Nonstimulatory peptides contribute to antigen-induced CD8–T cell receptor interaction at theimmunological synapse

Pia P Yachi1, Jeanette Ampudia1, Nicholas R J Gascoigne1 & Tomasz Zal1,2

It is unclear if the interaction between CD8 and the T cell receptor (TCR)–CD3 complex is constitutive or antigen induced. Here,

fluorescence resonance energy transfer microscopy between fluorescent chimeras of CD3f and CD8b showed that this interaction

was induced by antigen recognition in the immunological synapse. Nonstimulatory endogenous or exogenous peptides presented

simultaneously with antigenic peptides increased the CD8-TCR interaction. This finding indicates that the interaction between

the intracellular regions of a TCR-CD3 complex recognizing its cognate peptide–major histocompatibility complex (MHC)

antigen, and CD8 (plus the kinase Lck), is enhanced by a noncognate CD8-MHC interaction. Thus, the interaction of CD8 with a

nonstimulatory peptide-MHC complex helps mediate T cell recognition of antigen, improving the coreceptor function of CD8.

The ab T cell receptor (TCR) is responsible for the affinity andspecificity of antigen recognition1,2, whereas the coreceptors CD8 andCD4 enhance the sensitivity of TCR recognition3. Disruption ofinteractions between the coreceptor and peptide–major histocompat-ibility complex (pMHC) inhibits or changes the quality of the T cellresponse3–5. Coreceptors act in two main ways. First, they bind tononpolymorphic regions of the MHC4. This can aid in adhesion, butthe main function is generally believed to be increasing the sensitivityof T cell activation through the entropic facilitation of TCR-pMHCbinding rather than through energetic stabilization of the trimolecularcomplex6–9. Second, CD4 and CD8 are associated with the kinase Lck.Coreceptor binding to pMHC recruits Lck close to the TCR, enablingit to phosphorylate components of the signaling complex of the TCR(CD3), thus enhancing signal transduction3.

There are conflicting data on the interaction between the TCR-CD3complex and coreceptors, in particular regarding whether any inter-action between these molecules is constitutive or is induced by antigenrecognition. Various coimmunoprecipitation and flow cytometryfluorescence resonance energy transfer (FRET) experiments havesuggested constitutive interaction between some TCR-CD3 complexesand coreceptors in unstimulated T cells7,10–13. Others have showninteraction only after T cell activation14–17. Cocapping and comodula-tion experiments also support the idea of an interaction of coreceptorwith TCR18–21. CD8ab and CD4 reside on glycolipid-enriched micro-domains or rafts, whereas TCR association with rafts is greatlyincreased after stimulation22,23. It is therefore questionable whetherall these assays measure direct interaction or simply colocalization ofthe molecules on the same rafts.

Whether the interaction between CD8 and TCR is constitutive orantigen induced has important consequences for the function of CD8.Constitutive association indicates that CD8 acts as a ‘universalamplifier’ in antigen recognition. Inducible interaction suggests anextra level of fine tuning whereby CD8 could sharpen and amplify thesensitivity and specificity of recognition. To definitively addresswhether CD8 and TCR interact and to what extent the interactionis induced after antigen recognition, we have used microscopy tomeasure FRET between CD8b–yellow fluorescent protein (YFP) andCD3z–cyan fluorescent protein (CFP) fusion proteins during antigenrecognition in live and fixed OT-I T hybridoma cells. FRET can beused to address the question of proximity between molecules, becauseits effective range is less than 10 nm.

T cells acquire a polarized morphology after antigen recognitionin which certain surface and intracellular molecules become concen-trated into the contact area between the T cell and antigen-presentingcell (APC) known as the immunological synapse9,24. This is a dynamicstructure whose exact function in T cell activation is controversial,although it is without doubt the site of antigen recognitionand signaling. The coreceptor is brought into the synapse quicklyafter T cell–APC contact25–27. CD8–MHC class I interaction isabsolutely required for synapse formation, because an alteration inthe CD8-binding site of MHC class I renders T cells unable to formT cell–APC conjugates28.

FRET experiments done by flow cytometry cannot relate molecularinteractions between TCR and coreceptors to the formation of theimmunological synapse. Microscopic evaluation of FRET betweenfluorescent chimeric proteins therefore has a great advantage in that

Published online 26 June 2005; doi:10.1038/ni1220

1Department of Immunology, IMM1, The Scripps Research Institute, La Jolla, California 92037, USA. 2Present address: Department of Immunology, MD Anderson CancerCenter, University of Texas, Houston, Texas 77030, USA. Correspondence should be addressed to N.R.J.G. (gascoigne@scripps.edu).

NATURE IMMUNOLOGY VOLUME 6 NUMBER 8 AUGUST 2005 785

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

it allows spatiotemporal localization ofmolecular interactions with subcellular reso-lution in living cells. This makes it possibleto determine whether the molecular interac-tion between TCR and CD8 drives CD8recruitment to the synapse or if the interac-tion is induced in the synapse. FRET micro-scopy experiments using living cells haveshown that interaction between fluorescentchimeras of CD4 and CD3z is stronglyinduced by antigen recognition and occursonly in the synapse26.

Certain APCs support non-antigen-specificrecruitment of coreceptor to the synapse bypMHC complexes bearing nonstimulatoryor even antagonist peptide9,26,29. Nonstimu-latory as well as antigenic peptide–MHC classII complexes can contribute to T cell activa-tion and synapse formation30,31. This effecthas been noted at low antigen concentrationand is not caused by antagonist pMHCligands30. It involves interaction between theTCR and the endogenous peptide–MHCclass II complex31.

Using FRET microscopy, we show here thatinteraction between CD3z and CD8b wastransiently induced at the synapse betweena T cell and an APC loaded with agonistpeptide but not by a nonstimulatory peptide.The presence of nonstimulatory peptidesenhanced antigen recognition and increasedCD8-CD3z interaction. Thus, most endoge-nous pMHC complexes enhance recognitionof the rare antigenic pMHC complex throughnoncognate interaction between CD8 andMHC class I.

RESULTS

Biological activity of fluorescent chimeric proteins

To study the interaction between the TCR and coreceptors duringantigen recognition in living cells, we developed chimeras of CD3z-CFP and CD8b-YFP, expressing them in a T cell hybridoma. FRETbetween CFP and YFP ‘reports’ when these molecules are broughtwithin 10 nm of each other26,32. The OT-I hybridoma recognizes H-2Kb with an ovalbumin peptide (OVA)2 and its TCR binds to H-2Kb

bound to OVA (H-2Kb–OVA) with relatively high affinity comparedwith the binding of nonstimulatory peptides for this TCR, such as apeptide derived from vesicular stomatitis virus nucleoprotein (VSV)33.We fused the C termini of CD8b and CD3z with YFP or CFP,respectively, using peptide linkers to allow proper folding and flex-ibility. We retrovirally transduced the chimeric genes encoding CD8b-YFP and CD3z-CFP into OT-I hybridomas34 expressing wild-typeCD8a to obtain stable transfectants (OT-I.ZC.8bY) with surface CD8bexpression similar to that of CD8+ splenocytes (staining intensity withantibody to CD8b (anti-CD8b): 738 7 200 versus 622 7 136 abovebackground, respectively). CD8b must pair with CD8a for expressionon the cell surface35. CD8b-YFP did not reach the surfaces of cellslacking CD8a (data not shown) but was expressed on the surfaces ofcells expressing wild-type CD8a (Fig. 1a). Anti-CD8a (or anti-CD8b)staining showed complete colocalization and cocapping with CD8b-YFP (Fig. 1b), confirming that CD8b-YFP formed a complex with

CD8a. Release of interleukin 2 was undetectable in OT-I hybridomaslacking CD8. Transfection of wild-type CD8a partially restored anti-gen reactivity, whereas CD8b-YFP plus wild-type CD8a producedresponsiveness that was 1,000-fold stronger than that produced bywild-type CD8a alone, similar to that of wild-type CD8ab (Fig. 1c),showing that CD8b-YFP retained the coreceptor properties of CD8b.Immunoprecipitation with anti-CD8b coprecipitated Lck (Fig. 1d),showing that the YFP moiety did not interfere with binding of Lck tothe cytoplasmic domain of CD8a. This also confirmed the CD8a–CD8b-YFP interaction, because Lck is associated with CD8a. There-fore, the CD8b-YFP was biologically functional.

Antigen-dependent and antigen-independent cell coupling

We quantified the ability of antigenic or nonstimulatory peptides toinduce stable conjugate formation to assess the contribution ofantigen recognition to the formation of stable conjugates26,29. Forthis we used RMA-S cells, which are Tap2-deficient mouse tumor cellsthat can be used to express H-2Kb or H-2Db in complex withexogenously added peptides in the absence of endogenous pMHC.We labeled RMA-S cells with indodicarbocyanine (Cy5) and ‘loaded’them with OVA or VSV peptides at concentrations that caused equalexpression of stabilized H-2Kb on the cell surface36 (Fig. 2a). Weincubated OT-I.ZC.8bY cells with RMA-S cells at 37 1C. At varioustime points, we pipetted the cells to separate any weakly conjugatedcells and fixed the cells. We assessed by flow cytometry the percentageof OT-I.ZC.8bY cells forming conjugates with RMA-S cells (Fig. 2b,c).

140

0100 101 102 103 104 106

0

2

4

6

8

10

12

107 108 109 1010 1011 1012 0

No CD8CD8αCD8α + CD8βCD8α + CD8β-YFP

CD8β-YFP

αCD8β

αCD8α

IP CD8β

1(kDa)120

85605040

25

2 3 4 5 6 7 8 9

p56Lck

Cell lysates

Streptavidin-APC

Anti-CD8β (fluorescence)

Cel

ls

IL-2

-dep

ende

nt C

TLL

prol

ifera

tion

(×10

4 c.

p.m

.)

OVA (pM)

a

b

c

d

Figure 1 Function of fluorescent chimeric proteins. (a) Flow cytometry of OT-I hybridomas with

wild-type CD8a (broken line) or with wild-type CD8a plus CD8b-YFP (solid line), stained with

anti-CD8b. (b) OT-I.ZC.8bY cells labeled with biotinylated anti-CD8a (aCD8a) or anti-CD8b (aCD8b)and crosslinked with streptavidin-allophycocyanin. Cells were imaged for YFP or allophycocyanin

fluorescence to demonstrate the relative positions of the CD8b-YFP chimera and the proteins that were

stained and crosslinked by the antibodies. Original magnification, �63. (c) Proliferation of the IL-2-

dependent T cell line CTLL cultured with supernatants of OT-I hybridomas lacking CD8 (No CD8) or

expressing wild-type CD8a, wild-type CD8a plus CD8b-YFP, or wild-type CD8a plus wild-type CD8b.

The OT-I hybridomas had been cultured with irradiated B6 splenocytes loaded with titrated amounts of

OVA for 24 h (details, Supplementary Methods online). (d) Interaction of p56Lck with CD8b versus total

p56Lck expression in OT-I hybridomas lacking CD8 (lanes 5 and 9) or with wild-type CD8a (lanes 4 and

8), wild-type CD8a plus wild-type CD8b (lanes 3 and 7) or wild-type CD8a plus CD8b-YFP (lanes 2

and 6), immunoprecipitated with anti-CD8b (IP CD8b; lanes 1–5; lane 1, anti-CD8b alone) and

analyzed by immunoblot with anti-Lck, detected with anti-MigG–horseradish peroxidase. Lanes 6–9,

cell lysates without immunoprecipitation. Data are representative of three or more experiments except

d, which was done twice.

786 VOLUME 6 NUMBER 8 AUGUST 2005 NATURE IMMUNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

In the absence of exogenously added peptides, less than 1% of theOT-I.ZC.8bY cells formed conjugates with RMA-S cells (Fig. 2b).However, when we sorted these few putative conjugates and visualizedthem with a microscope, 90% were single cells (data not shown).Therefore, RMA-S cells without exogenously added peptides rarelyformed any stable conjugates with OT-I.ZC.8bY cells.

T cell–APC conjugate formation with RMA-S cells loaded with thenonstimulatory peptide VSV was greater than that of RMA-S withoutpeptide (Fig. 2b). However conjugate formation was much greaterwhen the RMA-S cells expressed antigenic H-2Kb–OVA (Fig. 2b).Thus, conjugate formation was strongly induced by an antigenrecognition–dependent mechanism. The weak but distinct activity ofH-2Kb–VSV suggested that nonstimulatory (endogenous) pMHCcomplexes could be important for initial formation of T cell–APCconjugates. We sorted and imaged the conjugates and grouped themaccording to whether they had recruited to the T cell–APC interfaceCD8b-YFP with or without CD3z-CFP. About 70% of T cells inconjugates recruited CD8b-YFP regardless of the peptide (Fig. 2d).This was in contrast to CD3z-CFP recruitment, which occurredmainly in cells forming conjugates with OVA-loaded APCs,peaking around 15 min (Fig. 2d; representative images, Supplemen-tary Fig. 1 online).

CD8 recruitment is not peptide specific

We assessed the relationship between CD8 recruitment and pMHCdensity by analyzing the synapses formed between OT-I.ZC.8bY cellsand RMA-S cells expressing H-2Kb–OVA or H-2Kb–VSV. We com-pared the intensity of CD8b-YFP in the contact area with CD8b-YFPexpression in the rest of the membrane (Fig. 3a). We quantifiedpMHC molecule expression on RMA-S cells by comparing anti-H-2Kb–phycoerythrin staining with that of reference beads loaded withknown quantities of phycoerythrin. When we added RMA-S cells withequally high expression of H-2Kb–OVA or H-2Kb–VSV (26,000molecules/cell) to the OT-I.ZC.8bY cells, a much smaller percentageof T cells formed conjugates with VSV (Fig. 2b). However, in T cell–APC conjugates, we noted the same 2.5-fold increase of CD8 in the

synapse relative to that in the rest of the cell membrane with bothOVA and VSV. Therefore, conjugate formation depends on an earlysignal that is infrequently generated in the absence of antigen. Theequivalent ability of nonstimulatory and agonist peptides to induceCD8b-YFP clustering showed that CD8 recruitment itself was notpurely an antigen-dependent process but instead was driven by theCD8–MHC class I interaction. When we ‘titrated’ H-2Kb–OVA(Fig. 3a) or H-2Kb–VSV (Supplementary Fig. 2 online), CD8b-YFPrecruitment decreased. Thus, the amount of CD8 recruitment dependson MHC class I density, whereas the cell-to-cell frequency of T cell–APC conjugate formation is antigen specific.

In contrast, TCR-CD3z recruitment to the synapse is antigendependent, as shown before24,26,37 (Figs. 2d and 3b). In the presenceof antigenic pMHC, CD3z expression in the synapse was increased

Figure 2 T cell–APC conjugate formation and

CD3z and CD8b recruitment to the synapse.

(a) H-2Kb density of RMA-S cells loaded with

OVA peptide (36 mM; gray shading), VSV peptide

(80 mM; solid line) or no peptide (dotted line).

These concentrations had been determined by

titration to produce equal H-2Kb density, as

shown by staining with anti-H-2Kb. (b) Timecourse of conjugate formation with antigen and

nonstimulatory pMHC complexes. OT-I.ZC.8bY

cells were allowed to interact with Cy5-labeled

RMA-S cells expressing the same MHC density

of OVA or VSV peptide or no peptide. At various

time points, cells were fixed and conjugate

formation was assessed by flow cytometry.

(c) Representative dot plot of the flow cytometry

in b. Results in b,c are representative of three

independent experiments. (d) Proportion of

conjugates that had recruited CD8b-YFP or

CD3z�CFP at various time points. Conjugates

were sorted by flow cytometry and visualized by

microscopy and randomized samples were

categorized into three groups: increased CD8b-

YFP and CD3z-CFP in synapse compared with the

rest of the cell surface (gray bars); increased CD8b-YFP only (filled bars); or no increase in either protein (open bars). 0 min, cells before conjugate

formation. The results are averages of 34–108 cells from two independent experiments. Representative images are in Supplementary Figure 1 online.

350

0

12

10

8

6

4

2

00 10 20 30 40 50

104

103

102

101

100

100 101 102

YFP

0 min 5 min 15 min 30 min 60 min

Cy5

103 104

60 70H-2Kb (log10 fluorescence)Time (min)

OVAVSV

No peptide

T c

ell c

onju

gate

s (%

)T

cel

l con

juga

tes

(%)

Cel

ls

100

80

60

40

20

0

OVAVSV

OVAVSV

OVAVSV

OVAVSV

OVAVSV

a b

c d

3

2.6

2.2

1.8

1.4

2

1.8

1.6

1.4

1.2

1126,000OVA

'Fol

d in

crea

se' C

D8β

-YF

P

'Fol

d in

crea

se' C

D3ζ

-CF

P

26,000VSV

7,300OVA

3,200OVA

26,000OVA

26,000VSV

7,300OVA

H-2Kb molecules/cell H-2K

b molecules/cell

3,200OVA

a b

Figure 3 CD8 clustering to the synapse is MHC driven, whereas CD3zclustering is antigen driven. OT-I.ZC.8bY cells were allowed to interact

with RMA-S cells with equivalent expression of H-2Kb–OVA or H-2Kb–VSVcomplexes or different expression of H-2Kb–OVA. Numbers below bars

indicate the number of pMHC molecules expressed by the RMA-S cells.

OT-I.ZC.8bY cells were assessed for their ‘fold increase’ in CD8b-YFP (a)

or CD3z-CFP (b) in the synapse by comparison of the fluorescence intensity

in the contact area to that in the rest of the cell membrane. Results are

presented as mean + s.e.m. (n 4 10). (a) P o 0.05 for all sample

comparisons except for 26,000 OVA versus 26,000 VSV and 7300 OVA

versus 3200 OVA (P 4 0.05). (b) P o 0.05 for VSV versus all OVA

samples, but not between different OVA samples. Results are representative

of three independent experiments.

NATURE IMMUNOLOGY VOLUME 6 NUMBER 8 AUGUST 2005 787

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

1.5- to 1.7-fold compared with expression in the rest of the cellmembrane. This increased CD3z-CFP was not reduced, as H-2Kb–OVA was reduced (below about 3,000 H-2Kb molecules per cell, toofew T cell–APC conjugates were formed to analyze). The lower CD3z-CFP at the highest concentration of OVA was probably due to faster,stronger TCR downregulation at such high antigen density (discussedbelow). Thus, above a certain threshold, the degree of CD3z recruit-ment was not dependent on MHC density itself.

Antigen-induced CD8b-CD3f interaction

The interaction between two molecules (defined as proximity of lessthan 10 nm) can be assessed by FRET imaging32. After CFP excitation,FRET leads to decreased fluorescence of CFP and increased fluores-cence of YFP as energy is transferred from CFP to YFP. We allowedOT-I.ZC.8bY cells to interact with OVA- or VSV-loaded RMA-S cellsat 37 1C and assessed the interaction of CD3z-CFP and CD8b-YFP byFRET efficiency imaging26,38 (Fig. 4). After stimulation with OVA-loaded cells, both CD3z-CFP and CD8b-YFP were recruited to thesynapse, where the FRET signal between CD8b-YFP and CD3z-CFPincreased (Fig. 4a,b). FRET intensity did not correlate with donor/acceptor ratio (Fig. 4i), YFP intensity (Fig. 4g,h) or CFP intensity(Fig. 4j). Thus, FRET between CD8b-YFP and CD3z-CFP was not dueto nonspecific, diffusion-driven interactions caused by increasedcrowding of the molecules.

We quantified the FRET signal over time(Fig. 4e). To obtain accurate time points andequivalent experimental conditions for FRETanalysis, we fixed the conjugates, resulting inlower FRET efficiency than that of live cells.FRET peaked at 10–12 min after initiation ofstimulation, showing that interaction betweenCD8 and TCR was transiently induced afterantigen recognition. The nonstimulatory pep-tide VSV did not induce an increase in the

FRET signal (Fig. 4c–e). Basal CD3z-CFP expression was high enoughto give a FRET signal (Fig. 4j), indicating that the lack of increasedFRET signal was not caused by insufficient CFP26. A low basal FRETsignal (FRET efficiency of about 2.5%) was detectable on the cellsurface in the absence of any APCs. Whether this was due to lowconstitutive interaction between CD8b and CD3z is unclear. Weassessed TCR downregulation after stimulating OT-I.ZC.8bY cellswith RMA-S cells loaded with OVA or VSV. OVA induced rapidTCR downregulation, whereas this was absent when we used VSV(Fig. 4f). The TCR downregulation and FRET signal downmodulationhad similar kinetics.

Immobilized pMHC induces CD8-CD3f interaction

To study whether the interaction between CD8b and CD3z could beinduced by pMHC complexes alone, we refolded soluble H-2Kb withpeptide in vitro, biotinylated the H-2Kb molecule and bound thecomplex to avidin-coated nonfluorescent beads. Both H-2Kb–VSV–and H-2Kb–OVA–coated beads readily interacted with the T cells at37 1C, causing accumulation of CD8b-YFP at the T cell–bead interface(Fig. 5a,b). CD3z-CFP became concentrated at the interface withH-2Kb–OVA– but not H-2Kb–VSV–coated beads. Recognition ofH-2Kb–OVA caused rapid engulfment of the beads, as noted before39.The FRET signal was increased on the contact interfaces of H-2Kb–OVA but not H-2Kb–VSV beads (Fig. 5). Thus, CD8-CD3z interaction

Merge CD8β-YFP CD3ζ-CFP FRET

14%

0%

14%

0%

7

6

5

4

3

2

1

00

140

120

100

80

60

40

20

00 20 40 60 80 100 120 140

10 20 30 40 50 60 70

Time (min)

Time (min)

TC

R (

%)

OVAVSV

OVAVSV

FR

ET

effi

cien

cy (

%)

*** ****

10987654321010 15 20 25 30 35 40

YFP intensity (×102)

10 15 20 25 30 35 40YFP intensity (×102)

YFP/CFP

CFP intensity (×102)

VSV

OVA

OVA

OVA

FR

ET

effi

cien

cy (

%)

109876543210

FR

ET

effi

cien

cy (

%)

109876543210

2

2 4 6 8 10

3 4 5 6 7 8

FR

ET

effi

cien

cy (

%)

109876543210

FR

ET

effi

cien

cy (

%)

a

f

g

h

i

j

b

c

d

e

Figure 4 Agonist peptide induces colocalization

and interaction between CD8b-YFP and

CD3z-CFP. (a–d) Live OT-I hybridomas expressing

CD3z-CFP or wild-type CD8a plus CD8b-YFP

(OT-I.ZC.8bY) incubated with Cy5-stained APCs

loaded with OVA (a,b) or VSV (c,d). (a,c) Mid-cell

sections. (b,d) En face projections. Far left,

merged images: red, CD8b-YFP; green, CD3z-CFP; blue, Cy5. Far right, FRET efficiency images

(‘scale bar’, right margin). Contact areas were

defined in three dimensions by the T cell–APC

fluorescence overlap. Original magnification,

�63. (e) FRET efficiency of OT-I.ZC.8bY cells

incubated with OVA- or VSV-loaded RMA-S cells

(time, horizontal axis) and fixed. Average FRET

efficiency + s.e.m. (n ¼ 18–38, except OVA

30 min, n ¼ 13; VSV 20 min, n ¼ 11; VSV

30 min, n ¼ 15) was assessed from the synapse.

***, P o 0.00002, and **, P o 0.01, OVA

versus VSV. For VSV, there were too few 5-minute

T cell–APC conjugates to measure. (f) Cell surface

TCR, assessed using anti-Vb5, in cells treated as

in e. (g–j) FRET efficiency of OT-I.ZC.8bY cells

incubated for 12 min with RMA-S cells loaded

with OVA (g,i,j) or VSV (h); FRET efficiency was

assessed from the synapses. (g,h) FRET efficiency

versus YFP intensity. (i) FRET efficiency versusYFP/CFP (molar ratio ¼ intensity ratio/2.2).

(j) FRET efficiency versus CFP intensity.

788 VOLUME 6 NUMBER 8 AUGUST 2005 NATURE IMMUNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

was induced by pMHC complexes alone in the absence of costimu-latory molecules.

Nonstimulatory peptides enhance antigen recognition

Endogenous pMHC complexes accumulate in the synapse between aCD4 T cell and APC in the presence of agonist pMHC. This enhancesT cell activation, indicating that simultaneous recognition of self andforeign peptide on the same APC can affect the response to a foreignligand30. To study how a nonstimulatory peptide influences antigenrecognition and CD8-CD3z interaction, we used monoclonal antibody25-D1.16, which specifically recognizes H-2Kb–OVA40. This enabledcomparison of T cell responses to RMA-S cells with various expressionof H-2Kb–OVA in the presence or absence of additional nonstimula-tory ligand (H-2Kb–VSV; Fig. 6a–c). We ‘titrated’ the OVA peptideand varied the amount of VSV peptide so that the total MHC density

on cells loaded with OVA plus VSV was kept constant. We plottedconjugate formation, TCR downregulation and FRET efficiency as afunction of anti–H-2Kb–OVA fluorescence. Therefore, any differencein the dose-response curves between the two groups was caused by thepresence of the nonstimulatory pMHC. TCR downregulation inresponse to OVA was enhanced by the presence of VSV (Fig. 6a).Similarly, the formation of cell conjugates was greatly enhanced whenthe VSV peptide was also present (Fig. 6b). We obtained similarresults at time points between 6 and 30 min (data not shown).Therefore, very small amounts of agonist peptide were needed toinduce conjugate formation in the presence of excess nonstimulatorypeptides. In the absence of nonstimulatory peptides, very largeamounts of the cognate antigen were required for conjugate forma-tion. Most notably, the nonstimulatory peptide was able to increaseinteraction between CD8 and CD3z, as measured by FRET (Fig. 6c).

Figure 5 Close molecular proximity of CD8band CD3z can be induced by pMHC molecules

alone. (a,b) OT-I.ZC.8bY cells incubated with

H-2Kb–OVA–coated beads (a) or H-2Kb–VSV–

coated beads (b). The beads are not themselves

fluorescent. H-2Kb–OVA induces the recruitment

of both CD8b-YFP and, to a lesser extent,

CD3z-CFP (a), whereas H-2Kb–VSV allows onlythe recruitment of CD8b-YFP to the contact

interface with the bead (b). The cells almost

engulf the beads, as noted before39. The FRET

signal is specifically induced at the interface with H-2Kb–OVA beads but not with H-2Kb–VSV beads. Original magnification, �63. (c) FRET efficiency

(average + s.e.m.) of fixed cells in a,b. n ¼ 36, OVA-coated beads (OVA bead); n ¼ 36, VSV-coated beads (VSV bead); n ¼ 19, OVA-loaded RMA-S cells

(OVA RMA-S); and n ¼ 10, VSV-loaded RMA-S cells (VSV RMA-S). P ¼ 0.0000027, OVA versus VSV bead; P ¼ 0.032, OVA bead versus VSV RMA-S;

P ¼ 0.123, OVA bead versus OVA RMA-S.

5Merge CD8β-YFP CD3ζ-CFP FRET

14%

0%

4

3

FR

ET

effi

cien

cy (

%)

2

1

0 OVAbead

OVARMA-S

VSVbead

VSVRMA-S

a c

b

*

*

* * ****

* *

OVAOVA + VSV

OVAOVA + P815OVA + Mapk1

100

80

60

40

20

0

6

5

4

3

2

1

0

0

1

2

3

4

5

6

7

1 10 102 103 104 1 10 102 103 104

1 10 102 103 104 1 10 102 103 104

1 10 102 103 104 1 10 102 103 104

TC

R e

ndoc

ytos

is (

%)

100

80

60

40

20

0

TC

R e

ndoc

ytos

is (

%)

Con

juga

tes

(%)

109876543210

Con

juga

tes

(%)

FR

ET

effi

cien

cy (

%)

H-2Kb–OVA H-2Kb–OVA

0

1

2

3

4

5

6

7

FR

ET

effi

cien

cy (

%)

a d

b e

c f

Figure 6 Nonstimulatory peptides increase

antigen-induced interaction between CD8 and

TCR. RMA-S cells were loaded with various

amounts of OVA alone or OVA plus VSV (a–c), or

OVA alone, OVA plus P815 or OVA plus Mapk1

(d–f). The amount of H-2Kb–OVA was determined

with anti–H-2Kb–OVA (antibody 25-D1.16

(ref. 40); horizontal axes). The total number

of H-2Kb molecules for the OVA-only group was

estimated from staining with phycoerythrin-labeled anti-H-2Kb (AF6-88.5), as in Figure 3:

26,000 H-2Kb molecules corresponds to 3,515

mean fluorescence intensity (MFI) on the H-2Kb–

OVA scale; 7,300 H-2Kb, to 987 MFI; and 3,200

H-2Kb, to 433 MFI. RMA-S cells loaded with

OVA plus VSV, plus P815 or plus Mapk1-all

expressed about 26,000 H-2Kb molecules.

(a,d) TCR endocytosis by OT-I.ZC.8bY cells

incubated for 1 h with RMA-S cells loaded with

OVA or with OVA plus VSV, plus P815 or plus

Mapk1 and stained with anti-Vb5. (b,e) Flow

cytometry of conjugate formation by OT-I.ZC.8bY

cells incubated for 12 min with Cy5-labeled

RMA-S cells and fixed. (c,f) FRET efficiency

assessed from synapses between OT-I and RMA-S

cells. Data are presented as average FRET

efficiency + s.e.m. (c, n ¼ 17–96; f, n ¼ 12–48;

average n ¼ 30). %, FRET efficiency values withsignificant difference (P o 0.05) for OVA versus

OVA plus VSV (c) or for OVA versus OVA plus

P815 (f). $, significant difference (P o 0.05)

for OVA versus OVA plus Mapk1 (f). For clarity,

not all significant differences are marked.

NATURE IMMUNOLOGY VOLUME 6 NUMBER 8 AUGUST 2005 789

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

We did similar experiments using six endogenous H-2Kb-bindingpeptides that do not activate OT-I T cells41: P815, Mapk1p (Fig. 6d–f),STAT3, Ndufa4, Slc2a3 and Hcph (Supplementary Fig. 3 online). LikeVSV (Figs. 2 and 3), these peptides presented on RMA-S cells in theabsence of OVA, induced more T cell–APC conjugates than did RMA-Scells without peptide and were as active as OVA in recruiting CD8 tothe synapse (Supplementary Fig. 3 online). When used in combina-tion with titrated amounts of OVA, these peptides, like VSV, enhancedconjugate formation, TCR downmodulation and FRET. Thus, theability of a nonstimulatory peptide to increase antigen recognitionwhen presented simultaneously with OVA was not limited to the virus-derived VSV peptide but also occurred with endogenous peptides.Thus, noncognate binding of CD8 to MHC class I molecules bearing anonstimulatory ligand enhanced recognition of antigenic ligands.

DISCUSSION

There is considerable disagreement in the literature about whether theinteraction between CD8 and the TCR-CD3 complex is constitutive oris induced during antigen recognition. Interaction between TCR-CD3and CD8 is detectable by coimmunoprecipitation from unstimulatedcells and therefore has been judged constitutive7,10–13. Other data haveindicated constitutive TCR-CD3-CD8 interaction that is increased byT cell activation15,16. Immunoprecipitation analyses have detectedinteractions between small proportions of the molecular species invol-ved. These may not be representative of most molecules. Our FRETanalysis here showed that there was little interaction, as measured byFRET, between CD3z-CFP and CD8b-YFP on unstimulated T cells.After antigen recognition by the TCR, interaction between CD3z-CFPand CD8b-YFP was triggered in the synapse and there was anincreased FRET signal. Thus, most of the TCR-CD3 and CD8molecules do not interact in the absence of TCR stimulation, althoughwe cannot say that there is absolutely no interaction. The low back-ground FRET signal may represent such an interaction, but even so,the strong TCR-CD3-CD8 interaction induced in the immunologicalsynapse by antigen recognition is notable.

Two published studies used FRET in flow cytometry to investigateTCR-CD8 interactions. In one study, monomeric or tetrameric pMHCinduced FRET between anti-CD3e and anti-CD8 bound to T cells17.In the absence of pMHC, there was no FRET, indicating that theinteraction of TCR-CD3 with CD8 was induced by the interaction ofpMHC with both TCR and CD8. In the other study, FRET betweenanti-CD3e and anti-CD8b with or without saturating amounts ofsoluble monomeric pMHC was interpreted as showing constitutiveTCR-CD3-CD8 interaction13. Potential reasons for the discrepanciesbetween those and our results include the use of antibodies (large,flexible molecules) as the fluorescent species in FRET experiments.These carry the inherent risk that the distance between the targetmolecules is much greater than the 10 nm suggested by a FRET signal.Also, antibodies could potentially induce cellular signaling. Theprevious experiments13,17 measured FRET between antibodies to theextracellular portions of CD8 and TCR, whereas here we measuredFRET between intracellular portions of the molecules. The previousstudies reflect the ability of pMHC to couple TCR and CD8, whereasour results reflect more the ability of cytoplasmic portions importantin signal transduction to be brought together. Induction of FRETbetween fluorescent CD3 and CD8 chimeras by purified pMHC onbeads confirms that pMHC is sufficient to induce TCR-CD3-CD8interaction17 and shows that this applies to the intracellular portionsof the molecules.

The interaction between CD8 and CD3z peaked in the first 10 minand then gradually ‘faded’ with time. The transient nature of the

TCR-CD3-CD8 interaction may be important in regulating signaltransduction. The similar kinetics of the FRET response, TCR down-regulation and number of conjugates suggest that the loss of the TCR-CD3-CD8 interaction is due to loss of TCR from the cell surface andthat as stimulation becomes lower, the T cells dissociate from the APC.

We found non-antigen-specific recruitment of CD8 into thesynapse, as reported for CD4 (ref. 26). Antigen-independent synapseshave been noted between naive T cells and dendritic cells26,29. RMA-Scells presenting nonstimulatory peptide formed more T cell–APCconjugates than did RMA-S cells without peptide (and therefore withlittle or no MHC class I). The frequency of conjugate formation wasgreater in the presence of antigenic than nonstimulatory peptide.However, the nonstimulatory peptides induced CD8 recruitment tothe synapse as efficiently as the antigenic peptide in the conjugates.This indicated that CD8 clustering was driven by pMHC density andsuggested that the weak CD8–MHC class I interaction6 was none-theless sufficient to pull this ligand-receptor pair into the interfacebetween two cells, as discussed before42. This would have the effect ofconcentrating CD8 and pMHC molecules in the synapse where theycan affect responses to antigen stimulation. Clustering of CD8 in thesynapse may explain published findings showing increased adhesionbetween CD8 and MHC class I molecules on cytotoxic T lymphocytesafter stimulation with soluble anti-TCR or antigen43.

The TCR must recognize a few antigenic ligands in a vast excess ofself-derived peptides. Such endogenous peptides mediate importantsurvival signals and can enhance antigen recognition in CD4 T cellswhen presented simultaneously with an antigenic peptide31. So far,endogenous peptides have not been shown to aid antigen recognitionby CD8+ T cells. Indeed, a previous attempt to demonstrate functionalenhancement by endogenous peptides in CD8+ T cells failed44. Herewe found that the simultaneous presence of nonstimulatory (exoge-nous or endogenous) and agonist pMHC on the same APC enhancedCD8+ T cell responses. Nonstimulatory peptides presented simulta-neously with an antigenic peptide increased interaction between CD8and TCR-CD3, specifically between their cytoplasmic domains. Thus,our FRET data demonstrate that the noncognate CD8-MHC interac-tion enhances the interaction between CD8 and TCR during recogni-tion of cognate pMHC.

In the simplest model for coreceptor function, TCR and CD8 bindto the same antigenic pMHC, with the cytoplasmic region of thecoreceptor providing Lck to phosphorylate CD3. However, thatclassical model cannot explain how endogenous or other nonstimu-latory peptides enhance antigen recognition. The classical model wasinitially challenged by structural studies suggesting that it might beimpossible for a coreceptor to interact with the same MHC moleculecontacted by the TCR45 (although too little is known of the CD8 stalkto state this with any certainty). This suggested that a coreceptormight bind pMHC with its extracellular domain yet interact intra-cellularly with a TCR recognizing a different pMHC, thus bridgingtwo TCRs. This ‘pseudodimer’ model could explain how T cellactivation requires TCR crosslinking to at least a dimer, yet anindividual antigenic pMHC is sufficient to induce activation46. Thecoreceptor must interact with the MHC molecule presenting antigen,suggesting that a coreceptor interacts intracellularly with a TCRbinding to endogenous pMHC31. Experiments using MHC class II–restricted cells have shown involvement of TCR recognition of theendogenous pMHC, because recruitment of endogenous pMHC isCD4 independent30 and because only some endogenous pMHCcomplexes work to enhance T cell activation31. One model hassuggested that endogenous pMHC can amplify the T cell responseto a few antigenic pMHC complexes by virtue of Lck phosphorylation

790 VOLUME 6 NUMBER 8 AUGUST 2005 NATURE IMMUNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

of TCRs binding transiently to the endogenous pMHC47. The fewcognate TCR-pMHC interactions would recruit coreceptor plusLck. Thus, the Lck would be in the vicinity of the frequent (butvery short-lived) TCR interactions with endogenous pMHC andwould be ready to phosphorylate these TCR complexes47. An artificialheterodimer of agonist and certain endogenous pMHC complexes wasshown to be sufficient to activate MHC class II–restricted T cells,suggesting that a TCR recognizing antigenic pMHC and one recogniz-ing endogenous pMHC are bridged by the coreceptor CD4 (ref. 31).

These models for amplification of recognition by endogenouspMHC could explain our data, except that we do not see a require-ment for TCR interaction with endogenous pMHC. In our studyusing CD8+ cells, all seven nonstimulatory peptides enhanced antigenrecognition. Only two of six peptides enhanced CD4+ antigen recog-nition strongly (and two enhanced it weakly)31. CD8 and CD4 arestructurally very different45, so the higher-order structures formed bythese molecules may be different. It is therefore possible that the MHCclass I–restricted response is not dependent on interaction of the TCRwith endogenous pMHC in the same way as the MHC class IIresponse. The noncognate interaction between CD8 and MHC classI could enhance the cognate TCR-CD3-CD8 interaction (and thereforeantigen recognition) by concentrating pMHC and CD8 to the synapse.Thus, CD8-Lck is concentrated in the vicinity of any cognate inter-action of TCR with antigenic pMHC, making it easier for CD8 to bindantigenic pMHC and thus to phosphorylate CD3 and start the T cellactivation cascade. Also, pMHC is concentrated in the synapse so thatthe TCR has a more concentrated set of pMHC species to sort throughto find the few antigenic pMHC complexes. These mechanisms wouldresult in enhancement of coreceptor function, whether CD8 binds tothe same pMHC complex as the TCR or whether the CD8 associatedLck interacts in trans with immunoreceptor tyrosine-based activationmotifs of CD3z.

METHODSPeptides and antibodies. Peptides OVA (SIINFEKL), VSV (RGYVYQGL),

P815 (HIYEFPQL), Mapk1(19–26) (amino acid range in parentheses;

VGPRYTNL), STAT3(53–60) (ATLVFHNL), Ndufa4(61–68) (VNVDYSKL),

Slc2a3(314–321) (VNTIFTVV) and Hcph(503–510) (AQYKFIYV) were synthe-

sized and purified as described33,41. Anti-CD8b (H35-17.2), anti-H-2Kb

(AF6-88.5), anti-Vb5 (to the TCR variable region; MR9-4) and horse-

radish peroxidase–conjugated goat anti-mouse immunoglobulin G were

from BD Pharmingen.

Constructs and cells. Hybridomas expressing the OT-I TCR with or without

CD8a and CD8b were made by retroviral transfection of TCR-deficient 58a–b–

cells34. Chimeric genes encoding CD8b-YFP (Supplementary Methods online)

and CD3z-CFP26 were constructed, were inserted into retroviral vector pBMN-

Z (S. Kinoshita and G. Nolan, www.stanford.edu/group/nolan) and were

expressed in Phoenix packaging cells. Supernatants of these cells were used

to transduce OT-I hybridomas (Supplementary Methods online). For optimal

FRET sensitivity and to avoid false negative results, the acceptor (YFP) should

be in excess relative to the donor (CFP). OT-I cell clones were selected based on

the optimal YFP/CFP molar ratio for FRET; that is, an acceptor/donor molar

ratio of 1:1 to 3:1.

Microscopy. A dual-camera system specifically designed for FRET imaging was

used for imaging, allowing simultaneous acquisition of donor emission and

acceptor emission during donor excitation and fast changes between donor and

acceptor excitation. This consisted of two CoolSnapHQ cameras (Roper)

attached to a Zeiss 200M microscope through a beam splitter (custom

510LPXR; Chroma) and stationary emission filters. A DG4 galvo illuminator

customized with a 300-W xenon lamp (Sutter) was used for rapid wavelength

switching. YFP excitation was attenuated to 20% by appropriate positioning of

the exit mirror. The system was run by Slidebook 4.0.3.9tz software (3I). The

optical filters were as follows (center/bandpass): YFP excitation, 510/20 nm;

YFP emission, 550/50 nm; CFP excitation, 430/25 nm; CFP emission, 470/30

nm; Cy5 excitation, 622/36 nm; Cy5 emission, 700/75 nm. Beamsplitting was

achieved with a JP4 dichroic mirror (Chroma). Exposure times were 0.2–0.5 s,

with 2 � 2 binning and a 63�, 1.4–numerical aperture oil objective, and

software flatfield correction was used. Three-dimensional images were recon-

structed from 15 z-sections located 0.468 mm apart, collected with a 63�, 1.4–

numerical aperture oil objective with 1 � 1 binning. Live cells were imaged in

HEPES-buffered 199 medium (low riboflavin autofluorescence; Gibco) with

5% FBS and without antibiotics and were maintained at 37 1C by the FCS2 live

imaging chamber and objective heater (Bioptechs). T cells and APCs were

mixed and were added to a prewarmed imaging chamber coated with poly-D-

lysine (Sigma). For fixed-cell imaging, cells were mounted in Slowfade Light

antifade mounting media (Molecular Probes).

FRET analysis. A three-filter set algorithm for the crosstalk compensation and

extraction of donor-normalized FRET was used as described earlier26,38. Details

are available (Supplementary Methods online). The FRET image was masked

to accept only regions in which CFP intensity was more than four times above

background noise. Cells with CD8b-YFP/CD3z-CFP ratios outside the stoi-

chiometric range of 1:1 to 3:1, as well as movement artifacts, were excluded

from analysis. The average FRET was calculated from the synapse. Statistical

differences were calculated using the mean difference hypothesis of Student’s

two-tailed t-test assuming different variances and a confidence level of 95%.

APC preparation. RMA-S cells have a defect in Tap2 and cannot bind

endogenous peptides to newly synthesized MHC class I molecules. The

addition of synthetic peptides able to bind to H-2Kb or H-2Db at low

temperature leads to the assembly and stable expression of pMHC on the cell

surface. The peptide-loaded molecules remain stable at physiological tempera-

ture48. Different peptides have different abilities to stabilize the cell surface

pMHC. To obtain equal numbers of pMHC complexes on the cell surface for

both OVA and VSV, we ‘titrated’ the amount of peptide required to produce

equal pMHC loading by both peptides, as described36. RMA-S cells were

stained with Cy5 20 h before experiments by incubation of cells for 5 min at

25 1C with 0.1 mg/ml of Cy5 monomeric succinimidyl ester (Amersham

Biosciences) in RPMI medium, then washing of cells with RPMI medium,

followed by quenching with 10% FBS in RPMI medium. RMA-S cells were

incubated at 29 1C overnight, were pulsed with peptides for 30 min at 29 1C,

were incubated for 3 h at 37 1C and were washed once. For pMHC

quantification, RMA-S cells were stained with anti-H-2Kb–phycoerythrin.

The QuantiBRITE phycoerythrin fluorescence quantification kit (Becton Dick-

inson) was used to calculate the number of molecules.

TCR downregulation and conjugate formation assays. OT-I hybridomas and

peptide-pulsed RMA-S cells (1 � 105 cells each in 50 ml) were incubated in flat-

bottomed wells at 37 1C. After incubation, cells were stained for Vb5 and were

analyzed by flow cytometry. Data are presented as the percentage of Vb5

expression on the surface of cells compared with that of cells incubated with

RMA-S without peptide or as the percentage of TCRs endocytosed. For the

conjugate formation assay, cells were pipetted up and down three times after

incubation, were fixed in 2% paraformaldehyde and were washed in PBS, and

paraformaldehyde was inactivated by 10 mM Tris, pH 7.4, in PBS. Cell

conjugates were analyzed by flow cytometry based on simultaneous expression

of YFP (OT-I hybridoma) and Cy5 (Cy5-labeled RMA-S cell).

Immunoprecipitation and immunoblots. Cells (1 � 107) were lysed in 1%

Brij96, 50 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM Na3VO4 and a protease

inhibitor ‘cocktail’, were precleared with Sepharose G beads and were immuno-

precipitated overnight with anti-CD8b-conjugated beads. After being washed,

samples were separated by 8% reducing SDS-PAGE, were blotted and were

‘developed’ with anti-Lck (3A5; Upstate Biotechnology) followed by anti-

MigG–horseradish peroxidase.

Preparation of beads coated with pMHC. H-2Kb with an Escherichia coli

biotin ligase (BirA) biotinylation sequence and human b2-microglobulin were

produced individually in E. coli, refolded and biotinylated as described33,36,49.

Biotinylated pMHC was added to streptavidin-coated 6-mm carboxylated

NATURE IMMUNOLOGY VOLUME 6 NUMBER 8 AUGUST 2005 791

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

microspheres (Polysciences) at twice the saturating concentration, and samples

were incubated for 16 h at 4 1C with constant agitation. The beads were washed

three times before use.

Accession codes. BIND (http://bind.ca): 300549 and 300550.

Note: Supplementary information is available on the Nature Immunology website.

ACKNOWLEDGMENTSWe thank S. Jameson (University of Minnesota, Minnesota) for the collectionof endogenous H-2Kb-binding peptides; E. Palmer, R. Germain, G. Nolan andJ. Altman for reagents; and W. Havran, L. Sherman and C. Lotz for criticalreading of the manuscript. Supported by the National Institutes of Health(GM065230, DK61329 and GM039476 to N.R.J.G. and T32 AI07290 to T.Z.)and Scripps NeuroAIDS Preclinical Studies (P30MH62261). This is manuscript16479-IMM from The Scripps Research Institute.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Received 1 April; accepted 19 May 2005

Published online at http://www.nature.com/natureimmunology/

1. Davis, M.M. et al. Ligand recognition by ab T cell receptors. Annu. Rev. Immunol. 16,523–534 (1998).

2. Gascoigne, N.R.J., Zal, T. & Alam, S.M. T-cell receptor binding kinetics in T-celldevelopment and activation. Exp. Rev. Mol. Med. [online] (http://www.expertreviews.org/01002502h.htm) (2001).

3. Zamoyska, R. CD4 and CD8: Modulators of T-cell receptor recognition of antigen and ofimmune responses? Curr. Opin. Immunol. 10, 82–87 (1998).

4. Potter, T.A., Rajan, T.V., Dick, R.F., II & Bluestone, J.A. Substitution at residue 227 ofH-2 class I molecules abrogates recognition by CD8-dependent, but not CD8-indepen-dent, cytotoxic T lymphocytes. Nature 337, 73–75 (1989).

5. Madrenas, J., Chau, L.A., Smith, J., Bluestone, J.A. & Germain, R.N. The efficiency ofCD4 recruitment to ligand-engaged TCR controls the agonist/partial agonist propertiesof peptide-MHC molecule ligands. J. Exp. Med. 185, 219–229 (1997).

6. Wyer, J.R. et al. Tcell receptor and coreceptor CD8aa bind peptide-MHC independentlyand with distinct kinetics. Immunity 10, 219–225 (1999).

7. Arcaro, A. et al. CD8b endows CD8 with efficient coreceptor function by coupling T cellreceptor/CD3 to raft-associated CD8/p56lck complexes. J. Exp. Med. 194, 1485–1495(2001).

8. Xiong, Y., Kern, P., Chang, H-C. & Reinherz, E.L. T cell receptor binding to a pMHCIIligand is kinetically distinct from and independent of CD4. J. Biol. Chem. 276, 5659–5667 (2001).

9. Gascoigne, N.R.J. & Zal, T. Molecular interactions at the T cell-antigen-presenting cellinterface. Curr. Opin. Immunol. 16, 114–119 (2004).

10. Gallagher, P.F., Fazekas de St Groth, B. & Miller, J.F.A.P. CD4 and CD8 molecules canphysically associate with the same T-cell receptor. Proc. Natl. Acad. Sci. USA 86,10044–10048 (1989).

11. Beyers, A.D., Spruyt, L.L. & Williams, A.F. Molecular associations between theT-lymphocyte antigen receptor complex and the surface antigens CD2, CD4, or CD8and CD5. Proc. Natl. Acad. Sci. USA 89, 2945–2949 (1992).

12. Suzuki, S., Kupsch, J., Eichmann, K. & Saizawa, M.K. Biochemical evidence of thephysical association of the majority of CD3 d chains with the accessory/co-receptormolecules CD4 and CD8 on nonactivated T lymphocytes. Eur. J. Immunol. 22, 2475–2479 (1992).

13. Doucey, M.A. et al. CD3d establishes a functional link between the T cell receptor andCD8. J. Biol. Chem. 278, 3257–3264 (2003).

14. Mittler, R.S., Goldman, S.J., Spitalny, G.L. & Burakoff, S.J. T-cell receptor-CD4physical association in a murine T-cell hybridoma: Induction by physical antigenreceptor ligation. Proc. Natl. Acad. Sci. USA 86, 8531–8535 (1989).

15. Anel, A., Martinez-Lorenzo, M.J., Schmitt-Verhulst, A.M. & Boyer, C. Influence on CD8of TCR/CD3-generated signals in CTL clones and CTL precursor cells. J. Immunol. 158,19–28 (1997).

16. Osono, E., Sato, N., Yokomuro, K. & Saizawa, M.K. Changes in arrangement and inconformation of molecular components of peripheral T-cell antigen receptor complexafter ligand binding: analyses by co-precipitation profiles. Scand. J. Immunol. 45,487–493 (1997).

17. Block, M.S., Johnson, A.J., Mendez-Fernandez, Y. & Pease, L.R. Monomeric class Imolecules mediate TCR/CD3e/CD8 interaction on the surface of T cells. J. Immunol.167, 821–826 (2001).

18. Takada, S. & Engleman, E.G. Evidence for an association between CD8 molecules andthe T cell receptor complex on cytotoxic T cells. J. Immunol. 139, 3231–3235 (1987).

19. Anderson, P., Blue, M-L., Schlossman, S.F. & Comodulation of CD3 and CD4Evidence for a specific association between CD4 and approximately 5% of the

CD3:T cell receptor complexes on helper T lymphocytes. J. Immunol. 140, 1732–1737 (1988).

20. Rojo, J.M., Saizawa, K. & Janeway, C.A., Jr. Physical association of CD4 and the T-cellreceptor can be induced by anti-T-cell receptor antibodies. Proc. Natl. Acad. Sci. USA86, 3311–3315 (1989).

21. Kwan Lim, G.E., McNeill, L., Whitley, K., Becker, D.L. & Zamoyska, R. Co-cappingstudies reveal CD8/TCR interactions after capping CD8b polypeptides and intracellularassociations of CD8 with p56lck. Eur. J. Immunol. 28, 745–754 (1998).

22. Montixi, C. et al. Engagement of T cell receptor triggers its recruitment to low-densitydetergent-insoluble membrane domains. EMBO J. 17, 5334–5348 (1998).

23. Xavier, R., Brennan, T., Li, Q.Q., McCormack, C. & Seed, B. Membrane compartmenta-tion is required for efficient T cell activation. Immunity 8, 723–732 (1998).

24. Bromley, S.K. et al. The immunological synapse. Annu. Rev. Immunol. 19, 375–396(2001).

25. Krummel, M.F., Sjaastad, M.D., Wulfing, C. & Davis, M.M. Differential clustering ofCD4 and CD3z during T cell recognition. Science 289, 1349–1352 (2000).

26. Zal, T., Zal, M.A. & Gascoigne, N.R.J. Inhibition of T-cell receptor-coreceptor interac-tions by antagonist ligands visualized by live FRET imaging of the T-hybridomaimmunological synapse. Immunity 16, 521–534 (2002).

27. Purbhoo, M.A., Irvine, D.J., Huppa, J.B. & Davis, M.M. T cell killing does not requirethe formation of a stable mature immunological synapse. Nat. Immunol. 5, 524–530(2004).

28. Potter, T.A., Grebe, K., Freiberg, B. & Kupfer, A. Formation of supramolecular activationclusters on fresh ex vivo CD8+ T cells after engagement of the T cell antigen receptorand CD8 by antigen-presenting cells. Proc. Natl. Acad. Sci. USA 98, 12624–12629(2001).

29. Revy, P., Sospedra, M., Barbour, B. & Trautmann, A. Functional antigen-independentsynapses formed between T cells and dendritic cells. Nat. Immunol. 2, 925–931(2001).

30. Wulfing, C. et al. Costimulation and endogenous MHC ligands contribute to T cellrecognition. Nat. Immunol. 3, 42–47 (2002).

31. Krogsgaard, M. et al. Agonist/endogenous peptide-MHC heterodimers drive T cellactivation and sensitivity. Nature 434, 238–243 (2005).

32. Zal, T. & Gascoigne, N.R.J. Using live FRET imaging to reveal early protein-protein interactions during T cell activation. Curr. Opin. Immunol. 16, 418–427(2004).

33. Alam, S.M. et al. T cell receptor affinity and thymocyte positive selection. Nature 381,616–620 (1996).

34. Stotz, S.H., Bolliger, L., Carbone, F.R. & Palmer, E. T cell receptor (TCR) antagonismwithout a negative signal: Evidence from T cell hybridomas expressing two independentTCRs. J. Exp. Med. 189, 253–263 (1999).

35. Gorman, S.D., Sun, Y.H., Zamoyska, R. & Parnes, J.R. Molecular linkage of the Ly-3and Ly-2 genes. Requirement of Ly-2 for Ly-3 surface expression. J. Immunol. 140,3646–3653 (1988).

36. Holmberg, K., Mariathasan, S., Ohteki, T., Ohashi, P.S. & Gascoigne, N.R.J. TCRbinding kinetics measured with MHC class I tetramers reveal a positive select-ing peptide with relatively high affinity for TCR. J. Immunol. 171, 2427–2434 (2003).

37. Monks, C.R.F., Freiberg, B.A., Kupfer, H., Sciaky, N. & Kupfer, A. Three-dimensionalsegregation of supramolecular activation clusters in T cells. Nature 395, 82–86(1998).

38. Zal, T. & Gascoigne, N.R.J. Photobleaching-corrected FRET efficiency imaging of livecells. Biophys. J. 86, 3923–3939 (2004).

39. Freiberg, B.A. et al. Staging and resetting T cell activation in SMACs. Nat. Immunol. 3,911–917 (2002).

40. Porgador, A., Yewdell, J.W., Deng, Y., Bennink, J.R. & Germain, R.N. Localization,quantitation, and in situ detection of specific peptide–MHC class I complexes using amonoclonal antibody. Immunity 6, 715–726 (1997).

41. Santori, F.R. et al. Rare, structurally homologous self-peptides promote thymocytepositive selection. Immunity 17, 131–142 (2002).

42. Kupfer, A., Singer, S.J., Janeway, C.A., Jr & Swain, S.L. Coclustering of CD4 (L3T4)molecule with the T cell receptor is induced by specific direct interaction of helper Tcells and antigen-presenting cells. Proc. Natl. Acad. Sci. USA 84, 5888–5892(1987).

43. O’Rourke, A.M., Rogers, J. & Mescher, M.F. Activated CD8 binding to class Iprotein mediated by the T cell receptor results in signalling. Nature 346, 187–189(1990).

44. Sporri, R. & Reis e Sousa, C. Self peptide/MHC class I complexes have a negligibleeffect on the response of some CD8+ T cells to foreign antigen. Eur. J. Immunol. 32,3161–3170 (2002).

45. van der Merwe, P.A. & Davis, S.J. Molecular interactions mediating T cell antigenrecognition. Annu. Rev. Immunol. 21, 659–684 (2003).

46. Irvine, D.J., Purbhoo, M.A., Krogsgaard, M. & Davis, M.M. Direct observation of ligandrecognition by T cells. Nature 419, 845–849 (2002).

47. Li, Q.J. et al. CD4 enhances T cell sensitivity to antigen by coordinating Lckaccumulation at the immunological synapse. Nat. Immunol. 5, 791–799 (2004).

48. Ljunggren, H.G. et al. Empty MHC class I molecules come out in the cold. Nature 346,476–480 (1990).

49. Altman, J.D. et al. Phenotypic analysis of antigen-specific T lymphocytes. Science 274,94–96 (1996).

792 VOLUME 6 NUMBER 8 AUGUST 2005 NATURE IMMUNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y