Wild type and tailless CD8 display similar interaction with microfilaments

during capping

PASCALE ANDRE1, JEAN GABERT2'*, ANNE MARIE BENOLIEL1, CHRISTIAN CAPO1, CLAUDE

BOYER2, ANNE MARIE SCHMITT-VERHULST2, BERNARD MALISSEN2 and PIERRE BONGRAND1

1Laboratoire d'Immunologie, Hdpital de Sainte-Marguerite, BP 29, 13277 Marseille Cedex 09, France2Centre d'Immunologie de Marseille Luminy, Case 906, 13288 Marseille Cedex 9, France

* Present address: Institut Paoli-Calmettes, BP 156, 13273 Marseille Cedex 9, France

Summary

We examined the influence of the intracytoplasmicregion of CD8a on capping and interaction withmicrofilaments. We used cell clones obtained bytransfecting a CD4+ T-cell hybridoma with (a) T-cellreceptor (TCR) a and fi chains from a cytolytic cloneand (b) CD8« genes that were either native ormodified by extensive deletion of the intracytoplas-mic region or replacement of the transmembrane andintracytoplasmic domains with those of a class Imajor histocompatibility complex gene (Letourneuret al. (1990). Proc. natn. Acad. Sci. U.S.A. 87,2339—2343). Different cell surface structures werecross-linked with anti-T-cell receptor, anti-CD8 oranti-class I monoclonal antibodies and anti-immuno-globulin (Fab')2. Double labeling and quantitativeimage analysis were combined to monitor fluor-escence anisotropy and correlation between differ-ent markers. Microfilaments displayed maximalpolarization within two minutes. The correlation

between these structures and surface markers wasthen maximal and started decreasing, whereas theredistribution of surface markers remained stable orcontinued. Furthermore, wild type and altered CD8«exhibited similar ability to be capped and to induceco-capping of TCR and MHC (major histocompati-bility complex) class I: the fraction of cell surfacelabel redistributed into a localized cap rangedbetween 40% and 80%. Finally, cytochalasin Ddramatically decreased CD8 capping in all testedclones.

It is concluded that the transmembrane and/orintracellular domains of CD8 molecules are able todrive the extensive redistributions of membranestructures and cytoskeletal elements that are trig-gered by CD8 cross-linking.

Key words: membrane, molecular motion, cytoplasmic domain,cytoskeleton, CD8.

Introduction

Since the elaboration of the fluid mosaic model of theplasma membrane by Singer and Nicolson (1972), it hasbecome clear that many cell functions are affected bylateral movements of membrane molecules. Cell adhesionis thus highly dependent on the lateral mobility of ligandmolecules (Rutishauser and Sachs, 1974). Clearly, theformation of strong intercellular bonds may require thatligand molecules be concentrated in adhesion areas (Bellet al. 1984), and this concentration was indeed observed inseveral experimental studies using different method-ologies such as electron microscopy (Singer, 1975), micro-fluorometry (Poo and McCloskey, 1986) and conventional(Kupfer and Singer, 1989) or computer-assisted (Andre etal. 1990a,b) fluorescence microscopy. Also, the microaggre-gation of cell surface receptors was found to enhance theiradhesive capacity (Detmers et al. 1987).

However, the mechanisms of molecular motion in theplasma membrane remain incompletely understood.When the mobility of cell surface proteins was studiedwith the fluorescence photobleaching recovery (Schless-inger et al. 1976) or post electric field relaxation (Poo et al.1978) method, the diffusion coefficient of concanavalin AJournal of Cell Science 100, 329-337 (1991)Printed in Great Britain © The Company of Biologists Limited 1991

'receptors' (Schlessinger et al. 1976; Poo et al. 1978; Smithet al. 1979), class I histocompatibility molecules (Edidinand Zuniga, 1984; Bierer et al. 1987; Wier and Edidin,1986) or lymphocyte surface immunoglobulins (Dragstenet al. 1979; Barisas, 1984) varied between less than 10~12

and about 2xlO~9cm2s~1. The latter value was roughlyconsistent with theoretical estimates for free diffusion in athick viscous sheet (Wiegel, 1984). Thus, it appearedreasonable to view the motion of cell surface proteins as ahindered-diffusion process. In addition, these proteins mayundergo active displacements as exemplified by thecapping process (Taylor et al. 1971; Bourguignon andBourguignon, 1984).

Strong experimental evidence suggests that interac-tions between cytoskeletal elements and membraneintrinsic proteins are dynamic events (Flanagan andKoch, 1978; Burn et al. 1988) that may hinder randomdiffusion (Koppel et al. 1981; Tank et al. 1982). Further,active movements may be affected by cytoskeletal inhibi-tors (Taylor et al. 1971; Bourguignon and Bourguignon,1984). However, the precise molecular interactions in-volved in these phenomena remain incompletely under-stood and it is not well known whether cytoskeletalelements are bound to many membrane proteins or

329

whether a few membrane molecular species mediateindirect interactions between cytoskeleton and membranestructures, as previously suggested (Bourguignon andSinger, 1977).

Recently, several authors used mutated membraneproteins to assess the influence of cytoplasmic domains onlateral diffusion. Thus, deletion of the intracellulardomains of class I histocompatibility molecules (Edidinand Zuniga, 1984), epidermal growth factor receptor(Livneh et al. 1986) or viral proteins (Scullion et al. 1987)did not change lateral diffusion, whereas a similarmutation resulted in tenfold increase of the diffusioncoefficient of MHC class II molecules (Wade et al. 1989).

The present work was planned to evaluate the influenceof the cytoplasmic domain of membrane proteins on theiractive movements and interactions with cytoskeletalelements and membrane proteins. We used a CD4+ T-cellhybridoma transfected with CD8cx genes bearing alteredcytoplasmic domains (Letourneur et al. 1990). CD8molecules were capped with specific antibodies and theredistribution of CD8, T-cell receptor (TCR) and MHCclass I molecules as well as microfilaments was studied bycombination of double labeling and image analysis. It isconcluded that the extracellular and/or transmembranedomains of CD8 molecules are sufficient to mediate theextensive interactions observed between these moleculesand microfilaments or membrane proteins.

Materials and methods

CellsKB5C20 (Albert et al. 1982) is an alloreactive H-2Kb-specificcytotoxic T lymphocyte (CTL) clone of B10.BR origin (H-2k). ACD4+ anti-ovalbumin T-cell hybridoma (Marrack et al. 1983) wastransfected with a- and /3 chain genes encoding KB5C20 T-cellreceptor (Letourneur et al. 1990) and (i) a CD8

derivatized streptavidin. The intensity of biotin-associated label-ing was 7 % and 6 % of that obtained with fluorescent (Fab')2 whenboth fluorophore combinations (fluorescein/rhodamine andrhodamine/fluorescein) were used.

Although this weak artefactual labeling was inapparent onvisual examination of fluorescent cells, it was important to knowwhether this could significantly alter the correlation between thefluorescence distributions corresponding to different antigens ondoubly labeled cells. This was tested by computer simulation: 100double fluorescence distributions were randomly generated. Themean standard deviation of sector relative fluorescences was 0.28.The mean normalized correlation (V3/2xln[(l+r)/(l-r)] wasthen calculated assuming 0 %, 5 %, 10 % and 15 % contaminationof each fluorescence by the other one. We obtained 0.09, 0.29, 0.51and 0.78, respectively. It was concluded that the correlationcoefficients measured on doubly labeled cells (see Tables) couldnot be ascribed to artefactual overlap between both labels.

Flow cytometryThe mean fluorescence of different populations of labeled cells wasdetermined with a Profile flow cytometer (Coulter, Hialeah, FL).

Results

Tailless CD8 molecules can be capped and interact withmicrofilamentsIn a first series of experiments, we compared the cappingproperties of wild-type and engineered CD8 molecules.

Typical fluorescence images are shown in Fig. 1. Cellswith limited asymmetry of CD8 are displayed inFig. 1 A,C. A typical redistribution of labeled molecules isshown in Fig. IE: in this case, sector analysis shows thatthe mean CD8 fluorescence of the right sector is about 3times higher than that of left CD8-depleted sectors.

There was a need for quantitative parameters tomonitor CD8 capping. The standard deviation of therelative fluorescence of the six sectors defined in testedcells (Fig. 1B,D,F) seemed a suitable choice: this was 0.45on the capped cell (Fig. 1E,F) as compared with 0.17 and0.13 of the two other cells (Fig. 1A,C).

The possible occurrence of a relationship between CD8and microfilament distributions was studied by calculat-ing the correlation coefficient between sector relativefluorescences on doubly labeled cells. This is 0.97 on thecell shown in Fig. 2, which is indeed indicative of a highlysignificant corrrelation (P=0.001). Since the statisticaldistribution of the correlation coefficient r in a randompopulation is expected to be strongly non-Gaussian, weused its normalized transform (i.e. V3/2xln[(l+r)/(l-r)])in order to allow simple averaging of parametersmeasured on individual cells. This parameter was 3.62 onthe cells shown in Fig. 2. A value higher than 2 may beconsidered as significantly different from zero at the 0.05significance level.

The kinetics of fluorescence redistribution triggered bycross-linking normal or engineered CD8 molecules isshown in Fig. 3. The anisotropy of CD8 and cytoskeletalfluorescences was nearly maximal within 2-5 min. Wildtype and modified CD8 molecules exhibited markedredistribution with comparable kinetics. However, themaximum standard deviation of CD8 relative fluor-escences was higher on DC41.1.4 clone bearing native CD8(0.87±0.06 S.E.) than on DC142 clone with chimeric CD8(0.61+0.07) and DC 136 clone with tailless CD8(0.52±0.07).

Also, with all tested cells, cross-linking CD8 induced amarked redistribution of cytoskeletal elements, with amaximal correlation between both fluorescence distri-

Fig. 1. Principle of sector analysis. Cells from the DC 41.1.4clone were labeled with anti-CD8 monoclonal antibodies, thenexposed for 30 min at 4°C (A-C) or 37 °C (E-F) to rhodamine-labeled anti-mouse immunoglobulin (Fab')2. Three typicalfluorescence images are shown (A, C, E). These images weredivided into six sectors of angle fc/3 and the relativefluorescence of each sector is shown (B, D, F). The asymmetryof each fluorescence distribution was quantitated bycalculating the standard deviation of sector fluorescences,yielding 0.17 (B), 0.13 (C) and 0.45 (F). The latter value maybe considered as indicative of capping. Bar, 5 //m.

butions of 2.10±0.14 (DC142 clone) and 2.22±0.30 (DC136clone). Hence, no alteration of the ability to interact withcytoskeletal elements was detected in tailless CD8molecules.

Another finding was that the correlation between CD8and microfilament distributions was nonzero in theabsence of any cross-linking of surface molecules. Thismay be ascribed to asymmetrical morphology (i.e. devi-ation from a spherical shape) and permanent anisotropy(Andre et al. 1990a). This emphasizes the interest ofquantitative methods of fluorescence analysis.

In order to assess the selectivity of cytoskeletal proteinredistribution, we looked for a possible reorganization oftubulin molecules after CD8 capping: as shown on Table 1,CD8 cross-linking induced no substantial redistribution of

CDS cytoplasmic domain and capping 331

B

Fig. 2. Principle of correlation studies. Cells from the DC41.1.4 clone were labeled with an anti-CD8 monoclonal antibody, thenexposed for 30min at 37 °C to rhodamine-labeled anti-immunoglobulin (Fab')2 and fixed before cytoskeleton labeling withNBD-phallacidin. The distribution of CD8 (A and B) and NBD-phallacidin (C and D) was studied by quantitative fluorescencemicroscopy. Cell images were divided into six sectors of angle x/Z and the relative fluorescence of each sector was calculated. Theasymmetry of each fluorescence distribution was monitored by calculating the standard deviation of sector fluorescences, yielding0.50 (A,B) and 0.25 (C,D). The correlation between two fluorescence distributions was evaluated by calculating the correlationcoefficient r between sector fluorescences and using the normalized coefficient V3/2xln[(l + r)/(l-r)].

tubulin molecules, and no significant correlation wasfound between CD8 and tubulin markers.

The capping of wild type and modified CD8 molecules isdependent on microfilament integrityIt was important to determine to what extent cytoskeletalreorganization was a cause or a consequence of the cappingof normal and altered CD8 molecules. This point wasaddressed by studying the influence of cytochalasin D, a

fairly selective inhibitor of actin polymerization (Cooper,1987), on the capping of CD8. As shown in Table 2, 5minafter the onset of capping, cytochalasin D-treated cellsdisplayed 3-4-fold decrease of the CD8 capping parameterin all tested clones. Thus it was concluded that the cellcytoskeleton plays an active role in CD8 redistribution.

In order to further explore the involvement of microfila-ments in the capping process cells were labeled withNBD-phallacidin at various times after the onset ofcapping. Mean cell fluorescence was assayed with flow

332 P. Andre et al.

Correlation3-

0.5-

0 10

B S.D.

1-

0.5-

20 Time (min) 0

Correlation3-

10 20

1-

0 10

0.5

2 0 Time (min) °Correlation

3"

10 20

10 2 0 Time (min) ° 10 20

Fig. 3. Capping of CD8. Cells expressing intact CD8 (CD41.1.4clone, A) or CD8 modified by replacement of intramembraneand intracytoplasmic residues with those of MHC class I(DC142 clone, B) or by extensive deletion of intracytoplasmicdomains (DC136 clone, C) were exposed to an anti-CD8monoclonal and anti-immunoglobulin (Fab')2. They were thenfixed at regular intervals and processed for quantitativeanalysis of the distribution of CD8 and microfilaments. Thestandard deviation of relative sector fluorescences of CD8 label(full lines) and microbilament stain (broken lines) is shown onleftwards plots. The normalized correlation between both labelsis shown on rightward plots. Each point represents a mean of10-30 determinations. Vertical bar length is twice thestandard error.

Table 1. Absence of substantial redistribution of tubulinmolecules after cross-linking CD8

Cell treatment Control Ten minute capping

Standard deviation of sectorfluorescence

CD8Tubulin

Correlation between markers

0.13±0.014 (9)0.068±0.004 (9)0.66±0.18 (9)

0.55±0.03 (20)0.10±0.008 (20)

l.l±0.31 (20)

Cells from the DC41.1.4 clone were exposed to anti-CD8 monoclonalantibody and rhodamine-labeled anti-mouse immunoglobulin (Fab')2 at4°C. They were then fixed immediately (control) or after 10minincubation at 37°C (capped cells) and stained for tubulin withNBD-colcemid. The fluorescence distribution was determined for bothmarkers and standard deviations of sector fluorescences were calculatedas well as the correlation between both markers. Mean values areshown±standard error of the mean. Number of studied cells inparenthesis.

Table 2. Effect of cytochalasin D on the redistribution ofmicrofilaments and CD8 induced by CD8 cross-linking

Clone

DC41.1.4

DC142

DC136

Cytochalasin D

05/igml"1

05/igmr1

05/igml"1

Standard deviation ofsector fluorescence

Phallacidin

0.19±0.012 (60)0.18+0.014 (61)0.25±0.013 (57)0.17±0.019 (59)0.23±0.018 (37)0.16±0.016 (38)

CD8

0.80±0.039 (60)0.26±0.016 (61)0.79±0.034 (57)0.20±0.011 (59)0.91 ±0.056 (37)0.23±0.019 (38)

Cells from the DC41.1.4 clone (expressing wild-type CD8ff), or DC142(with hybrid CD8 bearing transmembrane and intracytoplasmicdomains of MHC class I molecules) or DC136 (with tailless CD8ff) wereexposed to anti-CD8 monoclonal antibodies and fluorescent anti-immunoglobulin (Fab')2> then incubated for 5min at 37°C with orwithout S^igml"1 cytochalasin D. They were then assayed withfluorescence microscopy for quantitative determination of the anisotropyof CD8 and cytoskeleton labeling. Five separate experiments wereperformed and each value shown as a mean±standard error of themean. The number of examined cells is shown in parenthesis.

Table 3. Effect of capping on the microfilament contentofDC41.1.4 cells

Mean relativefluorescence

Standard error

0

1.0

1

1.02

0.03

Time

2

1.08

0.03

(min)

5

1.12

0.06

10

1.13

0.06

30

0.94

0.08

Cells were exposed at 4°C to anti-CD8 monoclonal antibodies andanti-mouse immunoglobulins. They were then warmed to 37 °C andincubated for different periods of time before being fixed and stainedwith NBD-phallacidin for determination of total fluorescence with aflow cytometer. Values shown are means of 8 separate experiments.

cytometry and quantitative fluorescence microscopy. Asshown in Table 3, the former method suggested theoccurrence of a transient and moderate increase of actinpolymerization following capping induction. Fluorescencemicroscopy did not reveal any modification of phallacidinstaining during the studied process (data not shown), dueto the lower size of assayed cell populations and sub-sequent higher dispersion of results.

Intact and modified CD8 molecules display similar co-redistribution with other cell membrane proteinsThe simplest explanation for the ability of tailless CD8molecules to interact with cytoskeletal elements was thatthis association was mediated by other cell membranemolecules that would interact with CD8 through theirextracellular or intrabilayer domains and with microfila-ments through their intracytoplasmic region in accord-ance with a model suggested by Bourguignon and Singer(1977). The possibility of extensive interactions betweenCD8 and other membrane molecules was tested bytreating CD8 positive clones with anti-CD8 and studyingthe redistribution of the TCR and MHC class I molecules.As shown in Tables 4 and 5, a close correlation was foundbetween the membrane distribution of CD8, TCR andMHC class I in all tested clones, supporting the possibilitythat CD8 might drag cell surface glycoproteins towards acell pole during the capping process, thus allowing indirectinteractions between CD8 and microfilaments.

CD8 cytoplasmic domain and capping 333

Table 4. Co-capping of MHC class I molecules by anti-CD8 antibodies

Clone

DC41.1.4 (n=23)DC142 (n=14)DC136(n=14)

Standard deviation ofsector fluorescence

CD8

1.02±0.070.58±0.090.66±0.09

MHC class I

0.34±0.030.23±0.040.41±0.06

Correlationbetweenmarkers

1.56±0.301.85±0.503.23±0.34

Cells were treated with anti-CD8 monoclonal antibodies and exposedfor 20min at 37 °C to fluorescein-conjugated anti-murineimmunoglobulin (Fab')2- They were then cooled and labeled withbiotinylated anti-MHC class I antibodies and rhodamine-conjugatedstreptavidin. They were then examined with quantitative fluorescencemicroscopy for determination of fluorescence asymmetry (as expressedby the standard deviation of sector fluorescence) and correlationbetween both antigens (as expressed with normalized correlation). Meanvalues are shown ± standard error. Number of analyzed cells inparenthesis.

Table 5. Co-capping of T cell receptor by anti-CD8antibodies

Standard deviation ofsector fluorescence

Clone CD8 T cell receptor

Correlationbetweenmarkers

DC41.1.4 (n=20)DC136(n=14)

0.60 ±0.060.76±0.04

0.58±0.050.34±0.02

3.75±0.282.83±0.16

Cells were treated with anti-CD8 monoclonal antibodies and exposedfor 20min at 37 °C to fluorescein-conjugated anti-immunoglobulin(Fab')2- They were then cooled and labeled with biotinylated anti-T cellreceptor antibodies and rhodamine-derivatized streptavidin. They werethen examined with quantitative fluorescence microscopy fordetermination of fluorescence asymmetry (as expressed by the standarddeviation of sector fluorescence) and correlation between markers (asexpressed with normalized correlation). Mean values are shown±standard error. Number of analyzed cells in parenthesis.

Capping may involve an early microfilament-dependentand a late micro filament-independent stage

As shown in Fig. 3A,B,C, the correlation between cyto-skeleton and CD8 distributions was maximum withinabout 2 min after exposure to the capping stimulus. Thiscorrelation then decreased, whereas the anisotropy offluorescence distributions kept on increasing. This sugges-ted that CD8 capping involved some kind of associationbetween extra- and intra-cellular structures only duringthe earliest phase of the capping process. Indeed,10-30 min after the triggering of redistribution, weobserved cells displaying redistributed CD8 and microfila-ment markers, with a different intracellular localization(Fig. 4). It was of interest to know whether these findingsapplied to other cell surface markers, in addition to CD8.As shown in Fig. 5, when anti-TCR or anti-MHC class Iantibodies were used to induce capping, a co-redistributionof cell surface markers and microfilaments was also found,and the correlation between extra- and intra-cellularmarkers was maximum within two minutes, then de-creased, as was found with CD8.

Discussion

The purpose of our work was to assess the role of theintracellular domains of membrane molecules on theactive movements of these structures. Capping was chosenas a suitable model for this study.

1-

0.5

Correlation31

S.D.

1-

10 2 0 Time (min)0

Correlation

10 20

0.51-

10 20 Time (min) 10 20

Fig. 5. Capping of TCR and MHC class I. Cells from theDC41.1.4 clone were exposed to an anti-TCR (A) or anti-MHCclass I (B) monoclonal antibody and anti-immunoglobulin(Fab')2. They were incubated at 37°C for various periods oftime, then processed for quantitative analysis of thedistribution of TCR or MHC and polymerized actin. Thestandard deviation of relative sector fluorescences for TCR orMHC (full lines) or microfilament (broken lines) is shown onleft plots. The correlation between both markers is shown inthe righthand plots. Each point represents a mean of 20-30determinations. Vertical bar length is twice the standard error.

A first requirement was to achieve a quanti tat ivedescription of the tested phenomenon. The sector analysiswe performed was felt a satisfactory way of monitoring thecapping process since in most cases the value of thestandard deviation parameter matched the intuitivefeeling of the extent of antigen redistribution provided byvisual examination of fluorescence images (Fig. 1).

A second conclusion made clear by quanti tat ive fluor-escence determinations is tha t only a limited fraction ofcross-linked cell surface molecules were capped. Indeed,using the derivation shown in the Appendix, the fractionof capped molecules was found to range between 40 % and80% of total label. Obviously, it is difficult to evaluatefluorescence intensities using a grey scale (Figs 1, 2). Asshown in Fig. 4, a coded color display gives a much moreprecise description of the distributions of light intensities.

Now, the most clearcut conclusion of our experiments istha t wild type and altered CD8 chains were all able to beredistributed in response to a cross-linking stimulus, andthey induced similar redistribution of polymerized actin(Fig. 3). The simplest interpretation for these findings isthat the redistribution of cytoskeletal elements as well asTCR or MHC class I molecules induced by bridging'tailless' CD8 on DC 136 clone was mediated by molecularinteractions occurring in the bilayer or the extracellularregion. This might involve some third-party structure, inaccordance with a model previously reported by Bourguig-non and Singer (1977). Direct CD8-cytoskeleton interac-tion did not play a quantitatively significant role in thisrespect. This conclusion is consistent with the finding that

334 P. Andre et al.

B

Fig. 4. Coded color display of immunofluorescence distributions. Cells from the CD41.1.4 clone were exposed to anti-CD8monoclonal antibodies and anti-immunoglobulin (Fab')2 for 5min (A and B) or 30min (C and D) at 37°C before fixation andstaining for quantitative analysis of the distribution of CD8 (A and C) and polymerized actin (B and D). Fluorescence intensitieswere represented with coded color display using a 16-level scale. The normalized correlation between both fluorescencedistributions is 0.81 (A and B) and -0.42 (C and D).

complement decay-accelerating factor, a lipid-anchoredmembrane protein, could be made to cap and interact withcytoskeletal elements as efficiently as CD3, CD4 or CD8(Kammer et al. 1988). It must be emphasized that specificintermolecular associations may not be required tomediate interactions between microfilaments and cellsurface proteins, since a bulk displacement of a givenmolecule may generate a lipid flow in the viscous bilayeror drive extracellular glycocalyx elements entangled withthe oligosaccharide chains of moving molecules.

Another finding is that capping CD8 molecules induceda significant co-redistribution of TCR (Table 5) and, to alesser extent, MHC class I (Table 4) molecules. It must bepointed out that this co-redistribution was sometimeslimited (see 1st row of Table 4 and 2nd row of Table 5).Perhaps co-capping is not an all-or-none phenomenon, asmight be suggested by qualitative observation. This viewis not at all inconsistent with the finding that antibody-induced co-capping of different cell membrane antigensmay be dependent on the affinity and specificity of cappingantibodies (Kupfer and Singer, 1988).

A final point is to understand the importance of activecytoskeletal movements in the redistribution process. Asshown on Table 2, the capping of CD8 measured 5 minafter exposure to cross-linking antibodies was drasticallydecreased by cytochalasin D, suggesting an involvement ofmicrofilaments in the first steps of capping. Interestingly,capping-associated cytoskeleton redistribution was quan-titatively less important and less sensitive to cytochalasinD than that of surface molecules (Table 2). Two non-exclusive explanations may be offered for this finding.First, only a minor part of cell microfilaments might beinvolved in capping, as compared to the fraction of CD8molecules affected by this process. Second, cytochalasinreduction of microfilament length might at the same timeincrease their capacity to rearrange and decrease theirability to drag cell surface structures, which would beconsistent with a minimal overall effect of cytochalasintreatment on microfilament distribution.

Since the redistribution of surface antigens went onafter the peak of microfilament reorientation (Fig. 3)without any concomitant increase of the correlationbetween extra- and intra-cytoplasmic structures (Fig. 1),the simplest hypothesis would be that the clustering ofstudied surface molecules was completed through thermaldiffusion. This is consistent with quantitative estimatesbased on known values of the diffusion constant ofmembrane glycoproteins. As estimated by Berg andPurcell (1977), the mean time required for a molecule withdiffusion constant D to be trapped on a spherical cell ofradius a by a disc of radius s (representing the cap) isabout:

t= 1.1 a2/'Dln(1.2 a2/d2).

Using 5xlO~10cm2s~1 as an estimate for the diffusioncoefficient of the most mobile fraction of cell surfacemolecules (Dragsten et al. 1979), 5,um for a and ljun for s,we find:

t = 30 min.

Hence, it seems reasonable to suggest that the reorgan-ization of cell surface antigens was triggered by acytochalasin D-inhibitable motile mechanism. This wouldresult in the formation of a localized molecular clusterwithin a few minutes. Free molecules might then gettrapped within this cluster when encountering it onrandom diffusion. The decrease of the correlation between

cell surface molecules and cytoskeletal elements observed5 min after the onset of capping may be ascribed toindependent drift of extracellular submembranar struc-tures. Hence, the main role of the intracytoplasmic tail ofmembrane proteins such as CD8 might be related totransmembrane signalling (Zamoyska et al. 1989) orinternalisation (Boyer et al. 1989; Capps et al. 1989) ratherthan lateral motility.

This work was supported by the INSERM (CJF 89-07).

AppendixSignificance of quantitative fluorescence indices

It was important to assess the significance of theparameters used to describe fluorescence distributions.The basic difficulty is that a measured fluorescencepattern depends on both the cellular label distribution andthe relative orientation between the cell and the opticalaxis of the microscope. This problem was addressed byconsidering model fluorescence distributions with randomorientation as shown below.

Significance of the standard deviation of relative sectorfluorescenceLet us consider a labeled sphere with total fluorescence T.We assume that: a fraction xxT (O^x^l) of total labelingis concentrated on a point of the sphere. The fluorescencemeasured on sector i (i=1...6) is kxTxpj, where k is aconstant representing the fraction of emitted fluorescentlight detected by the measuring apparatus and p{ valuesare positive real numbers such that:

6

I P i = l .

A fraction ( l -x )xT of total labeling is uniformly spreadon the sphere. The fluorescent light intensity falling onsector i is kx(l-x)xT/6.

Hence, the relative fluorescence of sector i is:

The standard deviation of relative sector fluorescences istherefore:

S.D. = x x (E(6pi - 1)2/5)J = xxf.

Hence, parameter S.D. is proportional to the fraction ofredistributed label. In order to evaluate /"numerically, weconsidered 4800 points uniformly spread on the surface ofa sphere, and we used a previously described (Andre et al.1990a, b) geometrical model of fluorescence images tocalculate the average value of f. We found:

f= 1.23 ± 0.58 (standard deviation).

Significance of correlationThe significance of the correlation between fluorescencedistributions was assessed with a similar procedure to thatperformed for the standard deviation. Since the corre-lation coefficient between two random variables x and y isthe same as that between ajc+b and cy+d, where a, b, c andd are real constants, the correlation between the fluor-escence distributions of two spheres with a combination ofuniform label and a point fluorescence (as described above)is only dependent on the location of both concentrations offluorescent points. Thus, we considered 4800 points

CD8 cytoplasmic domain and capping 335

3-

1-

0-

-1

-0.5 0cos(A)

0.5

Fig. 6. Significance of the correlation parameter. Ten thousandcouples of fluorescent spheres with a randomly located pointfluorescence concentration were generated for determination of(A) fluorescence correlation parameters of each couple offluorescence distributions and (B) cosine of the angle betweenthe directions of fluorescent points. Results are shown for 10classes of cosine values (vertical bar length is twice thestandard deviation of correlation parameter).

uniformly spread on a sphere. Since actual averaging ofthe correlation coefficients between all couples of pointswould require unreasonably lengthy calculation, we useda Monte-Carlo-like method (Metropolis et al. 1953) tocalculate an average correlation: 10 000 random couples ofpoints on the sphere surface were generated (using theBASIC RND function of an IBM-compatible desk com-puter). Sector fluorescences were calculated in order toobtain the correlation parameter. The average relation-ships between this parameter and the cosine of the anglebetween the radius vectors of these points is shown inFig. 6.

References

ALBERT, F., BUFERNE, M., BOYER, C. AND SCHMITT-VERHULST, A. M.(1982). Interaction between MHC-encoded products and cloned T-cells.I — Fine specificity of induction of proliferation and lysis.Immunogenetics 16, 533-549.

ANDRE, P., BENOLIEL, A. M, CAPO, C, FOA, C, BUFERNE, M., BOYER, C,SCHMITT-VERHULST, A. M. AND BONGRAND, P. (1990a). Use ofconjugates made between a cytolytic T cell clone and target cells tostudy the redistribution of membrane molecules in cell contact areas.J. Cell Sci. 97, 335-347.

ANDR£, P., CAPO, C, BENOLIEL, A. M, BUFERNE, M. AND BONGRAND, P.(19906). Analysis of the topological changes induced on cells exposedto adhesive or mechanical stimuli. Cell Biophys. 16, 13-34.

BARAK, L. S., YOCUM, R. R., NOTHNAGEL, E. A. AND WEBB, W. W. (1980).Fluorescence staining of the actin cytoskeleton in living cells with 7-nitrobenz-2-oxa 1,3-diazole phallacidin. Proc. natn. Acad. Sci. U.S.A.77, 980-984.

BARISAS, G. (1984). Photobleaching recovery studies of the mobility ofpolymeric antigens on B cell surface. In Cell Surface Dynamics (eds A.S. Perelson, C. DeLisi and F. W. Wiegel), Marcel Dekker Inc., NewYork. 167-202.

BELL, G. I., DEMBO, M. AND BONGRAND, P. (1984). Cell-Cell adhesion:competition between nonspecific repulsion and specific bonding.Biophys. J. 45, 1051-1064.

BERG, H. C. AND PURCELL, E. M. (1977). Physics of Chemoreception.Biophys. J. 20, 193-219.

BIERER, B. E., HERMANN, S. H., BROWN, C. S., BURAKOFF, S. J. ANDGOLAN, D. E. (1987). Lateral mobility of class I histocompatibilityantigens in B lymphoblastoid cell membranes: modulation by cross-linking and effect of cell density. J. Cell Biol. 105, 1145-1152.

BOURGUIGNON, L. Y. W. AND BOURGUIGNON, G. J. (1984). Capping andthe cytoskeleton. Int. Rev. Cytol. 87, 195-224.

BOURGUIGNON, L. Y. W. AND SINGER, S. J. (1977). Transmembraneinteractions and the mechanism of capping of surface receptors bytheir specific ligands. Proc. natn. Acad. Sci. U.S.A. 74, 5031-5035.

BOYER, C, AUPHAN, N., GABERT, J., BLANC, D., MALISSEN, B. ANDSCHMITT-VERHULST, A. M. (1989). Comparison of phosphorylation andinternalization of the antigen receptor/CD3 complex, CD8 and class IMHC-encoded proteins on T cells. J. Immun. 143, 1905-1914.

BURN, P. A., KUPFER, A. AND SINGER, S. J. (1988). Dynamic membrane-cytoskeletal interactions: specific association of integrin and talinarises in vivo after phorbol ester treatment of peripheral bloodlymphocytes. Proc. natn. Acad. Sci. U.S.A. 85, 497-501.

CAPPS, G. G., VAN KAMPEN, M., WARD, C. L. AND ZUNIGA, M. L. (1989).Endocytosis of the class I histocompatibility antigen via a phorbolmyristate acetate inducible pathway is a cell-specific phenomenon andrequires the cytoplasmic domain. J. Cell Biol. 108, 1317-1329.

COOPER, J. A. (1987). Effect of cytochalasin and phalloidin on actin.J. Cell Biol. 105, 1473-1478.

DETMERS, P. A., WRIGHT, S. D., OLSEN, E., KIMBALL, B. AND COHN, Z. A.(1987). Aggregation of complement receptors on human neutrophils inthe absence of ligand. J. Cell Biol. 105, 1137-1145.

DRAGSTEN, P., HENKART, P., BLUMENTHAL, R., WEINSTEIN, J. ANDSCHLESSINGER, J. (1979). Lateral diffusion of surface immunoglobulin,Thy-1 antigen and a lipid probe in lymphocyte plasma membrane.Proc. natn. Acad. Sci. U.S.A. 76, 5163-5167.

EDIDIN, M. AND ZUNIGA, M. (1984). Lateral diffusion of wild-type andmutant Ld antigens in L cells. J. Cell Biol. 99, 2333-2335.

FLANAGAN, I. AND KOCH, G. L. E. (1978). Cross-linked surface Igattaches to actin. Nature 273, 278-281.

GOLSTEIN, P., GORIDIS, C, SCHMITT-VERHULST, A. M., HAYOT, B.,PIERRES, A., VAN AGHTOVEN, A., KAUFMANN, Y., ESHAR, Z. ANDPIERRES, M. (1982). Lymphoid cell surface interaction structuresdetected using cytolysis-inhibiting monoclonal antibodies. Immunol.Rev. 68, 5-42.

HUA, C, BOYER, C, BUFERNE, M. AND SCHMITT-VERHULST, A. M. (1986).Monoclonal antibodies against an H-2Kb-specific cytotoxic T cell clonedetect several clone-specific molecules. J. Immun. 136, 1937-1944.

KAMMER, G. E., WALTER, E. AND MEDOF, M. (1988). Association ofcytoskeletal reorganization with capping of the complement decayaccelerating factor on T lymphocytes. J. Immun. 141, 2924-2928.

KOPPEL, D. E., SHEETZ, M. P. AND SCHINDLER, M. (1981). Matrix controlof protein diffusion in biological membranes. Proc. natn. Acad. Sci.U.S.A. 78, 3576-3580.

KUPFER, A. AND SINGER, S. J. (1988). Molecular dynamics in themembranes of helper T cells. Proc. natn. Acad. Sci. U.S.A. 85,8216-8220.

KUPFER, A. AND SINGER, S. J. (1989). Cell biology of cytotoxic and helperT cell functions: immunofluorescence microscopic studies of singlecells and cell couples. A. Rev. Immunol. 7, 309-339.

LETOURNEUR, F., GABERT, J., COSSON, P., BLANC, D., DAVOUST, J. ANDMALISSEN, B. (1990). A signaling role for the CD8tv cytoplasmicsegment revealed under limiting stimulatory conditions. Proc. natn.Acad. Sci. U.S.A. 87, 2339-2343.

LIVNEH, E., BENVENISTE, M., PRYWES, R., FELDER, S., KAM, Z. ANDSCHLESSINGER, J. (1986). Large deletions in the cytoplasmic kinasedomain of the epidermal growth factor receptor do not affect itslateral mobility. J. Cell Biol. 103, 327-331.

MARRACK, P., SHIMONKEWITZ, R., HANNUM, R., HASKINS, C. ANDKAPPLER, J. (1983). The major histocompatibility complex-restrictedantigen receptor on T cells. IV - An anti-idiotypic antibody predictsboth antigen and I specificity. J. exp. Med. 158, 1635-1646.

METROPOLIS, M., ROSENBLUTH, M. N., TELLER, A. H. AND TELLER, E.(1953). Equation of state calculation by fast computing machines. J.Chem. Phys. 21, 1087-1092.

NAQUET, P., MARCHETTO, S. AND PIERRES, M. (1983). Dissection of thepoly (Glu60Ala30Tyr10)(GAT)-specinc T cell repertoire in H-2Ik mice. II- The use of monoclonal antibodies to study the recognition of laantigens by GAT-reactive T-cell clones. Immunogenetics 18, 559-574.

Poo, M. M. AND MCCLOSKEY, M. A. (1986). Contact-inducedredistribution of specific membrane components: local accumulationand development of adhesion. J. Cell Biol. 102, 2185-2196.

Poo, M. M., Poo, W. J. H. AND LAM, J. W. (1978). Lateralelectrophoresis and diffusion of concanavalin A receptors. J. Cell Biol.76, 483-501.

RUTISHAUSER, U. AND SACHS, L. (1974). Receptor mobility and themechanism of cell-cell binding induced by concanavalin A. Proc.natn. Acad. Sci. U.S.A. 71, 2456-2460.

SCHLESSINGER, J., KOPPEL, D. E., AXELROD, D., JACOBSON, K., WEBB, W.W. AND ELSON, E. L. (1976). Lateral transport on cell membranes:

336 P. Andre et al.

mobility of concanavalin A receptors on myoblasts. Proc. natn. Acad.Sci. U.S.A. 73, 2409-2413.

SCULLION, B. F., HOU, Y., PUDDINOTON, L., ROSE, J. K. AND JACOBSON, K.(1987). Effect of mutations in three domains of the vesicularstomatitis viral glycoprotein on its lateral diffusion in the plasmamembrane. J. Cell Biol. 105, 69-75.

SINGER, S. J. (1975). The fluid mosaic model of membrane structures:some applications to ligand-receptor and cell-cell interactions. InSurface Membrane Receptors (eds R. A. Bradshaw, W. A. Frazier, R.C. Merrell, D. I. Gottlieb and R. A. Hogue-Angeletti), Plenum Press.New York. 1-20.

SINGER, S. J. AND NICOLSON, G. L. (1972). The fluid mosaic model of thestructure of cell membranes. Science 175, 720-731.

SMITH, B. A., CLARK, W. R. AND MCCONNELL, H. M. (1979). Anisotropicmolecular motion on cell surfaces. Proc. natn. Acad. Sci. U.S.A. 76,5641-5644.

SNEDECOR, W. W. AND COCHRAN, W. G. (1980). Statistical Methods. IowaState University Press, Ames. pp. 186-187.

TANK, D. W., WU, E. S. AND WEBB, W. W. (1982). Enhanced moleculardiffusibility in muscle membrane blebs: release of lateral constraints.J. Cell Biol. 92, 207-212.

TAYLOR, R. B., DUFFUS, W. P. H., RAFF, M. C. AND DEPETRIS, S. (1971).Redistribution and pinocytosis of lymphocyte surfaceimmunoglobuline molecules induced by anti-immunoglobulinantibodies. Nature New Biol. 233, 225-229.

WADE, W. F., FREED, J. H. AND EDIDIN, M. (1989). Translationaldiffusion of class II major histocompatibility complex molecules isconstrained by their cytoplasmic domains. J. Cell Biol. 109,3325-3331.

WIEGEL, F. W. (1984). Diffusion of proteins in membranes. In CellSurface Dynamics (ed. A. S. Perelson, C. DeLisi and F. W. Wiegel),Marcel Dekker, New York. 135-150.

WIER, M. AND EDIDIN, M. (1986). Effects of cell density and extracellularmatrix on the lateral diffusion of major histocompatibility antigens incultured flbroblasts. J. Cell Biol. 103, 215-222.

ZAMOYSKA, R. P., DERHAM, P., GORMAN, S. D., VON HOEGEN, P., BOLEN,J. B., VEILLETTE, A. AND PARNES, J. R. (1989). Inability of CD8a?polypeptides to associate with p56 lck correlates with impairedfunction in vitro and lack of expression in vivo. Nature 342, 278-281.

(Received 4 March 1991 - Accepted, in revised form, 1 July 1991)

CD8 cytoplasmic domain and capping 337