Video 1). From the trajectories three types of complexes were discernible as follows: (i) immobile (Fig. 1D); (ii) mobile (Fig. 1E); and (iii) those undergoing transitions between i and ii (Fig. 1F).
Diffusion of K d -Peptide P1 Complexes on the Lower Plasma Membrane of L-K d Cells-The diffusion coefficients (D) of K dpeptide complexes on the lower, support-adhered plasma membrane of L-K d cells were calculated from the short time regime of the trajectories of individual complexes (supplemental Fig. S1). Populations of mobile and immobile/confined complexes were clearly distinguishable in the histogram of the diffusion coefficients D and distinct from those of ATTO molecules immobilized on a glass slide ( Fig. 2A). Because of insufficient time resolution, D poorly discriminated between complexes that were immobile or confined in small nanometersized domains. We therefore calculated confinement lengths from the long time regime of the trajectories, which scale with sizes of domains explored by single complexes. In the histogram of the confinement lengths L the calculated confinements within regions of varying lengths resulted either from real or apparent confinements, caused by the stochastic nature of the diffusion of mobile complexes (Fig. 2B). Confinement lengths of Ͻ360 nm resulted from real confinements, as they did not occur in simulated trajectories of complexes following Brownian diffusion. Using 360 nm as threshold showed that 42 Ϯ 10% of complexes were immobile or confined in small domains (Fig.  2C). The distribution of L was shifted compared with immobilized ATTO molecules (Student's t test, p Ͻ 0.05), indicating that these complexes were confined in very small domains. By subtracting the noise from the average confinement length, an average domain size of 90 Ϯ 40 nm was obtained (supplemental Fig. S1).
A hallmark of the diffusion of K d -peptide complexes was the mobility transitions between mobile and immobile/confined periods. To quantify this, we used Monte Carlo simulations to calculate the number of stochastic mobility transitions occurring during Brownian diffusion. The simulated trajectories were screened for time periods for which the square displacements remained below 0.016 m 2 , indicating that none exceeded six frames. Screening of the trajectories of K d -peptide complexes for immobile periods exceeding seven frames showed real mobility transition for 4 molecules of the 66 analyzed; in three cases, mobile complexes became immobile, and in one case an immobile one became mobile (supplemental Fig. S2A). The probability of occurrence of immobile periods, P immob , as calculated by dividing the number of immobile periods (4) by the total number of steps (1174), was 0.34 Ϯ 0.33%, which was significantly higher than predicted by Brownian diffusion ( 2 test, p Ͻ 0.05). There was a probability of P mob-Ͼimmob ϭ 0.26 Ϯ 0.29% for a mobile molecule to immobilize and of P immob-Ͼmob ϭ 0.19 Ϯ 0.37% for an immobile molecule to become mobile (Fig. 2G). From these values estimates for the lifetime of the mobile periods, mob , of 38 s and of the immobile periods, immob , of 53 s were calculated. Eight immobile periods were detected among the trajectories of 64 mobile molecules (2085 steps), corresponding to a probability of entering immobilization periods of P immob periods ϭ 0.38 Ϯ 0.26% (supplemental Fig. S2C). Although the probabilities of entering immobilization periods were statistically similar ( 2 test, p Ͼ 0.8) in both membranes, the trajectories on the upper surface showed shorter immobilization periods than those on the lower (supplemental Fig. S2C). This allowed some K d -peptide complexes to undergo two mobility transitions in the same trajectory (supplemental Fig.  S2B). In total six transitions from mobile to immobile and four from immobile to mobile were observed, corresponding to transition probabilities of P mob-Ͼimmob ϭ 0.29 Ϯ 0.23% ( mob Ϸ 34 s) and P immob-Ͼmob ϭ 1.4 Ϯ 1.3% ( immo Ϸ 7.1 s), respectively (Fig. 2G). P mob-Ͼimmob values were not statistically different ( 2 test, p Ͼ 0.8) on the lower and upper surfaces, indicating that K d -peptide complexes became immobile with the same frequency. By contrast, P immob-Ͼmob was statistically higher on the upper surface ( 2 test, p Ͻ 0.05), resulting in shorter periods of immobilization and hence in a larger average fraction of mobile molecules (Fig. 2G).
The differences in the diffusion of K d -peptide complexes between the upper and the lower surfaces correlated with differences in the density of the actin cytoskeleton; the lower substrate-adhered membrane displayed a considerably higher density of actin as compared with a median section (Fig. 2, H and I).
The finding that K d coimmunoprecipitated with ICAM-1 and vice versa (supplemental Fig. S3, lanes 1 and 4) argues that the transient immobilizations of K d -peptide complexes are explained, at least in part, by their interaction with actin cytoskeleton-associated ICAM-1 (21).

Disruption of the Actin Cytoskeleton Increases the Mobility of K d -Peptide P1 Complexes and Decreases T Cell Recognition-To
find out whether the confinements of the K d -peptide complexes were related to the cytoskeleton, their diffusion was assessed on L-K d cells pretreated with latrunculin, to block actin polymerization. Because latrunculin-treated L-K d cells detached from the substrate, tracking experiments were performed only on the upper cell surface. Remarkably, virtually all complexes were mobile with diffusion coefficients roughly 10 times larger than on untreated cells (Fig. 3, A-C). Also the proportion of molecules exhibiting confinements of less than 1 m clearly decreased upon latrunculin treatment. These results argue that the transient confinements of K d -peptide complexes were because of interaction with the cytoskeleton or cytoskeleton-associated entities.
The proportion of conjugate formation was considerably decreased when target cells were pretreated with latrunculin, especially at low peptide concentrations (Fig. 3D), indicating that the increased mobility of K d -peptide complexes considerably inhibited the conjugate formation of S14 CTL with L-K d cells.

GPI-anchored K d -Peptide P1 Complexes Display Increased
Mobility-To generalize these findings we likewise examined K dpeptide complexes on K d -transfected RMA-S cells (RMA-S-K d ). Because RMA-S cells are weakly adherent, the diffusion of K d -peptide P1 complexes was only assessed on the upper cell surface. As on L-K d cells transient confinements were observed, but the confinement lengths were considerably larger with an average domain size of 240 Ϯ 30 nm (Fig. 4, A and B). 16 Ϯ 8% of the complexes were confined, which is similar as on L-K d cells ( 2 test, p Ͼ 0.3) (Fig. 4C).
To find out to what extent the confinements of K d -peptide P1 complexes were determined by their transmembrane and/or cytoplasmic portion(s), we replaced these by a GPI anchor. K d GPI -peptide complexes displayed diffusion coefficients in average 3-fold larger than K d -peptide complexes (Fig. 4,  A and D). Nearly all complexes (99 Ϯ 2%) were mobile (Fig. 4, E and  F). Because GPI-anchored molecules partition into lipid rafts (34), the transient confinements observed for wild type K d -peptide complexes were not accounted for by partitioning in rafts. Similar findings were obtained for K d -peptide P1 complexes that were incorporated into the membrane of L cells (H-2 k ) via the lipid anchor di-palmitoylphosphatidylethanolamine (39). On the lower membrane only one immobile period was detected in the trajectories of 50 complexes, demonstrating that the diffusion of full-length K d -peptide P1 complexes on the lower membrane was not influenced by interactions of their extracellular portion with the substrate (data not shown). Moreover, K d, , but considerably less K d GPI , coimmunoprecipitated with ICAM-1 and vice versa (supplemental Fig. S3, lanes 2, 3, 5, and 6), which correlated with the different motilities of K d and K d GP -peptide complexes (Fig. 4).

GPI-anchored K d -Peptide P1 Complexes
Are Inefficiently Recognized by S14 CTL-RMA-S cells transfected with GPI-linked K d (RMA-S-K d GPI ) expressed considerably higher levels of K d and bound more peptide than RMA-K d cells, when tested under identical conditions (Fig. 5A). To conclusively compare the CTL recognition of wild type and GPI-linked K d -peptide complexes, we determined loading conditions that provided equal peptide binding. To this end we used biotinylated PbCS peptide (PbCS(biotin)), which binds to K d as well as PbCS-(ABA) (31) and flow cytometric detection of cell-associated PbCS(biotin) via phycoerythrin-labeled streptavidin (36). By varying the incubation time and temperature, equal peptide binding was observed when RMA-S-K d GPI cells were incubated at 37°C for 5 min and RMA-S-K d cells at 32°C for 90 min in medium containing dialyzed serum (Fig. 5B). As assessed in a 51 Cr release assay, RMA-S-K d cells were recognized by S14 CTL several hundredfold more efficiently than RMA-S-K d GPI cells loaded with equal amounts of PbCS(ABA) peptide (Fig. 5C).
Cognate K d -Peptide P1 but Not Noncognate K d -Peptide P2 Complexes Are Immobilized in the Immunological Synapse-For tracking of K d -peptide complexes in the immunological synapse formed between peptide P1-pulsed L-K d cell and S14 CTL, newly formed conjugates were selected that exhibited strong and stable calcium flux in the CTL (Fig. 6A and  data not shown). K d -peptide P1 complexes were exclusively found in the contact site and colocalized with high density of actin (Fig. 6, B and C) and were distributed over most of the extended contact site, typically forming bright clusters (supplemental Fig. S4 and Videos 2 and 3). All complexes were immobile/confined (100 Ϯ 6%) (Fig.  6F). The average confinement length was slightly larger as compared with ATTO dye immobilized on a coverslip, which was explained, at least in part, by the reduced signal-to-noise ratio in the synapse (Fig. 6E). We cannot, however, rule out that some of the K d -peptide P1 complexes were confined within very small domains.
A strikingly different situation was observed for noncognate K dpeptide P2 complexes. Because peptide P2 was not recognized by S14 CTL (Fig. 1A), L-K d cells were pulsed first with a low concentration of peptide P2, followed by a high concentration of unlabeled PbCS(ABA) peptide. As in the previous experiment S14 CTL were selected that just formed stable conjugates exhibiting high calcium flux (Fig. 7A). Importantly, although cognate K d -peptide P1 complexes exclusively accumulated in the contact site, noncognate K d -peptide P2 complexes were enriched there but were also found outside (Fig. 7B, supplemental Fig. S5, and Video 4). Whereas in the synapse cognate complexes were immobile (or confined in very small domains), noncognate ones retained mobility (Fig. 7F, and Video 4). As shown in Fig. 7C, about half of the K d -peptide P2 complexes in the synapse were confined in domains of various sizes (trajectories 4 and 5), others were mobile (trajectory 3), and some exhibited reversible confinements, hopping from one domain to another (trajectories 1 and 2). The diffusion coefficients (D) of K d -peptide P2 complexes were larger than those of K d -peptide P1 complexes (Figs. 6D and 7D). Using the confinement length (L) of 360 nm as threshold showed that 56 Ϯ 15% of K d -peptide P2 complexes were immobile/confined and 44 Ϯ 15% mobile (Fig. 7, E and F). The average size of the domains for K d -peptide P2 complexes was 130 Ϯ 60 nm.
Essentially the same findings were obtained on conjugates between RMA-S-K d cells and S14 CTL. In the synapse formed between RMA-S-K d cells and S14 CTL cognate K d -peptide P1 complexes were all localized in the synapse and immobile or confined in small domains (supplemental Fig. S6 and Video 5). By contrast, noncognate K d -peptide P2 complexes were only enriched in the synapse and rapidly diffused in and out of the synapse (supplemental Fig. S7 and Video 6). In the synapse formed between peptide P2-pulsed RMA-S-K d GPI cells and S14  AUGUST 29, 2008 • VOLUME 283 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 24259
CTL essentially all complexes were mobile and recruitment into the synapse was scant (supplemental Fig. S8 and data not shown).

DISCUSSION
A key finding of the present study is that MHC I-peptide complexes reversibly immobilize in cell plasma membranes and that the duration of these confinements increased considerably in membranes engaged in adhesion (Figs. 2 and 7 and Videos 4 and 6). Substantially more K d -peptide P1 complexes were confined in small domains on the lower membrane of L-K d cells than on the upper membrane ( Fig. 2 and Video 1). Moreover, the diffusion of noncognate K d -peptide P2 complexes in the synapse between L-K d cells and S14 CTL resembled the one in the lower membrane of L-K d cells, except that the confinement lengths were smaller (Figs. 2 and 7 and Videos 1 and 4). Transitions between different diffusion modes within the trajectory of single MHC I molecules have been observed previously in single particle tracking experiments (16). However, this study lacks information on the size and frequency of the confinements, making a comparison with our results difficult. The dynamic and reversible confinements of K d -peptide complexes observed in our study explain why MHC I molecules were found to follow anomalous diffusion in the time scale of minutes (Figs. 2 and 4 and Videos 1-3 and 5) (15). Reversible confinements can give rise to anomalous diffusion by restrict-ing the apparent diffusion of initially mobile molecules over long distances as compared with free Brownian diffusion (40).
Our results on the diffusion of K d -peptide P1 complexes are consistent with previous studies using FRAP. The diffusion coefficient of K d -peptide P2 complexes on the upper cell membrane of 0.1 m 2 /s (Figs. 2 and 4) lies within the range of 0.06 -0.3 m 2 /s reported for MHC-I molecules (12)(13)(14). Smaller diffusion coefficients (0.004 -0.0067 m 2 /s) were reported in studies using SPT (14 -16), a finding often observed when comparing SPT and FRAP data (13,41). With a diffusion coefficient of 0.1 m 2 /s, it takes K d -peptide P2 complexes 3.27 min to explore the 78.5-m 2 area of plasma membrane of a cell with a diameter of 5 m. This is at variance with the observations that nonspecific CTL target cell conjugates are short lived and that cognate MHC I-peptide complexes are rapidly concentrated in nascent CTL target cell contact sites (Fig. 6, supplemental Fig. S4 and Videos 2 and 3) (2). Thus, mechanisms other than purely diffusion controlled transport of MHC I-peptide must exist to make possible efficient TCR scanning.
The observation that MHC I molecules dynamically associate with ICAM-1 explains how they can become transiently linked to the actin cytoskeleton (supplemental Fig. S3) (18 -20, 42). Because nonspecific conjugates of CTL and target cells are primarily mediated by LFA-1 and ICAM-1, these molecules are highly enriched in pre-synaptic contact sites (3-5, 18, 42). Dynamic mobility modulation of MHC I-peptide complexes via association with ICAM-1 therefore well explains their efficient diffusional trapping in nascent contact sites (Figs. 2 and 7 and supplemental Fig. S3). The thus greatly increased density of MHC I-peptide complexes in pre-synaptic contact sites in turn permits efficient TCR scanning. This view is consistent with the finding that K d GPIpeptide P1 complexes, which exhibit increased mobility and poorly associated with ICAM-1, were recognized by S14 CTL far less efficiently than wild type complexes (Figs. 4 and 5 and supplemental Fig. S3). It is also consistent with the finding that inhibition of the actin cytoskeleton of L-K d cells greatly increased the mobility of K d -peptide P1 complexes and inhibited conjugate formation with S14 CTL (Fig. 3). Stable conjugate and IS formation heavily depend on MHC I-peptide triggered TCR signaling, which activates LFA-1 for high avidity binding to ICAM-1 on target cells (3)(4)(5)42). Moreover, target cell recognition by CTL is independent of costimulation by CD28, but it does rely on costimulation by ICAM-1/LFA-1 (43). ICAM-1 (and possibly other ICAMs) therefore are uniquely suited for the here described mobility modulation-based concentration of MHC I-peptide complexes in newly formed contact sites.
Our assessment of the tempo-spatial dynamics of individual K d -peptide complexes revealed for the first time that cognate complexes were immobilized in small clusters scattered over most of the extended, nascent synapse with CTL ( Fig. 6 and supplemental Figs. S4 and S6 and Videos 2, 3, and 5). Similar microclusters have been observed on CD4 ϩ T cells rapidly after their interaction with APC or ICAM-1 and MHC II-peptide reconstituted planar membranes; they were shown to contain TCR and diverse signaling molecules and to be sites of intense signaling (5,(7)(8)(9). Initially these microclusters contain ICAM-1/LFA-1, but as they migrate toward the center to form the cSMAC, adhesion molecules become segregated from antigen recognition molecules (4,6,9,11).
In the nascent synapse between S14 CTL and L-K d cells, cognate K d -peptide P1 complexes were immobilized, whereas noncognate K d -peptide P2 complexes retained mobility and eventually diffused out again (Figs. 6 and 7 and supplemental Figs. S5-S8 and Videos 2-6). The difference between the two complexes is that one strongly and the other barely detectably interacts with S14 TCR, 5 arguing that the retention of MHC I-peptide complexes in the synapse depends on the strength of their interaction with TCR, which upon triggering are cytoskeleton linked (5,7,8,11). Noncognate K d -peptide P2 complexes were not recognized by S14 CTL (Fig. 1A), yet such ligands can participate in T cell recognition and increase its sensitivity (32,44). Considering the dimeric recognition modes proposed by these studies and the observation that MHC I-peptide complexes can form compact aggregates, it is conceivable that noncognate complexes partially become immobilized by association with cognate ones (45,46). This is consistent with the finding that the diffusion coefficients (D) and confinement lengths (L) for cognate and noncognate complexes partially overlap, and both exhibit the same cluster formation (Figs. 6 and 7 and supplemental Figs. S4 -S7 and Videos 2-6).
Other factors accounting for the remarkably slow diffusion of noncognate complexes in nascent synapses include the following. 1) the diffusion of noncognate complexes was considerably slower in the synapses of S14 CTL with L-K d cells than in the synapses with RMA-S-K d cells and especially with RMA-S-K d GPI cells (supplemental Videos 4 and 6). 5 Because L-K d cells express more ICAM-1 than RMA-S cells (see "Experimental Procedures") and K d GPI associates poorly with ICAM-1 (supplemental Fig. S3), this argues that the density of cytoskeletonassociated ICAM-1 modulates the mobility of noncognate complexes in the synapse. 2) CD8 in the synapse can bind to and retain noncognate MHC I-peptide complexes (47). These authors proposed that CD8 increases the local density of MHC I-peptide complexes in the synapse and vice versa and that this enhances the sensitivity of antigen recognition. Costimulation by irrelevant peptides was most pronounced on thymocytes, moderate on naive T cells, and mild on effector T cells (48).
The biological significance of dynamic mobility modulation of MHC I-peptide complexes was evidenced by the adverse effects on antigen recognition by measures that interfere with it, such as disruption of the cytoskeleton or replacement of the  For RMA-S-K d GPI cells (closed symbols), the incubations were performed at 37°C for 1 min (filled triangles), 5 min (filled diamonds), and 10 min (filled squares). One out of four experiments is shown. B, RMA-S-K d cells (squares) were incubated at 32°C for 90 min and RMA-S-K d GPI cells (triangles) at 37°C for 5 min with graded concentrations of PbCS(ABA) peptide and their recognition by cloned S14 CTL assessed in a 51 Cr release assay. One out of three experiments is shown. tact sites formed by CTL and target cells, this results in local diffusional trapping and sorting of complexes. This in turn increases the efficiency of TCR scanning and TCR triggering and hence the sensitivity of antigen recognition.