seminars in IMMUNOLOGY, Vol 12, 2000: pp. 5]21doi: 10.1006rsmim.1999.0203, available onlineat http:rrwww.idealibrary.com on

Cytoskeletal polarization and redistribution of cell-surface molecules during T cell antigen recognition

,

P. Anton van der Merwe1U, Simon J. Davis2

Michael L. Dustin3

T cell antigen recognition is accompanied by cytoskeletalpolarization towards the APC and large-scale redistributionof cell surface molecules into ‘supramolecular activation

( )clusters’ SMACs , forming an organized contact interface( )termed the ‘immunological synapse’ IS . Molecules are

arranged in the IS in a micrometer scale bull’s eye pattern(with a central accumulation of TCRrpeptide]MHC the

)cSMAC surrounded by a peripheral ring of adhesion( )molecules the pSMAC . We propose that segregation of cell

surface molecules on a much smaller scale initiates TCRtriggering, which drives the formation of the IS by activetransport processes. IS formation may function as a check-point for full T cell activation, integrating information onthe presence and quality of TCR ligands and the natureand activation state of the APC.

Key words: T cell receptor r T cell antigen recognition rSMAC r immunological synapse r segregation

Q2000 Academic Press

Introduction

LYMPHOCYTES ARE PROPELLED through the blood asrelatively undifferentiated spheres uniformly coveredwith short microvilli, but without obvious asymmetryor polarity. These cells also relatively non-adhesive,partly because integrins such as LFA-1 are im-mobilized by attachment to the actin cytoskeletonand maintained in a low avidity state.1,2 Activation of

From the 1Sir William Dunn School of Pathology, University ofOxford, Oxford OX1 3RE, UK, 2Nuffield Department of ClinicalMedicine, John Radcliffe Hospital, University of Oxford, Heading-

3

ton OX3 9DU, UK and Centre for Immunology and the Depart -ment of Pathology, Washington University School of Medicine,660 South Euclid Avenue, St. Louis, MO 63110, USA. UCorre-sponding author.

Q2000 Academic Press1044-5323r00r010005q17 $35.00r0

5

Andrey S. Shaw 3 and

lymphocytes by, for example, chemotactic factors orligation of antigen receptors results in increased ad-hesiveness,3 cytoskeletal polarization and the devel-opment of morphological asymmetry.4 It has longbeen recognized that cytoskeletal changes contributeto lymphocyte effector functions such as polarizedcytokine secretion and target cell killing.5,6 This re-view focuses on the role of cytoskeletally-mediatedprocesses in T cell antigen recognition.

Asymmetry of activated lymphocytes also extends tothe distribution of molecules on their surface. Inmigrating lymphocytes, chemokine receptors areconcentrated at the leading edge, whereas moleculessuch as CD43, CD44 and CD50 are found at thetrailing edge.4 Dramatic changes in cell-surfacemolecule distribution are also observed when the Tlymphocytes recognise antigens on the antigen pre-

Ž . 5,7 ] 10senting cells APCs . In this review we first de-scribe the redistribution that occurs during T cellantigen recognition, placing it in the context of othercellular events such as changes in cell motility andmorphology, increases in intracellular calcium, andpolarization of the cytoskeleton. We then discuss indetail the molecular mechanisms underlying redis-tribution. Finally, we speculate on the functional sig-nificance of redistribution, and attempt to integratethe experimental data into a general model of T cellantigen recognition.

Observations

Snapshots of cytoskeletal polarization and cell-surfacemolecule redistribution

Studies performed over a decade ago showed that,following recognition of the antigen on target cells orAPCs, T cells become polarized with respect to the

Ž .target cell or APC reviewed in ref 5 . The micro-Ž .tubule organising centre MTOC , which is usually

P. Anton van der Merwe et al

6

found in the trailing edge of a migrating T cell,repositions itself between the T cell nucleus and thetarget cell.11 This is accompanied by, and is probablyresponsible for, the movement of the Golgi apparatusto the same side of the cell.11 There is also anincrease in actin polymerization adjacent to the con-tact interface, and the actin-binding protein talinaccumulates in this area.12 The T cell receptor and

Ž .its CD3 signalling module TCRrCD3 , the TCR co-receptor CD4, and the adhesion receptorrligand pairLFA-1rICAM-1 all accumulate at the contact inter-face between CD4 T cells and APCs presenting speci-fic antigens.5 Much more recently, Wulfing andDavis13,14 have reported evidence for a more general-ized, cytoskeletally-driven movement of cell surfacemolecules towards the T cellrAPC contact interface.

Recently the redistribution of molecules within thecontact interface has been visualized both in fixed TcellrAPC conjugates7,8 as well as in live T cells inter-acting with planar, glass-supported lipid bilayers intowhich fluorescently-labelled, lipid-anchored mole-cules have been incorporated.10,15 These bilayers arecapable of stimulating T cell proliferation, suggestingthat they can function as surrogate APCs. While lessphysiological than APCs, the planar bilayer systemhas provided higher resolution images and the firstquantitative and dynamic data on this redistribution.Both approaches reveal that cell-surface moleculessegregate within T cellrAPC contacts into distinctregions or clusters, referred to by Monks et al8 as

Ž .supramolecular activation clusters SMACs , therebyforming an organized interface which we refer to as

10 Ž .an immunological synapse Figure 1 . The phrase

Ž .immunological synapse IS was coined several years

ago6 to draw attention to some of the similaritiesbetween these contact areas and neuronal synapses.16

We have extended the term to describe the charac-

Ž .Figure 1. Structure of the mature immunological synapse. ashown schematically above. The structure of the IS is shownthat some of the molecules are expressed on T cells and otheFluorescent images revealing the position of the CD28rCD80

Ž . Žimage shows the positions of peptide]MHC green , CD80 rand planar bilayers. The right image shows the positions ofinterface between 2B4 T cells and planar bilayers. The CD2whereas the CD2rCD48 complex segregates to an inner ring bGrakoui et al.10

Ž .Figure 2. Steps in the formation of the IS. A Fluorescent ibrought into contact with planar lipid bilayers at ts0. The

Ž .ICAM-1 red at the indicated times after initial contact are sh10 Ž .Reproduced with permission from Grakoui et al. B The

Reproduced with permission from Grakoui et al.10

7

Redistribution of T cell surface molecules

teristic molecular patterns formed in the process ofphysiological T cell activation.10

ŽIS formation passes through several stages Figure. Ž .2 , culminating in the ‘mature’ IS Figure 1 . In the

mature IS the TCRrCD3 and several molecules in-volved in TCR triggering congregate at the centre ofthe interface.7,8,10 These include peptide]MHC, theaccessory molecule CD28 and its ligand CD80, and

Ž .cytoplasmic signalling molecules e.g. lck, fyn, PKC-u .This central area has been termed the central clusteror cSMAC.8 A second group of molecules, includingLFA-1, ICAM-1 and talin, form a ring surroundingthe cSMAC, which has been termed the ‘outer adhe-

Ž . 7,8,10sion ring’ or peripheral SMAC pSMAC . In theplanar bilayer system10 the accessory molecule CD2and its ligand CD48 appear to be confined to an‘inner adhesion ring’. This surrounds the centralzone but lies within the outer LFA-1 adhesion ring.Finally, CD43, a large, highly-abundant mucin-likemolecule, is mostly excluded from the contact inter-face.9

Dynamics of polarization and redistribution

Studies in live cells have provided insights into thetiming and the sequence of events described above. Tcell clones and hybridomas are highly mobile in cellculture, migrating at speeds of 4]10 mmrmin.4,17

ŽWhen they encounter surrogate APCs usually cell.lines transfected with suitable MHC molecules that

are not presenting specific antigens, they show nochanges in motility or morphology.17,18 In contrast,when they encounter APCs presenting specific anti-

gens they arrest, undergo shape changes, and expandtheir contact with the APC.17,18 T cells also migraterapidly on artificial substrates coated with ICAM-1,but stop migrating when they encounter regions

Schematic view of the mature IS. The position of the IS isbelow as viewed perpendicular to the contact interface. Not

Ž .rs on APCs. CD43 is excluded from the entire interface. BŽ . Ž .left and CD2rCD48 right complexes in the IS. The left. Ž .ed and ICAM-1 blue at the interface between 3A9 T cells

Ž . Ž . Ž .peptide]MHC green , CD48 red and ICAM-1 blue at8rCD80 complex colocalizes with the TCRrpeptide]MHCetween TCRrpeptide]MHC and LFA-1rICAM-1. Data from

mages illustrating the dynamics of IS formation. T cells areŽ .positions of engaged peptide]MHC green and engaged

Ž .own superimposed on an image of the contact area white .three stages of IS formation in planar lipid bilayer system.

P. Anton van der Merwe et al

coated with ICAM-1 and specific peptide]MHC com-plex.19 Together these data suggest that an en-counter with specific antigens delivers a stop signal tomigrating T cells, which serves to keep them associ-ated with the APC that is presenting specific antigen.19

Naive T cells are less mobile but also show rapidchanges in morphology and motility upon contactwith B cells presenting specific antigens.20,21 Thesechanges in motility and morphology are followed in a

Ž .variable time seconds to minutes later by increases,sometimes oscillating, in intracellular calcium con-centration.17,18,22

A consistent feature in all the above studies is thatchanges in T cell motility, morphology and intracel-lular calcium concentration are only seen if the APCor planar bilayers present specific antigens, indicat-ing that they require TCR triggering. In contrast,when dendritic cells are used as APCs, morphologicalchanges, conjugate formation, and increases in cal-cium are seen in the absence of antigens,21 even inthe absence of MHC class II.23 This suggests that,unlike surrogate APCs or B cells, dendritic cells havethe ability to engage and activate naıve T cells prior¨to any TCR engagement, perhaps through the ex-pression of high levels of the CD28 ligands CD80 andCD86 or chemokines.24,25

Technical innovations such as the use of greenfluorescent protein chimeras13 and the planar bilayersystem15 have made it possible to visualise the redis-tribution of cell surface molecules in live T cells. Arecent study10 using primary T cells in the planarlipid bilayer system has revealed that the formation ofthe IS takes several minutes following initial contactand TCR triggering, and can be divided into at least

Ž .three stages Figure 2 .Following contact with planar bilayers presenting

specific antigens, the T cells arrest and within 30 s acentral zone containing the LFA-1 ligand ICAM-1 isformed. Analysis by interference reflectance micros-copy indicated that an area of close contact sur-rounds the central zone. An unexpected finding wasthat clusters of engaged peptide]MHC first appearedin this outer close contact ring. Since no other acces-

Žsory molecules were present in this system we A.S..and M.L.D. proposed that the cells were using the

LFA-1rICAM-1 interaction as an ‘anchor’ to force

the membranes into close proximity. This raises theinteresting possibility that active processes contributeto the membrane approximation that is necessary for

Žthe TCR to engage peptide]MHC complexes see.below .

8

Over the next 1.5]30 min the clusters ofpeptide]MHC in the outer ring moved towards thecentre of the interface, forming a compact centralcluster that is surrounded by a ring of ICAM-1 en-gaged to LFA-1. This organized interface, which istermed the ‘mature’ IS,10 is very similar to the pat-tern of SMACs described by Monks et al.8 The latterwere visualized at the interface of T cells and specificantigens presenting APCs that had been fixed after30 min of coculture. This transport of peptide]MHCis inhibited by cytochalasin D, suggesting that ISformation is dependent on the actin cytoskeleton.10

Using peptide variants it was shown that the finaldensity of peptide]MHC complexes in the centralcluster correlated best with the half-life of theTCRrpeptide]MHC interaction.10

A final stage of IS formation was analyzed by fluor-Ž .escent photobleaching recovery FPR experiments.

This revealed that, after formation, the central clus-ter of the engaged peptide]MHC complexes doesnot exchange with free peptide]MHC outside thebleached area. The reasons for this are currentlyunclear, and it is not known whether the TCR is alsostably associated with this central cluster. One possi-ble explanation for this stability is that peptide]MHCcomplexes, possibly with bound TCR, oligomerisedirectly with each other, forming a two-dimensionallattice.26 A second possibility is that a complex net-work of associations between TCRrCD3, the co-re-ceptors CD4 or CD8, accessory molecules, and theirvarious extracellular and intracellular ligands, isstabilized through the cooperative effects of multipleweak interactions. How can this stability be recon-ciled with other data27,28 suggesting that TCRrpeptide]MHC complexes are continuously formingand dissociating during T cell antigen recognition?One proposed explanation10 is that there is a highturnover of TCRrpeptide]MHC complexes duringthe formation of the mature IS in the steps precedingstabilization of the central cluster.

A key finding to emerge from these dynamic stud-ies is that TCR triggering is a very early event, andthat changes in morphology, motility, calcium con-centration and, in particular, IS formation all follow,and are dependent on initial TCR triggering. Thisraises three important questions that will be ad-

dressed in the remainder of this review. How doesthe initial TCR triggering occur? How does this trig-gering drive formation of the IS? Given that it is notrequired for initial TCR triggering, what is the pur-pose of IS formation?

Mechanisms

Overview

The processes contributing to the redistribution andsegregation of cell surface molecules fall into twogroups. Firstly there are passive mechanisms whichare dependent on properties intrinsic to themolecules and their interactions. These include in-teractions with counter-receptors at the contact inter-face, physical dimensions, and partition into lipidrafts. Although not active in the sense of consumingcellular energy, these mechanisms do produce physi-cal forces. Secondly, redistribution can be driven byactive cellular processes, such as polarized secretionor remodelling of the cortical actin cytoskeleton.

Biophysical considerations

Molecular dimensions, membrane alignment and two-di-mensional affinity

The molecular composition of the T cell surfacehas been intensively studied and is now reasonablywell characterized.29 One striking observation30,31 is

Žthat these molecules vary enormously in size Figure.3 . The extracellular portions of the TCR and its

Ž .ligand peptide]MHC are small ;7 nm comparedwith several highly-abundant cell surface molecules

Ž . Ž .such as CD43 ;45 nm and CD45 ;28]50 nm , aswell as important adhesion molecules such as inte-

Ž .grins ;21 nm . It is conceivable that undulations inthe plasma membrane could accommodate the sizedifferences of interacting molecules. However, a sec-ond striking observation is that interactions betweenmembrane-tethered adhesion molecules generateprecise alignment of apposing membranes.15,32

Consequently, molecular interactions spanningdifferent dimensions tend to segregate at the inter-face into distinct, homogenous zones.33 While theexistence and significance of membrane alignmenthad been proposed earlier,34 ] 36 it was demonstratedfor the first time by a comparison of the solution

Ž .affinity and the membrane or two-dimensional af-finity of the CD2rCD58 interaction, using solubleand membrane-tethered forms of these molecules,respectively.15,32 The membrane affinity constant istwo-dimensional because the concentration of mem-

brane-tethered molecules is measured as surface den-

Ž 2 .sity, a two-dimensional parameter moleculesrmm .This comparison revealed that, by confining both

Žinteracting binding sites to a small volume ;5 nm. Žhigh between apposing membranes the ‘confine-

9

Redistribution of T cell surface molecules

.ment region’ , precise alignment of membranes en-ables the CD2rCD58 interaction, which has a low

Ž .solution affinity e.g. K ;2 mM , to achieve a highdŽtwo-dimensional affinity 2D K ; 1 moleculerd

2 . 15,32mm .The presence of large molecules at the cell]cell

interface would be expected to act as a constraint onTCR engagement of peptide]MHC by preventingsufficiently close membrane approximation andmembrane alignment.37 Indeed, T cell antigenrecognition is enhanced when CD43 is absent from Tcells.38 Notably, however, many T cell ‘accessory’molecules which contribute to T cell antigen recogni-

Ž .tion are also small Figure 3 . These observation ledto the suggestion30,39,40 that, in order for the TCR toengage peptide]MHC, molecules at the interfacebetween T cells and APCs would have to segregate

Ž .according to the size Figure 4 , with small accessorymolecules contributing to the formation of ‘close-contact zones’, within which the inter-membrane sep-

Ž .aration distance ;15 nm matched the dimensionsof the TCRrpeptide]MHC complex, and from whichlarge molecules such as CD43, CD45 and integrinsare excluded. Structural41 ] 44 and mutagenesis36 stud-ies of CD2 and its ligands CD48 and CD58 haveshown that these molecules bind in a head-to-headorientation, forming a complex that spans the same

Ž .dimensions ;14 nm as the TCRrpeptide]MHCcomplex. On this basis it was proposed that onefunction of CD2 is to facilitate the formation ofclose-contact zones, thereby enhancing TCR engage-ment of peptide]MHC.36,39,40 The importance ofclose membrane approximation is strongly supportedby the recent observation that, whereas formation of

Ž .the wild-type ;14 nm CD2rCD48 complex stronglyenhances TCR engagement of peptide]MHC, an

Ž .elongated ;21 nm CD2rCD48 complex is stronglyinhibitory.37

Receptor lateral mobilityWhen two cells make contact, complementary sur-

face receptors will usually need to move laterallywithin the plane of the membrane in order to en-counter each other. This movement will depend ontheir lateral mobility,45 a term which refers collec-tively to two distinct properties; the proportion of the

Ž .molecule that is mobile the mobile fraction , and the

diffusion rates of the mobile molecules within the

Žplane of the membrane the lateral diffusion coeffi-.cient .

There is currently only limited information on thelateral mobility of T cell surface molecules. Approxi-

P. Anton van der Merwe et al

mately 75% of the CD2 molecules expressed in Jurkatcells were shown to be mobile,46 similar to observa-tions with other molecules.47 This agreed with the

10

observation that a similar proportion of the CD2accumulates in contact areas with CD58-bearing pla-nar bilayers.15 In contrast, only 10]20% of the TCR

on mature T cells is mobile.10 The reasons for thislow mobility are not clear but may include cytoskele-tal associations48,49 or diffusion barriers.50 The latter

Ž .may be in the cytoplasm e.g. cytoskeletal ‘fences’ orŽin the membrane itself e.g. other membrane pro-

.teins or lipid rafts . This low mobility needs to bereconciled with experiments suggesting that up to80% of cell surface TCRrCD3 engages peptide]MHCduring antigen recognition.27 One possible explana-tion is that the immobile TCRrCD348,49 is activelytransported to the contact interface over a long pe-riod. Alternatively, the T cell may form multiple

Ž .contacts simultaneous or sequential with severalAPCs.

Active processes

There is compelling evidence that active processescontribute to the redistribution of cell surfacemolecules during T cell antigen recognition. Firstly,much of the observed redistribution requires andfollows TCR triggering.5,8 ] 10,13,14,51 Secondly, cy -toskeletal inhibitors inhibit redistribution.10,13,14

Ž .Thirdly, a number of studies see, e.g. refs 5,52 haveshown that cross-linking of the TCRrCD3 complexwith antibodies can trigger redistribution of

Ž .TCRrCD3 ‘capping’ as well as other molecules

Ž . Ž .‘co-capping’ to one region the ‘cap’ of the T cellsurface. And finally, a cytoplasmic protein has beenidentified that interacts with the cytoplasmic domain

Ž .of CD2 CD2AP and influences its distribution withina T cellrmembrane interface.33

Figure 3. Dimensions of molecules involved in T cell antigen remolecules involved in T cell antigen recognition. Sizes are estim

Ž .with the exception of CD148. The size of CD148 ;55 nm isŽ .end-to-end ;40 nm to an ;80 amino acid mucin-like regio

Figure 4. The kinetic]segregation model of TCR triggering aŽ .cell surface molecules depicted as in Figure 3 in T cell activat

Ž .of the boxes on the right. i Before cell]cell contact thphosphorylation and dephosphorylation. Dephosphorylation

Ž .phosphorylation. ii Initial T cell contact with an APC leadcontact area as a result of small-scale segregation of molecules.Within these close-contact zones tyrosine phosphorylation i

Ž .excluded, and tyrosine kinases are concentrated. iii Engageclose-contact zones and results in the trapping of the TCRrCDdetermined primarily by the stability of the TCRrpeptide]MHphorylated, initiating the multi-step signalling process required

Ž .interaction complex II allows these signalling steps to be comlead to no or partial triggering because, on leaving the clmolecules, are re-exposed to high levels of tyrosine-phosphsegregation of molecules, culminating in the formation o‘co-stimulatory signals’ such as chemokines or the expressionthe IS results in full T cell activation.

11

Redistribution of T cell surface molecules

Two distinct transport processes have been impli-cated in the redistribution of cell-surface molecules;polarized secretion of cytoplasmic vesicles containingthe molecules, and redistribution of molecules al-ready at the surface.

Polarized secretionBoth CD8 T cells5,53 and CD4 T cells5,54,55 are

capable of polarized secretion, in which cytoplasmicvesicles fuse to the plasma membrane at the contactinterface with target cells or APCs. This allows the

Žcontents of these vesicles e.g. perforin, granzymes,.and cytokines as well as the vesicle membranes to be

directed to the interface.5,53 ] 55 Recent data suggestthat the T cell surface molecules FasL56 and CTLA-452

reside predominantly in cytoplasmic vesicular stores,and that these molecules are targeted to the site ofTCR ligation by polarized secretion.

There is some evidence that lipid rafts and theirŽ .associated molecules see below are concentrated in

intracellular, presumably vesicular, stores in naıve T¨cells.57,58 Polarized secretion is known to target raftsto the apical membrane of epithelial cells.59 Thisraises the question as to whether lipid rafts andassociated molecules are transported to the IS bypolarized secretion.

Movement of molecules on the cell surfaceSeveral lines of evidence implicate the cortical actin

Ž .cytoskeleton CAC in the redistribution of surfacemolecules to the contact interface. Firstly, treatmentof T cells with cytoskeletal inhibitors abrogates the

cognition. A schematic view of the dimensions of cell-surfaceated from structural studies of these or related molecules29

estimated by adding 10 fibronectin type III domains strungŽ .n ;15 nm .

nd immunological synapse formation. The redistribution ofion. The green and red boxes on the left are enlarged viewse TCRrCD3 complex is subject to constitutive tyrosinedominates, leading to a low steady-state level of tyrosine

s the formation of multiple close-contact zones within theThis is a passive process driven by size and ligand binding.

Ž .s favoured because tyrosine phosphatases e.g. CD45 arement by TCR of specific-peptide]MHC occurs within these3 complex within a zone for a period, the length of which isC complex. While in the zone, TCRrCD3 is tyrosinephos-

Ž .for triggering. iv A sufficiently stable TCRrpeptide]MHCŽ .pleted, leading to triggering. Unstable complexes I and III

ose-contact zone, TCRrCD3, and any recruited signallingŽ .atase activity. v TCR triggering initiates active large-scale

Ž . Ž . Ž .f the mature IS. vi Steps ii to v are enhanced byŽ .of CD28 ligands on APCs. vii Sustained signalling through

P. Anton van der Merwe et al

redistribution of ICAM-1 and other molecules to theinterface13,14 and inhibits transport of peptide]MHCwithin the interface.10 Secondly, a large number ofcell surface molecules which migrate towards andaway from the interface have been shown to associatewith the cytoskeleton or with linker molecules which

Ž .themselves associate with the cytoskeleton see below .Finally, ligation of the TCR leads to localized talinaccumulation5 and actin polymerization.12,60,61

How might the CAC organise the redistribution ofcell surface molecules upon T cell activation? TheCAC moves in a centripetal direction within thecellular processes that extend from cells, as a result ofa combination of actin filament remodelling andmyosin motor proteins.62 Following recognition, Tcells extend analogous processes around the APC,forming a cup-shape interface. Centripetal movementof the CAC within this ‘cup’ might move moleculescoupled to the CAC, such as TCRrCD3,48,49 radiallytowards the centre of the contact area. How are some

Ž .molecules e.g. CD2 and LFA-1 prevented from mov-Ž .ing into the centre and others e.g. CD43 actively

excluded? Possible explanations include the absenceof links to the CAC, physical exclusion on the basis ofsize, and attachment to other cytoskeletal structures.

Lipid rafts

The plasma membrane of many cells,63 including Tcells,64,65 contains microdomains which are enrichedin glycosphingolipids and cholesterol, termed lipidrafts. Some T cell surface molecules important in

Ž .TCR triggering e.g. lck, fyn, LAT, CD48 and CD4are preferentially associated with these structures64 ] 66.Others, most notably CD4564,67 appear to be depletedfrom lipid rafts. This being so, the redistribution oflipid rafts could contribute to the segregation of cellsurface molecules at the T cellrAPC interface. Re-cent data suggest that triggering of the TCR caninduce active redistribution of lipid rafts towards thesite of TCR triggering.57 If cell-surface molecules thatassociate with lipid rafts, such as the GPI-anchoredmolecule CD48, bind to a ligand on APCs, this couldprovide a passive mechanism for the accumulation ofrafts at the interface.66 The importance of lipid raftsin TCR triggering is highlighted by the finding that

chemical disruption of rafts abrogates TCR trigger-ing.64,68 Furthermore, mutations which disrupt theassociation with rafts of the essential signallingmolecules LAT69 or lck 51 also disrupt TCR trigger-ing. The role of lipid rafts and LAT in TCR signalling

12

are discussed in detail in the accompanying reviewsby Janes et al and Zhang and Samelson, respectively.

Linking TCR triggering to cytoskeletal polarization andreceptor redistribution

Role of TCR triggeringAnti-TCR antibodies can induce both re-orienta-

tion of the MTOC,61,70 polarized exocytosis52 andlocalized actin polymerization,61,70 suggesting thatTCR triggering may be sufficient for these processes.However, an important caveat is that these experi-

Žments were performed in the presence of serum a.source of extracellular matrix proteins or intact tar-

get cells, leaving open the possibility that integrinengagement is an additional requirement.71 As far asis known, TCR triggering is necessary for the re-orien-tation of the MTOC and Golgi apparatus towards theT cellrAPC contact interface.5,19,70 In contrast, local-ized actin polymerization does not require TCR trig-gering.70 An interaction between LFA-1 on T cellsand ICAM-1 on APCs is sufficient to induce corticalactin polymerization and talin accumulation at thecontact site.70

Anti-TCR antibodies can also induce redistributionof cell surface molecules, suggesting that TCR trig-gering is also sufficient for this to occur.5,52 It isunclear whether TCR triggering is absolutely neces-sary for active redistribution of cell surface molecules.The active redistribution described to date at T cellcontact interfaces has always been preceded by TCRligation.5,7,8,10 However, the various antigen-indepen-dent responses seen when T cells contact dendriticcells raise the possibility that TCR triggering may notbe essential for redistribution at T cellrdendritic cellinterfaces.21,23,24

Role of accessory molecules: a mechanism of co-stimulation?Recent studies have implicated accessory molecules

in the active redistribution of T cell surface molecules.Wulfing and Davis showed that bulk redistribution ofT cell surface molecules to the T cellrAPC interfaceis inhibited when CD28rCD80 or LFA-1rICAM-1 in-teractions are blocked.14 Viola et al found that theredistribution of lipid rafts to the area in contact withanti-CD3 antibody-coated beads is dramatically en-

hanced when the beads are also coated with anti-CD28antibodies.57 These findings suggest that accessorymolecules may enhance T cell activation by con-tributing to this redistribution process. The ability ofthe CD28rligand interaction to activate Rho family

Ž .GTPases see below independently of TCR trigger-ing support this notion.72 As noted above, the LFA-1rICAM-1 interaction can trigger changes in theCAC at the contact site.70

The finding that cytoskeletal polarization andlarge-scale receptor redistribution depend on TCRtriggering, and are enhanced by engagement of ac-cessory molecules such as CD28, suggests that ISformation is the point at which the T cell integratesinformation about the presence of the antigen on the

ŽAPC and its activation state e.g. expression of.chemokines or CD28 ligands . Or, to rephrase in the

terminology of the two-signal hypothesis of T cellactivation, IS formation may integrate signal 1Ž . Žstimulus from antigen with signal 2 co-stimulus

.from the APC . This is supported by the finding thatdendritic cells, the most effective APCs, are uniquelycapable of adhering to and inducing antigen-inde-pendent responses in T cells.21,23 ] 25

Signalling pathwaysRecently a tentative outline has emerged of the

signal transduction pathways that couple TCR trig-gering to cytoskeletal polarization and receptor redis-tribution.72,73 Attention has focused on the Rho family

Žof small GTP binding proteins especially RhoA,.Cdc42Hs and Rac1 since they have been shown to

regulate cytoskeletal and vesicle transport processesin other cells.72,74 Using a T cell hybridoma, Stower etal60 have shown that a dominant-negative mutant ofCdc42Hs inhibits actin and microtubule polarizationtowards the T cellrAPC interface.

How does TCR triggering activate rho GTPases?The most likely candidate is the GTPrGDP exchangefactor vav1, which activates Rac1 and has been re-ported to be essential for TCR-induced actin po-lymerization and capping.75 ] 77 Vav1 is activated bytyrosine phosphorylation andror by binding to po-lyphosphoinositide products generated following ac-

Ž . 73tivation of phosphatidylinositol PI 3-kinase.The mechanisms by which the rho GTPases regu-

late the cytoskeleton is not well-understood but islikely to involve effector proteins which regulateactin-binding proteins, such as the protein kinasesROCK and LIM-kinase.72,78 An effector protein al-ready implicated in coupling Rho GTPases to thecytoskeleton in T cells is WASp, so named because it

is defective in patients with Wiscott-Aldrich im-munodeficiency syndrome.79 WASp interacts withCdc42Hs and is required for capping induced byanti-CD3 antibodies, suggesting that it couplesCdc42Hs activation to the actin cytoskeleton.

13

Redistribution of T cell surface molecules

Thus, the picture emerging is that TCR triggeringactivates Rho family GTPases through activation ofproteins such as vav, and that Rho GTPases regulatevarious effector proteins, including protein kinasesand WASp, which regulate in turn actin bindingproteins. Direct links between vav1 and the actin73

and microtubule80,81 cytoskeleton have also beenidentified. Since the cytoskeleton and associated mo-tor proteins control the active transport processesimplicated in receptor redistribution, the vavrRhopathway is likely to couple TCR triggering to cy-toskeletal polarization and the redistribution of cell-surface molecules.

Linking the cytoskeleton to T cell surface moleculesFor TCR-induced changes in the CAC to con-

tribute the redistribution of cell-surface molecules,these molecules need to associate with the CAC.Linkages have been reported for LFA-1, which associ-ates via talin,82 CD2, which associates with CD2AP33

and tubulin,83 the TCR, which associates via theCD3z chain,48,49 and CD45, which associates with thelinker protein fodrin.84 Furthermore, members of

Ž .the ezrin-radixin-moesin ERM family of proteinscouple CD43, CD44, and ICAM-1, -2, and -3 to the

85 Ž .CAC. With the exception of CD2AP see below ,the role of these linker molecules in surface moleculeredistribution is not known. However, their diversityprovides a possible explanation for the distinct local-ization of cell-surface molecules within the IS.

CD2AP

CD2AP was first identified as a molecule that binds tothe cytoplasmic domain of CD2 and influences itsdistribution at the contact interface between a T celland planar lipid bilayers.33 The finding thatCD2rCD48 complexes occupy a special position inthe IS10 is further evidence that CD2 is being activelymoved at the interface, and it seems likely that thismovement is also mediated by CD2AP.

A recent study on CD2AP-deficient mice suggeststhat this molecule has a wider role in lymphoid andnon-lymphoid tissues.86 These mice die from compli-cations of excessive leakage of serum proteins into

Ž . 86the urine nephrotic syndrome . This leakage ap-

pears to result from defects in the epithelial cellŽ .podocyte processes that contribute to the glomeru-lar filtration barrier. Interestingly, CD2AP-deficient Tcells show a substantial reduction in response toanti-CD3 antibodies,86 whereas CD2-deficient mice

P. Anton van der Merwe et al

are normal in this respect, suggesting that CD2APplays a broader role in T cell function than CD2.

What might be the purpose of locating CD2rCD48complexes in an inner ring surrounding theTCRrpeptide]MHC-containing central zone? Oneintriguing possibility, consistent with a generalizedrole for CD2AP in forming filtration junctions, is thatthis ring of CD2rCD48 functions as a molecular filterwithin the IS keeping larger molecules out of thecentral zone in the face of centripetal actinomyosin-based transport processes.10,14

Functions

What is the purpose of cytoskeletal polarization andredistribution of cell surface molecules? Oneprobable role for cytoskeletal polarization is to facili-tate the targeting by T cells of cytokines and thecontents of lytic granules to the site of contact withAPCs or target cells, respectively.5,55 As argued above,cytoskeletal polarization is also likely to contribute tothe large-scale redistribution of cell surface moleculesleading to the formation of the IS. Here we speculateon the role of IS formation in T cell activation,placing it in the context of a recently-proposed modelof TCR triggering.39,40 It is important to re-emphasisehere that the formation of the IS follows and depends onTCR triggering, indicating that it cannot be the mechanismof initial TCR triggering. Nevertheless IS formationdoes appear to be required for full T cell activation.What is the evidence for this? Firstly, in the fewstudies performed to date, IS formation correlatedperfectly with full T cell activation.8,10 Secondly, dis-ruption of the actin cytoskeleton20,87 or the regula-tion thereof,75 ] 77 manipulations which would be ex-pected to disrupt IS formation, leads to disruption of

Žfull T cell activation proliferation andror cytokine.production and negative selection. While not con-

clusive, these findings are consistent with an essentialrole for the IS in full T cell activation.

Small-scale segregation provides a novel mechanism forsignal transduction across membranes

TCR triggering requires binding of the TCR to speci-fic peptide presented on MHC class I or II molecules.

One of the key, unresolved questions in T cell bi-ology is how this binding generates a signal. In otherwords how does the T cell ‘know’ that a particularTCR has bound peptide]MHC? Based on the waythey address this question, current models of TCR

14

triggering can be divided into aggregation,88,89 con-formational change,90,91 or segregation39,40,92,93 mod-els. Segregation models have in common the pro-posal that peptide]MHC binding leads to triggeringby holding the TCRrCD3 complex in a membraneenvironment in which tyrosine kinases have beensegregated from tyrosine-phosphatases. Kineticproof-reading,94 kinetic discrimination95 and serialtriggering27 models do not fall neatly into the aboveclassification because they do not address the centralquestion of how binding transduces a signal. Insteadthey try to account for particular features of TCRtriggering, such as the ability of the TCR to discrimi-nate between very similar ligands94,95 and the abilityof a few peptide]MHC complexes to engage manyTCRs.27 Consequently they can and have been incor-porated into aggregation, conformational change, orsegregation models.

Problems with aggregation and conformational changemodels

Aggregation models postulate that peptide]MHCengagement leads to clustering or aggregation ofTCRrCD3 complexes. At first glance this is an attrac-tive notion since dimerizationraggregation is a well-established mechanism of signal transduction,96 andit is well known that aggregation of TCRrCD3, witheither antibodies or multivalent peptide]MHC,97 willtrigger TCRs. However, there are reasons for doubt-ing that this is the physiological mechanism for TCRtriggering.92 In systems in which dimerizationraggre-gation lead to signalling, the mechanism of multimer-

96 Žization is usually clear; a multivalent ligand either.a monomer with two binding sites or a multimer

binds two or more receptors, thereby bringing theminto close proximity. In the case of the TCR theligand, peptide]MHC, is monovalent, and present atan exceptionally low density, with as few as 10]100molecules of a specific peptide]MHC on an entirecell. Thus, only a very small number of TCRs will bebound at any one time. Furthermore, each interac-tion is likely to be very short-lived, with a half-time ofa few seconds.89 This suggests that simultaneous en-

Ž .gagement of adjacent TCRs i.e. aggregation will beexceedingly rare. Furthermore, if TCR triggering de-pended on aggregation it should be very sensitive tothe surface density of specific peptide]MHC, whereas

Ž .the converse is true see, e.g. refs 10,98 . A refine-ment of the aggregation model, which attempts toaddress these problems, proposes that the binding ofTCR to peptide]MHC generates a new binding sitethat enables TCRrpeptide]MHC complexes to self-

associate.26 However, given the huge diversity in TCRand peptide]MHC structures, and the resulting di-versity in the structure of TCRrpeptide]MHC com-plexes,99 it is difficult to envisage how such bindingsites could be conserved in all TCRrpeptide]MHCcomplexes. A further difficulty is understanding howdirect self-association occurs on the cell surface wheneach TCR is surrounded by several other moleculesŽ .e.g. CD3 subunits and a co-receptor , all of whichare heavily glycosylated.

A variant of the aggregation model is that trigger-ing is a consequence of heterodimerization of the

Ž .TCR with its co-receptor CD4 or CD8 . While thismay be sufficient for TCR triggering in certain cir -cumstances,100 it cannot be the universal mechanismof triggering, since TCR triggering in vitro and invivo101,102 can occur in the complete absence of CD4or CD8.

Conformational change models propose that pep-tide]MHC binding leads to a conformational changein the TCR, which is somehow transduced into thecell interior. This model requires that every TCR isable to undergo the same binding-induced conforma-tional change in the face of huge diversity in thestructure of the antigen binding sites andpeptide]MHC, which would seem highly implausible.Furthermore, structural data available to date showthat there are no conformational changes in the TCRinduced by binding peptide]MHC except for adjust-ments within the antigen-binding site.91,99 Taking thelatter data into account, Wiley et al91 recently pro-posed a variant of the conformational change modelbased on studies of signalling through the erythropo-etin receptor.103 In this model the TCR exists as apreformed dimer. When both TCRs in the dimerbind to peptide]MHC the two TCRs, reposition withrespect to each other, and this repositioning is de-tected in some way. Unfortunately this model has thesame drawback as the aggregation models when ap-

Ž .plied to TCR triggering see above . A further dif-ficulty is envisaging how the independent binding oftwo specific peptide]MHC complexes could changethe relative positions of the two TCRs?

The kinetic-segregation model of TCR triggeringThe arguments outlined above against conforma-

tional change or dimerizationraggregation models

led to the proposal that an entirely novel mechanism,based on molecular segregation in contact areas, wasresponsible for TCR triggering.39 A related ‘topologi-cal model’, based on the implications of new data ontwo-dimensional affinity,32 subsequently converged on

15

Redistribution of T cell surface molecules

the same mechanism.40 These models emerged fromseveral key observations described below.39,40

Firstly, the TCR and T cell accessory molecules aremuch smaller than other abundant cell surfacemolecules, including CD43, CD45, and integrins30,31

Ž .Figure 3 . Taken together with the observation thatmembrane alignment is important for optimal bind-ing of membrane-tethered molecules,15,32 this sug-gested that these molecules must segregate accordingto size at the contact interface.36,39,40

Secondly, tyrosine phosphatase inhibitors can acti-vate T cells,104,105 producing tyrosine phosphoryla-tion patterns very similar to those seen during TCRtriggering.106,107 Furthermore, an inhibitor of src ki-

Ž .nases PP1 rapidly and dramatically decreases thelevel of basal tyrosine phosphorylation in T cells.57

These findings show that the tyrosine kinases impli-cated in TCR triggering are constitutively active andunder tonic inhibitory control by tyrosine phos-phatases, and imply that altering the local balancebetween these two competing, constitutive processeswould be sufficient to trigger intracellular signals.

Thirdly, one of the most active and abundant tyro-sine phosphatases is the large cell surface moleculeCD45.108 Furthermore, CD45 may associate directly

Ž .or indirectly via CD45-associated protein with theTCR and its signalling apparatus.109 ] 112 This sug-gested that, in addition to its generally accepted rolemaintaining lck in a primed state,93 CD45 may alsotonically inhibit TCR triggering by dephosphorylat-ing key tyrosine kinase substrates such as CD3z ,109

ZAP-70106 and lck itself.113 There is clear evidencethat membrane-associated tyrosine phosphatasesother than CD45 also reverse constitutively activetyrosine kinases.107 One possible candidate is CD148,which has recently been shown to inhibit TCR trig-gering.114 Interestingly, like CD45, CD148 is pre-

Ždicted to have a very large extracellular domain Fig-.ure 3 .

Finally, the TCR is able to discriminate between,and produce very different responses to, very similarligands.115 This is difficult to explain purely on thebasis of differences in affinity; all else being equal, atwofold change in affinity will produce a less thantwofold change in the total number of engaged TCRs.In contrast, small kinetic differences can readily beamplified to produce very different biochemical re-

sponses.94,95

These observations were synthesized into a newmodel of TCR triggering,39,40 which we now refer toas the kinetic]segregation model to emphasise its twokey elements, molecular segregation and binding ki-

P. Anton van der Merwe et al

netics. The model postulates the following sequenceŽ .of events in TCR triggering Figure 4 .

When T cells encounter target cells or APCs,molecules at the contact interface will segregate ac-cording to size. Small molecules such as CD2, CD4,CD8 and CD28 will contribute to the formation ofmultiple, small close-contact zones, from which largemolecules such as CD45, CD43, CD148 and LFA-1are excluded.

As a result of this segregation, close-contact zoneswill be a ‘pro-signalling’ environment, relatively de-

Žpleted of tyrosine phosphatases such as CD45 and.perhaps CD148 and enriched in tyrosine kinases

such as lck. The close membrane approximationwithin the close-contact zone will facilitate TCR en-counters with specific peptide]MHC. The binding ofTCR to specific peptide]MHC will result in the TCRand its associated CD3 complex being held within aclose contact zone, where they show a net increase inphosphorylation as a result of exposure to tyrosinekinases such as lck in the absence of tyrosine phos-phatases such as CD45. It is important to emphasisehere that ligation of a single TCRrCD3 can generate

Ž .a signal no dimerization or aggregation is required .Furthermore, tyrosine phosphorylation of TCRrCD3is not strictly dependent on its localization to a ‘pro-signalling’ close contact zone, given that it occurs

Ž .constitutively see above . It is the downstream sig-nalling events that are crucially dependent on local-ization to close-contact zones because this prolongsthe half-life of tyrosine phosphorylated species bysegregating the TCRrCD3 and associated moleculesfrom tyrosine phosphatases. This tyrosine phosphory-lation initiates a cascade of events including recruit-ment and activation of the tyrosine kinase ZAP-70and phosphorylation and activation of ZAP-70 subs-trates.116 This sequence will be interrupted when theTCRrCD3 complex and associated molecules leavessuch a close-contact zone. Because the TCRrCD3 isheld within the contact zone through binding topeptide]MHC, tyrosine phosphorylation and TCRtriggering will depend crucially on the half-life of theTCRrpeptide]MHC interaction117 and a decrease inthe half-life will lead to interruption of the cascade ofsignalling events.118,119

To provide kinetic discrimination the model re-quires that initial close-contact zones be relatively

small or short-lived so that inappropriately phospho-rylated TCRrCD3 complexes do not have to diffusefar prior to their dephosphorylation by CD45 orother excluded phosphatases. This may be a keydifference between the respective signalling roles of

16

the initial close-contact zones and the IS, which formsover a much larger area. It also seems likely thatthere will be many such close-contact zones formingwithin the contact area between T cells and APCs.Unfortunately, given the limited resolution of thelight microscope, the existence of such small close-contact zones remains speculative.

Recent studies relevant to the kinetic]segregation modelSince the kinetic]segregation model was first pro-

posed a number of findings have provided supportfor key elements of the model. First, mice with re-duced levels of CD45 expression show evidence ofenhanced signalling through the TCR,120 consistentwith an inhibitory effect of CD45. Second, CD45 hasbeen shown to inhibit the function of constitutively-active lck by dephosphorylating its substrates.121

Third, studies on planar bilayer systems have shownthat engaged CD2 and LFA-1, which differ consider-

Ž .ably in size Figure 3 , segregate in the contact inter-face independent of cytoskeletal interactions.33 Thisstrongly supports the proposal that size differencesdrive the segregation of molecules within the contactinterface. Fourth, it has been demonstrated that in-creasing the dimensions of an accessory moleculestrongly inhibits TCR engagement.37 Fifth, a form ofCD45 with a truncated extracellular domain has been

Žshown to inhibit TCR triggering by antigens C. Irles,A. Symons, F. Michel, P.A. van der Merwe, and O.

.Acuto, unpublished observations . Sixth, T cells areable to discriminate between peptide]MHC ligandswith similar affinities but with differences in half-lifefor TCR binding.89,117 Finally, as with early TCRsignalling events,117 the number of peptide]MHCcomplexes that accumulated in the central region ofthe IS correlated strongly with the half-life of theTCR binding.10 This provided the first direct evi -dence that IS formation is dictated by kinetics of theTCRrpeptide]MHC interaction,10 most likelythrough the dependence of large-scale segregationon early TCR signalling events that are triggered, wesuggest, by small-scale segregation.

TCR triggering is inhibited by cell-surface recep-tors such as CTLA-4 and Killer cell Immunoglobulin

Ž .Receptors KIRs , which associate with the cyto-plasmic tyrosine phosphatases SHP-1 and SHP-2.122,123

Interestingly, the complexes formed by CTLA-4 and

ŽKIRs and their ligands CD80rCD86 and MHC class

.I, respectively are predicted to have dimensionscompatible with their co-localization with the TCR in

Ž .contact areas Figure 3 . Furthermore, CTLA-4 in-hibits the earliest tyrosine phosphorylation events fol-

lowing TCR triggering.124,125 This suggests that in-hibitory receptors may function by altering thebalance between tyrosine phosphorylation and de-phosphorylation in the immediate vicinity of theTCRrCD3. As noted earlier, CTLA-4 is found pre-dominantly in vesicular stores and is directed to thesurface at the site of TCR triggering by polarizedsecretion.52 This suggests that CTLA-4 may be tar-geted to the centre of the IS as a means of modulat-ing or terminating TCR signalling.

The kinetic]segregation model is compatible withevidence that lipid rafts play a crucial role in TCR

Ž .triggering see above . Lipid rafts may accumulate inor adjacent to the close-contact zones through eitheractive57 or passive mechanisms,66 helping to con-tribute to the ‘pro-signalling’ environment by recruit-ing tyrosine kinases and other pro-signalling

Žmolecules and excluding CD45 see accompanyingreview by Janes et al.67 Based on these recent data,new models of TCR triggering have been proposedwhich postulate that TCR binding to peptide]MHCbrings the TCR and lipid rafts together.92,93 These‘lipid raft models’ are similar in principle to thekinetic]segregation model in that they also proposethat triggering is a consequence of the redistributionof the TCR into an environment enriched in tyrosinekinases and depleted of tyrosine phosphatases. Thesemodels may, however, need to be adjusted sincerecent evidence suggests that TCRs may be constitu-

Ž .tively associated with rafts see Janes et al . The maindrawback of lipid raft models is that they do notsuggest a clear mechanism, independent of aggrega-tion or conformational change, by which engage-ment of the TCR with peptide]MHC either enhancesits association with lipid rafts or promotes their ag-gregation.

The immunological synapse as an activation checkpointand signal integrator

In addition to being structurally plausible, any modelof TCR triggering has to explain a central paradox.The TCR detects peptide]MHC ligands with a widerange of affinities and surface densities but can nev-ertheless discriminate between peptide]MHC com-plexes with very subtle differences in binding proper-ties.115,117,126 The ability of a given TCR to respond to

a wide range of affinities is clearly illustrated by theability of thymocytes to undergo positive and negativeselection using the same TCR. More recently it hasbeen shown that mature peripheral T cells also

Žrecognise and respond to self-peptide]MHC re-

17

Redistribution of T cell surface molecules

.viewed in ref 126; see also refs 127,128 . These sameperipheral T cells require a much higher affinityTCR ligand in order to become fully activated. Howis the TCR able to respond to ligands with such awide range of affinities and yet retain the ability todiscriminate between ligand which differ only slightlyin their binding properties?

The kinetic]segregation model would allow veryweak interactions to be detected because TCRrCD3complexes need only be transiently detained inclose-contact zones to become phosphorylated. Thus,tyrosine phosphorylation could occur even with veryweak TCR binding to peptide]MHC. Perhaps this issufficient to generate the signals required for positiveselection in the thymus or survival in the periphery.The finding that ZAP-70 is constitutively associatedwith partially-phosphorylated TCR-z in peripheral Tcells, but that this is reduced in T cells from micelacking self-peptide]MHC and absent in T cells grownin culture, is consistent with continuous weak sig-nalling through the TCR.126

Because it would be inappropriate for peripheral Tcells to be fully-activated by self-peptide]MHC, amechanism must exist that enables a T cell to clearlydistinguish between the signals resulting from weakversus strong self-peptide]MHC engagement. Wesuggest that the IS provides such a mechanism be-cause it appears only to form when TCR binds pep-tide]MHC above a threshold affinity or half-life.10

Nevertheless the mechanisms that leads to IS forma-tion are exquisitely sensitive since these structuresform at very low surface densities of peptide]MHC.10

Taken together, these findings suggest that IS forma-tion may function as a checkpoint for full T cellactivation, facilitating discrimination between low

Ž .affinity presumably self- peptide]MHC engagementŽ .and high affinity presumably foreign- peptide]MHC

engagement.Why might IS formation be required for full T cell

activation? There are at least two possibilities. ISformation may lead to the generation of a uniquesignal. A candidate for such a signal is activatedprotein]kinase C-u , which selectively accumulates in

7 Žthe central portion of the IS or cSMAC see accom-.panying review by Kativar and Mochly-Rosen . A sec-

ond possibility is that IS formation may be necessaryŽ .for the prolonged )1 h triggering necessary for

129

full T cell activation.If the IS is required for full T cell activation, it

follows that any factor that contributes to IS forma-tion will contribute to full T cell activation. Like fullT cell activation, IS formation requires that the

P. Anton van der Merwe et al

TCRrpeptide]MHC interaction affinity or half-life isabove a threshold value.10 There is also intriguingevidence that ligation of accessory molecules such asCD28 and LFA-1 enhances some of the transportprocesses that lead to IS formation.14,57 Interactionswith these molecules are regulated by soluble media-

Ž .tors e.g. chemokines will activate LFA-1 and willdepend on the activation or maturation state of theAPC.24,25 Finally, there is evidence that dendritic cellsare distinct from other APCs in their ability adhere toand evoke changes in naıve T cells in the absence of¨and prior to antigen recognition.21,23,24 These datasuggest that the formation of the IS serves to inte-grate information relating to the presence and qual-

Žity of the antigen e.g. surface density and binding.properties and information on the nature and acti-

vation state of the APC.24,130

Based on these considerations we suggest the fol-Žlowing integrated model of full T cell activation Fig-

.ure 4 . Small scale-segregation, driven primarily byligand binding and steric factors, leads to initial TCRtriggering, by the mechanism outlined in the ki-netic]segregation model. This is sufficient to stimu-late certain T cell functions, the nature of whichŽ .positive selection or survival depends on the devel-opmental state of the T cell. Full activation, however,requires the formation of the IS, which involves ac-tive transport processes. IS formation requires athreshold level of triggering through the TCR and isinfluenced by the nature and activation state of theAPC.

Acknowledgements

P.A.V. is supported by the Medical Research Council andS.J.D. by the Wellcome Trust. M.L.D. is supported by theArthritis Foundation, The Whitaker Foundation and TheNational Institutes of Health. A.S. is supported by theNational Institutes of Health. We thank Don Mason, NeilBarclay, Marion Brown, Matt Thomas, and Talitha Bakkerfor critical comments on the manuscript. We thank S.Bromley and C. Sumen for the use of images in Figure 2.

References

Ž .1. Kucik DF, Dustin ML, Miller JM, Brown EJ 1996 Adhesionactivating phorbol ester increases the mobility of leukocyteintegrin LFA-1 in cultured lymphocytes. J Clin Invest97:2139]2144

Ž .2. Dustin ML 1998 Making a little affinity go a long way: atopological view of LFA-1 regulation. Cell Adhes Commun6:255]262

18

Ž .3. Dustin ML, Springer TA 1989 T-cell receptor cross-linkingtransiently stimulates adhesiveness through LFA-1. Nature341:619]624

Ž .4. Sanchez-Madrid F, del Pozo MA 1999 Leukocyte polariza-tion in cell migration and immune interactions. EMBO J18:501]511

Ž .5. Kupfer A, Singer SJ 1989 Cell biology of cytotoxic andhelper T cell functions: immunofluorescence microscopicstudies of single cells and cell couples. Annu Rev Immunol7:309]337

Ž .6. Paul WE, Seder RA 1994 Lymphocyte responses and cy-tokines. Cell 76:241]251

Ž .7. Monks CR, Kupfer H, Tamir I, Barlow A, Kupfer A 1997Selective modulation of protein kinase C-theta during T-cellactivation. Nature 385:83]86

8. Monks CR, Freiberg BA, Kupfer H, Sciaky N, Kupfer AŽ .1998 Three-dimensional segregation of supramolecular ac-tivation clusters in T cells. Nature 395:82]86

Ž .9. Sperling AI et al 1998 Cutting edge: TCR signaling inducesselective exclusion of CD43 from the T cell]antigen-present-ing cell contact site. J Immunol 161:6459]6462

Ž .10. Grakoui A et al 1999 The immunological synapse: a molec-ular machine controlling T cell activation. Science 285:221]227

Ž .11. Geiger B, Rosen D, Berke G 1982 Spatial relationships ofmicrotubule-organizing centers and the contact area of cyto-toxic T lymphocytes and target cells. J Cell Biol 95:137]143

12. Ryser JE, Rungger-Brandle E, Chaponnier C, Gabbiani G,Ž .Vassalli P 1982 The area of attachment of cytotoxic T

lymphocytes to their target cells shows high motility andpolarization of actin, but not myosin. J Immunol 128:1159]1162

Ž .13. Wulfing C, Sjaastad MD, Davis MM 1998 Visualizing thedynamics of T cell activation: intracellular adhesion molecule1 migrates rapidly to the T cellrB cell interface and acts tosustain calcium levels. Proc Natl Acad Sci USA 95:6302]6307

Ž .14. Wulfing C, Davis MM 1998 A receptorrcytoskeletal move-ment triggered by costimulation during T cell activation.Science 282:2266]2269

15. Dustin ML, Ferguson LM, Chan P, Springer TA, Golan DEŽ .1996 Visualization of the CD2 interaction with LFA-3 anddetermination of the two-dimensional dissociation constantfor adhesion receptors in a contact area. J Cell Biol 132:465]474

Ž .16. Norcross MA 1984 A synaptic basis for T-lymphocyte acti-Ž .vation. Ann Immunol Paris 135D:113]134

17. Negulescu PA, Krasieva TB, Khan A, Kerschbaum HH, Ca-Ž .halan MD 1996 Polarity of T cell shape, motility, and

sensitivity to antigen. Immunity 4:421]430Ž .18. Donnadieu E, Bismuth G, Trautmann A 1994 Antigen

recognition by helper T cells elicits a sequence of distinctchanges of their shape and intracellular calcium. Curr Biol4:584]595

19. Dustin ML, Bromley SK, Kan Z, Peterson DA, Unanue ERŽ .1997 Antigen receptor engagement delivers a stop signalto migrating T lymphocytes. Proc Natl Acad Sci USA 94:3909]3913

Ž .20. Delon J, Bercovici N, Liblau R, Trautmann A 1998 Imag-ing antigen recognition by naive CD4q T cells: compulsorycytoskeletal alterations for the triggering of an intracellularcalcium response. Eur J Immunol 28:716]729

21. Delon J, Bercovici N, Raposo G, Liblau R, Trautmann AŽ . 2q1998 Antigen-dependent and -independent Ca respon-

ses triggered in T cells by dendritic cells compared with Bcells. J Exp Med 188:1473]1484

Ž .22. Wulfing C et al 1997 Kinetics and extent of T cell activa-tion as measured with the calcium signal. J Exp Med185:1815]1825

Ž .23. Montes M, McIlroy D, Hosmalin A, Trautmann A 1999

Calcium responses elicited in human T cells and dendriticcells by cell]cell interaction and soluble ligands. Int Im-munol 11:561]568

Ž .24. Steinman RM 1991 The dendritic cell system and its rolein immunogenicity. Annu Rev Immunol 9:271]296

Ž .25. Banchereau J, Steinman RM 1998 Dendritic cells and thecontrol of immunity. Nature 392:245]252

Ž .26. Reich Z et al 1997 Ligand-specific oligomerization of T-cellreceptor molecules. Nature 387:617]620

27. Valitutti S, Muller S, Cella M, Padovan E, Lanzavecchia A¨Ž .1995 Serial triggering of many T-cell receptors by a fewpeptide]MHC complexes. Nature 375:148]151

Ž .28. Beeson C et al 1996 Early biochemical signals arise fromlow affinity TCR]ligand reactions at the cell]cell interface. JExp Med 184:777]782

Ž .29. Barclay AN et al 1997 The leucocyte antigen factsbook.Academic Press, London

Ž .30. Springer TA 1990 Adhesion receptors of the immunesystem. Nature 346:425]434

Ž .31. Barclay AN et al 1993 The leucocyte antigen factsbook.Academic Press, London

Ž .32. Dustin ML et al 1997 Low affinity interaction of human orrat cell adhesion molecule CD2 with ligands aligns adheringmembranes to achieve high physiological affinity. J BiolChem 272:30889]30898

Ž .33. Dustin ML et al 1998 A novel adaptor protein orchestratesreceptor patterning and cytoskeletal polarity in T cell con-tacts. Cell 94:667]677

Ž .34. Singer SJ 1992 Intercellular communication and cell]celladhesion. Science 255:1671]1677

Ž .35. Karjalainen K 1994 High sensitivity, low affinity}paradoxof T-cell receptor recognition. Curr Opin Immunol 6:9]12

36. van der Merwe PA, McNamee PN, Davies EA, Barclay AN,Ž .Davis SJ 1995 Topology of the CD2]CD48 cell-adhesion

molecule complex: implications for antigen recognition byT cells. Curr Biol 5:74]84

Ž .37. Wild MK et al 1999 Dependence of T cell antigen recogni-tion on the dimensions of an accessory receptor]ligandcomplex. J Exp Med 190:31]41

Ž .38. Manjunath N, Correa M, Ardman M, Ardman B 1995Negative regulation of T-cell adhesion and activation byCD43. Nature 377:535]538

Ž .39. Davis SJ, van der Merwe PA 1996 Structure and ligandinteractions of CD2: implications for T-cell function. Im-munol Today 17:177]187

Ž .40. Shaw AS, Dustin ML 1997 Making the T cell receptor gothe distance: a topological view of T cell activation. Immu-nity 6:361]369

Ž .41. Jones EY, Davis SJ, Williams AF, Harlos K, Stuart DA 1992Crystal structure of a soluble form of the cell adhesion

˚molecule CD2 at 2.8 A. Nature 360:232]239Ž .42. Bodian DL et al 1994 Crystal structure of the extracellular

˚region of the human cell adhesion molecule CD2 at 2.5 Aresolution. Structure 2:755]766

Ž .43. Ikemizu S et al 1999 Crystal structure of the CD2-bindingŽ .domain of CD58 lymphocyte function-associated antigen 3

at 1.8-A resolution. Proc Natl Acad Sci USA 96:4289]4294Ž .44. Wang JH et al 1999 Structure of a heterophilic adhesion

Ž .complex between the human CD2 and CD58 LFA-3 coun-terreceptors. Cell 97:791]803

Ž .45. Jacobson K, Ishihara A, Inman R 1987 Lateral diffusion ofproteins in membranes. Annu Rev Physiol 49:163]175

Ž .46. Liu SJ, Hahn WC, Bierer BE, Golan DE 1995 Intracellular

mediators regulate CD2 lateral diffusion and cytoplasmiccalcium ion mobilization upon CD2-mediated T cell activa-tion. Biophys J 68:459]470

Ž .47. Saxton MJ, Jacobson K 1997 Single-particle tracking: appli-cations to membrane dynamics. Annu Rev Biophys BiomolStruct 26:373]399

19

Redistribution of T cell surface molecules

Ž .48. Caplan S, Zeliger S, Wang L, Baniyash M 1995 Cell-surface-expressed T-cell antigen]receptor zeta chain is asso-ciated with the cytoskeleton. Proc Natl Acad Sci USA92:4768]4772

Ž .49. Rozdzial MM, Malissen B, Finkel TH 1995 Tyrosine-phos-phorylated T cell receptor zeta chain associates with theactin cytoskeleton upon activation of mature T lymphocytes.Immunity 3:623]633

Ž .50. Jacobson K, Sheets ED, Simson R 1995 Revisiting the fluidmosaic model of membranes. Science 268:1441]1442

Ž .51. Kabouridis PS, Magee AI, Ley SC 1997 S-acylation of LCKprotein tyrosine kinase is essential for its signalling functionin T lymphocytes. EMBO J 16:4983]4998

Ž .52. Linsley PS et al 1996 Intracellular trafficking of CTLA-4and focal localization towards sites of TCR engagement.Immunity 4:535]543

Ž .53. Griffiths GM 1995 The cell biology of CTL killing. CurrOpin Immunol 7:343]348

Ž .54. Kupfer A, Mosmann TR, Kupfer H 1991 Polarized expres-sion of cytokines in cell conjugates of helper T cells andsplenic B cells. Proc Natl Acad Sci USA 88:775]779

Ž .55. Kupfer H, Monks CR, Kupfer A 1994 Small splenic B cellsŽ .that bind to antigen specific T helper Th cells and face the

site of cytokine production in the Th cells selectively prolif-erate: immunofluorescence microscopic studies of Th-Bantigen-presenting cell interactions. J Exp Med179:1507]1515

Ž .56. Bossi G, Griffiths GM 1999 Degranulation plays an essen-tial part in regulating cell surface expression of Fas ligand inT cells and natural killer cells. Nat Med 5:90]96

Ž .57. Viola A, Schroeder S, Sakakibara Y, Lanzavecchia A 1999 Tlymphocyte costimulation mediated by reorganization ofmembrane microdomains. Science 283:680]682

Ž .58. Bachmann MF et al 1999 Developmental regulation of Lcktargeting to the CD8 coreceptor controls signaling in naiveand memory T cells. J Exp Med 189:1521]1530

Ž .59. Brown DA, Rose JK 1992 Sorting of GPI-anchored proteinsto glycolipid enriched membrane subdomains duringtransport to the apical cell surface. Cell 68:533]544

Ž .60. Stowers L, Yelon D, Berg LJ, Chant J 1995 Regulation ofthe polarization of T cells toward antigen-presenting cells byRas-related GTPase CDC42. Proc Natl Acad Sci USA 92:5027]5031

Ž .61. Lowin-Kropf B, Shapiro VS, Weiss A 1998 Cytoskeletalpolarization of T cells is regulated by an immunoreceptortyrosine-based activation motif-dependent mechanism. J CellBiol 140:861]871

Ž .62. Lin CH, Thompson CA, Forscher P 1994 Cytoskeletalreorganization underlying growth cone motility. Curr OpinNeurobiol 4:640]647

Ž .63. Simons K, Ikonen E 1997 Functional rafts in cell mem-branes. Nature 387:569]572

Ž .64. Xavier R, Brennan T, Li Q, McCormack C, Seed B 1998Membrane compartmentation is required for efficient T cellactivation. Immunity 8:723]732

Ž .65. Montixi C et al 1998 Engagement of T cell receptortriggers its recruitment to low density detergent-insolublemembrane domains. EMBO J 17:5334]5348

Ž .66. Moran M, Miceli MC 1998 Engagement of GPI-linkedCD48 contributes to TCR signals and cytoskeletal reorgani-zation: a role for lipid rafts in T cell activation. Immunity9:787]796

Ž .67. Rodgers W, Rose JK 1996 Exclusion of CD45 inhibits

activity of p56lck associated with glycolipid-enriched mem-brane domains. J Cell Biol 135:1515]1523

Ž .68. Stulnig TM et al 1998 Polyunsaturated fatty acids inhibit Tcell signal transduction by modification of detergent-insolu-ble membrane domains. J Cell Biol 143:637]644

Ž .69. Zhang W, Trible RP, Samelson LE 1998 LAT palmitoyla-

P. Anton van der Merwe et al

tion: its essential role in membrane microdomain targetingand tyrosine phosphorylation during T cell activation. Im-munity 9:239]246

Ž .70. Sedwick CE et al 1999 TCR, LFA-1, and CD28 play uniqueand complementary roles in signaling T cell cytoskeletalreorganization. J Immunol 162:1367]1375

Ž .71. Giancotti FG, Ruoslahti E 1999 Integrin signaling. Science285:1028]1033

Ž .72. Reif K, Cantrell DA 1998 Networking Rho family GTPasesin lymphocytes. Immunity 8:395]401

Ž .73. Fischer KD, Tedford K, Penninger JM 1998 Vav linksantigen]receptor signaling to the actin cytoskeleton. SeminImmunol 10:317]327

Ž .74. Hall A 1998 Rho GTPases and the actin cytoskeleton.Science 279:509]514

Ž .75. Fischer KD et al 1998 Vav is a regulator of cytoskeletalreorganization mediated by the T-cell receptor. Curr Biol8:554]562

Ž .76. Holsinger LJ et al 1998 Defects in actin-cap formation inVav-deficient mice implicate an actin requirement for lym-phocyte signal transduction. Curr Biol 8:563]572

Ž .77. Kong YY et al 1998 Vav regulates peptide-specific apoptosisin thymocytes. J Exp Med 188:2099]2111

Ž .78. Maekawa M et al 1999 Signaling from rho to the actincytoskeleton through protein kinases ROCK and LIM-kinase.Science

Ž .79. Snapper SB, Rosen FS 1999 The Wiskott-Aldrich SyndromeŽ .Protein WASP : roles in signaling and cytoskeletal organiza-

tion. Annu Rev Immunol 17:905]929Ž .80. Ley SC et al 1994 Tyrosine phosphorylation of alpha tubulin

in human T lymphocytes. Eur J Immunol 24:99]106Ž .81. Fernandez JA et al 1999 Phosphorylation- and activation-

independent association of the tyrosine kinase Syk and thetyrosine kinase substrates Cbl and Vav with tubulin in B-cells.J Biol Chem 274:1401]1406

Ž .82. Burn P, Kupfer A, Singer SJ 1988 Dynamic membrane-eytoskeletal interactions: specific association of integrin andtalin arises in vivo after phorbol ester treatment of periph-eral blood lymphocytes. Proc Natl Acad Sci USA 85:497]501

Ž .83. Offringa R, Bierer BE 1993 Association of CD2 with tubulin.Evidence for a role of the cytoskeleton in T cell activation. JBiol Chem 268:4979]4988

Ž .84. Lokeshwar VB, Bourguignon LY 1992 Tyrosine phos-Ž .phatase activity of lymphoma CD45 GP180 is regulated by

a direct interaction with the cytoskeleton. J Biol Chem267:21551]21557

Ž .85. Bretscher A 1999 Regulation of cortical structure by theezrin]radixin]moesin protein family. Curr Opin Cell Biol11:109]116

Ž .86. Shih NY et al 1999 Congenital nephrotic syndrome in micelacking CD2AP. Science 286:312]314.

87. Valitutti S, Dessing M, Aktories K, Gallati H, Lanzavecchia AŽ .1995 Sustained signalling leading to T cell activation re-sults from prolonged T cell receptor occupancy. Role of theT cell actin cytoskeleton. J Exp Med 181:577]584

Ž .88. Germain RN 1997 The importance of receptor clustering.Curr Biol 7:R640]R644

Ž .89. Davis MM et al 1998 Ligand recognition by alpha beta Tcell receptors. Annu Rev Immunol 16:523]544

Ž .90. Janeway CA 1995 Ligands for the T-cell receptor: hardtimes for avidity models. Immunol Today 16:223]225

91. Ding YH, Baker BM, Garboczi DN, Biddison WE, Wiley DCŽ .1999 Four A6-TCRrpeptiderHLA-A2 structures that gen-

erate very different T cell signals are nearly identical. Immu-nity 11:45]56

Ž .92. Lanzavecchia A, Lezzi G, Viola A 1999 From TCR engage-ment to T cell activation: a kinetic view of T cell behavior.Cell 96:1]4

Ž .93. Thomas ML 1999 The regulation of antigen]receptor sig-

20

naling by protein tyrosine phosphatases: a hole in the story.Curr Opin Immunol 11:270]276

Ž .94. McKeithan TW 1995 Kinetic proofreading in T-cell recep-tor signal transduction. Proc Natl Acad Sci USA 92:5042]5046

95. Rabinowitz JD, Beeson C, Lyons DS, Davis MM, McConnellŽ .HM 1996 Kinetic discrimination in T-cell activation. Proc

Natl Acad Sci USA 93:1401]1405Ž .96. Klemm JD, Schreiber SL, Crabtree GR 1998 Dimerization

as a regulatory mechanism in signal transduction. Annu RevImmunol 16:569]592

Ž .97. Boniface JJ et al 1998 Initiation of signal transductionthrough the T cell receptor requires the multivalent engage-ment of peptiderMHC ligands. Immunity 9:459]466

Ž .98. Viola A, Lanzavecchia A 1996 T cell activation determinedby T cell receptor number and tunable thresholds. Science273:104]106

Ž .99. Garcia KC, Teyton L, Wilson IA 1999 Structural basis of Tcell recognition. Annu Rev Immunol 17:369]397

Ž .100. Delon J et al 1998 CD8 expression allows T cell signalingby monomeric peptide]MHC complexes. Immunity 9:467]473

101. Locksley RM, Reiner SL, Hatam F, Littman DR, Killeen NŽ .1993 Helper T cells without CD4: control of Leishmaniasisin CD4-deficient mice. Science 261:1448]1451

Ž .102. Schilham MW et al 1993 Alloreactive cytotoxic T cells candevelop and function in mice lacking both CD4 and CD8.Eur J Immunol 23:1299]1304

Ž .103. Remy I, Wilson IA, Michnick SW 1999 Erythropoietinreceptor activation by a ligand-induced conformationchange. Science 283:990]993

104. O’Shea JJ, McVicar DW, Bailey TL, Burns C, Smyth MJŽ .1992 Activation of human peripheral blood T lymphocytesby pharmacological induction of protein-tyrosine phospho-rylation. Proc Natl Acad Sci USA 89:10306]10310

105. Secrist JP, Burns LA, Karnitz L, Koretzky GA, Abraham RTŽ .1993 Stimulatory effects of the protein tyrosine phos-phatase inhibitor, pervanadate, on T-cell activation events. JBiol Chem 268:5886]5893

Ž .106. Mustelin T et al 1995 Regulation of the p70zap tyrosineprotein kinase in T cells by the CD45 phosphotyrosinephosphatase. Eur J Immunol 25:942]946

Ž .107. Jin YJ, Friedman J, Burakoff SJ 1998 Regulation of tyrosinephosphorylation in isolated T cell membrane by inhibitionof protein tyrosine phosphatases. J Immunol 161:1743]1750

Ž .108. Trowbridge IS, Thomas ML 1994 CD45}an emergingrole as a protein-tyrosine-phosphatase required for lympho-cyte-activation and development. Annu Rev Immunol 12:85]116

Ž .109. Furukawa T, Itoh M, Krueger NX, Streuli M, Saito H 1994Specific interaction of the CD45 protein-tyrosine phos-phatase with tyrosine-phosphorylated CD3 zeta chain. ProcNatl Acad Sci USA 91:10928]10932

Ž .110. Leitenberg D, Boutin Y, Lu DD, Bottomly K 1999Biochemical association of CD45 with the T cell receptorcomplex: regulation by CD45 isoform and during T cellactivation. Immunity 10:701]711

Ž .111. Motoya S et al 1999 Interaction between CD45-AP andprotein-tyrosine kinases involved in T cell receptor signal-ing. J Biol Chem 274:1407]1414

112. Veillette A, Soussou D, Latour S, Davidson D, Gervais FGŽ .1999 Interactions of CD45-associated protein with the anti-gen receptor signaling machinery in T-lymphocytes. J Biol

Chem 274:14392]14399

Ž .113. D’Oro U, Sakaguchi K, Appella E, Ashwell JD 1996 Muta-tional analysis of Lck in CD45-negative T cells: dominantrole of tyrosine 394 phosphorylation in kinase activity. MolCell Biol 16:4996]5003

Ž .114. Tangye SG et al 1998 Negative regulation of human T cell

activation by the receptor-type protein tyrosine phosphataseCD148. J Immunol 161:3803]3807

Ž .115. Kersh GJ, Allen PM 1996 Essential flexibility in T-cellrecognition of antigen. Nature 380:495]498

Ž .116. Weiss A, Littman DR 1994 Signal transduction by lympho-cyte antigen receptors. Cell 76:263]274

Ž .117. Kersh GJ, Kersh EN, Fremont DH, Allen PM 1998 High-and low-potency ligands with similar affinities for the TCR:the importance of kinetics in TCR signaling. Immunity9:817]826

Ž .118. Sloane-Lancaster J, Shaw AS, Rothbard JB, Allen PA 1994Partial T cell signalling: altered phospho-z and lack ofZap70 recruitment in APL-induced T cell anergy. Cell79:913]922

Ž .119. Madrenas J et al 1995 z phosphorylation without ZAP-70activation induced by TCR antagonists or partial agonists.Science 267:515]518

Ž .120. Wallace VA et al 1997 Alterations in the level of CD45

surface expression affect the outcome of thymic selection. JImmunol 158:3205]3214

Ž .121. Gervais FG, Veillette A 1997 Reconstitution of interactionsbetween protein-tyrosine phosphatase CD45 and tyrosine-

Ž .protein kinase p56 lck in nonlymphoid cells. J Biol Chem272:12754]12761

21

Redistribution of T cell surface molecules

Ž .122. Thompson CB, Allison JP 1997 The emerging role ofCTLA-4 as an immune attenuator. Immunity 7:445]450

Ž .123. Long EO 1999 Regulation of immune responses throughinhibitory receptors. Annu Rev Immunol 17:875]904

Ž .124. Marengere LEM et al 1996 Regulation of T cell receptor´signalling by tyrosine phosphatase SYP association withCTLA-4. Science 272:1170]1173

Ž .125. Lee KM et al 1998 Molecular basis of T cell inactivation byCTLA-4. Science 282:2263]2266

Ž .126. Germain RN, Stefanova I 1999 The dynamics of T cellreceptor signaling: complex orchestration and the key rolesof tempo and cooperation. Annu Rev Immunol 17:467]522

Ž .127. Goldrath AW, Bevan MJ 1999 Low-affinity ligands for theTCR drive proliferation of mature CD8q T cells in lym-phopenic hosts. Immunity 11:183]190

Ž .128. Ernst B, Lee D, Chang JM, Sprent J, Surh CD 1999 Thepeptide ligands mediating positive selection in the thymuscontrol T cell survival and homeostatic proliferation in the

periphery. Immunity 11:173]181

Ž .129. Crabtree GR 1989 Contingent genetic regulatory events inT lymphocyte activation. Science 243:355]361

Ž .130. Janeway CA Jr 1989 Approaching the asymptote? Evolutionand revolution in immunology. Cold Spring Harb SympQuant Biol 54:1]13