Structure-function studies of CD2 by nmr and ... - Semantic Scholar

Immunoglobulin Superfamily Interactions

~~

~

Structure-function studies of CD2 by n.m.r. and mutagenesis Paul C. Driscoll,*t Jason G. Cyster,$ Chamorro Somoza,$ D. Arthur Crawford,t Peter Howe,t Timothy S. Harvey,t Bruno Kieffer,t lain D . Campbellt and Alan F. Williams$ tDepartment of Biochemistry, University of Oxford, South Parks Road, Oxford 0x1 3QU, U.K., and $MRC Cellular Immunology Unit, Sir William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford 0x1 3RE, U.K.

Introduction One of the more familiar examples of cell-cell recognition is that between T cells and an antigenpresenting cell, which consists of a complex array of interprotein interactions that are just beginning to be unravelled [ 1, 21. These interactions promote and support the recognition of foreign antigenic peptides bound to major histocompatibility proteins by specific T-cell receptor molecules. One component of this interface is the adhesion of the T-cell antigen CD2 to lymphocyte function-associated antigen-3 (LFA-3) (in the case of humans) [3, 41 or to CD48 (in the case of rodents) [S, 61 on the presenting cell. Human CD48 may also be a ligand for human CD2, although with lower affinity than LFA-3 [7]. The extracellular portions of LFA-3, CD2 and CD48 are all predicted to have similar structures [8]. Monoclonal antibodies to CD2 or to LFA-3 can block conjugate formation [9], and certain pairs of anti-CD2 mAbs can effect an antigenindependent stimulation of the T cell, implicating CD2 in a signal-transduction role [lo-121. The adhesion role of CD2 has been shown to be mediated by the first N-terminal domain [ 131. Hefore our structural work on the rat T-cell CD2 antigen, there were conflicting reports in the literature concerning the assignment of the protein to the immunoglobulin superfamily (IgSF) of cellsurface molecules [ 141. Williams and colleagues proposed, on the basis of sequence pattern analysis, that the extracellular portion of the CD2 antigen consisted of two domains with limited similarity to proteins with presumed immunoglobulin-like folds [8, 151. In contrast, an alternative analysis, based on theoretical structure prediction and on limited biophysical studies (including circular dichroism) of the first domain of human 0 2 , led to the proposal Abbreviations used: IXA-3, lymphocyte function-associated antigen-.?; KMSD, root-mean-square difference; IgSF, immunoglobulin superfamily; rCD2.D 1, rat CD2 domain 1; NOESY. nuclear Overhauser effect spectroscopy; HOHAHA, homonuclear Hartmann-Hahn; HSQC, heteronuclear single quantum coherence; HMQC, heteronuclear multiple quantum coherence.

'To whom correspondence should be addressed.

that CD2 is a member of a class of proteins containing both a-helical and /?-sheet secondary structure [ 16, 171.

Three-dimensional solution structure of rat CD2 domain I W e have determined the solution structure of the first domain of the T-cell antigen CD2 from rats (rCD2.D 1) using isotope labelling and multidimensional heteronuclear n.m.r. spectroscopy [ 181. A construct consisting of glutathione S-transferase [19] fused to a thrombin cleavage site (Leu-ValPro-Arg-Gly-Ser) and to residues 1-99 of rat CD2 was expressed as a soluble fusion protein in Escherichia coli at a level of around 40 mg/l. The fusion protein product was isolated from lysed cells by affinity chromatography on a glutathione-agarose column. After cleavage with thrombin, the rCD2.D 1 protein was obtained by affinity chromatography and gel filtration. The rCD2.Dl protein (consisting of residues 1-99 with an additional N-terminal GlySer dipeptide from the thrombin cleavage site) bound to two non-competitive monoclonal antibodies (MRC OX34 and OX55 [ l l ] ) with similar affinity to that of the extracellular domain of rat CD2 expressed in Chinese-hamster-ovary cells, indicating that the rCD2.Dl protein was correctly folded. The protein was also produced by expression in bacteria on a minimal medium with "Nlabelled NH,CI as the sole nitrogen source, yielding material suitable for two- and three-dimensional 15N-'H heteronuclear n.m.r. spectroscopy. An additional sample was also prepared, again with the bacteria on a minimal medium, but with 10% of the supplied glucose provided in a fully C6-'jc-labelled form. The biosynthetic processing of this substrate leads to stereospecific carbon-labelling patterns in the leucine and valine residues of the protein [20, 211. The labelling pattern allows stereospecific assignments of the methyl groups of leucine and valine residues to be made, leading to an improvement in the quality of the n.m.r. structures obtained. The pattern of labelling also permits the application of I3C relaxation measurements to investigate the side-chain dynamics in the folded protein (B. Kieffer and P. C. Driscoll, unpublished work).

I993

947

Biochemical Society Transactions

Sequence-specific assignments of the amide

NII region of 'H n.m.r. spectrum of rCD2.Dl were

948

obtained through the analysis of three-dimensional "N-'H nuclear Overhauser effect spectroscopy (N0ESY)-heteronuclear multiple quanta coherence (HMQC) and homonuclear Hartmann-Hahn (I 10HAHA)-HMQC n.m.r. spectra [22] recorded on the "N-labelled protein sample. Side-chain resonance assignments were completed by the analysis of two-dimensional 'H HOHAHA, NOESY and double quantum filtered correlated spectra of an unlabelled sample of rCD2.Dl in 'H,O solution [23, 241. Stereospecific assignments of the methyl-group resonances of valine and leucine residues were obtained by analysis of twodimensional 'C-'H heteronuclear spectra [heteronuclear single quantum coherence (HSQC), HSQC-HOHAHA and HSQC-NOESY] of the sample from bacteria grown on the 10% C,-"Cglucose-containing medium. Examination of the two-dimensional and three-dimensional NOESY spectra yielded a total of 728 NOE distance spin-spin couplrestraints. Measurements of ing constants allowed for the incorporation of (4 angle dihedral torsion restraints for 64 residues and NI I/NL€Iexchange experiments identified 34 backbone amide NI1 groups involved in hydrogen bonds.

'

'rHha

The three-dimensional solution structure of rCI12.Dl was calculated on the basis of the NOEderived distance restraints, H bond distance restraints and angle torsion-angle restraints. The protocol adopted was dynamical simulated annealing starting from random-coil configurations [ 25, 261, implemented in the program X-PLOR [27). Figure 1 shows a stereo picture of a representative selection of 21 rCD2.Dl structures selected from a set of 85 converged structures derived from different random conformations. The structures are bestfit to the backbone atoms of residues 3-98 and the pairwise root-mean-square difference (RMSD) between the co-ordinates of the backbone atoms (N, C", C, 0) of these residues is 0.073 nm (0.73 A). The Gly-Ser dipeptide leader, residues 1-2 and residue 09 are relatively ill-defined by the n.m.r. data, and their positions are ill-determined, as shown by the fraying of the N- and C-termini of the structures in Figure 1. Nuclear relaxation measurements on 'N-labelled rCI 12.1) 1 have shown that there is little evidence of internal mobility of the polypeptide backbone except for the Gly-Ser dipeptide leader and residues 1-5 at the N-terminus, and residue 99 at the C-terminus (I). A. Crawford, €3. Kieffer and P. C. I)riscoll, unpublished work). The definition of the polypeptide fold of the family of structures in Figure 1 represents a

Figure I

Results of the calculation of the three-dimensional solution structure of rat

CD2 domain I from n.m.r.-derived restraints. A stereo picture of the bestfit superposition (residues 5-98) of a family of 21 backbone folds from independent calculations starting from different random coil conformations Note the lack of definition of the structure for a stretch of residues a t the N-terminus leading into p-strand A and the C-terminal residue at the tail of /%strand G.

Volume 2 I

Immunoglobulin Superfamily Interactions

considerable improvement over the structures shown originally in [ 181 for which the corresponding pairwise backbone RMSI) figure was 0.133 nm (1.33 A). The improvement in the structures is derived in part from a rescaling of the distance restraints used previously, and in part from the incorporation of stereospecific assignments of leucine and valine methyl groups, due to the elimination of the requirement for several large pseudoatom correction factors. The structure of rCD2.I)l obtained from the n.m.r. solution structure calculations is clearly closely related to the V-type immunoglobulin fold that is seen for the hypervariable-loop-containing variable domains of immunoglobulins [28], the first domain of CD4 [29, 301, and the extracellular globular portion of CIIX [17]. Thus the fold consists of a major anti-parallel P-sheet consisting of strands G, F, C and C' with strand A lying in a parallel fashion t o strand G, stacked upon a smaller threestranded anti-parallel P-sheet (strands D, E and H), as shown in schematic form in Figure 2(a). The C" loop runs anti-parallel to strand C' and displays only limited characteristics of a regular anti-parallel P-sheet structure. The backbone core B-sheet region of the rCD2.D 1 structure superimposes very closely the core regions of a number of other IgSF V-type domains [18] with RMSD values for the superposition of 28 C" atoms that are typically in the range 0.08-0.15 nm (0.8-1.5 A). This correspondence in the core structures occurs in spite of the absence from the rCIl2.1) 1 structure of the consensus disulphide bond that is invariably formed between cysteine residues in strands 13 and F of the immunoglobulin fold. For rCD2.D 1, the consensus cysteine residues are substituted by Ile-18 in /3strand €3 and Val-78 in /?-strand F. The side chains of these two residues form close hydrophobic van der Waals contacts and their C" atoms are separated by approx. 0.7 nm (7.0 A), within the range that is seen for the Cys C"-Cys C" distances for IgSF domains with the consensus disulphide bond [31]. In contrast to the similarity of the core structures, the loop configurations of rCD2.D 1 differ significantly from those observed for other members of the IgSF. These differences are due to the different numbers of amino acids in the loops in the case of rat CD2. Thus the 1)-E loop in rCD2 is very short, but this appears to be compensated by a longer €3-C loop, which adopts a rather convoluted structure with bulky methionine and phenylalanine side chains (Phe-21 and Met-23) filling the intersheet void where the D-E loop is truncated. The &strand is cut short in rCD2.Dl compared with,

Figure 2 ( 0 ) Schematic representation of the secondary structure of the rat CD2 domain I structure. ( b ) Spacefilling representation of the computer model of human CD2 domain I based on the r a t CD2

structure In ((I) the domain i s oriented with the major GFCC' /--sheet face towards the front, with the smaller sheet consisting of p strands B, E and D at the rear In ( b ) the molecule i s oriented in the same direction as in ( 0 ) The shading indicates the effects of mutations on the neuraminidase-treated erythrocyte-rosetting function of COS-cell-expressed human CD2 constructs Those residues for which non-conservative mutations block the rosetting are shaded dark grey and are labelled, while those for which a non-conservative mutation had no effect on rosetting are shaded light grey Both figures were created using the program MOLSCRIPT [38]

(4

(b)

D87

G90

K49

Wfor example, human CD4 domain 1 [29, 301 because of a kink introduced by a proline residue (Pro-19). The only other IgSF molecule for which a proline residue is found at this position (that is, at the residue position immediately following the con-

I993

949

Biochemical Society Transactions

950

sensus position for the B-strand cysteine residue) is human LFA-3, the adhesion partner of CD2. Domain 1 of LFA-3 also lacks the consensus disulphide bond. Whilst there is no obvious sequence similarity between the sequences of rat CD2, human CD2 and human LFA-3 (beyond those residue patterns that allow it to be assigned to the IgSF class) it is intriguing to speculate that these structural similarities are related to an evolutionary path in which the heterotypic adhesion displayed between CD2 and LFA-3 is derived from prehistoric homotypic adhesion of a single ancestral IgSF molecule [ 321.

Investigating the ligand-binding surface of CD2 The determination of the three-dimensional fold of rCD2.D 1 resolves the contradition between predictions of the structure [13, 15, 161 in favour of the method based on sequence pattern matching. The structure also allows the assignment of residues that are known to be important for the LFA-3 adhesion function in the case of human CD2 to the major p-sheet face of the domain. Thus several point mutations of human CD2 that lead to the elimination of the erythrocyte-rosetting capability of CD2transfected COS cells have been identified previously [33-351. The side chains of these residues correspond to side chains in the homologous rat CD2 structure that are on the external surface of P-strands C’ and F, as well as the F-G loop. Before the determination of the rCD2.Dl structure, it had been expected that these residues would be in positions corresponding to the hypervariable loops of the putative IgSF fold [33]. The interpretation of the mutations in terms of the rCD2.Dl structure presented here strongly implicates the face of the psheet consisting of strands C, C‘, F and G, and prompted us to initiate a program of targeted mutagenesis to explore further the adhesion surface of CD2 [36]. In contrast to human CD2, rat CD2 does not cause rosetting of either sheep erythrocytes or neuramidase-treated human erythrocytes when transfected into COS cells. Thus, whereas human CD2 can adhere to either human or sheep LFA-3, rat CD2 shows no affinity for either form of LFA-3. Indeed the ligand for rat CD2 has now been identified as rat CD48 [6], which has a human homologue that is distinct from LFA-3. The strategy of mutagenesis was to substitute residues from rat CD2 into the corresponding positions in human CD2, and then to search for human CD2 mutants that failed to support the rosetting of neuramidase-

Volume 21

treated human erythrocytes when transfected into COS cells. A panel of 21 human CD2 mutants was constructed, including mutants at locations in both /?-sheet surfaces and the loops of domain 1 of human CD2 [36]. The results of the targeted mutagenesis study are shown in Table 1. Non-conservative mutations in the C-C’ and C” loops, and P-strand C”, as well as in p-strands I? and D on the opposite P-sheet surface, have no effect on the rosetting capability of human CD2, and are therefore not expected to be involved in contacts with the CD58 molecule. In contrast, the K34D, E36R, R48K/K49R and K91T/N92R mutants, which involve substitutions of residues on the surface of the F-G loop and of p-strands C and C‘, lead to blocking of the rosetting function of human CD2. The K82N mutation (P-strand F) also stops rosetting, but in this case the mutation leads to the introduction of a consensus N-linked glycosylation site (K82V83S84 becomes N82V83S84). The double mutant K82N/S84Q, which eliminates the glycosylation site through second-site mutation forms erythrocyte rosettes normally. This result strongly suggests that for the K82N mutant, the attachment of a bulky oligosaccharide to the asparagine residue leads to steric occlusion of the adhesion surface of human CD2. A computer model of domain 1 of human CD2 was built using the Protein Design facility of the program QUANTA (Molecular Simulations Inc.) starting from the average co-ordinates of the family of n.m.r. rCD2.Dl structures [ 181. The only insertion in the backbone that was required was for two extra residues in the C-C’ loop of human CD2. These were included by searching a library of loop structures derived from high-resolution protein X-ray crystal structures. Other residues with different side chains were substituted with maintenance of side-chain torsion angles where appropriate, and the model was subjected to two rounds of molecular dynamics and energy minimization to remove bad van der Waals contacts. Figure 2( b ) shows a space-filling illustration of the surface of human CD2 that is implicated in adhesion to LFA-3 by the results of both our mutagenesis studies and those reported previously [33-361. Using the computer model of domain 1 of human CD2, the side chains of residues that knock out the adhesion to LFA-3 when mutated are shown with dark shading. Residues for which mutation had no effect on erythrocyte rosetting are shown with light shading. The space-filling model is shown in the same orientation as the ribbon diagram of the rCD2.Dl structure that is shown in

Immunoglobulin Superfamily Interactions

Table I

Erythrocyte-rosetting ability of human CD2 mutants

~361 All substitutions are exchanges of the named residue in the human CD2 sequence for the residue in the corresponding position in the rat CD2 structure An asterisk indicates mutations selected t o produce non-conservative substitutions of amino acids in human CD2 AK41K42 denotes that residues K41 and K42 were deleted from the human CD2 construct, which shortens the C-C' loop to the same length as the loop in rat CD2 See the text for a description of the K82N mutant Location of mutant Erythrocyte Mutant

N 18TID2ON D32E K34R K37N38G AK4 I K42 E50WK5 I MiE52K F54L* K55UE56WE57K K61E 180T S84T S84Q* K82N/S84Q K89N E95D 197A

K34D* E36R* R48WK49R K82N K9 I T/N92R

Strand

Loop

B

rosetting Yes Yes

C C

Yes

C-C' C-C' Cf-C"

Yes Yes Yes

C"

Yes

C" D F F

Yes Yes

Yes Yes

F

Yes Yes

F F-G

Yes Yes

G G

Yes

C C C' F F-G

No No No No NO

Figure 2(a). It is clear that the surface of the domain that is involved in the adhesion of LFA-3 consists of the upper parts of B-strands C, C', F and G and the F-G loop. Recently, the X-ray crystal structure of the complete extracellular portion of rat CD2 consisting of domains 1 and 2 has been determined [37]. The structure of the first domain is essentially identical to that determined by n.m.r. spectroscopy. Interestingly, in the crystal packing the rat CD2 molecules form a extensive dimer contact involving the side chains of the major GFCC' B-sheet surface as well as the main-chain backbone atoms of the C-C' and F-G loops. Whilst the details of this intermolecular

interface are not entirely consistent with the results of our mutagenesis study of human CD2, it provides an enticing model for the rat CD2/CD48 and human CD2/LFA-3 complexes. N.m.r. experiments are in progress to demonstrate the effect of binding of the extracellular portion of rat CD48 to lSNlabelled rCDZ.Dl, with the aim of investigating the adhesion interaction in solution directly.

This work is a contribution from the Oxford Centre for Molecular Sciences, which is funded by the SEKC and MRC. P.C.D. is supported by a Royal Society Ciniversity Research Fellowship.

1. Springer, T. A. (1990) Nature (London) 346, 425-434 2. Rarclay, A. N., Rirkeland, M. I,., Brown, M. H., Reyers, A. D., Davis, S. J., Somoza, C. S. and Williams, A. F. (1993) The Leucocyte Antigen Factsbook, Academic Press, 1,ondon 3. Hunig, T. (1985) J. Exp. Med. 162,890-901 4. Selvaraj, P., Plunkett, M. I,., Dustin, M., Sanders. M. E., Shaw, S. and Springer, T. A. (1987) Nature (London) 326, 400-403 5. Kato, K., Koyanaga, M., Okada, H., Takanashi, T., Wong, Y. W., Williams. A. F., Okuniura, K. and Yagita, H. (1992) J. Exp. Med. 176, 1241-1249 6. van der Merwe, P. A,, Mcl'herson, D. C., Brown, M. H., Rarclay, A. N., Cyster, J. G., Williams, A. F. and Davis, S. J. (1993) Eur. J. Immunol. 23, 1373-1377 7. Arulanandam, A. K. N., Moingeon, I]., Concino, M. F., Recny, M. A,, Kato, K., Yagita, H., Koyasu, S. and Reinherz, E. I,. (1993) J. Exp. Med. 177, 1439-1450 8. Killeen, N., Moessner, K., Arvieux, J., Willis, A. and Williams, A. F. (1988) EMRO J. 7, 3087-309 1 9. Dustin, M. I,. and Springer, T. A. (1991) Annu. Rev. Immunol. 9 , 2 7 4 5 10. Meuer, S. C., Hussey, K. E., Fabbi, M., Fox. D., Acuto, K. A,, Fitzgerald, J. C., Hodgdon, J. P., Protentis, J. P., Schlossman, S. F.and Keinherz, E. I,. (1984) Cell 36,897-906 11. Heyers, A. D., Harclay, A. N.. Law, D. A,, He. Q. and Williams, A. F. (1989) Immunol. Rev. 111, 59-77 12. Spruyt, I,. I,., Glennie, M. J., Heyers, A. I). and Williams, A. F. (199 1) J. Exp. Med. 174, 1407- 14 15 13. Recny, M. A., Neidhart, E. A., Sayre, P. H., Ciardelli, T. I,. and Keinherz, E. I,. (1990) J. Riol. Chem. 265, 8542-8549 14. Williams, A. F. and Harclay, A. N. (1988) Annu. Rev. Immunol. 6,38 1-405 15. Williams, A. F., Rarclay, A. N., Clark, S. J., Paterson, D. J. and Willis, A. C. (1987) J. Exp. Med. 165, 368-380

I993

95 1

Biochemical Society Transactions

952

10 Clayton, I,. K., Sayre, I? H., Novotny. J. and Keinherz, E. I,.(1987) Eur. J. Immunol. 17, 1367-1 370 17. I,eahy, 1). J., Axel, K. and Hendrickson, W. A. (1992) Cell 68, 1145-1 168 18. L)riscoll, 1’. C., Cyster, J. G., Campbell, I. L). and Williams, A. F. (1091) Nature (London) 353, 762-705 19. Smith, 1). H. and Johnson, K. S. (1988) Gene 67,

3 1-40 20. Senn, H., Werner, H., Messerle, H. A., Weber, C., Traber, K. and Wiithrich. K. (1 989) FEHS Lett. 249, 113-118 21. Neri, L)., Szyperski, T., Otting, G., Senn, H. and Wuthrich, K. ( I Y X Y ) Biochemistry 28, 75 10-75 16 22. I)riscoll, P. C., Clore, G. M., Marion, L)., Wingfield, 1’. T. and Gronenborn, A. M. (1990) Biochemistry 29, 3542-3550 23. Ernst, K. K., Hodenhausen, G. and Wokaun, A. (19x7) Principles of Nuclear Magnetic Resonance in One and T w o Dimensions, Clarendon Press, Oxford 24. Wuthrich, K. (1970) NMK of Proteins and Nucleic Acids, Wiley, New York 25. Nilges, M., Kuszewski, J. and Hrunger, A. T. (1Y9 1) in Computational Aspects of the Study of Biological Macromolecules by Nuclear Magnetic Kesonance Spectroscopy (Hoch, J. C., I’oulsen, F. M. and Redfield, C., eds.), pp. 451-455, I’lenum I’ress, New York 20. Clore. G. M. and Gronenborn, A. M. (1089) Crit. Rev. I3iochem. Mol. Hiol. 24, 479-564 27. Hrunger, A. T. (1990) X-PLOK Manual, Version 2.1, Yale llniversity I’ress New Haven

28. Amzel, I,. M. and I’oljak. K. J. (1979) Annu. Rev. Hiochem. 48,961-967 29. Wang, J., Yan, Y., Garrett, T. PJ., Liu, J.? Kodgers, L). W., Garlick, K. I,., Tarr, G. E., Husain, Y., Keinherz, E. I,. and Harrison, S. C. (1990) Nature (London) 348,4 11-4 18 30. Kyu, S.-E., Kwong, P. D., Truneh, A,, Porter, T. G., Arthos, J., Kosenberg, M., Dai, X., Xuong, N., Axel, K., Sweet, K. W . and Hendrickson, W . A. (1000) Nature (London) 348,419-426 31. Richardson, J. S. (1981) Adv. Protein Chem. 34, 167-339 32. Springer, T. A. (1991) Nature (London) 353, 704-705 33. Peterson, A. and Seed, H. (1981) Nature (London) 329,842-846 34. Hierer, H., Peterson, A,, Harbosa, J., Seed. H. and Hurakoff, S.J. ( 1 988) I’roc. Natl. Acad. Sci. U.S.A. 85, 1194-1198 35. Wolff, H. I,., Burakoff, S. J. and Hierer, H. E. (1990) J. Immunol. 144,1215-1220 36. Somoza, C. S., Driscoll, P. C., Cyster, J. G. and Williams, A. F.(1993) J. Exp. Med. 178, 549-558 37. Jones. E. Y., Davis, S. J.? Williams, A. F., Harlos, K. and Stuart, D. I. (1992) Nature (London) 360, 232-239 38. Kraulis, 1’. J. (199 1) J. Appl. Crystallogr. 24,94h-Y50

Keceived 26 July 1993

Analysis of the structure and interactions of CD2 Simon J. Davis,” E. Yvonne Jones,tf Dale L. Bodian,t A. Neil Barclay* and P. Anton van der Merwe* *MRC Cellular Immunology Unit, Sir William Dunn School of Pathology, University of Oxford, Oxford 0x1 3RE, U.K.,+Laboratory of Molecular Biophysics, The Rex Richards Building, South Parks Road, Oxford OX I 3QU, U.K., and $Oxford Centre for Molecular Sciences, University of Oxford, Oxford OX I 3QU. U.K.

The cell adhesion molecule CD2 is a member of the immunoglobulin superfamily (IgSF) of proteins that is expressed on T lymphocytes and natural killer cells of the mammalian immune system. A ligand for human CD2 is the widely expressed CD58 (lymphocyte function-associated antigen-3, 1,FA-3) glycoprotein [ I , 21, and early studies with monoclonal antibodies showed that this interaction contributes to the adhesion of T lymphocytes to target and accessory cells [3, 41. In mice and rats, CD2 has been shown to bind to CD48. and in mice it Abbreviations used: IgSF, immunoglobulin superfamily; I,FA-.3? lymphocyte function-associated antigen-3; sCL12, soluble form of CD2; sc‘1>48-CD4. soluble chimeric form of rat CD4X.

Volume 2 I

appears that this is the principal ligand for CD2 [ 5, 61. Human CD2 also binds weakly to CD48, but the physiological significance of this interaction has not yet been determined [7]. The sequences of CD2, CD48 and CD58 are closely related [ X I and this, together with the genomic organization of the genes encoding this family of molecules, implies that the genes arose by duplication from a common precursor that was involved in homophilic interactions

~91. The physical basis of interactions between cell adhesion molecules is poorly understood. However, recent progress in studies of the structure and interactions of CD2 indicate that this system will be a useful model for understanding these processes. The initial data suggest that binding is not mediated